research paper on yellow fever

• Open access
• Published: 10 January 2020

Yellow fever control: current epidemiology and vaccination strategies

• Lin H. Chen   ORCID: orcid.org/0000-0002-5684-8436 1 , 2 &
• Mary E. Wilson 3 , 4  

Tropical Diseases, Travel Medicine and Vaccines volume  6 , Article number:  1 ( 2020 ) Cite this article

14k Accesses

60 Citations

17 Altmetric

Metrics details

Yellow fever (YF) outbreaks continue, have expanded into new areas and threaten large populations in South America and Africa. Predicting where epidemics might occur must take into account local mosquito populations and specific YF virus strain, as well as ecoclimatic conditions, sociopolitical and demographic factors including population size, density, and mobility, and vaccine coverage. Populations of Aedes aegypti and Aedes albopictus from different regions vary in susceptibility to and capacity to transmit YF virus. YF virus cannot be eliminated today because the virus circulates in animal reservoirs, but human disease could be eliminated with wide use of the vaccine. WHO EYE (Eliminate Yellow Fever Epidemics) is a welcome plan to control YF, with strategies to be carried out from 2017 to 2026: to expand use of YF vaccine, to prevent international spread, and to contain outbreaks rapidly. YF vaccination is the mainstay in controlling YF outbreaks, but global supply is insufficient. Therefore, dose-sparing strategies have been proposed including fractional dosing and intradermal administration. Fractional dosing has been effectively used in outbreak control but currently does not satisfy International Health Regulations; special documentation is needed for international travel. Vector control is another facet in preventing YF outbreaks, and novel methods are being considered and proposed.

Introduction

Yellow fever (YF) is a mosquito-borne flavivirus that causes outbreaks with high fatality. In the early 1900’s, the Yellow Fever Commission identified mosquitoes as vectors for YF, and mosquito control programs ensued that curtailed YF disease. Today, 47 countries in Africa and Central and South America are considered endemic for YF, and the WHO estimates an annual burden of 200,000 severe cases of YF and up to 60,000 deaths [ 1 ]. The main vectors are Aedes species, but Haemagogus and Sabethes are important forest species in South America, and non-human primates are the reservoir.

In 2015 and 2016, Angola and Democratic Republic of the Congo (DRC) experienced large outbreaks, followed by Brazil and Nigeria in 2017 and 2018. Brazil has experienced increased YF outbreaks because an ongoing YF epizootic has expanded endemic zones to areas near the megacities of Rio de Janeiro and Sao Paulo [ 2 ]. Alarmingly, unvaccinated travelers visiting endemic areas have acquired YF and died from YF in higher numbers since 2015 compared to the previous several decades [ 3 ].

Infection with YF virus can manifest with fever, nausea, vomiting, and abdominal pain. The symptoms may progress in 20% to jaundice, hepatic and renal failure, and bleeding. The case fatality rate from symptomatic YF can reach 50% [ 4 , 5 ]. The mainstay of YF control involves vector control and YF vaccination.

Recent yellow fever epidemiology

YF outbreaks continue to occur in Africa and in South America. In Africa, outbreaks affect urban and rural populations. In South America, recent human cases reflect sylvatic transmission – virus circulation among nonhuman primates and spillover into the human population, transmitted by mosquitoes that are found in forested areas (such as Haemagogus and Sabethes spp). Human infections are largely in males who enter forested areas for work or recreation. Urban transmission by Aedes aegypti , the mosquito that now infests cities throughout tropical and subtropical South America, has not been documented in recent outbreaks [ 2 ]. The last outbreak of urban YF in Brazil was in 1942.

Virus genomes [ 6 ] from the South America outbreak that was recognized in December 2016 provided convincing evidence that infection was due to forest YF with spillover into human populations, leading to > 2000 human cases and > 700 deaths in 2016–2018 in Brazil. Early sylvatic transmission was followed by spatial expansion towards previously YF-free areas. Human cases lagged behind those in nonhuman primates by about 4 weeks. This was the largest epidemic in Brazil in decades. A more recent analysis [ 7 ] added other details. Researchers analyzed YF viruses from humans, nonhuman primates, and mosquitoes across 5 Brazilian states (YF endemic and nonendemic) between 2015 and 2018 to reconstruct virus spread. The outbreak in southeastern Brazil, unrecognized until late 2016, originated from a 2014 event in Goias state. The lineage from Goias state was introduced into Minas Gerais at least twice. Virus sub-lineages spread by different routes towards densely populated regions in eastern Brazil. At the start of the epidemic, an estimated 35 million unvaccinated people lived in YF risk areas.

Globally two scenarios cause deep concern: the potential for YF virus to reach the major population centers in eastern Brazil and cause explosive urban outbreaks, and the potential for YF to spread to large densely populated urban centers in Asia. In both places, urban areas are infested with Aedes aegypti and/or Aedes albopictus. In both areas most people are unvaccinated and susceptible to infection.

Elements necessary for YF to appear in an area

It is useful to review what elements are necessary for an YF outbreak to occur in a human population. A source of the virus must be present. The virus can potentially be carried into a nonendemic area by a traveler. In many parts of South America the virus circulates in nonhuman primates, which are widely distributed including in urban parks.

A competent mosquito vector must infest an area. The topic of competence is discussed below. The mosquito must have access to a source of the virus (such as an infected human or nonhuman primate). The ecoclimatic conditions, including temperature, rainfall, and humidity, must allow the mosquito to survive long enough for the virus to disseminate in the mosquito to allow onward transmission. The extrinsic incubation period for the virus in the mosquito (the time between taking a blood meal that contains virus until the virus disseminates and can be transmitted via saliva during feeding) is highly dependent on temperature and humidity [ 8 ]. In cool areas, the mosquito may die before the virus disseminates and reaches the saliva. The mosquito must then have access to a nonimmune animal or human.

Although YF virus is usually transmitted from viremic host to mosquito to animal or human, another mechanism can maintain the virus. When a female YF virus-infected Aedes aegypti mosquito lays eggs (produced in infected ovaries), the mosquitoes that develop from the eggs may carry transmissible YF virus. Vertical transmission (or transovarial transmission) occurs with some other virus-mosquito pairs as well. Aedes aegypti eggs are desiccation resistant – they can survive dry conditions. Even if months pass between their production and the next rain, the eggs can still yield viable and infectious progeny. How big a contribution vertical transmission of virus in mosquitoes makes to the overall YF epidemiology is unclear, but it allows virus persistence in the absence of a vertebrate host. Studies by Aitken [ 9 ] documented vertical transmission of YF virus in Aedes aegypti . More recently Diallo [ 10 ] showed vertical transmission of YF virus in two colonies of Aedes aegypti from Senegal. The percent of female progeny infected reached 5.2% for those with longer extrinsic incubation periods. Researchers also observed vertical transmission outside the laboratory, finding virus in recently emerged adults from larvae collected in the field.

Virus mosquito interactions

A complicated biological process occurs within a mosquito that successfully transmits a pathogen, such as a virus, from one host to another. The mosquito must first be attracted to a specific host. Mosquitoes vary greatly in their host preferences. Some, such as Aedes aegypti , strongly prefer human blood and will ignore other sources of blood meals if they can find a human. Others, such as Aedes albopictus , are more cosmopolitan in preferences and will feed on whatever blood source is available, whether animal or human. Aedes albopictus mosquitoes have been responsible for outbreaks of human infections (including dengue and chikungunya) where humans are the main source of blood available. The capacity of Aedes albopictus to feed on many different species also means that they can potentially serve as bridge vectors and carry viruses from one species to another, for example, from animals to humans. This attribute, combined with their broader habitat in forested and park areas raises concern that they could transmit viruses usually found in animals to humans.

More than 3000 different species of mosquitoes have been identified globally and > 150 species are found in the US. Most do not transmit human pathogens. Within a species of mosquito, populations are heterogeneous. This means that Aedes aegypti mosquitoes that infest Trinidad may not be the same as Aedes aegypti mosquitoes that are found in Memphis, Tennessee. Results of laboratory studies done with mosquitoes from one area may not be generalizable to other populations of the same mosquito species.

Viruses are also heterogeneous. YF viruses, which probably originated in Africa, have evolved over time and now are clustered in seven different genotypes, 5 in Africa and 2 in South America. These viruses may differ in important characteristics, such as virulence and transmissibility by specific mosquito populations [ 11 ].

Vector competence and vectorial capacity

Vector competence refers to the ability of a vector, such as a mosquito, to acquire and transmit a pathogen, such as viruses or the malaria parasite [ 12 ]. Mosquitoes are refractory to infection by many viruses. A virus must overcome multiple barriers to be transmitted by a mosquito. The virus must infect the epithelial cells of the mosquito midgut. To do this, it must overcome the digestive enzymes in the mosquito, internal microbiota, and the physical barrier of the midgut epithelium. The virus must infect, replicate, be shed and then cross the basal lamina into the hemolymph of the mosquito. To be transmitted, the virus must traverse the basal lamina surrounding the salivary gland and infect acinar cells. From that site, virus can be inoculated into hosts at the time of blood feeding. If the salivary glands are infected, the virus persists in saliva for the life of the mosquito – allowing the mosquito to transmit the virus to more than one host. A virus may infect the midgut but be unable to disseminate or unable to infect the salivary glands. In the laboratory, mosquitoes can be allowed to feed on blood containing viruses and can then be tested at various points in time to assess presence of virus at various sites and its presence in mosquito saliva.

As an interesting aside, the YF vaccine virus (which has been used since 1937) apparently is able to infect mosquitoes but because of the midgut barrier is unable to disseminate in mosquitoes – so it cannot be transmitted by mosquitoes from person to person even though the vaccine virus produces viremia in those vaccinated [ 11 ].

It is relevant to know whether a specific vector population is competent to transmit a specific virus. It is even more useful to know the vector efficiency or vectorial capacity. Just because a particular mosquito is competent to transmit a specific virus, does not mean that it can do so efficiently. But even a mosquito that is a relatively inefficient vector, if present in large numbers may be able to sustain an outbreak. Vectorial capacity takes into account the number of mosquitoes relative to the host, the daily blood-feeding rate on the specific host (animal or human), the vector competence (the transmission rate among virus-exposed mosquitoes), the daily survival of the mosquito species or population being studied, and the external incubation period (time it takes for mosquito to transmit virus after initial exposure). Thus the vectorial capacity takes into account environmental as well mosquito factors that affect mosquito behavior, longevity, and biting activity.

Predicting where YF outbreaks might occur is important in planning how to use vaccine but is complicated. To better understand potential for introductions of YF virus into new areas, many laboratory studies have been done to assess specific mosquito populations and their competence to transmit specific YF viruses. Amraoui [ 13 ] and colleagues at the Pasteur institute showed that Aedes albopictus mosquitoes collected from southeast France could be infected in the laboratory with a West African strain of YF virus. Infectious virus was found in mosquito saliva. Amraoui [ 14 ] also studied YF virus and Aedes albopictus mosquitoes from Manaus (Brazil). In the laboratory the virus adapted and was excreted in mosquito saliva after 4 passages in Aedes albopictus . Couto-Lima et al. [ 15 ] showed that Aedes aegypti and Aedes albopictus as well as forest mosquitoes from YF virus-free areas of Brazil were highly susceptible to American and African strains of YF viruses. This suggests that a traveler returning from YF-endemic parts of Africa could be a source of local transmission in South America. Infestation indexes for Aedes albopictus in Brazil are highest in southeastern and southern Brazil where recent outbreaks have occurred. During recent outbreaks in Brazil [ 6 ], YF virus was detected in Aedes albopictus caught in Minas Gerais in January 2017. An earlier study [ 16 ] assessed 23 Ae aegypti populations from 13 Brazilian states and found that the mosquitoes were highly susceptible to YF virus. Aedes aegypti was eliminated from Brazil in 1955 but control efforts faltered and the country was reinfested in the 1970s. Dengue outbreaks have occurred since the early 1980s and dengue is now endemic in many cities and severe epidemics are common. Earlier Tabachnick [ 17 ] tested 28 different populations of Aedes aegypti for susceptibility to YF viruses and found extensive variation. Yen [ 18 ] showed that Aedes aegypti mosquitoes from Guadeloupe (French West Indies) were able to transmit YF virus.

Dengue is endemic and epidemic in wide areas globally that are infested with Aedes aegypti and the size and intensity of outbreaks has increased in recent decades. Why is YF virus, which can be transmitted by the same mosquitoes, not similarly widespread in unvaccinated populations? What are the basic differences in the viruses and virus-vector pairs that affect epidemic potential? Both dengue and YF fever viruses produce viremias but the level in YF (10 5 to 10 6 ) [ 5 ] is significantly lower than in dengue (can reach 10 7 –10 9 ) [ 19 ]. Duration of viremia in humans is also longer for dengue than for YF (5 days (3–8) vs. 3 days (2–5)). The basic reproductive number is lower for YF than for dengue [ 20 ]. Dengue viruses may also be better adapted than YF viruses to Aedes aegypti .

Many groups have developed models to try to predict populations at risk, anticipate the geographic spread, size of outbreaks, and other findings that may be useful in planning interventions to prevent introductions or control spread.

Brent et al. [ 21 ] noted that in 2016, 923 million people lived in areas with endemic YF transmission and 45.2 million travelers departed YF-endemic countries/territories for international destinations. The highest volumes of travelers from YF-endemic countries arrived in Brazil, China, India, Mexico, Peru, and the US. Of note, among those headed for destinations with conditions suitable for transmission, 7.7 million were not required to show proof of YF vaccination upon arrival. Policy changes could help to reduce risk of introductions.

Dorigatti et al. [ 22 ] calculated the expected number of YF cases departing from Brazil during incubation or infectious periods during recent outbreaks. They used World Tourism Organization data on volume of air, water, and land border crossings, and suggested that the number of countries that may have received at least one case capable of seeding an epidemic included the US, Argentina, Uruguay, Spain, Italy, and Germany.

A modeling exercise by Shearer [ 23 ] found the highest predicted annual case numbers in Nigeria and South Sudan. They predicted high receptivity to transmission in parts of Malaysia, Indonesia, Thailand and noted areas with potential for importation and spread included Central America, eastern Brazil, and SE Asia. They also predicted the relative risk of YF across 47 countries in Americas and Africa. They estimated the number of cases that could be averted by vaccination and provide results for targeted vaccination campaigns.

Kraemer and colleagues [ 24 ] using data from recent outbreaks in Angola and Democratic Republic of Congo analyzed datasets for vector suitability, human demography, and mobility to infer district-specific YF infection risk during an epidemic. They provide estimates of the areas that could be prioritized for vaccination. In Africa population density and human movements are important in the spread of YF. Kraemer et al. [ 25 ] have also developed maps of the distribution of Aedes aegypti and Aedes albopictus – past distribution and projections for the future. The size of the populations living in infested areas is projected to increase in the future because of population growth in these regions.

Massad [ 20 ] has used models to calculate the critical proportion to vaccinate against YF to prevent epidemic urban YF in a dengue-endemic area. The investigators calculated the force of infection and the relative vector competence for dengue vs. YF viruses. They calculated the basic reproductive number for YF and for dengue for multiple cities in Brazil. The basic reproductive number for YF is lower than for dengue and varies by location for both. The critical percentage needed to be vaccinated to prevent an urban outbreak was as high as 88% in one location.

As a means of controlling YF outbreaks, Massad and colleagues also discussed the possibility of vaccinating monkeys against YF [ 26 ] given their role in human outbreaks in Brazil. They discuss feasibility of vaccinating monkeys in smaller green areas in urban centers. In Sao Paulo, where 202 human cases of YF occurred (79 deaths), none was attributed to transmission by Ae. aegypti , but most of the urban parks were closed to humans after deaths of monkeys from YF in the parks [ 27 ].

Diagnosis and surveillance

Several authors have emphasized the gaps in surveillance and diagnostic capacity. In both endemic and other countries, lack of good diagnostics can delay recognition of an outbreak. Johansson [ 28 ] and colleagues estimated that there may be between 1 and 70 infections that are asymptomatic or mild for every severe case. Early outbreaks may be missed allowing unrecognized spread. In the Democratic Republic of the Congo in 2016, diagnosis of YF was delayed. Most of YF cases had been acquired in Angola. Onset of jaundice, a valuable clinical clue, occurred a week or more after onset of infection [ 29 ]. An external quality assessment of European laboratories was conducted in response to the YF outbreak in Brazil and importations by travelers returning to Europe [ 30 ]. Adequate capability for diagnosing YF infections was lacking in 10 of 23 countries. Surveillance for YF must include study of nonhuman primates and mosquitoes as well as identification of human infections.

The recent importation of 11 documented cases of YF in Chinese workers from Angola again raises the specter of YFV spread in Asia where massive populations live in areas infested with Aedes mosquitoes [ 31 , 32 , 33 ]. It is estimated that half a million Chinese travelers have destinations in YF-endemic countries per year [ 33 , 34 ]. Use of vaccine needs to be expanded and International Health Regulations applied. The imported cases suggest failure of the system to require vaccination for individuals visiting areas of active transmission [ 32 ].

Many continue to ponder why YF epidemics have never occurred in Asia. Wasserman and colleagues [ 35 ] provide a thoughtful discussion of possible factors. They mention geographical variation in the susceptibility of mosquitoes to YF virus and suggest that wide dengue immunity may play a role. Studies done by Theiler [ 36 ] showed that monkeys immunized with pooled human sera from dengue-1 infected volunteers were relatively protected against a challenge with YF virus. Dengue-immune monkeys had lower YF viremia levels. The authors concluded that immunity to dengue may provide a barrier to YF introduction, but this has since been challenged. Dengue virus now circulates in many YF endemic areas in South America and some in Africa.

The Brazil outbreak led to YF being transported to 7 other countries in 2018, including 5 European countries [ 3 ]. Chinese working in Angola carried YF back to China. Phylogenetic analysis found the viruses imported to China were homologous with Angola strains [ 37 ]. This is the first time that YF has been documented in this region. Increasing travel between Asia and Africa heightens concern about possible outbreaks in Asia [ 38 , 39 ]. Asian populations are largely nonimmune and unvaccinated except for the tiny fraction who have received YF vaccination before travel.

YF has never emerged in the Pacific despite the presence of competent vectors and warnings of risk [ 40 ], especially in the wake of epidemics of chikungunya and Zika virus infections. Populations in the Pacific countries and territories, like Asian countries, have not been immunized against YF. Climate change and warming can potentially make some areas more able to sustain YF virus transmission. Over many decades the distribution of YF in Central America and northern South America shrank in response vaccination campaigns and vector control.

In response to recent outbreaks in Brazil, mass vaccination campaigns have begun. As of late 2018, 13.3 million people in Sao Paulo, 6.5 million in Rio de Janiero, and 1.85 million in Bahia states were vaccinated. Vaccine coverage in those states is now about 55% [ 41 ]. In 2018 Bolivia, Brazil, Colombia, French Guiana, and Peru were reporting new YF cases. Aedes albopictus mosquitoes naturally infected with YF virus were captured in 2 rural areas in Minas Gerais in 2017. As of April 2019, transmission of YF virus by Aedes aegypti had not been documented.

WHO has developed a global strategy to eliminate YF epidemics, EYE, to be carried out from 2017 through 2026 [ 42 ]. The three strategic objectives are to protect at-risk populations by expanding use of YF vaccine, to prevent international spread, and to contain outbreaks rapidly. The program will bring together multiple partners; it targets countries and regions most vulnerable to YF outbreaks. Countries are classified taking into account environmental factors, population density, and vector prevalence. Forty countries (27 in Africa and 13 in the Americas) are considered at highest risk. Increased access to YF vaccines is critical. An improved reference genome Ae aegypti may also accelerate work on vector control [ 43 ].

The YF vaccine is a live-attenuated vaccine developed from the wild-type Asibi strain in the 1930s, and passaged in embryonated chicken eggs [ 1 ]. All currently available vaccines derive from the substrains 17D-204 (China, France, Senegal, and the US), 17D-213 (Russia), and 17DD (Brazil); 95% of vaccinees become seropositive within 30 days [ 44 ]. Four vaccines (Brazil, France, Russia, and Senegal) are WHO prequalified and stockpiled for use in YF vaccination campaigns [ 1 , 44 ].

Proof of YF vaccination is required for entry into some countries according to the International Health Regulations (IHR). For some travelers visiting endemic countries, YF vaccination is recommended to protect the traveler [ 4 ]. The risk of YF illness among travelers to Africa for a 2-week stay is estimated to be 50/100,000 persons and for South America, 5/100,000 persons [ 4 ]. Unvaccinated travelers may import the infection to other countries. The concern for YF introduction to susceptible populations arose when persons infected in Angola traveled to/back to DRC, Mauritania, Kenya, and China [ 44 ].

Rare severe viscerotropic and neurologic adverse events have caused concern. Estimated rates are 0.3 and 0.8 per 100,000 doses, respectively, and the risk rises for persons aged 60 years and older [ 4 , 45 ]. Because YF vaccines are live-attenuated, they are contraindicated in immunocompromised persons including persons with immune compromising conditions, those on immune modulating medications, and HIV-infected persons with moderate- to severe immune compromise [ 4 ]. The decision about vaccination requires consideration of multiple factors including the traveler’s age, destination, health background, immune status, and whether future travel may benefit from YF vaccination [ 4 , 46 , 47 , 48 ].

The estimated global YF vaccination coverage from 1970 to 2016 based on sources such as WHO reports and health-service-provider registries that reported YF vaccination activities between May 1, 1939, and Oct 29, 2016 [ 49 ] concluded that in order to achieve 80% coverage in YF-endemic populations, between 393·7 to 472·9 million people still need to be vaccinated. This is between 43 and 52% of the population within YF risk zones [ 49 ]. The recent high numbers of YF cases was attributed to low vaccination coverage - lower than that needed to prevent outbreaks [ 23 ]; it was also estimated that vaccination coverage levels achieved by 2016 averted between 94 336 and 118 500 cases of YF annually within risk zones [ 23 ].

Duration of protection

YF vaccine was previously considered to be valid for 10 years. The WHO Strategic Advisory Group of Experts (SAGE) on Immunization and the World Health Assembly updated the duration to long-term protection, removing the 10-year booster requirement from the IHR [ 1 , 4 ]. The revision to life-long protection was partly based on the paucity of identified vaccine failures in vaccinated individuals, although post-marketing monitoring for break-through infections is lacking. Since the presence of YF neutralizing antibodies is associated with protection, this recommendation has caused debate about whether a single dose of YF vaccine can protect travelers whose neutralizing antibodies have declined and who are traveling to a high-risk area [ 50 , 51 , 52 , 53 , 54 , 55 ].

