The potential impact of expanding target age groups for polio immunization campaigns

Background

The global commitment to eradicate wild polioviruses (WPVs) and end poliomyelitis led to the launch of the Global Polio Eradication Initiative (GPEI) in 1988, which successfully reduced global polio incidence by 99% [1] and eradicated one of the three WPV serotypes (i.e., type 2 or WPV2) around 1999 [2]. By early 2014, all except three countries (i.e., Afghanistan, Nigeria and Pakistan) successfully interrupted indigenous transmission of WPV type 1 (WPV1) and no country has reported a case of WPV type 3 (WPV3) for over a year [3]. Despite these successes, the GPEI still needs to stop transmission of WPV1 everywhere contemporaneously and transition away from the use of oral poliovirus vaccine (OPV) to achieve the ultimate goal of ending all cases of poliomyelitis [4]. Partly due to importations into previously polio-free countries that caused outbreaks, the global annual incidence of paralytic polio remained around 1,000-2,000 cases between 2000 through 2010. The World Health Assembly in 2012 urged further intensification of the GPEI by declaring “the completion of global poliovirus eradication a programmatic emergency for global public health” ([5], p. 2). During 2012 the GPEI reported the fewest cases and smallest number of countries reporting cases in its history [6].

Successful polio eradication requires achieving and maintaining sustained high levels of population immunity until wild poliovirus transmission stops everywhere contemporaneously [4]. Although many high- and middle-income countries can maintain high population immunity by reaching and sustaining high routine immunization coverage, other countries must rely on the use of supplemental immunization activities (SIAs) that provide OPV during a short period of time regardless of immunization history to periodically and significantly boost population immunity [6]. Historically, SIAs emerged as a strategy to interrupt poliovirus transmission during the low season by “flooding the environment with vaccine virus so that a susceptible child has a lower probability of encountering wild virus and/or by rapidly increasing population immunity so that fewer susceptible children remain” ([7], p. 1332). The GPEI continues to rely heavily on SIAs to close immunity gaps resulting from sub-optimal routine immunization by targeting all children under 5 years of age [8]. Typically, polio SIAs focus on immunizing children younger than 5 years of age, because infants become susceptible once they lose protection from maternal antibodies and most people get exposed to circulating viruses during early childhood. Based on experience dating back to early OPV trials by Albert Sabin [9] and polio elimination from the Western Hemisphere [10-12], SIAs can dramatically reduce the prevalence of WPV and rapidly interrupt transmission. However, following the introduction of vaccination, the circulation of viruses may become more episodic, and those missed by vaccination may continue to remain susceptible and begin to accumulate.

With the passage of time since the interruption of WPV transmission in most places, importation outbreaks increasingly involve paralytic patients 5 or more years of age [13-17], which provides clear evidence that immunity gaps persist beyond early childhood in polio-free areas. Recognizing the need to protect older children and adults in these situations, the GPEI modified its outbreak response recommendations to explicitly expand the age group targeted in the first two SIAs up to 15 years of age or older depending on the context even in the absence of observed outbreak cases in older children and adults [8]. The failure of frequent SIAs targeting only children under 5 years of age to interrupt indigenous WPV transmission in the remaining endemic areas also raises the possibility that older children and adults may contribute significantly to the overall transmission of polioviruses in those places, even in the context of a substantial missed proportion of targeted children during SIAs. This would mean that expanding the target age groups for some SIAs in these areas may provide a strategy to more rapidly interrupt WPV transmission and maintain population immunity at a level high enough to prevent transmission. Alternatively, frequent SIAs targeting only young children may not prevent persistent transmission because they fail to protect enough children to completely stop transmission or they repeatedly miss relatively isolated subpopulations that sustain clusters of fully susceptible individuals. This would mean that reaching previously missed subpopulations may more rapidly interrupt transmission than expanding the target age range in the entire population.

