Modeling poliovirus transmission in Pakistan and Afghanistan to inform vaccination strategies in under-vaccinated subpopulations

Abstract

Due to security, access, and programmatic challenges in areas of Pakistan and Afghanistan, both countries continue to sustain indigenous wild poliovirus (WPV) transmission and threaten the success of global polio eradication and oral poliovirus vaccine (OPV) cessation. We fitted an existing differential equation-based poliovirus transmission and OPV evolution model to Pakistan and Afghanistan using four subpopulations to characterize the well-vaccinated and under-vaccinated subpopulations in each country. We explored retrospective and prospective scenarios for using inactivated poliovirus vaccine (IPV) in routine immunization or supplemental immunization activities (SIAs). The under-vaccinated subpopulations sustain the circulation of serotype 1 WPV and serotype 2 circulating vaccine-derived poliovirus. We find a moderate impact of past IPV use on polio incidence and population immunity to transmission mainly due to 1) the boosting effect of IPV for individuals with pre-existing immunity from a live poliovirus infection, and 2) the effect of IPV-only on oropharyngeal transmission for individuals without pre-existing immunity from a live poliovirus infection. Future IPV use may similarly yield moderate benefits, particularly if access to under-vaccinated subpopulations dramatically improves. However, OPV provides a much greater impact on transmission and the incremental benefit of IPV in addition to OPV remains limited. This study suggests that despite the moderate effect of using IPV in SIAs, using OPV in SIAs remains the most effective means to stop transmission, while limited IPV resources should prioritize IPV use in routine immunization.

1. INTRODUCTION

In 1988, when the world resolved to globally eradicate polio,⁽¹⁾ all three wild poliovirus (WPV) serotypes circulated, causing hundreds of thousands of paralytic poliomyelitis (polio) cases in over 125 countries each year.⁽²⁾ Since 2013, only serotype 1 WPV (WPV1) has caused any reported polio cases associated with WPVs⁽³⁾ and only 3 WPV1-endemic countries remained in 2017 (i.e., Pakistan, Afghanistan, and Nigeria).⁽⁴⁾ The Global Polio Eradication Initiative (GPEI) increasingly focuses on transitioning to the post-eradication era, including coordination of the 2016 global switch from oral poliovirus vaccine (OPV) containing all 3 attenuated poliovirus serotypes (trivalent OPV, or tOPV) to OPV containing only attenuated serotype 1 and 3 polioviruses (bivalent OPV, or bOPV).⁽⁵⁾ Following the certification of serotype 2 WPV (WPV2) eradication in 2015,⁽⁶⁾ this switch effectively withdrew all serotype 2-containing OPV except for emergency response use of monovalent OPV (i.e., mOPV2) from an outbreak response stockpile. OPV contains attenuated live poliovirus (LPV) that can, in very rare cases, mutate to neurovirulent vaccine-derived poliovirus (VDPV).^{(7, 8)} Elimination of all disease caused by LPVs requires the cessation of OPV use, with the switch to bOPV representing the first step. OPV risks include sporadic vaccine-associated paralytic poliomyelitis cases, circulating VDPV (cVDPV) outbreaks that can emerge in populations with low coverage and behave like WPVs, and immunodeficiency-associated VDPVs (iVDPVs) excreted by immunocompromised individuals with a long-term infection.⁽⁹⁾ OPV cessation will end the routine introduction of viruses that can cause vaccine-associated paralytic poliomyelitis and evolve to VDPVs. However, VDPVs that trace back to pre-cessation OPV use may persist and cause outbreaks, as may potential releases of viruses from laboratories or unauthorized use of OPV after cessation.^{(10, 11)}