Lindsey et al. found protective neutralizing antibody levels (PRNT 90  > 10) following one reported dose of YF vaccine in 146/150 individuals (94%) vaccinated within 10 years (median 4 months, interquartile range [IQR] 2 months–3 years) and 54/66 individuals (82%) vaccinated at 10 years or earlier (median 15 years, IQR 12–25 years) [ 56 ]. These findings are comparable to prior studies in vaccinees residing in non-YF-endemic areas [ 4 , 50 , 51 , 52 ].

There are some concerns regarding subsets of YF vaccine recipients who developed lower antibody responses and/or shorter duration of antibody persistence [ 4 , 50 , 51 , 57 ]. In Brazil, seroconversion rates in children were lower when YF vaccine was administered concurrently with measles, mumps, rubella, possibly due to interference from co-administration of these two live-attenuated virus vaccines [ 58 , 59 ]. Goujon et al. assessed 131 infants; 96% had protective YF antibody levels. All 4 infants without a protective titer of YF antibodies had concurrent MMR and YF administration [ 59 ]. YF seropositivity declined in Malian children from 96.7% at 28 days after vaccination to 50.4% at 4.5 years postvaccination; seropositivity also declined in Ghanaian children from 72.7% at 28 days after vaccination to 27.8% at 2.3 years postvaccination [ 60 ]. In a nonendemic area, 63.8% of YF vaccinees were seropositive at ≥10 years and seronegativity most likely occured from 3 to 12 years postvaccination [ 61 ]. For travelers from non-endemic areas, the findings of blunted response and shorter seroprotection period have led to more conservative recommendations for persons vaccinated during early childhood, during pregnancy, or who were HIV-infected [ 4 ]. ACIP also recommends 10-year booster doses for persons who received YF vaccination preceding a hematopoietic stem cell transplant, laboratory workers handling YFV, and travel to higher-risk settings including long stays and travel to areas experiencing outbreaks [ 4 ].

Host concerns

The risk-versus-benefit of immunizing at-risk immunocompromised persons (e.g. HIV-infected, rheumatoid arthritis on disease modifying anti-rheumatic drugs DMARD) against YF vaccine poses challenges. The proliferation of biologic agents and immune modulators has led to the question of whether withholding YF vaccination for immunocompromised travelers is reasonable [ 62 ]. Recent studies have been conducted on the safety and immunogenicity of YF vaccination in these groups. Ferreira et al. measured neutralizing antibodies by PRNT and cellular immunity by in vitro YF-specific peripheral blood lymphoproliferative assay [ 63 ]. They compared conventional synthetic DMARD (csDMARD) and conventional synthetic plus biological DMARD (cs + bDMARD) to controls and found that only the cs + bDMARD led to an earlier decline in the vaccine response; there was lower PRNT seropositivity between 5 and 9 years and lower effector memory in CD8+ T cells as early as 1–5 years after 17DD-YF vaccination. These finding suggest that a 10-year booster dose of YF vaccine should be administered for persons receiving bDMARD, if they are able to suspend bDMARD [ 63 ].

In HIV patients, a study found that patients who received primary YF vaccination while plasma HIV RNA was suppressed maintained high seropositivity 99% within 1 year, 99% at 5 years, and 100% at 10 years [ 64 ]. The control of HIV replication at the time of YF vaccination appeared to affect the 10-year immune response; those on successful combination antiretroviral therapy show immune response comparable to that of non-HIV-infected adults up to 10 years [ 64 ]. The authors recommend a 10-year YF vaccine booster for patients vaccinated on successful cART, and an early booster for those with uncontrolled HIV RNA.

Finally, 21 human stem cell transplant (HSCT) recipients immunized with YF vaccine at a median of 39 months after HSCT and a median of 33 months after withdrawal of immunosuppression reported no side effects [ 65 ]. Eighteen had protective immunity after YF vaccination; furthermore, a third of the recipients who had pre-HSCT YF vaccination had persistent protective immunity after HSCT. If practical, establishing YF seropositivity by measuring antibody titers can help to reassure of immunity in an immunocompromised host.

Global supply

In YF-endemic countries, WHO recommends that YF vaccine be administered concurrently with the first dose of measles-containing vaccine, and the EYE strategy strives to ensure adequate global vaccine supply. Worldwide there are four WHO-prequalified vaccine manufacturers, and usually there is a stockpile of six million YF vaccine doses to be used if YF epidemics occur, but recent epidemics depleted the stockpile, resulting in the global shortage [ 66 ]. The concurrent YF outbreaks in Angola and DRC led to shortage of vaccines and a need for an emergency YF dose-sparing vaccination strategy [ 67 , 68 , 69 ]. If YF were to be introduced and spread to other regions, the current global supply of YF vaccine would be insufficient to control outbreaks.

There is only one Food and Drug Administration (FDA)-licensed vaccine in the US, YF-VAX® (Sanofi-Pasteur, Swiftwater, Pennsylvania). In 2017, manufacturing issues resulted in a stockout of YF-VAX® [ 70 ]. Sanofi Pasteur worked with the FDA to import another 17D-204 vaccine, Stamaril®, produced in France, in order to supply YF vaccine to US travelers [ 70 ]. However, this has led to severe limitations in accessing YF vaccine and significant inconvenience for travelers and providers [ 71 ]. Japan experienced a similar YF-VAX® stockout that necessitated the creation of a clinical trial to use Stamaril® [ 72 ].

Dose-sparing strategies and their duration of protection

Two dose-sparing alternatives have been studied: fractional dosing of YF vaccine and intradermal vaccination [ 67 , 68 , 73 ]. The fractional dosing strategy was first used in the large 2016 YF outbreak in the DRC, but excluded pregnant women, children under 2 years of age, and HIV-infected persons [ 74 ]. Intradermal YF vaccination has only been used in research setting [ 73 ].

Fractional dosing is based on the recognition that the standard vaccine dose, typically contains much higher content of virus (≥10,000 international units (IU)), that exceeds the minimum amount of virus required to achieve a protective titer of neutralizing antibody (1000 IU) [ 68 , 75 , 76 ]. At lower vaccine virus concentrations of 587 IU and lower, there was slower onset of viremia and lower geometric mean titers (GMT), seroconversion rates, and seropositivity at 10 months post-vaccination [ 75 ]. However, fractional doses containing ≥587 IU of virus achieved GMTs and seroconversion rates similar to those from full doses, and doses containing ≥3013 IU achieved immune response measurements similar to full doses [ 75 ]. The WHO recommended that one-fifth of the 0.5 ml original Brazil-made 17DD vaccine be reconstituted to 0.5 ml and administered subcutaneously to persons aged over 2 years; a full dose was still administered to children aged 9–23 months and pregnant women [ 1 , 68 ]. Fractional dosing was implemented in the DRC to help control the outbreak and proved to be successful [ 74 ]. Seroconversion occurred in 98% of fractional dose recipients at 28 days [ 74 ]. de Menezes Martins et al. evaluated immunity to fractional-dose YF vaccine and found 85% to be seropositive 8 years after vaccination [ 77 ].

Currently receipt of a fractional dose of YF does not officially meet the IHR requirement. In Canada, YF-VAX® is also the only licensed YF vaccine and fractional dosing is used for international travelers during YF vaccine shortage [ 78 , 79 ]; CATMAT advised clinicians to document fractional dosing of YF-VAX® in a Certificate of Medical Contraindication to Vaccination provided by the Public Health Agency of Canada [ 78 , 79 ]. Documentation in this way may ease international travel during periods of YF vaccine shortage. In the US, fractional dose of YF-VAX® is not recommended because efficacy data are still considered limited [ 70 ].

Intradermal administration of YF vaccine is another dose-sparing strategy [ 73 , 80 , 81 ]. Previously intradermal administration of reduced influenza vaccine doses demonstrated protective responses that were non-inferior to standard intramuscular vaccination [ 81 ]. Intradermal administration offers the theoretical advantage of mimicking the route of natural flavivirus entry [ 73 , 82 , 83 , 84 ]. For instance, dengue virus is injected by feeding Aedes mosquitoes into the skin and infects dendritic cells in the dermis and epidermis. These host immune cells disseminate via lymph to regional nodes, resulting in viremia and systemic infection. YF, another flavivirus, is expected to follow a similar sequence.

The immunogenicity of reduced-dose 17D-YFV (Stamaril, Sanofi Pasteur, France) via intradermal administration (0.1 mL) was non-inferior to that of full-dose subcutaneous administration (0.5 mL) [ 73 , 83 ]. Follow up of a subgroup at 10 years illustrated 98% had protective YF neutralizing antibodies, also non-inferior to recipients of the standard-dose [ 83 ]. Although subject numbers are small, intradermal YF vaccine administration appears to be a feasible option for YF protection.

Other strategies: new vaccine, primate reservoir, and vector control

Improving YF vaccine supply has led to development of a plant-produced subunit vaccine candidate derived from YF virus envelope protein [ 85 ]. While a study found partial protective efficacy in mice, the plant-based vaccine achieved inferior efficacy compared to that of the live attenuated 17DD vaccine [ 85 ]. As noted earlier, another strategy that has been suggested is vaccinating monkeys against YF [ 26 ].

Vector control has also been a component of YF control strategies. Achee et al. reviewed alternative strategies for YF control [ 86 ]. The current vector control strategies include the use of pyrethroid insecticide spraying, larval control including larvicides, insect growth regulators, and bacterial toxins, and biologic agents such as predatory copepods, fish, and Toxorhynchites larvae. Many alternative strategies are being explored [ 86 ] (See Table 1 ).

Despite the availability of a highly effective vaccine, yellow fever outbreaks have continued and have expanded into new areas in recent years. Many populations remain vulnerable to outbreaks. Increasing global travel and population movements pose risks of introductions into large urban areas in tropical and subtropical areas that are infested with mosquitoes competent to transmit YF. Limited supplies of vaccine have hobbled control efforts. Resources, political will, and leadership will be needed to control YF. Even though the YF virus cannot be eliminated from the animal reservoir today, the tools are available to eliminate YF infections in humans.

Availability of data and materials

Not applicable.

WHO. Vaccines and vaccination against yellow fever. WHO position paper – June 2013. Wkly Epidemiol Rec. 2013;88(27):269–83.

Google Scholar  

WHO. Disease outbreak news: Yellow fever – Brazil. Available at https://www.who.int/csr/don/18-april-2019-yellow-fever-brazil/en/ Accessed 1 Oct 2019.

Hamer DH, Angelo K, Caumes E, van Genderen PJJ, Florescu SA, Popescu CP, et al. Fatal yellow fever in travelers to Brazil, 2018. MMWR. 2018;67(11):340–1.

PubMed   PubMed Central   Google Scholar  

Staples JE, Bocchini JA Jr, Rubin L, Fischer M. Centers for Disease Control and Prevention (CDC). Yellow fever vaccine booster doses: recommendations of the advisory committee on immunization practices, 2015. MMWR Morb Mortal Wkly Rep. 2015;64(23):647–50.

Monath TP. Yellow fever: an update. Lancet Infect Dis. 2001;1:11–20.

Article   CAS   PubMed   Google Scholar  

Faria NR, Kraemer MUG, Hill SC, Goes de Jesus J, Aguiar RS, Iani FCM, et al. Genomic and epidemiological monitoring of yellow fever virus transmission potential. Science. 2018;361:894–9.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Delatorre E, de Abreu FVS, Ribeiro IP, Gómez MM, Dos Santos AAC, Ferreira-de-Brito A, et al. Distinct UFV lineages co-circulated in the central-western and southeastern Brazilian regions from 2015 to 2018. Front Microbiol. 2019;10:1079.

Article   PubMed   PubMed Central   Google Scholar  

Johansson MA, Araba-Vizcarrondo N, Bj BI, Staples JE. Incubation periods for yellow fever virus. Am J Trop Med Hyg. 2010;83(1):183–8.

Aitken THG, Tesh RB, Beaty B, Rosen L. Transovarial transmission of yellow fever virus by mosquitoes (Aedes aegypti). Am J Trop Med Hyg. 1979;28(1):119–21.

Diallo M, Thonnon J, Fontenille D. Vertical transmission of the yellow fever virus by Aedes aegypti : dynamics of infection in F1 adult progeny of orally infected females. Am J Trop Med Hyg. 2000;62(1):151–6.

Barrett AD, Higgs S. Yellow fever: a disease that has yet to be conquered. Annu Rev Entomol. 2007;52:209–29.

Azar SR, Weaver SC. Vector competence: what has Zika virus taught us? Viruses. 2019;11:867. https://doi.org/10.3390/v11090867 .

Article   PubMed Central   Google Scholar  

Amraoui F, Vazeille M, Failloux AB. French Aedes albopictus are able to transmit yellow fever virus. Euro Surveill. 2016;21(39). https://doi.org/10.2807/1560-7917.ES.2016.21.39.30361 .

Amraoui F, Pain A, Piorkowski G, Vazeille M, Couto-Lima D, de Lamballerie X, et al. Experimental adaptation of the yellow fever virus to the mosquito Aedes albopictus and potential risk of urban epidemics in Brazil, South America. Sci Rep. 2018;8:14337.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Couto-Lima D, Madec Y, Bersot MI, Campos SS, Motta MA, Santos FBD, et al. Potential risk of re-emergence of urban transmission of yellow fever virus in Brazil facilitated by competent Aedes populations. Sci Rep. 2017;7:4848.

Lourenco-de-Oliveira R, Vazielle M, de Filippis AM, Faillox AB. Aedes aegypti in Brazil: genetically differentiated populations with high susceptibility to dengue and yellow fever viruses. Trans R Soc Trop Med Hyg. 2004;98(1):43–54.

Tabachnick WJ, Wallis GP, Aitken TH, Miller BR, Amato GD, Lorenz L, et al. Oral infection of Aedes aegypti with yellow fever virus: geographic variation and genetic considerations. Am J Top Med Hyg. 1985;34(6):1219–24.

Article   CAS   Google Scholar  

Yen PS, Amraoui F, Rua V, Failloux AB. Aedes aegypti mosquitoes from Guadeloupe (French West Indies) are able to transmit yellow fever virus. PLoS One. 2018;13(9):e0204710.

Vaughn DW, Green S, Kalayanarooj S, Innis BL, Nimmannitya S, Suntayakorn S, et al. Dengue viremia titer, antibody response pattern, and virus serotype correlate with disease severity. JID. 2000;181:2–9.

Massad E, Coutinho FAB, Burattini MN, Lopez LF. The risk of yellow fever in a dengue-infested area. Trans R Soc Trop Med Hyg. 2001;95:370–4.

Brent SE, Watts A, Cetron M, German M, Kraemer MU, Bogoch II, et al. International travel between global urban centres vulnerable to yellow fever transmission. Bull World Health Organ. 2018;96(5):343–354B.

Dorigatti I, Hamlet A, Aguas R, Cattarino L, Cori A, Donnelly CA, et al. International risk of yellow fever spread from the ongoing outbreak in Brazil, December 2016 to May 2017. Euro Surveill. 2017;22(28). https://doi.org/10.2807/1560-7917.ES.2017.22.28.30572 .

Shearer FM, Longbottom J, Browne AJ, Pigott DM, Brady OJ, Kraemer MUG, et al. Existing potential infection risk zones of yellow fever worldwide: a modelling analysis. Lancet Glob Health. 2018;6:e270–8.

Kraemer MUG, Faria NR, Reiner RC Jr, Golding N, Nikolay B, Stasse S, et al. Spread of yellow fever virus outbreak in Angola and the Democratic Republic of the Congo 2015-16: a modelling study. Lancet Infect Dis Lancet Infect Dis. 2017;17(3):330–8.

Article   PubMed   Google Scholar  

Kraemer MUG, Reiner RC Jr, Brady OJ, Messina JP, Gilbert M, Pigott DM, et al. Past and future spread of the arbovirus vectors Aedes aegypti and Aedes albopictus. Nat Microbiol. 2019;4(5):854–63.

Massad E, Miguel MM, Coutinho FAB. Is vaccinating monkeys against yellow fever the ultimate solution for the Brazilian recurrent epizootics? Epidemiol Infect. 2018;146(13):1622–4.

Massad E, Amaku M, Coutinho FAB, Struchiner CJ, Lopez LF, Coelho G, et al. The risk of urban yellow fever resurgence in Aedes-infested American cities. Epidem & Infect. 2018;146:1219–25.

Article   Google Scholar  

Johansson MA, Vasconcelos PFC, Staples JE. The whole iceberg : estimating the incidence of yellow fever virus infection from the number of severe cases. Trans R Soc Trop Med Hyg. 2014;108:482–7.

Ingelbeen B, Weregemere NA, Noel H, Tshapenda GP, Mossoko M, Nsio J, et al. Urban yellow fever outbreak – Democratic Republic of the Congo, 2016: towards more rapid case detection. PLoS Negl Trop Dis. 2018;12(12):e0007029.

Domingo C, Ellerbrok H, Koopmans M, Nitsche A, Leitmeyer K, Charrel RN, et al. Need for additional capacity and improved capability for molecular detection of yellow fever virus in European Expert Laboratories: External Quality Assessment, March 2018. Euro Surveill. 2018;23(28). https://doi.org/10.2807/1560-7917.ES.2018.23.28.1800341 .

Song R, Guan S, Lee SS, Chen Z, Chen C, Han L, et al. Late or Lack of Vaccination Linked to Importation of Yellow Fever from Angola to China. Emerg Infect Dis. 2018;24(7):1383–6.

Article   CAS   PubMed Central   Google Scholar  

Schlagenhauf P, Chen LH. Yellow fever importation to China – a failure of pre- and post-travel control systems? Int J Infect Dis. 2017;60:91–2.

Wilder-Smith A, Leong WY. Importation of yellow fever into China: assessing travel patterns. J Travel Med. 2017;24(4). https://doi.org/10.1093/jtm/tax008 .

Wilder-Smith A, Massad E. Estimating the number of unvaccinated Chinese workers against yellow fever in Angola. BMC Infect Dis. 2018;18(1):185.

Wassserman S, Tambyah PA, Lim PL. Yellow fever cases in Asia: primed for an epidemic. Int J Infect Dis. 2016;48:98–103.

Theiler M, Anderson CR. The relative resistance of dengue-immune monkeys to yellow fever virus. Am J Trop Med Hyg. 1975;24(1):115–8.

Cui S, Pan Y, Lyu Y, Liang Z, Li J, Sun Y, et al. Detection of yellow fever virus genomes from four imported cases in China. Int J Infect Dis. 2017;60:93–5.

Gubler DJ. Potential yellow fever epidemics in unexposed populations. Bull World Health Organ. 2018;96(5):299.

Wilder-Smith A, Lee V, Gubler DJ. Yellow fever: is Asia prepared for an epidemic? Lancet Infect Dis. 2019;19(3):241–2.

Musso D, Parola P, Raoult D. Yellow fever: the Pacific should be prepared. Lancet. 2018;392(10162):2347.

PAHO. Epidemiological update: yellow fever 25 January 2019. Available at https://www.paho.org/hq/index.php?option=com_topics&view=article&id=69&Itemid=40784&lang=en Accessed 1 Sept 2019.

WHO. A global strategy to eliminate yellow fever epidemics (EYE) 2017–2026. Available at http://apps.who.int/iris/bitstream/handle/10665/272408/9789241513661-eng.pdf?ua=1&ua=1 Accessed 1 Sept 2019.

Matthews B, Dudchenko O, Kingan SB, Koren S, Antoshechkin I, Crawford JE, et al. Improved reference genome of Aedes aegypti informs arbovirus vector control. Nature. 2018;563:501–7.

Collins ND, Barrett AD. Live attenuated yellow fever 17D vaccine: a legacy vaccine still controlling outbreaks in modern day. Curr Infect Dis Rep. 2017;19(3):14.

Lindsey NP, Rabe IB, Miller ER, Fischer M, Staples JE. Adverse event reports following yellow fever vaccination, 2007-13. J Travel Med. 2016;23(5). https://doi.org/10.1093/jtm/taw045 .

Lown BA, Chen LH, Wilson ME, Sisson E, Gershman M, Yanni E, et al. Vaccine administration decision making: the case of yellow fever vaccine. Clin Infect Dis. 2012;55(6):837–43.

Lown BA, Chen LH, Han PV, Jentes ES, Wilson ME, Benoit CM, et al. Preferences and decision needs of Boston-area travelers to countries with risk of yellow fever virus transmission: implications for health care providers. J Travel Med. 2014;21(4):266–71.

Leder K, Chen LH, Wilson ME. Aggregate travel vs. single trip assessment: arguments for cumulative risk analysis. Vaccine. 2012;30(15):2600–4.

Shearer FM, Moyes CL, Pigott DM, Brady OJ, Marinho F, Deshpande A, et al. Global yellow fever vaccination coverage from 1970 to 2016: an adjusted retrospective analysis. Lancet Infect Dis. 2017;17:1209–17.

Gotuzzo E, Yactayo S, Córdova E. Efficacy and duration of immunity after yellow fever vaccination: systematic review on the need for a booster every 10 years. Am J Trop Med Hyg. 2013;89:434–44.

Collaborative group for studies on yellow fever vaccines. Duration of post-vaccination immunity against yellow fever in adults. Vaccine. 2014;32:4977–84.

Amanna IJ, Slifka MK. Questions regarding the safety and duration of immunity following live yellow fever vaccination. Expert Rev Vaccines. 2016;15:1519–33.

Wieten RW, Jonker EF, van Leeuwen EM, Remmerswaal EB, Ten Berge IJ, de Visser AW, et al. A single 17D yellow fever vaccination provides lifelong immunity; Characterization of Yellow-Fever-Specific Neutralizing Antibody and T-Cell Responses after Vaccination. PLoS One. 2016;11:e0149871.

Vasconcelos PF. Single shot of 17D vaccine may not confer life-long protection against yellow fever. Mem Inst Oswaldo Cruz. 2018;113:135–7.

Visser LG, Veit O, Chen LH. Waning immunity after single-dose yellow fever vaccination: Who needs a second shot? J Travel Med. 2019;26(7). https://doi.org/10.1093/jtm/tay134 .