Achieving WPV eradication is the primary objective of the GPEI. Further delay of the achievement of global WPV eradication and subsequent OPV cessation carries significant economic costs [18,19]. Donor fatigue and competition for public health resources during difficult economic times imply real financial risks for the GPEI. This paper evaluates the potential role and benefits of expanded age group vaccine campaigns using a differential-equation based poliovirus transmission model [20]. We explore different conditions and timing with respect to the implementation of expanded age group campaigns by considering several scenarios. We consider both preventive expanded age group SIAs occurring before outbreaks in previously polio-free areas or expanded age group SIAs prior to surges in incidence in endemic areas, and more reactive expanded age groups SIAs in the context of outbreak response in previously polio-free areas or efforts to accelerate the interruption of WPV transmission in endemic areas. Specifically, we model the potential impact of including older age groups in SIAs for three specific situations: alternative responses to the 2010 outbreak in Tajikistan, retrospective examination of elimination in northern India, and retrospective and prospective strategies to accelerate elimination in endemic northwestern Nigeria. While numerous inputs of the poliovirus transmission model remain highly uncertain [20-22], we assume for this analysis that the process of model validation across a wide range of situations produced representative generic model inputs that apply to poliovirus transmission in any situation (e.g., immunity states, waning process, OPV evolution process) as well as realistic characterization of the conditions in each of the three situations examined [20].

Methods

We use a previously developed model of poliovirus transmission and consider three of the situations we used to fit generic model inputs [20]. Briefly, the deterministic, differential equation-based model includes immunity states to characterize the effect of recent (8 immunity states) or historic (24 immunity states) live poliovirus (LPV) infections and/or successful inactivated poliovirus vaccine (IPV) vaccinations. For this analysis, the 18 IPV-related immunity states remain inconsequential because we assume negligible IPV use in the three situations. Individuals in any active immunity state remain completely protected from paralysis if re-infected, but fully susceptible individuals and infants with insufficient maternal immunity (those who no longer possess protective levels of maternal antibodies) become subject to a serotype-specific, age-independent paralysis-to-infection ratio if infected [20]. The model characterizes the evolution process from OPV viruses towards WPV-like vaccine-derived polioviruses over 20 discrete stages, and explicitly considers serotype differences and both fecal-oral and oropharyngeal transmission [20].

The overall basic reproductive number (R₀) characterizes the average number of secondary infections generated by combined fecal-oral and oropharyngeal transmission from one infected previously fully susceptible individual introduced in an entirely susceptible population. R₀ for any population depends on hygienic and sanitary conditions, population density, mixing patterns, and climate, and may exhibit short-term oscillations due to seasonality as well as potential long-term trends. Given the different conditions that exist around the world, R₀ varies between populations. The model assumes space-homogeneous mixing within each modeled (sub)population but includes age-heterogeneous mixing by assuming that individuals tend to mix preferentially with individuals of similar ages [20]. For each situation, we separately determined the proportion of contacts of any mixing age group (i.e., < 5, 5–14, or > 15 years of age) reserved for individuals of the same mixing age group (κ) to produce age distributions of cases consistent with the data. Specifically, the model assumes κ of 30%, 35%, and 40% of contacts in Tajikistan, northern India, and northwestern Nigeria, respectively, implying moderate preferential mixing within each mixing age group, with the remaining proportion of contacts occurring proportionately with individuals of all ages (i.e., including the mixing age group of the contacted individual) [20]. Aging represents a multi-stage process with transitions between any two age groups occurring according to a first-order, exponential expiry process [23,24], as specified in the model description [20]. Similarly, we characterize waning mucosal immunity, which influences the ability to participate in fecal-oral and oropharyngeal transmission, as a multi-stage process from high immunity to poliovirus transmission immediately after recovery from an LPV infection to reduced immunity to poliovirus transmission in the last waning stage [20]. We assume previously infected individuals reach the last waning stage after 4 years (Types 1 and 2) or 3 years (Type 3) on average in the absence of further infections [20]. Additional file 1 provides the equations for the force-of-infection in the model with age-mixing and multiple subpopulations as well as the waning curves.