The GPEI must simultaneously ensure completion of global WPV eradication and manage the risks associated with OPV cessation. These goals sometimes lead to competing priorities for financial, human, and vaccine resources, which historically experienced supply challenges for both OPV and inactivated poliovirus vaccine (IPV).^{(12, 13)} For example, the available OPV filling lines must meet the competing priorities of filling enough bOPV to complete WPV1 eradication and filling enough mOPV2 to ensure sufficient mOPV2 reserves in the stockpile.⁽¹⁴⁾ Most higher-income countries currently use IPV because it does not come with the OPV-associated risks while still providing their populations, characterized by high immunization coverage and good hygiene and sanitation, with sufficient protection from poliovirus transmission to prevent large outbreaks.^{(15, 16)} IPV provides good individual protection from disease and oropharyngeal transmission but costs more and provides much less protection from fecal-oral poliovirus transmission than OPV,^{(16, 17)} as demonstrated by a recent episode of asymptomatic WPV1 circulation in Israel despite high IPV coverage.^{(18, 19)} Exclusive IPV use likely will not significantly reduce the probability of outbreaks in lower-income settings with conditions conducive to fecal-oral transmission.⁽²⁰⁾ However, IPV provides the only available vaccine to protect individual children born after the switch to bOPV from serotype 2 paralysis due to infection. Accordingly, the GPEI recommended the introduction in routine immunization (RI) of at least one IPV dose in all countries prior to the switch.⁽⁷⁾ However, supply limitations forced the GPEI to delay IPV introduction in 20 countries and caused stock-outs in other countries.⁽²¹⁾ Nevertheless, the GPEI prioritized IPV use in supplemental immunization activities (SIAs, i.e., campaigns that target children within a broad age range, typically through age 4 years, regardless of immunization history) based on evidence that IPV boosts the intestinal immunity of individuals with immunity from prior LPV (i.e., WPV, OPV, OPV-related, and VDPV) infections,^{(22, 23)} although modeling provided mixed evidence that this translates into a significant effect on transmission at the population level.^{(20, 24, 25)}

The geographic area of Pakistan and Afghanistan represents the only epidemiological region besides the Lake Chad Basin in which WPV1 eradication and post-switch risk management represent simultaneous critical priorities. We previously developed a deterministic, differential equation-based model of poliovirus transmission and OPV evolution (i.e., the DEB model)⁽²⁶⁾ that we tailored to settings like northwest Nigeria and northern India to help inform polio endgame strategies.^{(11, 19, 20, 25, 27–33)} We use the DEB model to reproduce the epidemiological evidence related to poliovirus transmission in Pakistan and Afghanistan, characterize population immunity to poliovirus transmission of all three serotypes, and explore IPV strategies aimed at WPV1 eradication and serotype 2 risk management.

2. METHODS

We adopt the generic structure and model inputs from the DEB model that remain the same in all settings, which represents the result of an extensive expert review ^{(15, 34, 35)} and model calibration ⁽²⁶⁾ process (see Appendix A1). This process ensured internally consistent assumptions to simultaneously reproduce the epidemiological evidence in 9 diverse situations. We express population immunity to transmission with the mixing-adjusted effective immune proportion (EIPM), which in a single number reflects the potential contribution to transmission of all individuals in the population, taking into account how prior infections and vaccination affects their potential to participate in transmission and how they mix with other age groups or subpopulations.⁽²⁹⁾ The model assumes that individuals with prior live poliovirus infections or who responded to IPV vaccination remain permanently protected from polio paralysis, although they can participate asymptomatically in transmission to different degrees (Appendix A1). If the EIPM remains above the threshold, i.e., EIP*=1–1/R0, where R₀ represents the basic reproduction number that quantifies the transmissibility of the virus,⁽³⁶⁾ then the average number of new secondary infections per infected individual remains below 1, leading to eventual die-out of the virus. For example, in a population with an R₀ of 10, transmission will eventually die out if EIPM remains above 0.9 (i.e., 1–1/10). Although EIP* differs by OPV reversion stage because R₀ differs by OPV reversion stage, in this study we focus on the EIP* for WPV, which equals that of fully-reverted VDPVs because we assume the same R₀ for both viruses.^{(26, 35)}