Lindsey NP, Horiuchi KA, Fulton C, Panella AJ, Kosoy OI, Velez JO, et al. Persistence of yellow fever virus-specific neutralizing antibodies after vaccination among US travellers. J Travel Med. 2018;25(1). https://doi.org/10.1093/jtm/tay108 .

Vasconcelos PFC, Barrett ADT. Are booster doses of yellow fever vaccine needed? Lancet Infect Dis. 2019. https://doi.org/10.1016/S1473-3099(19)30411-6 [Epub ahead of print].

Group for Studies of Yellow Fever Vaccine. A randomized double-blind clinical trial of two yellow fever vaccines prepared with substrains 17DD and 17D-213/77 in children nine-23 months old. Mem Inst Oswaldo Cruz. 2015;110:771–80.

Goujon C, Gougeon ML, Tondeur L, Poirier B, Seffer V, Desprès P, et al. CHRONOVAC VOYAGEUR: a study of the immune response to yellow fever vaccine among infants previously immunized against measles. Vaccine. 2017;35(45):6166–71.

Domingo C, Fraissinet J, Ansah PO, Kelly C, Bhat N, Sow SO, Mejía JE. Long-term immunity against yellow fever in children vaccinated during infancy: a longitudinal cohort study. Lancet Infect Dis. 2019;19(12):1363–70.

Kareko BW, Booty BL, Nix CD, Lyski ZL, Slifka MK, Amanna IJ, Messer WB. Persistence of Neutralizing Antibody Responses Among Yellow Fever Virus 17D Vaccinees Living in a Nonendemic Setting. J Infect Dis. 2019. https://doi.org/10.1093/infdis/jiz374 [Epub ahead of print].

de Jong W, de Man RA, Dalm VASH, Reusken CBEM, Goeijenbier M, van Gorp ECM. Yellow fever vaccination for immunocompromised travellers: unjustified vaccination hesitancy? J Travel Med. 2019;26(6). https://doi.org/10.1093/jtm/taz015 .

Ferreira CC, Campi-Azevedo AC, Peruhype-Magalhāes V, Coelho-Dos-Reis JG, Antonelli LRDV, Torres K, et al. Impact of synthetic and biological immunomodulatory therapy on the duration of 17DD yellow fever vaccine-induced immunity in rheumatoid arthritis. Arthritis Res Ther. 2019;21(1):75.

Veit O, Domingo C, Niedrig M, Staehelin C, Sonderegger B, Héquet D, et al. Long-term immune response to yellow fever vaccination in human immunodeficiency virus (HIV)-infected individuals depends on HIV RNA suppression status: implications for vaccination schedule. Clin Infect Dis. 2018;66(7):1099–108.

Sicre de Fontbrune F, Arnaud C, Cheminant M, Boulay A, Konopacki J, Lapusan S, et al. Immunogenicity and safety of yellow fever vaccine in allogeneic hematopoietic stem cell transplant recipients after withdrawal of immunosuppressive therapy. J Infect Dis. 2018;217(3):494–7.

Chen LH, Hamer DH. Vaccination challenges in confronting the resurgent threat from yellow fever. JAMA. 2017;318(17):1651–2.

Barrett AD. Yellow fever in Angola and beyond--the problem of vaccine supply and demand. N Engl J Med. 2016;375(4):301–3.

WHO. Yellow fever vaccine: WHO position on the use of fractional doses – June 2017. Wkly Epidemiol Rec. 2017;92(25):345–50.

Vannice K, Wilder-Smith A, Hombach J. Fractional-dose yellow fever vaccination - advancing the evidence base. N Engl J Med. 2018;379(7):603–5.

Gershman MD, Angelo KM, Ritchey J, Greenberg DP, Muhammad RD, Brunette G, et al. Addressing a yellow fever vaccine shortage - United States, 2016-2017. MMWR Morb Mortal Wkly Rep. 2017;66(17):457–9.

Chen LH, Kozarsky PE, Visser LG. What’s old is new again: the re-emergence of yellow fever in Brazil and vaccine shortages. Clin Infect Dis. 2019;68(10):1761–2.

Ujiie M, Terada M, Kobayakawa M, Ohmagari N. Clinical trial with unlicensed yellow fever vaccine in Japan. J Travel Med. 2019;26(6). https://doi.org/10.1093/jtm/tay151 .

Roukens AH, Vossen AC, Bredenbeek PJ, van Dissel JT, Visser LG. Intradermally administered yellow fever vaccine at reduced dose induces a protective immune response: a randomized controlled non-inferiority trial. PLoS One. 2008;3(4):e1993.

Ahuka-Mundeke S, Casey RM, Harris JB, Dixon MG, Nsele PM, Kizito GM, et al. Immunogenicity of fractional-dose vaccine during a yellow fever outbreak - preliminary report. N Engl J Med. 2018. https://doi.org/10.1056/NEJMoa1710430 [Epub ahead of print].

Martins RM, Maia Mde L, Farias RH, Camacho LA, Freire MS, Galler R, et al. 17DD yellow fever vaccine: a double blind, randomized clinical trial of immunogenicity and safety on a dose-response study. Hum Vaccin Immunother. 2013;9(4):879–88.

Campi-Azevedo AC, de Almeida EP, Coelho-Dos-Reis JG, Peruhype-Magalhães V, Villela-Rezende G, Quaresma PF, et al. Subdoses of 17DD yellow fever vaccine elicit equivalent virological/immunological kinetics timeline. BMC Infect Dis. 2014;14:391.

de Menezes MR, Maia MLS, de Lima SMB, de Noronha TG, Xavier JR, Camacho LAB, et al. Duration of post-vaccination immunity to yellow fever in volunteers eight years after a dose-response study. Vaccine. 2018;36(28):4112–7.

Committee to Advise on Tropical Medicine and Travel. Interim Canadian recommendations for the use of a fractional dose of yellow fever vaccine during a vaccine shortage. Available at https://www.canada.ca/en/public-health/services/publications/diseases-conditions/interim-recommendations-fractional-dose-yellow-fever-vaccine-shortage.html . Accessed 1 June 2018.

Teitelbaum P, Bui YG, Libman M, McCarthy A. Fractional dosing of yellow fever vaccine during shortages: perspective from Canada. J Travel Med. 2018;25(1). https://doi.org/10.1093/jtm/tay098 .

Roukens AH, Gelinck LB, Visser LG. Intradermal vaccination to protect against yellow fever and influenza. Curr Top Microbiol Immunol. 2012;351:159–79.

CAS   PubMed   Google Scholar  

Belshe RB, Newman FK, Cannon J, et al. Serum antibody responses after intradermal vaccination against influenza. N Engl J Med. 2004;351:2286–94.

Norton SA, Morens DM. Vaccination strategies during shortages of yellow fever vaccine. JAMA. 2018;319(12):1280.

Roukens AHE, van Halem K, de Visser AW, Visser LG. Long-term protection after fractional-dose yellow fever vaccination: follow-up study of a randomized, controlled, Noninferiority Trial. Ann Intern Med. 2018;169(11):761–5.

Roukens AHE, Visser LG. Fractional-dose yellow fever vaccination: an expert review. J Travel Med. 2019;26(6). https://doi.org/10.1093/jtm/taz024 .

Tottey S, Shoji Y, Jones RM, Chichester JA, Green BJ, Musiychuk K, et al. Plant-produced subunit vaccine candidates against yellow fever induce virus neutralizing antibodies and confer protection against viral challenge in animal models. Am J Trop Med Hyg. 2018;98(2):420–31.

Achee NL, Grieco JP, Vatandoost H, et al. Alternative strategies for mosquito-borne arbovirus control. PLoS Negl Trop Dis. 2019;13(1):e0006822.

Download references

Acknowledgements

Author information, authors and affiliations.

Mount Auburn Hospital, 330 Mount Auburn Street, Cambridge, MA, 02138, USA

Lin H. Chen

Harvard Medical School, Boston, MA, USA

Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, MA, USA

Mary E. Wilson

Department of Epidemiology and Biostatistics, School of Medicine, University of California, San Francisco, USA

You can also search for this author in PubMed   Google Scholar

Contributions

Both authors contributed equally to the manuscript. Both authors read and approved the final manuscript.

Authors’ information

Dr. Lin H. Chen is Director of the Travel Medicine Center at Mount Auburn Hospital, Cambridge, Massachusetts, and Associate Professor of Medicine at Harvard Medical School, Boston, Massachusetts, USA. Dr. Mary E. Wilson is Adjunct Professor in the Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, USA, and Clinical Professor in the Department of Epidemiology and Biostatistics, School of Medicine, University of California, San Francisco, USA.

Corresponding author

Correspondence to Lin H. Chen .

Ethics declarations

Ethics approval and consent to participate, consent for publication.

Authors consent for the article to be published in Tropical Diseases, Travel Medicine and Vaccines.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/ ), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Cite this article.

Chen, L.H., Wilson, M.E. Yellow fever control: current epidemiology and vaccination strategies. Trop Dis Travel Med Vaccines 6 , 1 (2020). https://doi.org/10.1186/s40794-020-0101-0

Download citation

Received : 18 October 2019

Accepted : 05 January 2020

Published : 10 January 2020

DOI : https://doi.org/10.1186/s40794-020-0101-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• 3–10: yellow fever
• Epidemiology
• Vaccination
• Dose-sparing
• Fractional dosing
• Vaccine supply

Tropical Diseases, Travel Medicine and Vaccines

ISSN: 2055-0936

• Submission enquiries: [email protected]

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• My Account Login
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Perspective
• Open access
• Published: 06 July 2020

Review of data and knowledge gaps regarding yellow fever vaccine-induced immunity and duration of protection

• J. Erin Staples   ORCID: orcid.org/0000-0002-1446-4071 1 ,
• Alan D. T. Barrett 2 ,
• Annelies Wilder-Smith 3 , 4 &
• Joachim Hombach 5  

npj Vaccines volume  5 , Article number:  54 ( 2020 ) Cite this article

6715 Accesses

38 Citations

10 Altmetric

Metrics details

• Public health
• Viral infection

Yellow fever (YF) virus is a mosquito-borne flavivirus found in Sub-Saharan Africa and tropical South America. The virus causes YF, a viral hemorrhagic fever, which can be prevented by a live-attenuated vaccine, strain 17D. Despite the vaccine being very successful at decreasing disease risk, YF is considered a re-emerging disease due to the increased numbers of cases in the last 30 years. Until 2014, the vaccine was recommended to be administered with boosters every 10 years, but in 2014 the World Health Organization recommended removal of booster doses for all except special populations. This recommendation has been questioned and there have been reports of waning antibody titers in adults over time and more recently in pediatric populations. Clearly, the potential of waning antibody titers is a very important issue that needs to be carefully evaluated. In this Perspective, we review what is known about the correlate of protection for full-dose YF vaccine, current information on waning antibody titers, and gaps in knowledge. Overall, fundamental questions exist on the durability of protective immunity induced by YF vaccine, but interpretation of studies is complicated by the use of different assays and different cut-offs to measure seroprotective immunity, and differing results among certain endemic versus non-endemic populations. Notwithstanding the above, there are few well-characterized reports of vaccine failures, which one would expect to observe potentially more with the re-emergence of a severe disease. Overall, there is a need to improve YF disease surveillance, increase primary vaccination coverage rates in at-risk populations, and expand our understanding of the mechanism of protection of YF vaccine.

Similar content being viewed by others

Long COVID: major findings, mechanisms and recommendations

Vaccination impairs de novo immune response to omicron breakthrough infection, a precondition for the original antigenic sin

A guide to vaccinology: from basic principles to new developments

Introduction.

Yellow fever (YF) virus, a mosquito-borne flavivirus, is present in tropical areas of Africa and South America. Infection in humans can produce a hemorrhagic fever and is fatal in 30–60% of persons with severe disease 1 , 2 . Recent decades have witnessed an unprecedented emergence of YF virus activity, including in highly urbanized areas where vaccination coverage was low 3 , 4 , 5 . It has been recently estimated that roughly 400 million individuals require vaccination within at-risk zones to potentially prevent epidemic of the disease though many more might be at risk due to the recent expansion of risk zones, particularly in Brazil 3 , 6

YF vaccine was first developed in the 1930s after successful attenuation of the Asibi strain of YF virus to generate the strain 17D 7 . Today, three substrains (17D-204, 17DD, and 17D-213) are used as vaccines and are manufactured by six companies, of which four are prequalified by the World Health Organization (WHO) 8 . The vaccine is given as one dose either by subcutaneous or intramuscular administration, with 80% of vaccine recipients develop neutralizing antibodies 10 days post immunization and close to 100% by one month post immunization in clinical trials 9 . However, it has been noted that children <2 years of age can have lower seroconversion rates following a single dose of YF vaccine 10 . No human efficacy studies have ever been performed with the vaccine, but protection has been robustly demonstrated. Evidence for this conclusion include (1) reduction of laboratory-associated infections in vaccinated workers, (2) observation following initial use of the vaccine in Brazil and other South American countries that YF occurred only in unvaccinated persons, (3) rapid disappearance of cases during YF vaccination campaigns initiated during epidemics, (4) very few vaccine failures detected in any endemic country, and (5) protection of rhesus monkeys against virulent wild-type (WT) YF virus challenge by neutralizing antibodies generated in response to YF vaccination 11 , 12 , 13 .

A booster dose requirement for YF vaccine was first put into place in 1959 under the precursor to International Health Regulations (IHR), International Sanitary Regulations, with booster doses initially being required every 9 years based on available data 14 , 15 . The booster dose interval was changed in 1965 to every 10 years based on limited evidence from two published studies that showed neutralizing antibodies were present in most vaccine recipients, including those who received the vaccine in childhood, for at least 10 years after vaccination 16 , 17 . Starting in late 2011, the WHO Strategic Advisory Group of Experts (SAGE) on Immunization YF working group conducted a systematic review of ~17 unpublished and published studies that identified a very low number of vaccine failures and high seropositivity rates following vaccination over time 18 , 19 . From these additional, albeit observational data, SAGE concluded that a single primary dose of YF vaccine is sufficient to confer sustained immunity and lifelong protection against YF disease, and that a booster dose is not needed, except for special populations (e.g., immunocompromised and immunosuppressed) 20 . In May 2014, the World Health Assembly adopted the recommendation to remove the 10-year booster dose requirement from the IHR, which was enacted in June 2016 21 . In 2014, the United States Advisory Committee on Immunization Practices (ACIP) YF vaccine working group conducted a similar systematic review of YF vaccine immunogenicity 10 . However, since SAGE’s recommendation removed the IHR requirement for boosters, ACIP working group reviewed the available data to determine whether or not booster doses were needed as ACIP had never recommended a booster dose of the vaccine before. Based on the available data, ACIP voted in 2015 that a single primary dose of YF vaccine provides long-lasting protection and is adequate for most travelers 22 . However, as a precautionary measure, it was noted that a booster dose may be given to travelers who received their last dose of YF vaccine at least 10 years previously and who will be in a higher-risk setting based on season, location, activities, and duration of their travel. This would include travelers who plan to spend a prolonged period in endemic areas or those traveling to highly endemic areas, such as rural West Africa during peak transmission season or an area with an ongoing outbreak.

Subsequent to SAGE and ACIP recommendations that a single dose of YF vaccine is sufficient to provide lifelong protection in most individuals, several have questioned this decision 23 , 24 , 25 , 26 . Furthermore, several recent studies have noted waning antibody titers after vaccination and potential vaccine failures 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 . Below we note what is known about vaccine immunity, review the additional data that have been generated using full-dose YF vaccine since the SAGE recommendation in 2013, and discuss next steps to determine if booster doses of YF vaccine are needed.

What constitutes YF vaccine immunity?

One of the key questions to know whether or not YF vaccine booster doses are needed is what constitutes protective vaccine immunity. The closest correlate of protection that exists for YF vaccination was established in one study of non-human primates vaccinated with YF vaccine and then challenged with virulent WT YF virus 11 . From this study, log 10 neutralization index (LNI) of ≥0.7 was established as a potential cut-off for protective immunity with 51 (94%) of 54 surviving monkeys having a LNI ≥ 0.7. In comparison, only one (8%) of 12 monkeys who died when challenged had a LNI above 0.7. Currently, plaque reduction neutralization tests (PRNTs) are used to establish the quantitative titers of YF virus-specific antibodies as it uses less serum and is typically easier to perform. Current studies typically report either 90% PRNT (PRNT 90 ), PRNT 80 , or PRNT 50 titers. Although a PRNT 90 titer is more specific as it reduces the likelihood of positive results due to cross-reactive neutralizing antibodies from other flaviviruses, it measures at the bottom of the S-shaped neutralization curve, which leads to less variability and can lead to false-negative results for lower virus-specific antibody titers 35 . PRNT 50 titer are at the midpoint or more linear portion of an S-shaped curve making them higher, more variable and sensitive, but less specific. Most clinical trials for flavivirus vaccines use a PRNT 50 assay with a titer of 1 in 10 as a correlate of protection 36 , 37 , 38 . However, LNI and PRNT have never been formally compared using standardized reagents to understand how they might relate. Furthermore, it is unclear if neutralizing antibodies as measured using current assays are the only correlate of protection. Our understanding of the role of cell-mediated immunity in both the initial immunologic response, as well as longer-term protection is advancing, but it also comes with the uncertainty of what might represent protective types and levels immunity that could prevent a person developing WT YF disease. However, there is general agreement that the pool of memory cells needs to be able to quickly proliferate when challenged to protect an individual as the incubation period of YF is typically short ranging from 3 to 6 days 24 , 39 , 40 .

The question of what constitutes vaccine immunologic memory is not unique to YF vaccine. Smallpox vaccine also was utilized before efficacy studies could be performed and the same questions about vaccine immunity are present for live-attenuated vaccines against vaccinia virus 41 . Although detection of antibodies is used to denote protective immunity following measles vaccination, it also has been documented that individuals lacking detectable neutralizing antibodies can develop secondary immune response with revaccination or exposure to measles virus suggesting that alternative types of immunity exist 42 .

Currently, whether or not the absence of detectable neutralizing antibodies represent an absence of protective immunity against WT YF disease is a critical knowledge gap for YF immunity. As noted above, it is also unclear what amount of antibody might be needed to protect someone against developing a symptomatic infection or viremia. Two studies have documented roughly one-third of individuals with preexisting YF virus-specific neutralizing antibodies fail to develop an anamnestic neutralizing antibody response (i.e., ≥4-fold or greater increase in neutralization titers) following a booster dose suggesting sterilizing immunity that is correlated with higher pre-vaccination titers 9 , 43 . If it is correct that an absence of detectable neutralizing antibodies following primary immunization or the development of an amnestic response following a booster vaccine dose means an absence of protection for YF in a primary vaccinee, one might have expected more cases of WT YF disease to be reported in children 4–10 years post-vaccination 33 . However, epidemiologic data from the recent outbreaks in Brazil indicate that very few cases of WT disease occurred in children, with a lower incidence of WT disease in children compared to adults 5 , 44 . Although this might be secondary to who is being exposed or differences in clinical attack rate, the recent outbreaks occurring near and in urban areas as well as the notable occurrence of cases in women tend to suggest children were likely exposed to the virus in these recent outbreaks. Finally, the development of an amnestic response might not equate to a lack of protection, particularly if the kinetics of the immunologic response is fast enough to blunt the viremia due to a WT infection.

Vaccine failures

Since 2013, there has been several reports of vaccine failures, one in peer-reviewed literature plus epidemiologic reports issued by public health authorities 45 , 46 , 47 . The published study, which has been cited by others in editorials and reviews to support the need for booster doses, came out in 2014 during the ACIP deliberations and describe individuals having a history of YF vaccination who later develop WT YF disease 24 , 26 , 45 . The ACIP YF vaccine working group contacted the Brazil Ministry of Health (MOH) to verify that, as stated, 459 (55%) of 831 YF cases in Brazil from 1973 to 2008 were vaccine failures, including 27 (3%) primary vaccine failures (e.g., occurring after the first 10 days of vaccination but within the first 10 years of vaccination) and 432 (52%) secondary vaccine failures (e.g., occurring more than 10 years after vaccination potentially due to waning antibody titers) 45 . The Brazil MOH provided data to the working group noting that there were seven vaccine failures in Brazil from 1973 to 2008; five constituting primary vaccine failures, and two secondary vaccine failures occurring at 20 and 27 years post vaccination 10 , 45 , 48 , 49 , 50 . Unfortunately, there has never been a publication to clarify that the data were not accurate and it continues to be cited as evidence to support the need for booster doses 33 .

From data reported to the Pan American Health Organization (PAHO) during 2000–2014 and published on their website, 83 (7%) of 1164 of sylvatic YF cases reported from Bolivia, Brazil, Colombia, and Peru occurred in individuals who reported receiving YF vaccine 46 . More recently during the large outbreaks of YF in Brazil, an epidemiologic bulletin noted at least 11 cases of WT YF in individuals who were previously vaccinated and several more cases have been noted during a recent meeting 47 , 51 . Unfortunately, the information about these additional cases is very limited. It is unknown if these cases represent primary or secondary vaccine failures, whether and what confirmatory laboratory testing was performed, and the underlying medical history of the cases (e.g., immunosuppressed or compromised) that might have impacted their initial immunologic response to the vaccine or longer-term immunologic memory. Critically, given that YF IgM antibodies can persist for years following vaccination 52 , obtaining information about how the diagnosis of WT YF disease was made is important to interpret these results. Furthermore, it is important to note that not all individuals respond to YF vaccination; there is a median seroconversion rate of 99% (range 81–100%) in clinical trials 8 . Critically, for a state like Minas Gerais in Brazil with a population over 20 million, this means that even with 100% vaccination coverage more than 200,000 individuals who were vaccinated would fail to develop an immune response to the vaccine and would be at risk for developing disease if exposed.

Seropositivity in vaccinated individuals

Since the SAGE recommendations in 2013, a number of articles have been published related to the immune response seen following YF vaccine, including cohorts of individuals in endemic and non-endemic locations, of different ages, and at different time points following vaccination. All studies used PRNT or microneutralization test for the detection of neutralizing antibodies against YF virus. However, the percent plaque reduction cut-off used and the definition of seropositivity or protection varied by study such that quantity of neutralizing antibodies measured in different studies are difficult to compare 35 . Furthermore, several of the studies did not use the international standard making comparison of seropositivity or antibody concentrations between studies further challenging 53 . The findings of these studies are summarized below.