To characterize the stochastic event of die-out (elimination) of transmission of each LPV virus (i.e., WPV or OPV-related virus in each reversion stage) in a population in our deterministic model, we fixed a transmission threshold of the effective infectious proportion (i.e., the number of infectious people in a population, weighted by their relative infectiousness to others, divided by the total number of people in the same population) below which the force-of-infection of a LPV becomes 0 in the model [20]. We define the threshold as a proportion rather than an absolute number to avoid model artifacts related to population size. Specifically, using a threshold of 1 infection in a large population would eventually lead to circulating vaccine-derived polioviruses (cVDPVs) following a SIA round in a WPV-free setting with low or no routine immunization coverage, because it would take a long time for the prevalence in the last reversion stage to drop below 1, while cVDPVs would not emerge in a very small population in which the prevalence would never exceed 1 in higher reversion stages. We apply the threshold for each mixing age group and subpopulation separately, which means that virus can continue to circulate in one of these groups while giving a force-of-infection of 0 to other groups. For mixing between sub-populations we assume highly preferential mixing so that the time of elimination can differ between subpopulations substantially (especially if their immunity levels differ significantly). For mixing between age groups we assume relatively much less preferential mixing, so that elimination times between age groups remain very similar. This approach provides a very crude method to approximate die-out that occurs in actual populations and depends on local heterogeneity and chance. However, the threshold we determined generally produced elimination times consistent with the evidence [20] and allows us to explore the relative differences in potential elimination times between scenarios, even if the actual elimination times in the model do not reflect the stochasticity of actual virus elimination.

We conducted extensive literature reviews with experts in the field [21,25] and elicited numerical values from experts to estimate initial ranges of values for model inputs related to poliovirus transmission and immunity [22]. We then calibrated the model within the ranges obtained from the expert review process by visually fitting historical poliomyelitis incidence data and other features of poliovirus transmission in 9 different situations, including the emergence of cVDPVs and WPV elimination times, the age distribution of cases, and estimates of zero-dose children derived from surveillance data (see Additional file 1). We adopt the resulting generic and situation-specific model inputs for this analysis [20] (specifically, Tables I, VII, XI, and XII for generic, Tajikistan, northwestern Nigeria, and northern India inputs, respectively in the prior publication). However, we modified the characterization of SIAs for prospective use of the model to allow it to directly specify the true coverage of individual rounds and the probability of children repeatedly receiving or missing doses (see Additional file 1). Briefly, the approach requires specifying the true coverage (TC) of each SIA, defined as the fraction of the targeted population receiving a dose, and the repeated missed probability (P _RM ), defined as the conditional probability that a targeted individual does not receive a dose in a round, given that the individual did not receive a dose in the previous round despite falling into the targeted population for that round. From these, we calculate the repeated reached probability (P _RR ), defined as the conditional probability that a targeted individual receives a dose in a round, given that the individual received a dose in the previous round. In the model, we then track which of the fully susceptible and maternally immune individuals (i.e., defined as children born with maternal antibodies transferred to them by an immune mother) at the time of an SIA are: (1) children born since the last SIA, subject to a coverage of TC, (2) children who did not receive a dose in the previous round, subject to a coverage of 1-P _RM , and (3) children who received a dose in the previous SIA but remain fully susceptible or maternally immune due to failure of the vaccine to take, subject to a coverage of P _RR . To ensure that the overall coverage matches the specified TC in each round, we evenly divide all remaining doses over all targeted individuals with active immunity (i.e., individuals not in the fully susceptible or maternally immune state). Fractional rounds target less than the whole modeled population target age range (e.g., only one district when the modeled population represents a state) and reduce each coverage by the same fraction targeted (F). For the under-vaccinated subpopulations that we characterize in northern India and northwestern Nigeria, we specify a relative coverage for this subpopulation compared to the general population (see Additional file 1).

For all three situations, we compared the starting point for the model to the best information available to us as of August 2013. For all analyses, we keep all model inputs constant except for those vaccination inputs that specify the different policy scenarios (e.g. target age groups of an SIA). For Tajikistan, we modeled a single population, while for northern India and northwestern Nigeria we modeled 90% of the population as the general population and 10% as a relatively isolated subpopulation. Following the publication of our earlier analysis [20], more complete data became available for analysis on the cases that occurred in the Tajikistan outbreak and the dose histories of all acute-flaccid paralysis (AFP) cases since 1997 for northern India and since 2003 for northwestern Nigeria. We also updated the original demographic data [26] with the 2012 revision from the United Nation’s Population Division World Population Prospects [27]. These data allowed us to better assess the age distribution of cases and outbreak kinetics for Tajikistan and the historic proportions of missed children for northern India and northwestern Nigeria. Applying the new characterization of SIAs to ensure consistency with the historic reported zero-dose proportions among non-polio AFP (NPAFP) cases over time, rather than estimates based on expert judgment, led to some adjustments of generic and situation-specific model inputs in order to produce more realistic historic poliomyelitis incidence (see Additional file 1, Table 1).