Humoral immunity in adults

There are data on longer-term humoral immunity for at least eight distinct cohorts of adults in both YF endemic and non-endemic areas of the world who received a full dose of YF vaccine (Table 1 ) 27 , 28 , 31 , 32 , 54 , 55 , 56 , 57 . Notably, there were no apparent differences between studies undertaken in endemic and non-endemic countries. In the first 5 years post-vaccination, seropositivity in the cohorts was >90%. At ≥10 years post-vaccination, the rates of seropositivity were generally lower ranging from 67% to 88% using PRNT 50 –PRNT 90 , except for a small cohort of healthy volunteers in the Netherlands where 97% (34/35) of individuals vaccinated with a full-dose of the vaccine were seropositive at 10 years when measured with PRNT 80 57 . Interestingly, several of the studies saw higher rates of seropositivity 30–35 years post-vaccination compared to rates at 10–20 years post vaccination 54 , 56 . However, the number of individuals in the later vaccination time points are quite limited and they likely received an older vaccination formulation, which have differing quantities of vaccine virus 8 , impacting the generalizability of these results. Several other factors likely impact the overall rates of seropositivity in these studies, such as (1) proof of vaccination 27 , (2) different seropositivity cut-offs 28 , 32 , 35 , (3) different individuals at each time point post-vaccination often with different demographic (e.g., age of vaccination) 27 , 28 , 30 , 56 , (4) potential natural boosting for residents and travelers to endemic areas, and (5) potentially receiving an additional doses of YF vaccine 31 .

Humoral immunity in children

There have been four additional published studies with short-term and long-term immunogenicity for children receiving a full dose of YF vaccine (Table 2 ). The published studies contain cohorts of children who received YF vaccination at 9–23 months of age. Of the two studies published evaluating the seroconversion rate following YF vaccination in children, the rates are highly variable within one of the studies and between the studies 58 , 59 . In a study of 595 children living in Colombia and Peru who received YF vaccine alone or with a tetravalent dengue vaccine on a YF vaccine backbone, the rate of seroconversion was noted to be 99.8–100% when measured by PRNT 50 and titer ≥10 58 . These rates were similar though slightly higher than the rates seen in Mali (95–98%) among children who received a meningococcal A (Men A) vaccine either concurrently or serially with YF vaccine 59 . However, in the same Men A vaccine study, children in Ghana only achieved 68–79% seroconversion rates following YF vaccination. This same trend in lower rates of detectable antibodies between the two populations in the Men A study was seen when the cohorts were followed up at 2–6 years post-vaccination 34 . Seropositivity rates as low as 28% were reported for children in Ghana at 2.3 years post-vaccination, though the rate increased at 6 years post-vaccination to 43%, compared to 50% seropositivity among the children in Mali at 4.5 years post-vaccination 34 . When demographic (age of vaccination, sex), vaccination and exposure history (season of vaccination and pre-vaccination titers), and nutritional status were compared between the children in Mali and Ghana, no significant differences were identified to explain the different rates of seropositivity between these two populations 60 . In the second study evaluating longer-term immunity in different cohorts of children in Brazil up to 10 years post-vaccination, a substantial decline was noted in the seropositivity rates over time 33 . Using a titer ≥10 with PRNT 50 , 54% of children were not seropositive at 7 years post-vaccination. Although the rates of seropositivity increased when using a lower titer cut-off (PRNT ≥ 5), 36% of children at 7 years post-vaccination lacked detectable neutralizing antibodies.

One potential explanation for the varying immune response both initially and potentially longer-term among the pediatric studies could be the age at which the children received their vaccine. Younger age groups might be expected to have a less robust initial immune response, potential immunologic interference from maternal antibodies, or more concomitant infections lead to a decreased immune response 61 , 62 . The cohorts in Mali, Ghana, and some of the children in the Brazil study received YF vaccine at 9 months of age. This is compared to children in Colombia and Peru who received the vaccine at 12 months of age and others in the Brazil cohort who were as old as 23 months when they were vaccinated. However, when the age of vaccination was assessed by the ACIP YF working group relative to the seroconversion rates, the analysis of results from aggregated studies found no difference in seroconversion rates when the children were vaccinated at 9 months of age compared to 12 months 10 , 22 .

With these new pediatric data, there are seemingly more questions than answers to the variability of the results between the pediatric cohorts. The authors of the studies and associated editorials question what contributes to the variability in results hypothesizing that it could be due to differences in immune microenvironment, vaccine substrains used, how the samples were handled, the test used, and potential difference in vaccine handling 33 , 61 , 63 , 64 . Furthermore, in both Ghana and Brazil, the authors questioned whether or not children had received another dose of the vaccine as the proportion seropositive was higher at later time points 33 , 34 .

Additional immunogenicity data

Since 2013, several studies have been published regarding cellular immunity, including CD8+, CD4+, and memory phenotypes, formed in response to YF vaccine 30 , 54 , 55 , 65 . However, the specific impact of alternative types of immunologic memory and their role in protecting persons against disease is not well-characterized or known.

The studies published since SAGE and ACIP made their recommendation that one dose of YF vaccine is sufficient to provide lifelong protection in most individuals provide additional data on YF vaccine immunity. Given the heterogeneity of results, in particular for the pediatric cohorts, further studies would be welcomed.

However, the basic questions that were debated in the discussions of both SAGE and ACIP still remain, how durable is the immunity elicited by YF vaccine and what constitutes protective immunity against YF virus infection and disease? To truly address these questions, additional research and data are needed. Increased transparency and sharing of information on potential vaccine failures are critical to better understand of the >800 million doses the vaccine that have been administered how many might have failed to provide both short-term and long-term protective immunity. With this is the need to continue improving and strengthening YF disease surveillance and laboratory testing 66 , not only to detect possible vaccine failures but also to obtain samples early enough to make a definitive diagnosis of WT disease by molecular testing. In addition, every effort must be made to ascertain the vaccination status of the patient. As noted above, using standards and evaluating the correlation between neutralization titers determined by LNI and PRNT would improve our ability to compare studies and begin to set thresholds as to what antibodies levels are needed to potentially prevent WT disease. Furthermore, additional research is needed to determine the kinetics of the immune response when a vaccinee receives a booster vaccine dose or has a WT infection (e.g., does an amnestic response mean a lack of adequate protection?) and to validate the immune correlate of protection following YF vaccination using more modern knowledge and techniques (e.g., assessing the role of cellular immunity). WHO currently plans to receive input from subject matter experts on how best to proceed with measuring YF vaccine immunity in a consistent manner to allow for comparability between studies.

Overall, we expect the debate of whether or not to give booster doses of YF vaccine to continue in lieu of more data. However, one clear public health action that can and should be taken now is to improve YF vaccination coverage among children living in at risk areas. Based on WHO and UNICEF estimates of vaccine coverage (WUENIC), YF vaccination rates among children living in YF endemic areas ranges from 42% to 97% (median of 85%) in the Americas and 29–94% (median: 68%) in Africa 67 . The current large outbreaks of measles throughout the world, including in YF endemic areas where the vaccines are often given at the same visit, reinforces poor YF vaccination rates that exist among children. If children do not even receive their first dose of YF vaccine, it is hard to focus on whether they might need a booster dose. We encourage researchers, clinicians, and public health officials to continue to evaluate and publish quality data on YF vaccine immunity and vaccine failures to inform public health policy related to YF vaccine use and optimize our ability to prevent YF.

Johansson, M. A., Vasconcelos, P. F. & Staples, J. E. The whole iceberg: estimating the incidence of yellow fever virus infection from the number of severe cases. Trans. R. Soc. Trop. Med. Hyg. 108 , 482–487 (2014).

PubMed   PubMed Central   Google Scholar  

Ho, Y. L. et al. Severe yellow fever in Brazil: clinical characteristics and management. J. Travel Med. 26 , https://doi.org/10.1093/jtm/taz040 (2019).

Rezende, I. M. et al. Persistence of Yellow fever virus outside the Amazon Basin, causing epidemics in Southeast Brazil, from 2016 to 2018. PLoS Negl. Trop. Dis. 12 , e0006538 (2018).

Pan American Health Organization. Epidemiologic Update: Yellow Fever. https://www.paho.org/hq/index.php?option=com_docman&view=download&category_slug=yellow-fever-2194&alias=43619-16-february-2018-yellow-fever-epidemiological-update-619&Itemid=270&lang=en (2018).

Moussallem, T. M. et al. Yellow fever outbreak in a rural–urban mixed community of Espirito Santo, Brazil: epidemiological aspects. Rev. Panam. salud publ. = Pan Am. J. Public Health 43 , e29 (2019).

Google Scholar  

Shearer, F. M. et al. Global yellow fever vaccination coverage from 1970 to 2016: an adjusted retrospective analysis. Lancet Infect. Dis. 17 , 1209–1217 (2017).

Barrett, A. D. Yellow fever vaccines. Biologicals 25 , 17–25 (1997).

CAS   PubMed   Google Scholar  

Staples, J. E., Monath, T. P., Gershman, M. D. & Barrett, A. D. T. In Plotkin’s Vaccines (eds Plotkin, S. A., Orenstein, W. A., Offit, P. A. & Edwards, K. M.) 1181–1265 (Elsevier Inc., 2018).

Casey, R. M. et al. Immunogenicity of fractional-dose vaccine during a yellow fever outbreak—final report. N. Engl. J. Med. 381 , 444–454 (2019).

Advisory Committee on Immunization Practices (ACIP). Grading of Recommendations, Assessment, Development, And Evaluation (GRADE) for Use of Yellow Fever Vaccine Booster Doses. https://www.cdc.gov/vaccines/acip/recs/grade/yf-vac-boost.html (2015).

Mason, R. A., Tauraso, N. M., Spertzel, R. O. & Ginn, R. K. Yellow fever vaccine: direct challenge of monkeys given graded doses of 17D vaccine. Appl. Microbiol. 25 , 539–544 (1973).

CAS   PubMed   PubMed Central   Google Scholar  

Staples, J. E., Gershman, M. & Fischer, M. Centers for Disease Control & Prevention. Yellow fever vaccine: recommendations of the Advisory Committee on Immunization Practices (ACIP). Morb. Mortal. Wkly. Rep. 59 (RR07), 1–27 (2010).

Elliott, M. Yellow fever in the recently inoculated. Trans. R. Soc. Trop. Med. Hyg. 38 , 231–234 (1944).

Courtois, G. [Duration of immunity after yellow fever vaccination]. Ann. Soc. Belg. Med. trop. 34 , 9–12 (1954).

CAS   Google Scholar  

Dick, G. W. & Gee, F. L. Immunity to yellow fever nine years after vaccination with 17D vaccine. Trans. R. Soc. Trop. Med. Hyg. 46 , 449–458 (1952).

Groot, H. & Riberiro, R. B. Neutralizing and haemagglutination-inhibiting antibodies to yellow fever 17 years after vaccination with 17D vaccine. Bull. World Health Organ. 27 , 699–707 (1962).

Rosenzweig, E. C., Babione, R. W. & Wisseman, C. L. Jr. Immunological studies with group B arthropod-borne viruses. IV. Persistence of yellow fever antibodies following vaccination with 17D strain yellow fever vaccine. Am. J. Trop. Med. Hyg. 12 , 230–235 (1963).

Gotuzzo, E., Yactayo, S. & Cordova, E. Efficacy and duration of immunity after yellow fever vaccination: systematic review on the need for a booster every 10 years. Am. J. Trop. Med. Hyg. 89 , 434–444 (2013).

World Health Organization. SAGE Yellow fever work. Background Paper Yellow Fever Vaccines. http://www.who.int/entity/immunization/sage/meetings/2013/april /1_Background_Paper_Yellow_Fever_Vaccines.pdf?ua=1 (2013).

World Health Organization. Vaccines and vaccination against yellow fever. WHO position paper–June 2013. Wkly. Epidemiol. Rec. 88 , 269–283 (2013).

World Health Organization. International and Traveler Health: World—Yellow Fever Vaccination Booster . http://www.who.int/ith/updates/20140605/en/ (2014).

Staples, J. E., Bocchini, J. A. Jr., Rubin, L. & Fischer, M. Centers for Disease Control and Prevention (CDC). Yellow fever vaccine booster doses: recommendations of the Advisory Committee on Immunization Practices, 2015. Morbid. Mortal. Wkly. Rep. 64 , 647–650 (2015).

Grobusch, M. P. et al. Yellow fever revaccination guidelines change—a decision too feverish? Clin. Microbiol. Infect. 19 , 885–886 (2013).

Plotkin, S. A. Ten yearly yellow fever booster vaccinations may still be justified. J. Travel Med. 25 , https://doi.org/10.1093/jtm/tay130 (2018).

Vasconcelos, P. F. Single shot of 17D vaccine may not confer life-long protection against yellow fever. Mem. Inst. Oswaldo Cruz 113 , 135–137 (2018).

PubMed   Google Scholar  

Amanna, I. J. & Slifka, M. K. Questions regarding the safety and duration of immunity following live yellow fever vaccination. Expert Rev. Vaccines 15 , 1519–1533 (2016).

Kareko, B. W. et al. Persistence of neutralizing antibody responses among yellow fever virus 17D vaccinees living in a nonendemic setting. J. Infect. Dis. 221 , 2018–2025 (2020).

Collaborative group for studies on yellow fever. Duration of post-vaccination immunity against yellow fever in adults. Vaccine 32 , 4977–4984 (2014).

Collaborative group for studies on yellow fever. Duration of immunity in recipients of two doses of 17DD yellow fever vaccine. Vaccine 37 , 5129–5135 (2019).

Campi-Azevedo, A. C. et al. 17DD yellow fever revaccination and heightened long-term immunity in populations of disease-endemic areas, Brazil. Emerg. Infect. Dis. 25 , 1511–1521 (2019).

Miyaji, K. T. et al. Prevalence and titers of yellow fever virus neutralizing antibodies in previously vaccinated adults. Rev. Inst. Med. Trop. Sao Paulo 59 , e2 (2017).

de Menezes Martins, R. et al. Duration of post-vaccination immunity to yellow fever in volunteers eight years after a dose-response study. Vaccine 36 , 4112–4117 (2018).

de Noronha, T. G. et al. Duration of post-vaccination humoral immunity against yellow fever in children. Vaccine 37 , 7147–7154 (2019).

Domingo, C. et al. Long-term immunity against yellow fever in children vaccinated during infancy: a longitudinal cohort study. Lancet Infect. Dis. 19 , 1363–1370 (2019).

Barrett, A. D. T. Yellow fever vaccine: the conundrum of 2 doses, one dose, or one-fifth dose to induce and maintain protective immunity. J. Infect. Dis. 221 , 1922–1924 (2020).

World Health Organization. Guidelines for Plaque Reduction Neutralization Testing of Human Antibodies to Dengue Viruses. https://apps.who.int/iris/bitstream/handle/10665/69687/who_ivb_07.07_eng.pdf;jsessionid=19BF765D9B3FDAD85EF3F217B0165953?sequence=1 (2007).

Hombach, J., Solomon, T., Kurane, I., Jacobson, J. & Wood, D. Report on a WHO consultation on immunological endpoints for evaluation of new Japanese encephalitis vaccines, WHO, Geneva, 2–3 September, 2004. Vaccine 23 , 5205–5211 (2005).

Roehrig, J. T., Hombach, J. & Barrett, A. D. Guidelines for plaque-reduction neutralization testing of human antibodies to dengue viruses. Viral Immunol. 21 , 123–132 (2008).

Johansson, M. A., Arana-Vizcarrondo, N., Biggerstaff, B. J. & Staples, J. E. Incubation periods of Yellow fever virus. Am. J. Trop. Med. Hyg. 83 , 183–188 (2010).

Visser, L. G., Veit, O. & Chen, L. H. Waning immunity after single-dose yellow fever vaccination: who needs a second shot? J. Travel Med. 26 , https://doi.org/10.1093/jtm/tay134 (2019).

Kennedy, R. B., Lane, J. M., Henderson, D. A. & Poland, G. A. in Plotkin’s Vaccine (eds Plotkin, S. A., Orenstein, W. A., Offit, P. A. & Edwards, K. M.) 1001–1030 (Elsevier Inc., 2018).

Erdman, D. D., Heath, J. L., Watson, J. C., Markowitz, L. E. & Bellini, W. J. Immunoglobulin M antibody response to measles virus following primary and secondary vaccination and natural virus infection. J. Med. Virol. 41 , 44–48 (1993).

Hepburn, M. J. et al. Neutralizing antibody response to booster vaccination with the 17d yellow fever vaccine. Vaccine 24 , 2843–2849 (2006).

Brasil Ministerio da Saude. [Yellow Fever Seasonal Monitoring Brasil—2017/2018] https://portalarquivos2.saude.gov.br/images/pdf/2018/outubro/08/Informe-FA.pdf (2018).

Camara, F. P., de Carvalho, L. M. & Gomes, A. L. Demographic profile of sylvatic yellow fever in Brazil from 1973 to 2008. Trans. R. Soc. Trop. Med. Hyg. 107 , 324–327 (2013).

Pan American Health Organization. Yellow Fever: Vaccination Status Among Confirmed Cases in the Americas, 2000–2014. http://ais.paho.org/phip/viz/ed_yellowfever.asp (2019).

Minas Gerais Secretaria de Estado de Saude. [Epidemiologic Information on Yellow Fever]. http://www.saude.mg.gov.br/fornecimento-de-medicamentos/story/10284-informe-epidemiologico-da-febre-amarela-27-02 (2018).

Tuboi, S. H., Costa, Z. G., da Costa Vasconcelos, P. F. & Hatch, D. Clinical and epidemiological characteristics of yellow fever in Brazil: analysis of reported cases 1998–2002. Trans. R. Soc. Trop. Med. Hyg. 101 , 169–175 (2007).

Filippis, A. M. et al. Isolation and characterization of wild type yellow fever virus in cases temporally associated with 17DD vaccination during an outbreak of yellow fever in Brazil. Vaccine 22 , 1073–1078 (2004).

Saraiva, M. et al. Historical analysis of the records of sylvan yellow fever in the State of Amazonas, Brazil, from 1996 to 2009. Rev. Soc. Bras. Med. Trop. 46 , 223–226 (2013).

Fereguetti, T. Novel clinical findings associated with YF in rural Brazil in Symposium 69: American Committee on Arthropod-Borne Viruses (ACAV) Symposium II: Everything Old Is New Again - The Re-Emergence of Yellow Fever. 68th Annual Meeting of American Society of Tropical Medicine and Hygiene (National Harbor, MD, USA, 2019).

Gibney, K. B. et al. Detection of anti-yellow fever virus immunoglobulin m antibodies at 3–4 years following yellow fever vaccination. Am. J. Trop. Med. Hyg. 87 , 1112–1115 (2012).

Ferguson, M. Collaborative study to assess the suitability of a candidate International Standard for yellow fever vaccine. Biologicals 32 , 195–205 (2004).

Wieten, R. W. et al. A single 17D yellow fever vaccination provides lifelong immunity; characterization of yellow-fever-specific neutralizing antibody and T-cell responses after vaccination. PLoS ONE 11 , e0149871 (2016).

Campi-Azevedo, A. C. et al. Short-lived immunity after 17DD yellow fever single dose indicates that booster vaccination may be required to guarantee protective immunity in child. Front. Immunol. 10 , 2192 (2019).

Lindsey, N. P. et al. Persistence of yellow fever virus-specific neutralizing antibodies after vaccination among US travellers. J. Travel Med. 25 , https://doi.org/10.1093/jtm/tay108 (2018).

Roukens, A. H. E., van Halem, K., de Visser, A. W. & Visser, L. G. Long-term protection after fractional-dose yellow fever vaccination: follow-up study of a randomized, controlled, noninferiority trial. Ann. Intern. Med. 169 , 761–765 (2018).

Lopez, P. et al. Immunogenicity and safety of yellow fever vaccine (stamaril) when administered concomitantly with a tetravalent dengue vaccine candidate in healthy toddlers at 12–13 months of age in Colombia and Peru: a randomized trial. Pediatr. Infect. Dis. J. 35 , 1140–1147 (2016).

Roy Chowdhury, P. et al. Immunogenicity of yellow fever vaccine coadministered with MenAfriVac in healthy infants in Ghana and Mali. Clin. Infect. Dis. 61 , S586–593 (2015).

Idoko, O. T. et al. Antibody responses to yellow fever vaccine in 9 to 11-month-old Malian and Ghanaian children. Expert Rev. Vaccines 18 , 867–875 (2019).

Muyanja, E. et al. Immune activation alters cellular and humoral responses to yellow fever 17D vaccine. J. Clin. Investig. 124 , 3147–3158 (2014).

Albrecht, P., Ennis, F. A., Saltzman, E. J. & Krugman, S. Persistence of maternal antibody in infants beyond 12 months: mechanism of measles vaccine failure. J. Pediatr. 91 , 715–718 (1977).

Vasconcelos, P. F. C. & Barrett, A. D. T. Are booster doses of yellow fever vaccine needed?. Lancet Infect. Dis. 19 , 1275–1276 (2019).

Domingo, C. & Niedrig, M. Safety of 17D derived yellow fever vaccines. Expert Opin. Drug Saf. 8 , 211–221 (2009).

Kongsgaard, M. et al. Adaptive immune responses to booster vaccination against yellow fever virus are much reduced compared to those after primary vaccination. Sci. Rep. 7 , 662 (2017).

Berkley, S. Health security’s blind spot. Science 359 , 1075 (2018).

Adrien, N. et al. Differences between coverage of yellow fever vaccine and the first dose of measles-containing vaccine: a desk review of global data sources. Vaccine 37 , 4511–4517 (2019).

Wieten, R. W. et al. 17D yellow fever vaccine elicits comparable long-term immune responses in healthy individuals and immune-compromised patients. J. Infect. 72 , 713–722 (2016).

Martins, R. M. et al. 17DD yellow fever vaccine: a double blind, randomized clinical trial of immunogenicity and safety on a dose-response study. Hum. Vaccines Immunother. 9 , 879–888 (2013).

Download references

Acknowledgements

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention and World Health Organization.

Author information

Authors and affiliations.