Table 1 summarizes the inputs that we updated based on improved information and better characterization of SIAs, with all other input values in the model remaining the same as reported earlier (see Additional file 1) [20]. In the context of implementing the refined approach to SIA characterization, we assume the same true coverage by subpopulation for the expanded age group SIAs as for children under 5 years of age (i.e., expanded age groups SIAs imply the same coverage applied to a larger target age range and the corresponding increase in vaccine doses used). The assumption about the repeated missed probabilities for older children and adults results in negligible impact because a high proportion of individuals in the expanded age groups benefit from actively acquired immunity, while we distinguish repeated missed and repeated reached probabilities only for fully susceptible and maternally immune individuals.

We compute the number of doses administered for any given SIA as the sum over each population of F × TC × N _target , where the size of the target population N _target comes directly from the model. The number of doses distributed differs from the doses administered, depending on the wastage for the modeled situations. While we considered data from the GPEI on doses planned nationally for each SIA, we do not know how many doses the modeled states actually administered, so we estimated these in the model.

Results

Discussion

The results from this analysis suggest that expanded age group SIAs may prevent polio cases and shift WPV elimination forward in time, in some cases significantly and in other cases moderately, depending on the circumstances. If conducted preventively before a possible surge in cases (e.g., prior to an introduction into a polio-free area, a time of poor vaccination intensity in an endemic area), then we expect the expanded age group round to prevent a significant number of cases. This supports the GPEI policy since 2012 of expanding the age group targeted in the first two SIAs in response to an outbreak in a polio-free country independent of the initial age distribution of cases, as long as the outbreak response remains large enough to include areas not yet affected by the outbreak. The ability of reactive expanded age group SIAs to stop transmission in endemic areas depends on the target age range and how well the expanded age group round reaches the key reservoirs of transmission. Consequently, the existence of a preferentially mixing subpopulation consisting of clusters of un(der)-vaccinated individuals in large enough numbers to sustain transmission represents a key limiting factor for expanded age group rounds if these rounds do not reach these populations better than regular SIAs targeting children under 5 years of age. However, the model results show that as new under-vaccinated communities get identified with improved micro-planning, targeting age groups older than only children under 5 years of age in these communities may under some circumstances help accelerate the interruption of WPV transmission. Moreover, the widespread social mobilization of a massive expanded age group round may help to reach more children under 5 years of age as well. Vaccinating expanded age groups immunizes some of the remaining fully susceptible individuals in these age groups, which although they represent only a small proportion of all individuals in these age groups [33], could add up to large numbers due to the size of the population of individuals over 4 years of age.