Arboviral Diseases Branch, U.S. Centers for Disease Control and Prevention, Fort Collins, CO, USA

J. Erin Staples

Department of Pathology and Sealy Institute for Vaccine Sciences, University of Texas Medical Branch, Galveston, TX, USA

Alan D. T. Barrett

Institute of Public Health, University of Heidelberg, Heidelberg, Germany

Annelies Wilder-Smith

London School of Hygiene and Tropical Medicine, London, UK

World Health Organization, Geneva, Switzerland

Joachim Hombach

You can also search for this author in PubMed   Google Scholar

Contributions

J.E.S., A.D.T.B., and J.H. contributed to the conception of the manuscript; J.E.S., A.D.T.B., A.W.-S., and J.H. contributed to reviewing and interpreting available literature; J.E.S., A.D.T.B., and A.W.-S. contributed to drafting the manuscript; and J.E.S., A.D.T.B., A.W.-S., and J.H. contributed to reviewing and editing the manuscript. All authors approved the submitted version and agreed both to be personally accountable for the author’s own contributions and to ensure that questions related to the accuracy or integrity of any part of the work.

Corresponding author

Correspondence to J. Erin Staples .

Ethics declarations

Competing interests.

J.E.S., A.W.-S., J.H. declares no competing interests. A.D.T.B. is Editor-in-Chief of npj Vaccines.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Staples, J.E., Barrett, A.D.T., Wilder-Smith, A. et al. Review of data and knowledge gaps regarding yellow fever vaccine-induced immunity and duration of protection. npj Vaccines 5 , 54 (2020). https://doi.org/10.1038/s41541-020-0205-6

Download citation

Received : 13 January 2020

Accepted : 29 May 2020

Published : 06 July 2020

DOI : https://doi.org/10.1038/s41541-020-0205-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

This article is cited by

Genomic diversity contributes to the neuroinvasiveness of the yellow fever french neurotropic vaccine.

• Florian Bakoa
• Christophe Préhaud
• Nolwenn Jouvenet

npj Vaccines (2021)

A single-dose live-attenuated YF17D-vectored SARS-CoV-2 vaccine candidate

• Lorena Sanchez-Felipe
• Thomas Vercruysse
• Kai Dallmeier

Nature (2021)

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Loading metrics

Open Access

Peer-reviewed

Research Article

Yellow Fever in Africa: Estimating the Burden of Disease and Impact of Mass Vaccination from Outbreak and Serological Data

Affiliation MRC Centre for Outbreak Analysis, Department of Infectious Disease Epidemiology, Imperial College London, United Kingdom

Affiliation World Health Organization, Geneva, Switzerland

Affiliation Immunization and Vaccine Development, World Health Organization, Ouagadougou, Burkina Faso

Affiliation Ottawa Public Health, Ottawa, Ontario, Canada

Affiliation Arboviral Disease Branch, Centers for Disease Control and Prevention, Fort Collins, Colorado, United States of America

* E-mail: [email protected]

¶ Membership of the Yellow Fever Expert Committee is provided in the Acknowledgments.

• Tini Garske, 
• Maria D. Van Kerkhove, 
• Sergio Yactayo, 
• Olivier Ronveaux, 
• Rosamund F. Lewis, 
• J. Erin Staples, 
• William Perea, 
• Neil M. Ferguson, 
• for the Yellow Fever Expert Committee

• Published: May 6, 2014
• https://doi.org/10.1371/journal.pmed.1001638
• Reader Comments

Yellow fever is a vector-borne disease affecting humans and non-human primates in tropical areas of Africa and South America. While eradication is not feasible due to the wildlife reservoir, large scale vaccination activities in Africa during the 1940s to 1960s reduced yellow fever incidence for several decades. However, after a period of low vaccination coverage, yellow fever has resurged in the continent. Since 2006 there has been substantial funding for large preventive mass vaccination campaigns in the most affected countries in Africa to curb the rising burden of disease and control future outbreaks. Contemporary estimates of the yellow fever disease burden are lacking, and the present study aimed to update the previous estimates on the basis of more recent yellow fever occurrence data and improved estimation methods.

Methods and Findings

Generalised linear regression models were fitted to a dataset of the locations of yellow fever outbreaks within the last 25 years to estimate the probability of outbreak reports across the endemic zone. Environmental variables and indicators for the surveillance quality in the affected countries were used as covariates. By comparing probabilities of outbreak reports estimated in the regression with the force of infection estimated for a limited set of locations for which serological surveys were available, the detection probability per case and the force of infection were estimated across the endemic zone.

The yellow fever burden in Africa was estimated for the year 2013 as 130,000 (95% CI 51,000–380,000) cases with fever and jaundice or haemorrhage including 78,000 (95% CI 19,000–180,000) deaths, taking into account the current level of vaccination coverage. The impact of the recent mass vaccination campaigns was assessed by evaluating the difference between the estimates obtained for the current vaccination coverage and for a hypothetical scenario excluding these vaccination campaigns. Vaccination campaigns were estimated to have reduced the number of cases and deaths by 27% (95% CI 22%–31%) across the region, achieving up to an 82% reduction in countries targeted by these campaigns. A limitation of our study is the high level of uncertainty in our estimates arising from the sparseness of data available from both surveillance and serological surveys.

Conclusions

With the estimation method presented here, spatial estimates of transmission intensity can be combined with vaccination coverage levels to evaluate the impact of past or proposed vaccination campaigns, thereby helping to allocate resources efficiently for yellow fever control. This method has been used by the Global Alliance for Vaccines and Immunization (GAVI Alliance) to estimate the potential impact of future vaccination campaigns.

Please see later in the article for the Editors' Summary

Editors' Summary

Yellow fever is a flavivirus infection that is transmitted to people and to non-human primates through the bites of infected mosquitoes. This serious viral disease affects people living in and visiting tropical regions of Africa and Central and South America. In rural areas next to forests, the virus typically causes sporadic cases or even small-scale epidemics (outbreaks) but, if it is introduced into urban areas, it can cause large explosive epidemics that are hard to control. Although many people who contract yellow fever do not develop any symptoms, some have mild flu-like symptoms, and others develop a high fever with jaundice (yellowing of the skin and eyes) or hemorrhaging (bleeding) from the mouth, nose, eyes, or stomach. Half of patients who develop these severe symptoms die. Because of this wide spectrum of symptoms, which overlap with those of other tropical diseases, it is hard to diagnose yellow fever from symptoms alone. However, serological tests that detect antibodies to the virus in the blood can help in diagnosis. There is no specific antiviral treatment for yellow fever but its symptoms can be treated.

Why Was This Study Done?

Eradication of yellow fever is not feasible because of the wildlife reservoir for the virus but there is a safe, affordable, and highly effective vaccine against the disease. Large-scale vaccination efforts during the 1940s, 1950s, and 1960s reduced the yellow fever burden for several decades but, after a period of low vaccination coverage, the number of cases rebounded. In 2005, the Yellow Fever Initiative—a collaboration between the World Health Organization (WHO) and the United Nations Children Fund supported by the Global Alliance for Vaccines and Immunization (GAVI Alliance)—was launched to create a vaccine stockpile for use in epidemics and to implement preventive mass vaccination campaigns in the 12 most affected countries in West Africa. Campaigns have now been implemented in all these countries except Nigeria. However, without an estimate of the current yellow fever burden, it is hard to determine the impact of these campaigns. Here, the researchers use recent yellow fever occurrence data, serological survey data, and improved estimation methods to update estimates of the yellow fever burden and to determine the impact of mass vaccination on this burden.

What Did the Researchers Do and Find?

The researchers developed a generalized linear statistical model and used data on the locations where yellow fever was reported between 1987 and 2011 in Africa, force of infection estimates for a limited set of locations where serological surveys were available (the force of infection is the rate at which susceptible individuals acquire a disease), data on vaccination coverage, and demographic and environmental data for their calculations. They estimate that about 130,000 yellow fever cases with fever and jaundice or hemorrhage occurred in Africa in 2013 and that about 78,000 people died from the disease. By evaluating the difference between this estimate, which takes into account the current vaccination coverage, and a hypothetical scenario that excluded the mass vaccination campaigns, the researchers estimate that these campaigns have reduced the burden of disease by 27% across Africa and by up to 82% in the countries targeted by the campaigns (an overall reduction of 57% in the 12 targeted countries).

What Do These Findings Mean?

These findings provide a contemporary estimate of the burden of yellow fever in Africa. This estimate is broadly similar to the historic estimate of 200,000 cases and 30,000 deaths annually, which was based on serological survey data obtained from children in Nigeria between 1945 and 1971. Notably, both disease burden estimates are several hundred-fold higher than the average number of yellow fever cases reported annually to WHO, which reflects the difficulties associated with the diagnosis of yellow fever. Importantly, these findings also provide an estimate of the impact of recent mass vaccination campaigns. All these findings have a high level of uncertainty, however, because of the lack of data from both surveillance and serological surveys. Other assumptions incorporated in the researchers' model may also affect the accuracy of these findings. Nevertheless, the framework for burden estimation developed here provides essential new information about the yellow fever burden and the impact of vaccination campaigns and should help the partners of the Yellow Fever Initiative estimate the potential impact of future vaccination campaigns and ensure the efficient allocation of resources for yellow fever control.

Additional Information

Please access these websites via the online version of this summary at http://dx.doi.org/10.1371/journal.pmed.1001638 .

• The World Health Organization provides detailed information about yellow fever (in several languages), including photo stories about vaccination campaigns in the Sudan and Mali; it also provides information about the Yellow Fever Initiative (in English and French)
• The GAVI Alliance website includes detailed of its support for yellow fever vaccination
• The US Centers for Disease Control and Prevention provides information about yellow fever for the public, travelers, and health care providers
• The UK National Health Service Choices website also has information about yellow fever
• Wikipedia has a page on yellow fever that includes information about the history of the disease (note that Wikipedia is a free online encyclopedia that anyone can edit; available in several languages)

Citation: Garske T, Van Kerkhove MD, Yactayo S, Ronveaux O, Lewis RF, Staples JE, et al. (2014) Yellow Fever in Africa: Estimating the Burden of Disease and Impact of Mass Vaccination from Outbreak and Serological Data. PLoS Med 11(5): e1001638. https://doi.org/10.1371/journal.pmed.1001638

Academic Editor: Simon I. Hay, University of Oxford, United Kingdom

Received: June 7, 2013; Accepted: March 27, 2014; Published: May 6, 2014

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Funding: The research leading to these results has received funding from the Medical Research Council, the Bill & Melinda Gates Foundation, and the European Union Seventh Framework Programme [FP7/2007–2013] under Grant Agreement n°278433-PREDEMICS. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Abbreviations: AUC, area under the curve; BIC, Bayesian Information Criterion; EPI, Enhanced Programme for Immunization; EVI, enhanced vegetation index; GAVI, Global Alliance for Vaccines and Immunization; MCMC, Markov Chain Monte Carlo; MIR, middle infrared reflectance; YFSD, yellow fever surveillance database

Introduction

Yellow fever is a flavivirus infection that is transmitted primarily by mosquitoes of the species Aedes ssp. and Haemagogus spp., with humans and non-human primates being the main vertebrate hosts. It is endemic in tropical areas of Africa and Central and South America. The clinical course of infection in humans shows a wide spectrum of severity including asymptomatic infection, mild illness with flu-like symptoms, and severe disease including fever with jaundice or haemorrhage and death.

Several different transmission cycles have been defined, depending on which host and vector species are involved in transmission: in the sylvatic cycle, tree-dwelling mosquitoes of Aedes spp. (Africa) or Haemagogus spp. (Americas) transmit the virus to non-human primates. In this cycle, spill-over infection of humans occurs when they encroach on this jungle habitat. Conversely, in the urban transmission cycle, humans are the main hosts with transmission occurring via domestic mosquito species. The typical urban vector is Aedes aegyptii , which also serves as the main vector for dengue virus transmission. If yellow fever is introduced into urban areas, large explosive outbreaks can occur, which can be difficult to control. In Africa, there is also an intermediate transmission cycle that occurs in rural areas typically at the edges of forests with humans as well as non-human primates affected, and transmission driven by domestic and semi-domestic mosquito species [1] , [2] .

While eradication of yellow fever is not feasible due to the sylvatic reservoir, a high level of control is achievable owing to the availability of an efficacious and safe vaccine that confers long-lasting immunity from a single dose. Visas for many countries worldwide require proof of previous vaccination against yellow fever, particularly if travelers come from or have visited yellow fever endemic areas, in order to prevent the importation of the disease.

Quantifying the burden of disease caused by yellow fever is made challenging by the wide spectrum of clinical severity, with non-specific symptoms in the majority of infections making diagnosis difficult. In addition, there are considerable limitations in the surveillance and health care systems across much of the affected regions. However, it is clear that yellow fever is substantially underreported [3] , [4] . Previous estimates from the early 1990s placed the burden of disease at 200,000 cases and 30,000 deaths annually [5] , [6] . These estimates relied heavily on data from serological surveys performed in children in Nigeria between 1945 and 1971 [7] . These data still form the basis of more recent efforts to quantify disease burden or the cost-effectiveness of vaccines [8] , [9] . More recent approaches to quantify yellow fever circulation have focused on producing risk maps [10] – [12] , frequently employing regression techniques similar to the approach we adopt [10] , [12] , or relying on expert advice regarding local yellow fever distribution [11] , [12] . However, there are no recent estimates of the yellow fever burden that take into account more recent surveillance and serological data and that account for vaccination coverage.

In 2005, the Yellow Fever Initiative was launched as a collaboration between WHO and the United Nations Children's Fund (UNICEF) with support from the Global Alliance for Vaccines and Immunization (GAVI Alliance). The aim was to secure the precarious yellow fever vaccine supply by creating a vaccine stockpile to be used in outbreak response campaigns as well as to increase the vaccination coverage in the most affected areas by implementation of large preventive mass vaccination campaigns in 12 of the most affected countries in West Africa. Between 2006 and 2012, these campaigns have been implemented in all of these countries apart from Nigeria because of larger than anticipated vaccine needs and limited vaccine supplies. In the same time frame, the Central African Republic, though not covered under the Yellow Fever Initiative, also performed mass vaccination campaigns with support from the GAVI Alliance.

During the October 2011 meeting of the advisory committee on Quantitative Immunization and Vaccine Related Research ([QUIVER], currently named Immunization and Vaccines related Implementation Research [IVIR]), the Advisory Committee recommended that WHO establish a working group to generate updated yellow fever disease burden estimates for Africa. This paper reports the results of this activity, presenting new estimates of the disease burden caused by yellow fever in Africa and the impact of preventive vaccination campaigns carried out under the Yellow Fever Initiative. The estimates are derived from a coherent model framework that integrates all available data including incidence, serology, and vaccination coverage.

We fitted a generalised linear model to the locations where yellow fever was reported in the 25-year period between 1987 and 2011. This model estimated, for each location, the probability of at least one yellow fever report over the observation period. The number of infections required to give rise to these probabilities of occurrence was then estimated by taking into account the probability of detection of yellow fever cases in each country. Estimated numbers of infections were converted to estimates of the force of infection using data on the population size, age distribution, and age-specific vaccination coverage in the observation period. Again using demographic and vaccination coverage data, the burden in terms of the number of infections, severe cases presenting with fever and jaundice or haemorrhage, or deaths can then be obtained from the estimates of the force of infection for each location for any year in the past or future, given assumptions on population growth and size of future vaccination campaigns.

The model was fitted at a spatial resolution of the first sub-national administrative unit (which in many countries is called “province”; this is the terminology adopted throughout this manuscript), so all datasets were resolved or aggregated to this level as appropriate.

Yellow fever occurrence.

A database of the locations of reported outbreaks between 1987 and 2011 was compiled from various sources including the Weekly Epidemiological Record (WER) [13] , the WHO disease outbreak news (DON) [14] , an internal WHO database of outbreaks between 1980 and 2007, and the published literature. Locations were resolved to the province level, and data were recorded for each outbreak on the year of occurrence, size, and control measures implemented. Outbreak reports that could not be located at the province level were excluded.

In 2005, the African Regional Office of WHO established a yellow fever surveillance database (YFSD) of reports of suspected yellow fever cases (based primarily on a case definition of fever with jaundice) across 21 countries in West and central Africa. Data fields recorded for each case included age, gender, location, disease onset date, and the status of laboratory confirmation. The locations of all lab-confirmed cases between 2005 and 2011, resolved to the province level, were combined with the outbreaks dataset to generate an overall dataset of the areas of yellow fever occurrence, recording for each province whether or not there had been at least one yellow fever outbreak or case report in the period from 1987 to 2011.

Due to the very low proportion of suspected cases actually being attributed to yellow fever in the YFSD, the majority of cases reported likely had other causes (for instance viral hepatitis). Hence the national incidence of suspected cases is best interpreted as a measure of the effort put into yellow fever surveillance rather than a measure of yellow fever incidence itself. The incidence of suspected cases was aggregated at the country level and divided by the national population to be used as a covariate in the regression models fitted throughout.

Disease severity.

The proportion of infections presenting as severe cases and the proportion of severe cases resulting in death varies substantially between settings, depending on previous exposure to other flaviviruses, but also factors such as clinical care and importantly detection bias due to surveillance coverage or case definitions used [1] , [15] – [20] . Recent work by Johansson and colleagues [21] has estimated the proportion of infections that are asymptomatic, cause mild symptoms (excluding jaundice and haemorrhage), or severe symptoms (including jaundice, haemorrhage, or death), as well as the proportion of severe cases leading to death. We use these estimates of 13% (95% CI 5%–28%) of infections presenting as severe cases, and 46% (95% CI 31%–60%) of severe cases resulting in death to estimate the number of severe cases and deaths from the number of infections estimated by our model.

Vaccination coverage.

No comprehensive dataset of yellow fever vaccination coverage in the endemic area in Africa was available, so vaccination coverage was estimated using data on (i) large-scale mass vaccination activities in French West Africa during the 1940s to 1960s [22] ; (ii) outbreak response campaigns since 1970, as reported in outbreak reports in the WER or DONs [13] , [14] ; (iii) routine infant yellow fever vaccination occurring as part of the Enhanced Programme for Immunization (EPI) [23] ; and (iv) mass vaccination campaigns in 11 West African countries under the Yellow Fever Initiative and the Central African Republic from 2006 to 2012 [24] , [25] .

Information on yellow fever vaccination was compiled into a dataset of age-specific vaccination coverage at the second sub-national administrative level (district), taking into account the location and extent of each campaign as well as the demographics of the targeted populations. This dataset allowed the achieved coverage to be tracked through time for each birth cohort in each district.

The available information on vaccination activities varied greatly from country to country, sometimes specifying the coverage achieved in a certain area, sometimes the number of doses administered during a vaccination campaign, and sometimes both. If the area targeted by a campaign was well defined geographically we used information on the vaccination coverage achieved by that campaign in preference to the number of doses administered in order to avoid uncertainty in population size affecting our estimates. If no information on the coverage achieved was available or the target population was not sufficiently well defined, we calculated vaccination coverage as the number of doses administered divided by the population size, assuming that individuals from all targeted age groups had an equal chance of receiving the vaccine, and that vaccination was performed irrespective of previous vaccination or disease history.

From the vaccination coverage achieved in individual vaccination campaigns the coverage at the population level over time was obtained by tracking vaccination coverage in each birth cohort. In compiling the vaccination coverage dataset, population movements were ignored, and 100% vaccine efficacy was assumed, with lifelong protection. The last two assumptions are supported by data showing that 99% of individuals seroconvert within 30 days of vaccination [1] , [26] , and neutralising antibodies have been measured 35 years post vaccination [26] – [28] .

In estimating the impact of potential future vaccination campaigns we assumed that no further outbreak response vaccination campaigns would be undertaken and that the country-specific coverage in the infant immunization campaigns would be held constant at the levels estimated for 2011 (see Table S1 ) [23] .

Serological surveys.

Serological surveys have been used historically to assess overall levels of transmission. All literature on yellow fever serologic surveys conducted in Africa and published since 1980 were reviewed and the results collated [21] . For the analysis of transmission intensity, only surveys that had samples tested for yellow fever virus specific neutralising antibodies and were not part of an outbreak investigation were considered [29] – [34] , as surveys conducted in outbreak situations are typically not representative. Even if random population samples are obtained in an outbreak-associated survey, serology would be expected to yield information on the attack rate for that specific outbreak rather than the average force of infection over a longer time period.

Demographic data.

Demographic data on population size and age distribution at a sub-national level were used to interpret the data on vaccination campaigns as well as for estimating the burden. We used UN World population prospects (WPPs) [35] estimates of the population size by country in 5-year age bands for each year between 1950 and 2100. In order to achieve a higher spatial resolution of the population distribution, these estimates were combined with the LandScan 2007 dataset [36] , [37] , which gave estimates for the year 2007 of the total population on a grid of resolution of 1/120 degree latitude and longitude, which is approximately 1 km at the equator. By allocating each grid point to the second sub-national administrative unit (which in many countries is the district), the proportion of each country's population living in any particular district was estimated. In the absence of more detailed datasets, it was assumed that the age distributions were homogeneous within each country, neglecting local differences, for instance between rural and urban areas. We furthermore assumed that population growth was homogeneous within a country, and that the population proportions for each district obtained from the LandScan 2007 dataset were applicable to all other years. Thus we did not capture trends in urbanisation or other shifts in the relative population sizes of different districts over time.

We disaggregated the 5-year age bands of the UN WPP dataset into annual birth cohorts using the method described in Text S1 .

Population based variables for the regression model included the total population for each province, the logarithm of the population size and the proportion of the population living in urban areas (defined as LandScan 2007 dataset pixels with a population density of ≥386 people per sq km [38] ).

Environmental data.

Environmental datasets on rainfall [39] , day- and night-time air temperatures [40] , land cover classifications [41] , [42] , the enhanced vegetation index (EVI), the middle infrared reflectance (MIR) [43] , longitude, latitude, and altitude [44] , [45] were used as potential covariates in the generalised linear model. These data were available as gridded datasets of various spatial resolutions between about 1 km and 10 km, and were aggregated to province level by calculating the mean value for each variable, weighted by the population size attributed to each grid cell in order to obtain values representative of the conditions where human populations are concentrated.

For the land cover classification, the proportion of pixels (weighted by population size) for each category was aggregated for each province to obtain scalar variables. In the endemic zone, some of the 17 defined land cover classes occurred very scarcely or not at all, so we only considered those that accounted to over 5% of the area in at least one province as potential covariates. This resulted in the four categories of evergreen needleleaf forest, deciduous needleleaf forest, mixed forests, and snow and ice being excluded.

For each time-varying variable, the annual mean and the average annual minimum and maximum levels were considered, on the basis of 4-year time series obtained for the period from 2003 to 2006. To evaluate the average annual minimum and maximum, time series were smoothed using Fourier transforms as described by Garske and colleagues [40] . The minimum and maximum of these smoothed curves determined the typical annual minimum and maximum used here. The variable that varied with time were the night- and day-time air temperatures [40] , EVI, MIR [43] , and rainfall [39] .