Perhaps more importantly, vaccinating the expanded age groups boosts already immune individuals so that their ability to participate in transmission decreases. The importance of this effect depends on the inherent effect of waning on the ability to participate in poliovirus transmission. While some consensus suggests that even individuals with historic poliovirus infection or vaccination and serum antibodies that dropped below the detection limit still remain protected from paralytic poliomyelitis disease due to memory immunity, much uncertainty remains about the role of memory immunity in limiting participation in asymptomatic poliovirus transmission [21,22]. Several studies suggest that individuals with historic immunity become infected as readily after an OPV challenge as fully susceptible individuals [34-38], but data on duration and titers of excretion remain limited and no single study provides a direct comparison between known fully susceptible individuals and individuals with historic, waned immunity [21]. Moreover, the relationship between any reduction in excreted virus titers and infectiousness to others remains highly uncertain due to the practical challenges of measuring these directly [22]. Nevertheless, the epidemiology of polio strongly suggests that memory immunity reduces participation in transmission to some extent, because otherwise childhood vaccination would not sustain wild polio-free status in so many countries in the world for such a long time. Our assumptions about waning represent the result of an extensive model calibration process within uncertainty bounds obtained from expert assessments [22] based on careful consideration of the literature [21] that produced dynamic poliovirus transmission behavior consistent with the evidence across a large number of situations, including Tajikistan, northern India, and northwestern Nigeria [20]. Specifically, we assume that individuals with 2 or more historic poliovirus infections (from any live poliovirus infection) ultimately acquire a relative fecal-oral transmission potential compared to fully susceptible individuals of approximately 0.1 after starting out with a relative fecal-oral transmission potential of near 0 immediately after recovery from the infection. The assumptions differ by immunity state (e.g., for individuals with only 1 prior poliovirus infection, we assume approximately double these values; see Additional file 1). The expanded age group rounds move individuals from the historic to the recent immunity state, which explains the improvement in the overall population immunity even if very few fully susceptible individuals remain in the older age groups. We examined different assumptions about waning but those would produce behavior inconsistent with the evidence in some situations, unless we modified other assumptions. For example, in northwestern Nigeria, with more waning arguably under-vaccinated subpopulations matter less and the general population may sustain WPV transmission on its own, such that expanded age group SIAs could yield more benefits (i.e., reduced cases and shorter duration of transmission). However, such a scenario produces many more cases during the early 2000s, which appears inconsistent with the evidence, and would also imply that widespread WPV transmission continues throughout northwestern Nigeria even after 2010, rather than transmission confined to clusters of preferentially mixing under-vaccinated communities. Based on the extensive model calibration process, we believe that our model provides a realistic characterization of waning. While many existing poliovirus transmission models ignore the potentially important dynamics of reinfection and waning, one model that includes it also assumed increasing potential to participate in transmission as a function of time since the last infection [39]. Compared to our assumption of 0.1, Mayer et al. assumed that waning immunity eventually leads to a much higher relative transmission potential compared to fully susceptible individuals of between approximately 0.2 and 0.6, based on their interpretation of the evidence without a direct comparison to poliovirus incidence data [39].

Our prospective results for northwestern Nigeria offer some hope that the area we modeled may achieve sufficient population immunity to stop WPV1 elimination by the end of 2014 given the assumed current path. However, this part of Nigeria represents only one part of the country, which must achieve and maintain high levels of vaccine coverage and population immunity everywhere. The results shown here depend on the continuation of the aggressive immunization strategy that we modeled, and any change in strategy will change the results. Continued insecurity in the region, particularly in Kano state, represents an ongoing concern, which might reduce the ability to sustain immunization intensity in the area. Moreover, with only two tOPV SIAs per year the model did not project elimination of type 2 cVDPVs despite the assumed sustained improvement of SIAs going forward. This implies the need to actively manage population immunity for all three serotypes rather than focusing on a single serotype to reduce the immediate risks of importation outbreaks and cVDPV ermergence and the risks of type 2 cVDPV emergence following the planned cessation of type 2-containing OPV use [4,8,40,41]. Although we do not anticipate that our overall inferences related to the potential role of expanded age groups for SIAs would change significantly if a resurgence occurs, we emphasize that targeting expanded age groups may produce the largest reductions in expected cases and elimination times in areas with significant existing population immunity gaps. Although we focused on the northwestern part of Nigeria, the northeastern parts also warrant additional attention.

Our analyses yielded some important insights not directly related to SIAs targeting expanded age groups. First, the retrospective analysis of the 2010 Tajikistan outbreak makes a very strong case for a focus on prevention (i.e., more pSIAs), which could save both substantial health and financial costs. Given that an eradication initiative should focus on prevention of cases before they occur [4], pSIAs must represent an important programmatic priority. In the context of limited resources, continued assessment of population immunity in previously polio-free countries and identification of clusters of susceptible people should motivate efforts to conduct pSIAs [42]. Complementing serologic assessments and vaccination data collection, modeling may offer a tool to better identify population immunity gaps before they occur [4] and to determine the most important factors that contribute to low population immunity [43]. As long as WPV continues to circulate globally it can transmit across long distances [44], and the risk exists that the virus will find such clusters [45], countries will need to invest in both addressing these population immunity gaps and interrupting WPV globally as soon as possible. This suggests that if known immunity gaps exist, conducting preventive SIAs may offer a strategy that will save overall cases and doses of vaccine. Second, continued efforts to reach previously-missed or under-vaccinated populations remain essential to interrupting transmission, regardless of whether they focus only on children under 5 years of age or expanded age groups. However, reaching these groups with sufficient doses takes time and focusing only on these groups while ignoring the general population carries the risk of perpetually chasing the virus, without attaining enough general population immunity to prevent immunity gaps elsewhere. Thus, continued intense efforts to vaccinate the general populations everywhere remain critical while national programs in the remaining endemic countries continue to find and reach any remaining underserved populations.