Prior to fitting, all variables were scaled to unit variance in order to improve model convergence and make the fitted slope parameters comparable.

Model Structure and Fitting

The overall model consisted of several components that were fitted jointly using standard Markov Chain Monte Carlo (MCMC) techniques [46] , [47] .

Generalised linear model for the presence/absence of yellow fever reports.

As the occurrence of yellow fever certainly depends on environmental factors such as climate, land cover, but also the human population size, several environmental variables were considered as potential covariates. However, the number of such potential covariates was large, so the first step in variable selection was to fit univariate models to the dataset including each of the potential covariates in turn. Any variables that were not significantly associated with the data at the 10% confidence limit were excluded from further consideration. Some of the remaining variables were highly correlated, and inclusion of highly correlated variables in regression models can lead to instabilities in the parameter estimates. In order to avoid these problems, covariates were clustered into highly correlated groups, where the absolute pairwise correlation between any two variables within a group was above 0.75. A single variable from each group was then selected as a potential covariate in the regression modeling.

Multivariate models were fitted using the function glm in R version 2.14.2. These models included an intercept, the log surveillance quality indicator at the country level obtained from the YFSD and a factor for each country not included in that database as well as any possible combination of up to 12 additional environmental covariates. The model fit was evaluated using the Bayesian Information Criterion (BIC) [49] , and the 15 best models were further investigated in the full model framework.

From model predictions to transmission intensity.

Serological surveys and detection probability.

Estimating the burden from transmission intensity.

While the number of infections is the most relevant quantity for assessing the degree of transmission of yellow fever, morbidity and mortality estimates are required to assess the impact on populations and health care systems. In order to calculate the number of severe cases and deaths from the infections, we fitted beta distributions to the point estimates and 95% credibility intervals of the proportion of cases among infections and the case fatality ratio estimated by Johansson and colleagues [21] and generated samples from both distributions that we then multiplied by the number of infections estimated during each MCMC iteration. This approach allowed us to include the uncertainty of the severity spectrum in our burden estimates.

Model fitting.

The model fit of the full model was evaluated via BIC, as this takes into account both fit quality (measured by the log likelihood) while penalizing models with a large number of parameters. In addition, we calculated receiver operator characteristic (ROC) curves comparing the regression model predictions with the yellow fever presence/absence data to which the regression models were fitted, and the area under the ROC curve (AUC), which quantifies how well the regression model predictions matched the data [50] . A lower value of the BIC indicates a better model fit, whereas a value of the AUC of 0.5 indicates that model predictions are no better than chance, and a value of 1 corresponds to a perfect fit to the data.

Sensitivity analyses.

While the model inference framework adopted gives parameter estimates and credible intervals around these, there were however other sources of uncertainty that were more difficult to quantify, some of which were assessed in sensitivity analyses.

The impact of the choice of covariates included in the regression model was assessed by comparing the final burden estimates obtained for a number of the best fitting regression models. The model that ranked best in the initial fits of the linear regression model was used as the baseline model and is presented in the main paper, whereas results from the remaining models are shown in Text S2 .

Sensitivity to the magnitude of the standard deviation of the Gaussian prior distribution on the country factors was explored (see Text S3 ).

The vaccination coverage dataset compiled for this study suffers from a number of uncertainties in the input datasets that are difficult to quantify, including uncertain population sizes that impact directly the vaccination coverage achieved with a given number of doses, uncertainties about the completeness and accuracy of the records of past vaccination activities, and the influence of population movements on vaccination coverage. In order to explore the potential impact of these sources of uncertainty on the burden estimates, we generated five alternative vaccination coverage scenarios: assuming only 90% vaccine efficacy, alternative lower or higher population sizes, non-random vaccine allocation, and an alternative scenario of the historic mass vaccination campaigns based on different records [51] . We used these to assess the impact of uncertainty of coverage estimates on the overall estimates of disease burden (see Text S4 for further details).

Last, we also considered two refined model structures that relaxed the assumption that the probability of case detection via routine surveillance was constant through time (see Text S5 ).

Yellow Fever Occurrence

Between 1980 and 2012, 150 yellow fever outbreaks in 26 countries in Africa were reported to WHO ( Figure S1 ). A high number of large outbreaks occurred in the late 1980s and early 1990s, particularly in Nigeria, as well as a large number of relatively smaller outbreaks in West and later central Africa since the turn of the century.

The YFSD contained records of 29,237 suspected cases of yellow fever from 21 countries reported between 2005 and 2011, 302 of which were lab-confirmed, 231 classified as epidemiologically linked to a lab-confirmed case, and 416 as compatible with yellow fever based on symptoms and epidemiology, with the remaining cases considered not due to yellow fever after investigation. The locations of the lab-confirmed, linked, and compatible cases resolved to the province level are shown in Figure S2A , whereas the combined dataset of the presence or absence of yellow fever reports by province is shown in Figure 1A .

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

(A) Presence/absence of yellow fever over a 25-year period, by province. White, absence; red, presence of yellow fever reports. (B) Model predictions giving the estimated probability of at least one yellow fever report. (C) Estimates of the annual force of infection at the province level in the 32 countries considered endemic for yellow fever. (D) Estimates of the country-specific detection probability per infection. Countries not considered endemic for yellow fever are shown in navy (A, B, and D) or white (C).

https://doi.org/10.1371/journal.pmed.1001638.g001

The country-specific surveillance quality (defined as the mean annual number of reported suspected cases divided by the national population) is shown in Figure S2B . While there were suspect cases reported from 21 countries, the YFSD included only five suspect cases reported from Angola, none of which were confirmed. It was therefore assumed that this country did not participate effectively in the YFSD, reducing the number of countries included to 20.

Vaccination Coverage

The estimated vaccination coverage over time clearly shows the success of the mass vaccination campaigns in French West Africa between 1940 and 1960, and declining levels of immunity in the following decades caused by low vaccination levels, the birth of new unvaccinated cohorts, and the gradual depletion of the older protected cohorts through mortality. Between 1960 and 2000 there was limited vaccination activity across Africa resulting from disjointed reactive vaccination campaigns. Mass vaccination campaigns implemented since 2006 in the framework of the GAVI investment have achieved much higher coverage levels in West Africa ( Figure S3 ). The impact of infant immunization on coverage at the population level will take time to develop, but if this is pursued in the future and high coverage of new birth cohorts is achieved, it will eventually lead to a high coverage even in countries with currently low population-wide coverage. In countries that currently have high population-level coverage, infant immunization will prevent a repetition of the decline in vaccination coverage observed from the 1960s onwards.

Regression Model Fitting and Variable Selection

All models included log[surveillance quality] and country factors for those countries for which surveillance quality data were not available (due to non-participation in YFSD). In addition a total of 34 potential covariates were evaluated, nine of which were not significantly associated with the data at the threshold of p  = 0.1 (see Table S2 ). The remaining 25 variables were clustered into 18 groups (see Figure S4 for the correlations between variables and Figure S5 for maps of the 18 covariates considered in the multivariate regression models), leading to a total of 249,527 models fitted with standard regression software. The 15 best models further investigated in the full model framework included three to five additional covariates ( Table 1 ). These models were investigated further by MCMC, fitting simultaneously the regression parameters and the force of infection from the serological surveys. For identification, these models were indexed with their BIC rank from the initial model fit. Time series and autocorrelation plots of the model parameters for the baseline model (model 1) are shown in Figures S6 and S7, respectively.

https://doi.org/10.1371/journal.pmed.1001638.t001

One would expect the number of cases to be proportional to the population size, leading to a dependence of the model predictions (probability of detecting yellow fever) on the log[population size], and this covariate was indeed included in all of the 15 best fitting models with a regression parameter value around 1, indicating linear dependence of the number of cases on population size. Most models included longitude, mimicking the strong gradient in risk that is observed in yellow fever epidemiology. Latitude, mean EVI, mean MIR, and the land cover category indicating a mosaic of cropland and natural vegetation were included in about half of the models, with typically each model including either mean EVI or mean MIR. The land cover categories of deciduous broadleaf forest, open shrubland, and barren areas were only included in few models, and no further of the 18 potential covariates considered were included in the 15 best fitting models.

The differences in goodness of fit between the models were small compared with the uncertainty inherent in the BIC and AUC estimates ( Figure 2 ), although the BIC indicated a slightly better fit for the models with a smaller index in Table 1 compared with those with a larger index, mirroring the BIC rank in the pure regression models. AUC values were high, averaging just below 0.9, showing a good match between data and regression model predictions. The AUC indicated a better match between regression model predictions and data for models 1, 4, 13, and 15 than the other models, but again, the differences were small.

(A) BIC and (B) AUC values for the 15 models investigated with MCMC, with a prior standard deviation on the country factors of 2. Circles show posterior means, lines the 95% posterior range.

https://doi.org/10.1371/journal.pmed.1001638.g002

Outputs from the Baseline Model

The high values for the AUC seen for the model predictions testified to a good model fit, so it is unsurprising that the spatial distribution of the model predictions matched the dataset of presence or absence of yellow fever reports very well ( Figure 1A and 1B ). The model successfully captures the gradient of transmission intensity from west to east as well as the focus of transmission being in sub-Sahel and tropical latitudes, which is reflected in both the model predictions ( Figure 1B ) and the force of infection estimates ( Figure 1C ).

There was substantial uncertainty in the force of infection estimates, with the highest values of the coefficient of variation being in areas with the lowest force of infection estimates: Rwanda, Burundi, and western parts of Tanzania ( Figure S8 ). Due to the very low force of infection estimates in these areas, this uncertainty has little impact on the burden estimates.

The estimated country-specific detection probability per infection varied over nearly two orders of magnitude between countries. Countries with a higher estimated force of infection also had higher estimates of the case detection probability, with the highest values found in the Central African Republic and Togo and the lowest in Guinea-Bissau, Ethiopia, and Tanzania ( Figure 1D ), Notably the detection probability was estimated to be very low in Nigeria, which has a substantial impact on the burden estimates due to its large population.

The annual number of yellow fever infections, severe clinical cases, and deaths expected from the estimated force of infection were estimated for selected years ( Table 2 ). Between 1995 and 2005, the overall vaccination coverage remained roughly similar across the continent. The moderate increase in estimated burden between these years therefore reflects overall population growth. However, the large preventive mass vaccination campaigns performed between 2006 and 2012 increased the vaccination coverage in the participating countries, substantially outweighing population growth effects and resulting in a 2013 burden estimate of 180,000 (95% CI 51,000–380,000) severe cases presenting with fever and jaundice or haemorrhage including 78,000 (95% CI 19,000–180,000) deaths. We estimate that the recent preventive mass vaccination campaigns between 2006 and 2012 reduced the annual burden evaluated for 2013 by 27% (95% CI 22%–31%), which equates to an overall reduction of 57% (95% CI 54%–59%) in the 12 targeted countries. In these campaigns, the number of targeted provinces and districts and therefore the impact achieved varied by country, with the highest reductions achieved in Benin, Togo, and Cote d'Ivoire, where an estimated 82%, 77%, and 73%, respectively, of the burden was prevented in 2013. The reduction at the national level of participating countries reflects both vaccinated and non-vaccinated regions within each country.

https://doi.org/10.1371/journal.pmed.1001638.t002

Disease burden was estimated to be distributed very unevenly between countries, with by far the largest burden estimated for Nigeria, owing to the moderately high force of infection, low vaccination coverage, and a large population size ( Figure 3 ). The country contributing the next largest number of cases and deaths was the Democratic Republic of the Congo, followed by countries in West Africa with a high force of infection, some of which have recently benefited from the GAVI-funded mass vaccination campaigns ( Table 3 ).

Red bars show the number of deaths estimated assuming implementation of no mass vaccination campaigns between 2006 and 2012, orange bars show the number of deaths estimated for the actual vaccination. Lines show the 95% credibility intervals of the estimated number of deaths. Countries are ordered west to east.

https://doi.org/10.1371/journal.pmed.1001638.g003

https://doi.org/10.1371/journal.pmed.1001638.t003

Mass vaccination campaigns can be extremely effective at reducing the burden in populations with low immunity, with the effect being immediate and long lasting ( Figure 4 ). The impact wanes over the course of decades only as new birth cohorts join the populations ( Figure 4A and 4D ). In this context, routine infant immunization as performed in the EPI in many African countries since the 1980s ( Table S1 ) serves an important purpose by ensuring good vaccination coverage in new birth cohorts and thus preventing any long term decrease in population immunity. As the sole tool to increase population immunity infant immunization is less effective, as it takes decades for such a program to substantially increase the immunity of the whole population. Figure 4B and 4E shows the burden in Ghana and Liberia assuming no infant immunization ever in these two countries. The results illustrate how a high infant immunization coverage is crucial to sustaining low levels of burden (as in Ghana with 91% coverage), whereas low coverage levels (as in Liberia with 39% coverage) will reduce the burden a little but are too low to sustain a low level of burden in the future. A combination of mass vaccination campaigns and infant immunization at good coverage level is therefore likely to reduce the burden quickly and sustain it at low levels.

Thick lines show the point estimate, hashed areas the 95% credibility intervals. Baseline scenario (black) in (A–F) includes past mass vaccination and infant immunization, plus continuing infant immunization at 2011 coverage levels. Alternative scenario (red): (A and D): as baseline, but excluding the mass vaccination campaigns; (B and E): as baseline, but assuming no infant immunization at any time; (C and F): as baseline, but including mass vaccination campaigns targeting children under 5 every 5 years at a coverage of 90% instead of future infant immunization.

https://doi.org/10.1371/journal.pmed.1001638.g004

Some countries achieve high coverage in their routine infant immunization but the coverage in other countries is low. Conversely, the mass vaccination campaigns achieved high coverage levels in most countries targeted. If it is difficult to reach a substantial proportion of infants with routine immunization, one could instead consider repeated mass vaccination campaigns. Figure 4C and 4F shows the effect of repeating mass vaccination campaigns targeting children under 5 every 5 years is similar to what is achieved using routine infant immunization reaching a high proportion of infants. Such age-targeted campaigns would cost less than repeated mass vaccination campaigns targeting all age groups while being similarly effective.

We compared our estimates of mortality due to yellow fever to all-cause crude mortality estimates obtained from the UN WPP [35] for all endemic countries. For the period from 2005 to 2010, the estimates varied between eight and 18 deaths per year per 1,000 population, equating to 9.4 million deaths annually from any cause in the endemic region (calculated using 2010 population estimates). Our estimate of 78,000 deaths from yellow fever for 2013 therefore corresponds to 0.8% of all-cause mortality, but the proportion of the all-cause mortality that would be attributed to yellow fever based on our burden estimates varied substantially between countries ( Figure 5 ), ranging from close to zero in many east African countries to values typically between 1% and 3% in West Africa, with the highest values just under 6% in Mauritania and Guinea-Bissau.

Grey bars indicate the point estimates, black lines the range spanned by the 95% CIs of the burden estimates. Countries are ordered west to east.

https://doi.org/10.1371/journal.pmed.1001638.g005

In this study, we estimated the burden of yellow fever in terms of the number of infections, severe cases, and deaths across Africa by fitting generalised linear regression models to datasets of yellow fever reports between 1987 and 2011. We evaluated the impact of recent large-scale preventive mass vaccination campaigns undertaken between 2006 and 2012 under the Yellow Fever Initiative by estimating the burden expected had these vaccination campaigns not taken place.

We estimate that currently there are between 51,000–380,000 severe cases of yellow fever annually in Africa, resulting in an estimated 19,000–180,000 deaths. These figures are to be compared with previous global estimates of 200,000 cases and 30,000 deaths annually for the early 1990s, around 90% of which occur in Africa [5] , [6] , [52] . It is encouraging that both sets of estimates are broadly similar, particularly since the new estimates take into account all existing data on yellow fever that are currently available. The analysis provided here also gives a better understanding of the spatial and temporal distribution of yellow fever across Africa. The model framework developed takes into account a variety of different data sources, including information on population vaccination coverage over time, which can be used to evaluate the impact of past and potential future vaccination campaigns.

The average annual number of yellow fever cases officially reported to WHO by countries in the endemic zone [53] was 1,165 for the period from 1987 to 2011 considered in this analysis, and 656 for the period between 2005 and 2011 covered by the YFSD (note that this is a different dataset than the YFSD, containing only aggregate numbers). This was in contrast to the estimated annual burden of around 180,000 severe cases (which were defined as presenting with fever and jaundice or haemorrhage), meaning that for each officially reported case there might actually be as many as 50 to 500 severe cases. This is consistent with the 10–1,000-fold under-ascertainment of yellow fever morbidity and mortality recognized in past work [3] , [4] . Such levels of under-ascertainment highlight the difficulties inherent in yellow fever surveillance, which relies on clinical case definitions. Syndromic surveillance is challenging due to the variety of clinical manifestations seen in severe disease that do not include jaundice and therefore might be mistaken for other infections (notably malaria) [26] . In addition, not all jaundice is caused by yellow fever, with other causes including malaria, liver pathogens, and other conditions.

The detection probabilities fitted in our model are of the order of 10 −5 , but these describe the probability that an infection would be reported into either the YFSD or as an outbreak. In the YFSD, there were on average around 135 cases reported annually. Comparing this to our burden estimates of around 1.5 million infections annually in the time period covered by the YFSD, this would lead to an empirical detection probability of the order of 10 −4 across Africa, an order of magnitude larger than the values fitted in our model. However, the detection probabilities fitted in our model represent an average over 25 years, and detection was considerably poorer prior to the introduction of the YFSD.

The proportion of the all-cause mortality that would be attributed to yellow fever based on our burden estimates varied between countries with plausible estimates of less than 3% for most countries, with the exception of Mauritania and Guinea Bissau where nearly 6% of the all-cause mortality would be attributed to yellow fever on the basis of our estimates. The estimates for these two countries may appear unrealistically large, but it should be kept in mind that the uncertainty in the force of infection and consequently in the burden estimate is relatively high in Mauritania ( Figures 3 and S7 ), whereas for Guinea-Bissau, the estimated detection probability is the lowest estimated for any country ( Figure 1D ) due to its low rate of reporting suspected cases to the YFSD. If the overall surveillance quality in this country was not well represented by the participation in the YFSD, the burden estimate here would be over-inflated.

The datasets of yellow fever incidence used to fit the models rely on surveillance recognizing yellow fever cases. Typically the case definition is based on fever with jaundice and/or haemorrhaging symptoms, but of course the sensitivity and specificity of this case definition might vary between settings. In our analysis, we have allowed for the sensitivity to vary between countries by estimating the country-specific surveillance quality. The specificity of the case definition in our datasets should be high across the board, as only laboratory confirmed cases or cases closely linked epidemiologically were included in our analysis. There might however be substantial differences in the severity spectrum of yellow fever between settings, depending on factors such as previous exposure to other flaviviruses, the general immune status of the populations, or the access to health care facilities, although there is no treatment for yellow fever apart from general life support. While we were not able to include any of these effects, we used estimates of the severity with measures of uncertainty by Johansson and colleagues based on the limited available data [21] , capturing the variability seen across different settings. Our model estimates first the number of infections and infers the disease burden in terms of the number of severe cases and deaths from this, so the uncertainty of our burden estimates is inflated by the uncertainty of the severity spectrum. Nevertheless we have chosen to report mainly the number of deaths as cases and deaths are more relevant in terms of disease burden and health care needs than the number of infections, the majority of which are likely to be very mild or asymptomatic.

The credible intervals around the burden estimates presented here also reflect the fact that a range of values for the force of infection estimates yield a similarly good model fit. However, while the credible intervals represent the uncertainty in model parameter estimates, there are further potential sources of uncertainty that are not captured by credible intervals. Firstly, the choice of covariates included in the model could have an effect. However, the 15 models investigated here showed a similarly good fit to the dataset, and the burden estimates from all models were very similar (see Text S2 ).

Second, in order to prevent the country factors (which determine the detection probabilities in countries not participating in the YFSD) from taking infinite values, we assumed a Gaussian prior distribution for these within the Bayesian framework used for model fitting. The standard deviation of this prior distribution was chosen relatively arbitrarily; however, in the sensitivity analyses we have shown that the burden estimates again are fairly independent of the particular value chosen (see Text S3 ).

The dataset of vaccination coverage compiled from various sources reporting on vaccination activities in the last century contains a number of potential sources of uncertainty that are very difficult to quantify. Uncertainty in historical population sizes by age generates uncertainty in vaccination coverage estimates if those estimates are generated from records of the number of vaccine doses used. There are also concerns about the completeness and accuracy of the reports on vaccination activity. Furthermore, the effect of population movements on vaccination coverage could not be taken into account owing to lack of data. Our simplifying assumptions of a 100% vaccine efficacy and lifelong immunity conferred by the vaccine can also be questioned. To evaluate the impact of these uncertainties we undertook sensitivity analyses that carried vaccine effectiveness and the coverage achieved in historical campaigns, but found the effects on burden estimates to be slight (see Text S4 ). While we omitted reactive vaccination campaigns before 1970 in the generation of all vaccination coverage scenarios as these data were not routinely reported prior to this time, this is likely to have little impact on vaccination coverage levels due to the low yellow fever activity and resulting low number and extent of vaccination campaigns in this period.

Uncertainty in demographic data across Africa has a very direct impact on the burden estimates, as such estimates are directly proportional to the population size. This uncertainty is not captured in the confidence intervals given in this paper, as it was not possible to quantify the level of uncertainty.

There are substantial uncertainties regarding the spatial distribution of yellow fever occurrence, which were taken into account in our model by allowing infection risk and detection probabilities to vary between countries. However, the baseline model presented above did not allow for detection probabilities to vary over time, while activities such as the introduction of the YFSD in 2005 were clearly intended to improve surveillance. We therefore investigated two alternative model structures that both allow for a change in the detection probabilities at the time of introduction of the YFSD, both of which estimated an increased probability of case detection in the countries participating in the YFSD following its introduction. The overall burden estimates from these models were very similar to those obtained from the baseline model though there were subtle differences in the spatial distribution of the transmission intensity, with one of the alternative models showing a slightly less pronounced gradient in transmission strength from west to east (see Text S5 ).