As with any model, our results reflect the limitations of the model and the process used to generate model inputs. We rely on the approach of developing appropriate reference cases and making comparisons to these, to reduce the potential impact of the assumptions that we made to deal with inadequate data and multiple uncertainties as we calibrated the model. Our characterization of a subpopulation of un(der)-vaccinated individuals provides a tractable way to attempt to capture some of the complex heterogeneity that exists, but we recognize that it represents an abstract construction and that real populations are highly complex systems. Although we attempt to capture all of the important complexities that exist, our model represents a simplification of reality. With respect to the interpretation of the model results for the time to elimination, we caution that our simplistic approach to incorporate die-out in our deterministic model does not capture the stochastic processes and true heterogeneity that affect real elimination of polioviruses. Thus, the shifts in elimination times between scenarios that we reported indicate differences in the time when WPV prevalence in the last remaining mixing group reach very low levels, which may not fully capture all of the uncertainty and variability that underlies the actual process of die out. Other limitations related to the differential-equation based model discussed in detail elsewhere also apply to this analysis of expanded age groups [20].

Conclusions

This study suggests the need to carefully understand the epidemiological situation, including the role of subpopulations of susceptible individuals, and the potential for waning immunity and re-infection of older children and adults. Our results suggest relatively modest benefits of expanded age group SIAs on estimated paralytic cases and the time of interruption of WPV transmission, with the greatest potential to increase population immunity and accelerate WPV eradication for SIAs conducted preventively or reaching under-vaccinated sub-groups.

Abbreviations

AFP: Acute flaccid paralysis; bOPV: Type 1 and 3 bivalent OPV; cVDPV: Circulating vaccine-derived poliovirus; DTP: Difteria-tetanus-pertussis vaccine; GPEI: Global polio eradication initiative; IPV: Inactivated poliovirus vaccine; LPV: Live poliovirus; mOPV1,3: Type 1 and 3 monovalent OPV, respectively; NA: Not applicable; NPAFP: Non-polio AFP; OPV: Oral poliovirus vaccine (generic term); oSIA: Outbreak response SIA; POL: Any polio vaccine; pSIA: Preventive SIA; PV1,2,3: Type 1, 2, or 3 poliovirus (generic terms); R0: Basic reproductive number; SIA: Supplemental immunization activity; tOPV: Trivalent OPV; WPV: Wild poliovirus (generic term); WPV1,2,3: WPV type 1, 2, or 3, respectively; WUP: Western Uttar Pradesh.

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

RJDT led the development and implementation of the poliovirus transmission model and expanded age group scenarios, conducted the analyses for Tajikistan and northwestern Nigeria, and drafted the manuscript. DAK participated in the development of scenarios and implemented and conducted the analysis for northern India. SGFW, MAP, and SLC participated in the development of expanded age groups scenarios and helped with interpretation of the data for each situation. KMT conceived of the study, participated in its design and coordination and helped draft the manuscript. All authors revised the manuscript and read and approved the final manuscript.

Pre-publication history

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1471-2334/14/45/prepub

Supplementary Material

Acknowledgments

The authors thank the U.S. Centers for Disease Control and Prevention (CDC) for supporting this work under Cooperative Agreement U66IP000519-01. The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the CDC. The authors thank Drs. Gregory Armstrong, Johannes Everts, Marta Gacic-Dobo, Alex Gasasira, Benjamin Lopman, Patrick O’Connor, Roland Sutter, Rudi Tangermann, Iris Tetford, Linda Venczel, Jay Wenger, Chris Wolff, and the Global Polio Laboratory Network for information.