Similarly, while we allowed the force of infection for yellow fever to vary in space, we assumed it was constant throughout the 25-year observation period, as well as homogeneous by gender and age. While clearly there will be differences in exposure between age groups and genders, particularly in areas where non-human primates play an important role in transmission, the relatively crude nature of the yellow fever occurrence data did not support a model that would be able to estimate these differences. Our estimates are therefore representative of the overall population but do not reflect the age- and sex-specific exposure likely to be found in many places. The assumption of constant force of infection throughout time means we have not taken into account changes in transmission due to factors such as changed land use or climate change, which might influence the transmission intensity. Clearly yellow fever activity is not constant, but epidemic amplifications and reductions of transmission intensity happen over the timescale of decades. Epidemics are driven, at least in part, by the rapid removal and slow replenishment of susceptible hosts in both humans and wildlife, as illustrated by the widespread epidemics in much of western Africa, and particularly Nigeria, in the 1990s, and a reduction in epidemic activity since then. Furthermore, a serological survey in Central African Republic testing samples collected in 2006 and in 2009 found evidence of an increase in yellow fever exposure over this period [29] , mirroring the increasing number of cases reported from that region in recent years. Therefore our results should be seen as representative of the past 25 years, averaging over the large fluctuations that occur in reality, although the burden estimates for specific years do reflect the population size, age structure, and vaccination coverage pertaining to the time.

Burden estimates were strongly determined by the force of infection estimated from serological surveys [29] – [34] . However, the only surveys available were conducted in central Africa and Nigeria, with these results extrapolated to the remainder of the endemic zone in West and East Africa using the spatial distribution of transmission intensity estimated from the regression model. While all model structures reproduced the gradient in transmission intensity from west to east that is seen in yellow fever epidemiology, this gradient was more pronounced in the baseline model presented in the main paper than in the alternative model that was fitted to an annual dataset of yellow fever reports (see Text S5 ). In the absence of further reliable serological data outside central Africa it is presently not possible to distinguish which model better reflects reality. There are several serological surveys under way or close to completed in east African countries including Sudan, Rwanda, Uganda, Kenya, South Sudan, and Ethiopia. These data, once available, will substantially reduce model uncertainty, allowing us to discriminate between different model assumptions and resulting in more reliable estimates.

Cohort studies collecting data on case incidence and the severity spectrum of disease could also reduce the level of uncertainty. The relatively low incidence of yellow fever implies the need for large cohorts, which would be prohibitively expensive if performed for yellow fever alone. However, including yellow fever diagnostics into ongoing cohort studies (e.g., focused on HIV or malaria) might be a cost-effective way to improve basic understanding of yellow fever epidemiology. A further advantage of studies focusing on multiple diseases would be to understand interactions between infections (most notably cross-immunity between flaviviruses).

Our analysis does not take into account the epidemic character of yellow fever transmission, but rather assumes cases are distributed evenly over time according to a force of infection that is independent of the incidence of cases in the population. Consequently, the impact of vaccination campaigns will be underestimated, as lower transmission in a population due to vaccination also provides indirect protection to unvaccinated individuals (herd immunity). While the impact of herd immunity can be easily quantified in situations where there is only one type of host, this is currently impossible with yellow fever as it is unknown what proportion of cases arise through inter-human transmission via mosquito vectors, and what proportion through the sylvatic cycle. While this question cannot be answered with the methodology employed in the present study, it is an important topic for future work.

Keeping this limitation in mind, we conservatively estimate that the recent mass vaccination campaigns have reduced the yellow fever burden in the 12 participating countries for 2013 by 57% (95% CI 54%–59%) relative to a counterfactual scenario in which these campaigns were not conducted, by vaccinating 78 million people, who make up around 55% of the population of these countries. Across Africa, this amounts to a reduction of the total burden of yellow fever by 27% (95% CI 22%–31%), by vaccinating around 10% of the population in the endemic zone.

Partly as a result of the estimates presented here, in late 2013 the GAVI Alliance Board decided to make available support for additional yellow fever vaccination campaigns, targeting 144 million people across the endemic region in Africa [54] , [55] . Furthermore, the GAVI Alliance is now using our estimates for evaluating the past and future impact of their yellow fever vaccination activities.

The impact of both past and future mass vaccination campaigns will prevent a substantial proportion of yellow fever disease burden for years to come, with a gradual decrease in impact over the next decades as new birth cohorts that have not benefitted from these campaigns enter the population. This effect of slowly declining vaccination coverage following the abandonment of mass vaccination campaigns was seen since the 1960s, and was the cause of the gradual resurgence of yellow fever over the following decades. However, the achievements of the current mass vaccination campaigns could be sustained if a high level of immunization is achieved through a strong EPI program and preventive vaccination of populations that remain at-risk, such as migrants or populations from as yet unvaccinated districts. While the coverage achieved in the routine infant immunization is variable between countries, the coverage achieved in recent mass vaccination campaigns has generally been high. An alternative for countries struggling to reach high EPI coverage levels might therefore be to repeat mass vaccination campaigns targeted at children every few years, although the organizational and financial costs would probably be substantially higher than the existing EPI.

Yellow fever is a disease that is difficult to diagnose and confirm, whose symptoms can be mild and mistaken for other infections, and that occurs in some of the most resource-poor settings globally. Consequently surveillance data reflect patterns of endemicity and emergence of infection in new zones and provide sentinel data on imminent or ongoing outbreaks, but do not reflect the actual disease burden. The most recent estimates of the disease burden stemmed from the early 1990s and therefore an update taking into account the changes in demography, ecology, and vaccination coverage, such as the estimates provided in the present study, was long overdue. The framework for burden estimation developed here is also a useful tool for the evaluation and planning of effective vaccination campaigns. As such, it is being used by the partners of the Yellow Fever Initiative for planning their yellow fever vaccination strategy for the next decade.

Supporting Information

Map of the outbreaks recorded in Africa between 1980 and 2012. Outbreak size indicated by the symbol size, outbreak year coded by the colour.

https://doi.org/10.1371/journal.pmed.1001638.s001

(A) map of the number lab-confirmed, epi-linked, and compatible yellow fever cases reported in the YFSD by province. (B) Annual reporting rate of suspected cases per 100,000 population by country.

https://doi.org/10.1371/journal.pmed.1001638.s002

Estimated vaccination coverage at the first administrative level in the countries endemic for yellow fever on the African continent throughout the decades. Non-endemic countries are shown in grey. The estimate for 2015 is a projection that assumes infant immunization continues at the same levels as in 2011, and no other vaccination campaigns are implemented.

https://doi.org/10.1371/journal.pmed.1001638.s003

Absolute values of the pairwise correlations between the 25 potential covariates significant at the p  = 0.1 level from 0 (red) to 1 (white). Clusters are highlighted by a lack of separating lines, and variables not considered for the multivariate models printed in grey.

https://doi.org/10.1371/journal.pmed.1001638.s004

Maps of the 18 variables considered in the multivariate modeling as potential covariates. Colour scale from navy (low) to red (high). A, longitude; B, latitude; C, altitude; D LC, deciduous broadleaf forest; E LC, closed shrubland; F LC, open shrubland; G LC, woody savannas; H LC, urban and built-up; I LC, cropland/natural vegetation mosaic; J LC, barren or sparsely vegetated; K, mean day temperature; L, min day temperature; M, min night temperature; N, max night temperature; O, max EVI; P, min MIR; Q, min rainfall; R, max rainfall.

https://doi.org/10.1371/journal.pmed.1001638.s005

MCMC posterior trace plots of model parameter estimates for the baseline model, thinned by a factor 800.

https://doi.org/10.1371/journal.pmed.1001638.s006

Auto-correlation in posterior estimates of the model parameters for the baseline model. Posterior MCMC samples were thinned by a factor 800.

https://doi.org/10.1371/journal.pmed.1001638.s007

Coefficient of variation of the force of infection estimates. Countries not considered endemic for yellow fever are shown in white.

https://doi.org/10.1371/journal.pmed.1001638.s008

Coverage and year of introduction of the yellow fever vaccine into the routine Enhanced Programme of Immunization by country.

https://doi.org/10.1371/journal.pmed.1001638.s009

Covariates considered in the regression modeling, significance level in univariate models and cluster association.

https://doi.org/10.1371/journal.pmed.1001638.s010

Demographic data analysis.

https://doi.org/10.1371/journal.pmed.1001638.s011

Sensitivity analysis: impact of the covariates included.

https://doi.org/10.1371/journal.pmed.1001638.s012

Sensitivity analysis: impact of the standard deviation of the prior distribution on the country factors.

https://doi.org/10.1371/journal.pmed.1001638.s013

Sensitivity analysis: impact of alternative vaccination coverage scenarios.

https://doi.org/10.1371/journal.pmed.1001638.s014

Sensitivity Analysis: alternative model structures.

https://doi.org/10.1371/journal.pmed.1001638.s015

Acknowledgments

In addition to the author list, the Yellow Fever Expert Committee includes: Donald Burke, Fernando De La Hoz, Bryan Grenfell, Peter M Hansen, and Raymond Hutubessy.

The authors would like to thank Michael Johansson for helpful discussions and sharing his estimates on the yellow fever severity spectrum, Mark Kuniholm for the southern Cameroon dataset, Emily Jentes for helpful discussions, Kara Durski for providing data on several yellow fever vaccination campaigns, and Véronique Millot for support in collected data from countries. The Modis 12Q1 and 13A2 data were obtained through the online Data Pool at the NASA Land Processes Distributed Active Archive Center (LP DAAC), USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota ( http://lpdaac.usgs.gov/get_data ).

Author Contributions

Conceived and designed the experiments: TG MDVK SY WP NMF. Analyzed the data: TG NMF. Wrote the first draft of the manuscript: TG MDVK NMF. Contributed to the writing of the manuscript: TG MDVK SY OR RL JES WP NMF. ICMJE criteria for authorship read and met: TG MDVK SY OR RL JES WP NMF. Agree with manuscript results and conclusions: TG MDVK SY OR RL JES WP NMF. Provided input and advice on the methods: RL WP OR JES SY.

• View Article
• Google Scholar
• 3. World Health Organization (1986) Prevention and control of yellow fever in Africa. Geneva: World Health Organization.
• 5. World Health Organization. Yellow fever fact sheet no 100. Available: http://www.who.int/mediacentre/factsheets/fs100/en/ . Accessed 19 January 2014.
• 6. World Health Organization Expanded Programme on Immunization (1992) The Resurgence of deadly yellow fever: prevention using EPI. Geneva: WHO.
• 13. World Health Organization. Weekly Epidemiological Record. Available: http://www.who.int/wer/en/ . Accessed 11 June 2012.
• 14. World Health Organization. Disease Outbreak News. Available: http://www.who.int/csr/don/en/ . Accessed 11 June 2012.
• 21. Johansson MA, Staples JE (2013) The whole yellow fever iceberg: estimating the incidence of infection from the number of severe cases. Epidemics 4. Amsterdam: Elsevier.
• 22. Durieux C (1956) Mass yellow fever vaccination in French Africa south of the Sahara. Monograph Series World Health Organization. Geneva: WHO. pp. 115–122.
• 23. World Health Organization (2012) Immunization surveillance, assessment and monitoring. Available: http://www.who.int/immunization_monitoring/data/data_subject/en/index.html . Accessed 11 June 2012.
• 24. World Health Organization, unicef (2009) Yellow Fever Initiative Progress Report 2006 to 2009. Geneva: World Health Organization.
• 25. World Health Organization, unicef (2010) Yellow Fever Initiative Joint WHO and UNICEF 2010 Progress Report. Geneva: World Health Organization.
• 29. Diallo M, Janusz K, Lewis R, Manengu C, Sall AA, et al.. (2010) Rapid assessment of yellow fever viral activity in the Central African Republic. Geneva: World Health Organization.
• 35. United Nations, Department of Economic and Social Affairs, Population Division, Population Estimates and Projections Section. World Population Prospects: The 2010 Revision. Available: http://esa.un.org/wpp/Excel-Data/population.htm . Accessed 11 June 2011.
• 36. Bright EA, Coleman PR, King AL, Rose AN (2008) LandScan 2007. Oak Ridge (Tennessee). Available: http://www.ornl.gov/landscan/ . Accessed 10 December 2013.
• 42. NASA Land Processes Distributed Active Archive Center (LP DAAC) Land Cover Type Yearly L3 Global 1 km SIN Grid (12Q1). USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota. Available: http://lpdaac.usgs.gov/get_data . Accessed 13 July 2012.
• 43. NASA Land Processes Distributed Active Archive Center (LP DAAC) Vegetation Indices 16-Day L3 Global 1 km (13A2). USGS/Earth Resources Observation and Science (EROS) Center, Sioux Falls, South Dakota. Available: http://lpdaac.usgs.gov/get_data . Accessed 13 July 2012.
• 44. Hijmans RJ, Cameron SE, Parra JL. WorldClim. Available: http://www.worldclim.org/ . Accessed 12 January 2012.
• 46. Brooks S, Gelman A, Jones GL, Meng X-L, editors (2011) Handbook of Markov chain Monte Carlo. Boca Raton (Florida): Chapman & Hall/CRC.
• 47. Gilks WR, Richardson S, Spiegelhalter DJ (1996) Markov chain Monte Carlo in practice. Boca Raton (Florida): Chapman & Hall/CRC.
• 53. World Health Organization (2012) Yellow fever reported cases. Available: http://apps.who.int/immunization_monitoring/en/globalsummary/timeseries/tsincidenceyfe.htm . Accessed 24 April 2013.
• 54. GAVI Alliance. Vaccine investment strategy. Available: http://www.gavialliance.org/about/strategy/vaccine-investment-strategy/ . Accessed 18 January 2014.
• 55. GAVI Alliance (2013) Yellow fever vaccine investment strategy. Background document #5. GAVI Alliance. Available: http://www.gavialliance.org/library/gavi-documents/strategy/final-vis-analysis-2013—yellow-fever/ . Accessed 18 January 2014.

Yellow fever disease: density equalizing mapping and gender analysis of international research output

Affiliation.

• 1 Institute of Occupational, Social and Environmental Medicine, Goethe-University, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany. [email protected].
• PMID: 24245856
• PMCID: PMC3843536
• DOI: 10.1186/1756-3305-6-331

Background: A number of scientific papers on yellow fever have been published but no broad scientometric analysis on the published research of yellow fever has been reported.The aim of the article based study was to provide an in-depth evaluation of the yellow fever field using large-scale data analysis and employment of bibliometric indicators of production and quantity.

Methods: Data were retrieved from the Web of Science database (WoS) and analyzed as part of the NewQis platform. Then data were extracted from each file, transferred to databases and visualized as diagrams. Partially by means of density-equalizing mapping makes the findings clear and emphasizes the output of the analysis.

Results: In the study period from 1900 to 2012 a total of 5,053 yellow fever-associated items were published by 79 countries. The United States (USA) having the highest publication rate at 42% (n = 751) followed by far from Brazil (n = 203), France (n = 149) and the United Kingdom (n = 113). The most productive journals are the "Public Health Reports", the "American Journal of Tropical Medicine and Hygiene" and the "Journal of Virology". The gender analysis showed an overall steady increase of female authorship from 1950 to 2011. Brazil is the only country of the five most productive countries with a higher proportion of female scientists.

Conclusions: The present data shows an increase in research productivity over the entire study period, in particular an increase of female scientists. Brazil shows a majority of female authors, a fact that is confirmed by other studies.

Publication types

• Historical Article
• Bibliometrics*
• History, 20th Century
• History, 21st Century
• Internationality*
• Research / history*
• Sex Factors
• Yellow Fever / epidemiology*

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• BMC Infect Dis

The incidence and mortality of yellow fever in Africa: a systematic review and meta-analysis

Akuoma u. nwaiwu.

1 Division of Epidemiology & Biostatistics, Faculty of Medicine & Health Sciences, Stellenbosch University, Cape Town, South Africa

Alfred Musekiwa

2 School of Health Systems & Public Health, Faculty of Health Sciences, University of Pretoria, Pretoria, South Africa

Jacques L. Tamuzi

Evanson z. sambala.

3 Cochrane South Africa, South African Medical Research Council, Cape Town, South Africa

4 School of Public Health and Family Medicine, Kamuzu University of Health Sciences, Blantyre, Malawi

Peter S. Nyasulu

5 Division of Epidemiology & Biostatistics, School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa

Associated Data

All data generated or analysed during this study are included in this published article (Supplementary table 2).

Understanding the occurrence of yellow fever epidemics is critical for targeted interventions and control efforts to reduce the burden of disease. We assessed data on the yellow fever incidence and mortality rates in Africa.

We searched the Cochrane Library, SCOPUS, MEDLINE, CINAHL, PubMed, Embase, Africa-wide and Web of science databases from 1 January 1975 to 30th October 2020. Two authors extracted data from included studies independently and conducted a meta-analysis.

Of 840 studies identified, 12 studies were deemed eligible for inclusion. The incidence of yellow fever per 100,000 population ranged from < 1 case in Nigeria, < 3 cases in Uganda, 13 cases in Democratic Republic of the Congo, 27 cases in Kenya, 40 cases in Ethiopia, 46 cases in Gambia, 1267 cases in Senegal, and 10,350 cases in Ghana. Case fatality rate associated with yellow fever outbreaks ranged from 10% in Ghana to 86% in Nigeria. The mortality rate ranged from 0.1/100,000 in Nigeria to 2200/100,000 in Ghana.

The yellow fever incidence rate is quite constant; in contrast, the fatality rates vary widely across African countries over the study period. Standardized demographic health surveys and surveillance as well as accurate diagnostic measures are essential for early recognition, treatment and control.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-021-06728-x.

Yellow fever is an acute systemic illness caused by flavivirus transmitted by infected mosquitoes belonging to the Aedes and Haemogogus species [ 1 ]. Most of the yellow fever cases identified in Africa are seen in the unvaccinated population who live in the yellow fever belt. In severe cases, this viral infection causes high fever, bleeding into the skin and death of cells in the liver and kidney [ 2 ]. Currently, there is no treatment, cure or drugs for yellow fever but the disease can be prevented through vaccination. A yellow fever vaccine is available and recommended to laboratory workers, children over 9 months and people who are traveling to or living in yellow fever high-risk areas of Africa [ 3 ]. A diagnosis of yellow fever is difficult to make as the definition for suspected cases is based on similar signs and symptoms to other diseases like malaria, typhoid, dengue fever and other haemorrhagic fevers. Laboratory diagnosis exists, but the availability and lack of diagnostic capacity are major challenges in African countries. Consequently, it is a major public health problem often underreported in Africa.

Yellow fever has three modes of transmission cycles (Fig.  1 ). The first is sylvatic or jungle yellow fever. It happens when monkeys living in the tropical rainforest are infected through mosquitoes bites; and when humans visit or work in the jungle, the virus is transmitted from monkeys to humans by mosquitoes [ 4 ]. Other mosquitos that do not carry the virus become infected when they feed on the blood of the bitten monkeys, and the cycle continues [ 4 ]. In Africa, yellow fever is transmitted by five sylvatic vectors among which four are sylvatic ( Ae. africanus , Ae. furcifer , Ae. taylori and Ae. luteocephalus ) and one is intermediate savannah ( Ae. aegypti ) [ 5 ].

Transmission cycle of yellow fever virus

The second is the intermediate yellow fever (savannah) cycle, which includes the transmission of the virus from mosquitoes to humans living or working in the jungle border areas [ 4 ]. In this cycle, the virus can be transmitted from monkey to human or from human to human through mosquitoes [ 4 ]. While the third is urban yellow fever, which includes the transmission of the virus between humans and urban mosquitoes, mainly Ae. aegypti [ 6 ]. The virus spreads in the urban environment through humans who were infected through mosquito bites in the forest or savannah. Outbreaks are typically triggered by sylvatic and intermediate forms where the virus is transmitted when humans come into close contact with monkeys [ 6 ].

Yellow fever has a high case fatality rate, and there is no known cure for this disease [ 6 ]. Good in-hospital supportive treatment improves survival rate. In Africa and South America, 200,000 cases of yellow fever are recorded every year of which 90% occur in Africa, causing approximately 30,000 deaths [ 7 ]. It is estimated that as many as 610 million people in 32 African countries, including more than 219 million dwelling in urban settings remain at high risk of contracting yellow fever [ 8 ].

The World Health Organization (WHO) has reported few cases of yellow fever from 1980, until 2016, when an upsurge in cases, to as high as 1040 at highly irregular intervals were recorded compared to cases between 1986 and 1995 [ 9 , 10 ]. However, it is uncertain whether the recent increase in number of cases is due to increased surveillance or increased disease activity in these countries. Understanding yellow fever epidemiology as determined by its evolution is important to develop preventative measures such as immunization policies to mitigate yellow fever infection. Yellow fever cases are frequently reported in West Africa than anywhere else in the world, followed by East Africa [ 11 ]. Major epidemics have occurred in West Africa posing global threats. An historical outbreak first of its’ kind was recorded in Kenya, East Africa, in 1992–1993 [ 11 ]. It was reported that two cases of yellow fever were imported to Europe from West Africa in recent years which had fatal outcomes, suggesting intercontinental transmission [ 11 ].

Due to lack of reporting of yellow fever cases in Africa, the overall burden of the disease may potentially be underestimated. The WHO estimated the burden of and associated mortality of yellow fever in Africa to be between 84,000–170,000 severe cases and 29,000–60,000 deaths occurred in 2013 [ 12 , 13 ]. While some estimates of the burden of yellow fever exists in Africa, we could not find a published systematic review and meta-analysis on yellow fever for the entire Africa continent. Thus this review was aimed to quantify and summarize the overall burden of and fatality associated with yellow fever in Africa.

Eligibility criteria

In this review, we considered outbreak reports, cross sectional studies and other observational studies (case series, case report, epidemiological surveys, surveillance studies etc.) including published and unpublished studies carried out in Africa that reported the incidence, mortality and case fatality rate of yellow fever across all age groups. We included only patients who were considered to have suspected or confirmed cases of yellow fever. Suspected cases were defined as cases that were characterized by acute onset of fever followed by jaundice within 2 weeks of the onset of the first symptoms and confirmed cases as suspected cases that were laboratory-confirmed or epidemiologically linked to a laboratory-confirmed cases or outbreaks. Probable cases were defined as a suspected case and one of the following: (i) epidemiologically link to a confirmed case or an outbreak; (ii) positive post-mortem liver histopathology.

We considered probable cases as being confirmed cases when they included any of the following: a probable case and one of the following: (i) detection of yellow fever-specific IgM; (ii) detection of four-fold increase in yellow fever IgM and/or IgG antibody titres between acute and convalescent serum samples, (iii) detection of yellow fever virus-specific neutralizing antibodies.

All the different modes of transmission were considered and both hospital and field patients (population based) were included in this review.

Search method for identification of studies for inclusion

We exclusively searched the following databases: Cochrane library, MEDLINE, PUBMED, Embasse, SCOPUS, CINAHL (EBSCOhost), Africa-wide (EBSCOhost) and Web of science (SCI-EXPANDED) for studies published from 1 January 1975 to 30 October 2020 including unpublished studies. We selected the starting date of 1975 because of the generally increasing availability and quality of published studies from that date. The combination of keywords was used to search for relevant studies from the electronic databases. We included the following keywords: “outbreak”, “Burden of disease”, Incidence, Prevalence, Survey, Surveillance, Epidemic, Epidemiology, “yellow fever”, “yellow jack”, “yellow fever vaccines”, vaccination”, seroprevalence, “haemorrhagic fever”, “Jungle fever,” “cross sectional studies”. Boolean terms, AND/OR were used during the keywords search to identify relevant studies. Regional grouping such as sub-Saharan Africa and West Africa were also used to look for studies indexed under regional names. Medical Subject Heading (MesH) terms were used in PubMed and Medline search (Additional file 1 : Table S1). In consultation with a medical librarian, peer-reviewed journal papers were searched systematically in SCOPUS, Africa-wide (EBSCOhost) and Web of science (SCI-EXPANDED) using variations of MeSH terms (Additional file 1 : Table S1). Lastly, we also controlled vocabulary (subject headings) to search CINAHL, and Embase.

We identified three conferences related to yellow fever through a Google search: International Society for Infectious Diseases (ISID), International Congress on Infectious Diseases (ICID) and International Conference on Infectious Disease Dynamics (ICIDD). We searched for the following terms on the website of each of the three conferences identified: yellow fever, incidence, mortality, and Africa. We also consulted an expert librarian at Stellenbosch University, South Africa, to improve and sharpen the search strategy (Additional file 1 : Table S1). We identified other eligible studies by searching the reference list of included studies.

Data extraction

Two review authors independently and in duplicate screened titles and abstracts and selected studies for inclusion in this review using the set eligibility criteria. The identified studies were retrieved for full texts and included in the review after re-screening. Screening and data extraction was done using Covidence manager. Any discrepancies were resolved by consensus or by a third reviewer. We independently and in duplicate extracted the data on the following: burden of disease, characteristics of the participants, study setting, study design, date of study, study location. Risk of bias for each of the included studies was assessed using the validated quality appraisal tool developed by Hoy and colleagues (Additional file 1 : Table S3) [ 14 ]. We assessed each domain as either low or high risk of bias and regarded studies which were unclear as high risk of bias. We scored the overall risk of bias according to the number of high risk of bias parameters per study: low (1–3), moderate (4–6) and high (7–9).

Data synthesis and management

All included studies focused on incidence, case fatality rate (CFR), and mortality. We calculated incidence by dividing the number of confirmed and suspected cases by the total population in that region and expressed it per 100,000 populations. We calculated 95% CI for the incidence using the standard formula for calculating the standard error of a proportion, per 100,000, that is,

where p = incidence (per 100,000), and n = sample size. We assumed normality of the incidence statistic and used the critical value of 1.96 while calculating the 95% confidence intervals. We calculated CFR by dividing the number of deaths from yellow fever over a defined period of time by the number of individuals diagnosed with yellow fever then multiplied by 100 to yield a proportion. Mortality rate was calculated by dividing the number of deaths by the total population and then multiplied by 100,000.

We performed random-effects meta-analysis due to the variability in incidence estimates from different countries. We assessed heterogeneity using both the Chi-square test (p < 0.10 considered significant) and the I-square test statistic (> 50% considered significant) [ 15 ]. We investigated sources of heterogeneity through subgroup analysis with respect to the country of study [ 14 ]. Heterogeneity was also explored by examining the potential differences in the characteristics of the population such as the settings and other characteristics in the ‘Characteristics of included studies’ table. We performed meta-analyses using STATA version 15 and displayed results using forest plots. However, due to significant heterogeneity in meta-analyses for incidence rate and CFR, we performed systematic reporting of the results per study per country. The meta-analysis for mortality rate was not possible due to insufficient data; the results for the mortality rate were reported narratively.

Identification of studies for review

We identified 839 studies from electronic search of five databases. After removing duplicates, we screened the titles and abstracts of 493 published articles and excluded 464 studies. We retrieved the full texts of the remaining 29 studies and excluded 12 of these studies [ 1 , 11 , 16 – 25 ] because they either did not report on the yellow fever burden or were from non-African countries. From the 17 studies that reported on yellow fever incidence, we excluded five more studies [ 18 , 21 – 24 ] because they only reported on the evaluation of yellow fever vaccine without reporting data on yellow fever burden. We included a final total of 12 studies (Fig.  2 ) [ 2 , 26 – 36 ].

PRISMA flow diagram showing results of studies on yellow fever in Africa

Characteristics of included studies

A summary table of characteristics of the included studies is presented in the appendices (Additional file 1 : Table S2) We categorized studies based on year of assessment, test methods implemented for confirming the diagnosis of yellow fever, and clinical case definition used. The number of studies published each year increased modestly from 1975 to 2018. Although studies originated from eight African countries, three studies were from Nigeria [ 27 – 29 ], two studies each from Senegal [ 31 , 32 ] and Uganda [ 33 , 34 ] and one study from each of the five countries, Kenya [ 30 ], Ethiopia [ 34 ], Ghana [ 26 ], Gambia [ 2 ] and Democratic Republic of Congo [ 35 ]. The majority of studies ‘Seven’, were from West Africa [ 2 , 26 – 32 ] ‘Three’ studies from East Africa [ 30 , 33 , 36 ], ‘One’ from North Eastern Africa [ 34 ] and ‘One’ study from Central Africa [ 35 ]. All the three modes of yellow fever transmission considered in this review were reported by at least one study. Eight studies reported that the outbreak was due to sylvatic mode of transmission [ 2 , 27 , 30 – 34 , 36 ], two studies reported that it was due to an urban mode of transmission [ 28 , 35 ] and only one study reported the intermediate mode of transmission [ 36 ].

Ten of the studies [ 2 , 26 – 29 , 31 – 36 ] were reported from hospital and field-based surveillance studies while two studies [ 30 , 33 ] were only from hospital-based surveillance studies. Nine studies [ 2 , 26 – 29 , 31 , 32 , 36 ] included participants of all ages while four studies reported the disease in specific age groups such as from 10–70 years [ 30 ], 3 months–83 years [ 33 ], 3–64 years [ 36 ] and 10–72 years [ 35 ], respectively. Nine of the included studies [ 2 , 26 , 29 – 34 , 36 ] were conducted in rural areas, two [ 28 , 35 ] in urban areas and one [ 34 ] in mixed rural/urban setting. The population reported in all the studies lived on subsistence farming, growing several crops and rearing livestock. Most of the population under study practiced domestic water storage except for one study [ 34 ] that reported people practicing less livestock rearing and less domestic water storage while one study [ 35 ] did not report on any of the above. Rainy season and increased breeding sites were reported as risk factors for yellow fever epidemic as these can increase the mosquito population. However, two studies [ 31 , 36 ] did not attribute the outbreak to the rains rather reported that the study population were in contact with the forest for 2 years before returning from the Internally Displaced Persons camps. They found their homes dirty exposing them to multiple natural breeding sites for mosquitoes that could have led to the outbreak. All the included studies also reported that yellow fever was confirmed using viral serology by doing enzyme-linked immunosorbent assay (ELISA) to detect IgM antibodies to yellow fever virus, and to isolate the virus. Some of the studies for instance [ 28 , 30 ], reported that histopathology on liver specimens were done. One study, reported that test for antibody neutralization was not done [ 36 ].

The duration of the outbreak lasted for 11 months in Ethiopia [ 34 ] followed by Gambia which lasted for 8 months [ 2 ]. Outbreaks in Democratic Republic of Congo [ 35 ], Kenya [ 30 ], and Nigeria [ 27 ] lasted for 7, 6 and 5 months, respectively. Another outbreak in Nigeria [ 30 ] lasted for 4 months. Two studies from Senegal [ 31 ] lasted for 2 months and 1 month respectively [ 32 ] reported only 1 month. The other studies reported that the outbreak lasted for 3 months [ 28 , 31 , 32 ], while the study done in Ghana did not report the duration of the outbreak [ 26 ].

Assessment of risk of bias of included studies

We evaluated all the studies using the Hoy’s risk of bias tool [ 14 ]. Our summary assessment shows that ten studies were low risk of bias (83%) [ 2 , 27 – 34 , 36 ] and 2 studies were moderate risk of bias (17%) (Additional file 1 : Table S3).

Incidence of yellow fever

Meta-analysis of yellow fever incidence estimates from different studies and countries resulted in significant heterogeneity (I 2  = 99.4%, p < 0.001) and therefore we map the results per study/region in Africa (Fig.  3 ). The two studies from Uganda reported very low incidence of less than 3 and 13 cases per 100,000 population respectively, Kenya < 30 cases per 100,000, Ethiopian 40 cases per 100,000, In Gambia < 50 cases per 100,000, Nigeria the incidence ranged from < 1 to over 80 cases per 100,000 population. The two studies in Senegal reported incidence rates of approximately 1300 and 5900 cases per 100,000 population while Ghana reported the highest incidence which ranged between 320 to over 10,000 cases per 100,000 population (Fig.  3 ).

Map showing incidence rates (cases per 100,000) of yellow fever in Africa

Case fatality rate

Meta-analysis of CFR (in %) resulted in significant heterogeneity (I 2  = 95.6%, p < 0.001, Fig.  3 ) and therefore we report results narratively per study. The only study from Democratic Republic of Congo had the lowest CFR of just over 10%. The Ethiopian study reported a CFR of slightly over 30%, the two studies from Uganda reported CFRs of just over 30% each, while the one study from Gambia reported a CFR of less than 30%. The CFRs in different regions of Ghana ranged from just over 16% in Volta Region to almost 40% in Brong Ahafo. The two Senegal studies reported high CFRs of 28% and 42%, while the Kenyan study reported a higher CFR of over 60%. Lastly, the Nigerian studies reported varying CFRs from 11 to 85% (Fig.  4 ).

Forest plot showing case fatality rates associated with yellow fever outbreaks in Africa

Mortality rate

We could not perform meta-analysis of mortality rate because the population sizes were not reported by most included studies. The results also differed widely such that a meta-analysis would likely result in high heterogeneity. We therefore reported the results narratively for each study. The two studies from Senegal had high mortality rates ranging from over 500 to almost 1700 deaths per 100,000. The Ethiopian and Kenyan single studies had low mortality rates of 12 and 17 deaths per 100,000 population. In Uganda, the mortality rates were even lower, ranging from less than 1 to 4 deaths per 100,000 population. In Gambia, it was 12 deaths per 100,000 population. In Ghana, the mortality rate ranged from just over 50 in Volta Region to over 2200 deaths per 100,000 population in the Eastern Region. The one study in DRC reported a low mortality rate of 1 death per 100,000 population. Lastly, the studies from Nigeria reported mortality rates ranging from 0.2 deaths per 100,000 population in Kwara State to more than 44 deaths per 100,000 population in South Eastern Nigeria.

We carried out this systematic review with the objective of estimating the incidence and associated mortality of yellow fever in Africa. We found 12 studies from eight countries in Africa, suggesting that studies are scarce in yellow fever-endemic African countries. We found in studies in Nigeria [ 27 – 29 ], Uganda [ 33 , 36 ], Senegal [ 31 , 32 ] Kenya [ 30 ], Ethiopia [ 34 ], Gambia [ 2 ], Ghana [ 26 ] and Democratic Republic of Congo [ 35 ]. ‘Eight’ studies were undertaken more than 20 years ago, and with only four recent studies. This shows there is scarcity of data, thereby making the true estimation of yellow fever incidence difficult in Africa. However, data reported that yellow fever is a disease of the East and West African countries. We found that incidence of yellow fever varies in different countries. The incidence of yellow fever was highest in Senegal [ 31 , 32 ] (1300–5900/100,000 followed by 320/100,000 in Ghana [ 26 ] and lowest in Uganda with a range of 3–13/100,000 [ 33 , 36 ]. The reason for this increase could be due to an improved surveillance reporting system. Case fatality rate was highest (60%) in Kenya and ranged from 11 to 85% in Nigeria while Democratic Republic of Congo had the lowest case fatality rate of over 10% [ 28 , 30 , 35 ]. Finally, mortality rate was highest in Senegal [ 31 , 32 ] 500–1700 deaths/100,000, Uganda [ 33 , 36 ] had a low mortality rate of less than 1 to 4 deaths/100,000 while Nigeria ranged from 0.2 to 44 deaths/100,000 [ 27 – 29 ]. These results illustrated that the case fatality and mortality rates widely vary in the African yellow fever belt. In comparison, a 25-year study in Africa estimated that 13% (95% CI 5–28%) of yellow fever infections presented as severe cases, with 46% (95% CI 31–60%) of severe cases resulting in death [ 37 ]. These findings could also be as a result of better surveillance system and global awareness now compared to what we had previously. Some studies report that a combination of some factors like socioeconomic, demographic and ecological circumstances favoured the high incidence of yellow fever in Africa [ 36 ]. Other authors have indicated that Africa monkey are resistant to the yellow fever virus and if they become infected they usually do not die but rather become immune and humans become accidentally infected during their short forest activities [ 36 ]. Usually African population are at a higher risk of the disease because of very low immunity. This is due to low vaccination coverage and also linked to their daily activities and occupation such as farming, deforestation, and cattle grazing, which bring them in close proximity to the forest where the vectors live.

From the studies that we included in this review, the outbreaks that occurred were due to sylvatic mode of transmission which involves the virus passed on between non-human primates (e.g., monkeys) and mosquito species found in the forest canopy [ 26 – 29 , 32 – 36 ]. Yellow fever is endemic in the sylvatic settings in Africa predominantly in East and West Africa. The sylvatic mode of transmission gives a more regular epidemic pattern as opposed to urban transmission which often gives a periodic and more unpredictable outbreak. This information is ideal as it would help policy makers focus the strategies on how best to fight the regular clustering of yellow fever epidemic. Some of these strategies would include using an integrated mosquito management (IMM) [ 38 ] program steps such as: (i) surveillance: this will help detect the different mosquito species in a given area and with this data, they are able to effectively time larvicide and adulticide activities; (ii) public education: this includes educating the general public on how to control mosquito breeding sites at their backyard; (iii) larval and adult mosquito control: inspecting sources of standing water looking for mosquito larval so as to eliminate them before they become adults that can transmit the virus while the adult mosquitoes are treated using pesticides. We also noticed that yellow fever is endemic with the outbreaks occurring during rainy season. It is known that rainy season increases the breeding sites and the mosquito population which transmit the disease. Hence heavy and prolong rainfall result in high vector population due a favorable breeding environment for mosquitoes due to heavy rainfall during rainy season. This provides adequate ground to support continued circulation of the virus especially in the sylvatic cycle.

Previously, prevention of this disease was done through vector control which was very effective in reducing the occurrence of the disease. However, smaller outbreaks occur due to changes in the distribution of the disease and currently, vaccination strategies such as routine infant immunization, mass vaccination campaigns and vaccination of travelers going to yellow fever endemic regions are used to protect people against these outbreaks. Yellow fever vaccine called 17D is known for preventing the disease and just a single dose of the vaccine confers immunity and lifelong protection. The vaccine provides effective immunity within 10 days for 80–100% of people vaccinated, and within 30 days for more than 99% of people vaccinated [ 39 ]. Most African countries where yellow fever is endemic have included yellow fever vaccine in the Expanded Programme on Immunization (EPI) schedule for new-born babies. Most of the yellow fever cases identified in Africa are seen in the unvaccinated population who live in the yellow fever belt. Some countries have sustained epidemics across multiple years like Ghana (1977–83), Guinea (2000–2005), Nigeria (1986–1994) and Congo (2011–2013) [ 37 ]. Looking at previous data on yellow fever over a period of 25 years in Africa, showed that yellow fever has been persistent in West and Central Africa [ 37 ]. In East Africa, yellow fever has mostly been in Kenya [ 37 ]. However, cases were also reported in Uganda and Ethiopia. The estimated annual provincial incidence in 32 African countries known to be endemic for yellow fever varied from 0.7 to 10% [ 37 ] which is low compared to the overall incidence rates found by this systematic review. A recent study reported that in 2018 there were approximately 109,000 (95% CI 67,000–173,000) severe infections and 51,000 (95% CI 31,000–82,000) deaths due attributed to yellow fever in Democratic Republic of Congo and Amazon region [ 40 ].

In December 2016, cases of yellow fever were detected in Luanda, the capital of Angola, which were previously in the category of low-risk areas for yellow fever. The disease spread rapidly from Luanda to other urban communities in Angola and crossed into the neighbouring country of the Democratic Republic of Congo [ 16 , 41 ]. This observation empathizes the fact that yellow fever is a neglected tropical disease which needs more public health strategies to contain it coupled with more studies to be conducted in Africa to generate in-depth evidence for effective interventions.

Reviewing the vaccination coverage, WHO recommends population vaccination coverage of 80% or more to prevent and control outbreaks. However, all of the regions included in this review has recorded low vaccination coverage. A recent study showed the targeted versus the untargeted vaccination coverage of 54%/63% and 24%/24% in Democratic Republic of the Congo (Central Kongo region), 61%/75% in Senegal (Kaffrine region), zero coverage in Ethiopia (South Omo zone), 67%/67% in Ghana (Volta, Brong Ahafo and Eastern Regions), zero coverage in Kenya (Kerio Valley), zero coverage in Uganda (Northern Uganda, Masaka, Kalungu, Kalangala and Rukungire regions) and 56%/65% in Nigeria (Cross river, Lagos, Oyo, Ogun, Ondo, Kwara and Imo state) [ 42 ]. As shown, all of the regions with a high incidence of yellow fever and case fatality rate in this review had less than 80% vaccination coverage, with some regions having 0%.

The review has shown that yellow fever incidence seems to have been pretty constant throughout African countries over the inclusion period of the review. However, three of the studies included showed high incidence rates [ 26 , 31 , 32 ]. In contrast, the fatality rates varied widely across African countries over the same period of the review. Since most of the studies were conducted between 1984 and 1998, moderately high vaccination coverage rates across much of western and central Africa in the 1970s were the result of mass preventive campaigns in the 1940s to the 1960s [ 42 , 43 ], which reduced the number of outbreaks [ 43 ]. Coverage declined between 1960 and 2000 in most areas due to limited vaccination activity, the birth of new unvaccinated cohorts, a steady decline in the proportion of older covered cohorts through mortality [ 43 ] and vaccine stock shortages, which have been frequently reported in the African region [ 44 ]. Past public health successes led to a lax in maintaining local yellow fever vaccination coverage leading to waning herd immunity and an eventual re-emergence of large outbreaks in West Africa in the 2000s [ 42 , 45 ].

In fact, yellow fever may not easily be eliminated due to the presence of non-human wildlife reservoirs that sustain the sylvatic transmission cycle of the virus in non-urban settings, however the risk of a yellow fever outbreak can be eliminated if successful vector control, vaccination and surveillance of the disease are implemented and maintained [ 42 ]. The joint effort by the WHO, the United Nations International Children's Emergency Fund (UNICEF), the Global Alliance for Vaccinations and Immunization (GAVI) and yellow fever virus (YFV) endemic countries created the Yellow Fever Initiative (YFI) in 2006 [ 46 ], which focused primarily on widespread yellow fever vaccination initiatives and the implementation of childhood immunization vaccine. In the context of an emergency preparedness effort, this initiative has created an opportunity for global stock of yellow fever vaccines [ 41 , 47 ]. The 17D yellow fever vaccine is effective, safe, affordable, readily available, and can prevent the disease with just a single dose being sufficient enough to confer sustained immunity and lifelong protection.

The review has limitations the main one being paucity of yellow fever incidence/prevalence data from different African countries. The incidence estimates from the different studies were significantly heterogeneous and there were not enough data from the studies to determine the sources of this heterogeneity, but could have been due to the different geographical nature in Africa, seasons, settings (rural or urban), different populations, statistical methods and regional differences. Studies included in this review were fewer hence limiting the scope to generate more accurate estimates of the burden of yellow fever in African regions. Secondly most of the included studies were predominantly from East–West Africa, hence understanding the burden of yellow fever in other regions is restricted.

The risk of bias assessment showed that 83% of studies showed to have a low risk of bias, however included studies mighty not have been representative of a larger population. Furthermore, in most of the included studies, the number of confirmed and suspected cases were not clearly distinguished. As such this limits generalizability of the incidence and case fatality rate estimates to a larger African population.

This systematic review identified 12 observational studies assessing yellow fever incidence and fatality rates in Africa. Data shows that yellow fever incidence rate is quite constant across African countries with high incidence rates reported in three studies of the 12 studies. The case fatality rates for yellow fever varied widely across Africa. However, the lack of reliable epidemiological data on yellow fever in Africa compromises the public health priority that could give rise to yellow fever infection.

For that reason, it is essential to provide standardized demographic health surveys as a population surveillance strategy to track the burden of yellow fever as well as accurate diagnostic measures for early recognition and treatment of yellow fever. Knowing that yellow fever usually re-emerges in rural areas, rural facilities should be enforced in yellow fever prevention, management and reporting approach. Timely and accurate diagnosis of yellow fever could avoid untoward case fatality rates and minimize under-reporting. Accurate yellow fever data is substantial for good public health policy and guide planning of vaccine volumes and delivery system. In addition, public health control strategy should focus on strengthening yellow fever prevention including incorporating yellow fever immunization schedules in African endemic countries and mandatory reporting of cases in primary, secondary and tertiary levels. Future research should focus on evaluating yellow fever immunogenicity in children.

Acknowledgements

Abbreviations, authors’ contributions.

AN, PN, EZS designed the study. AN and EZS screened, selected studies and extracted data for the study. AM did data analysis and interpretation of results. AN, PN drafted the manuscript. JLT critically reviewed the manuscript and wrote other sections of the review. All authors reviewed the manuscript. All authors read and approved the final manuscript.

Availability of data and materials

Declarations.

Not required.

Not applicable.

The authors declare no competing interests.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.