An estimated 190 million and 33 million people lived in Pakistan and Afghanistan, respectively, in 2015.⁽³⁷⁾ RI with tOPV started in 1978 in both countries,^{(38, 39)} although coverage with the third dose remained below 10% until 1983.⁽⁴⁰⁾ Pakistan began tOPV SIAs in 1994 and acute flaccid paralysis (AFP) surveillance in 1995⁽³⁸⁾ while Afghanistan started AFP surveillance and national tOPV SIAs in 1997.⁽³⁹⁾ Only passive reporting of polio cases occurred before AFP surveillance started, probably resulting in substantial under-reporting of cases, particularly in Afghanistan, which did not report any polio incidence data to the World Health Organization between 1992–1996.⁽⁴¹⁾ As AFP surveillance continued to improve, Pakistan switched to the virological case definition for polio in 2000 and Afghanistan in 2003.^{(38, 39)} Despite considerable success in both countries by the mid-2000s, some areas remained difficult to access, and further deterioration of access due to increased conflict and floods led to a resurgence of reported cases from the late-2000s through 2014. Access-compromised areas concentrate around the Pakistan-Afghanistan border, although reported cases also continue to occur at a low rate throughout the two countries, in particular around Karachi. The areas reporting the most polio cases coincide with areas with a large fraction of people of Pashtun ethnicity, which represent approximately 15% of the total Pakistani population and 40% of the total Afghan population.^{(42, 43)} Precise numbers remain uncertain due to the high mobility of these populations and the lack of a recently conducted census in either country.

To model poliovirus transmission in Pakistan and Afghanistan, we divided both countries into a general population and an under-vaccinated subpopulation, consistent with prior situations modeled.⁽²⁶⁾ The under-vaccinated subpopulations represent historically under-vaccinated communities due to security and/or programmatic issues concentrated around, but not necessarily limited to, both sides of the Pakistan-Afghanistan border. This represents a conceptual construct characterized primarily by historically low vaccination levels rather than geography because under-vaccinated communities exist in different parts of Pakistan and Afghanistan and include mobile populations. Assuming that the under-vaccinated communities predominantly include people of Pashtun ethnicity, but that only a fraction of all Pashtun communities represent truly under-vaccinated communities, we estimate that the under-vaccinated subpopulation represents 5% of the total population of Pakistan and 10% of the total population of Afghanistan. Based on the model behavior in relation to the serotype 2 cVDPV (cVDPV2) transmission detected in Pakistan at the end of 2016, we considered two mixing scenarios that produced model behavior consistent with the observations as of early November, 2017 (see Appendix A2). The first mixing scenario (i.e., More isolated) assumes that both under-vaccinated subpopulations remain highly isolated, with only 0.3% of contacts occurring with individuals in other subpopulations. The second mixing scenario (i.e., Less isolated) assumes a ten times greater fraction of 3% of contacts occurring with individuals in other subpopulations. Given the shared ethnicity and high mobility of both under-vaccinated subpopulation, both mixing scenarios assume that between one-third and one-half of the contacts outside of the under-vaccinated subpopulation occur with contacts of the other under-vaccinated subpopulation across the border.

Table I lists other constant model inputs related to population structure, poliovirus transmission, and vaccination. We constrained the best estimates by values previously used in the DEB model in other settings and ensured that they did not yield model outputs inconsistent with the evidence. For example, the assumed average WPV1 R₀ of 11 remains between the previously assumed values of 7.5 for northwest Nigeria and 13 for northern India.⁽²⁷⁾ An R₀ value below 10 would result in premature die-out of WPV serotypes, while an R₀ above 12 would yield more cases after the first SIAs than plausible as well as unrealistic timing and number of cVDPV2 cases. Similarly, we assumed moderate seasonal variation in R₀ (i.e., annual maxima and minima of 15% above and below the average R₀ peaking in the summer) and adopted the same preferential mixing between broad age groups and a small contribution of oropharyngeal transmission to overall poliovirus transmission as we assumed in other high-risk settings (Table I). The average per-dose take rate of a vaccine represents the fraction of vaccine recipients that transfers to the corresponding higher immunity state. For susceptible vaccine recipients, this corresponds to a measurable antibody response or in the case of IPV, a priming response that manifests in faster serological response in the event of a subsequent exposure, both of which we assume lead to permanent and complete immunity from paralytic poliomyelitis. For already immune individuals, a subsequent vaccine take implies moving to an immunity state with a higher degree of immunity to poliovirus transmission (i.e., lower susceptibility to reinfection and a shorter duration and lower infectiousness to others if reinfected, as described elsewhere ^{(25, 26)}). For IPV, we assume that the same take rate applies regardless of prior immunity and that IPV take among individuals with pre-existing immunity from a live poliovirus infection leads them to the highest state of immunity to transmission in our model, consistent with trials that demonstrated an intestinal boosting effect of IPV on OPV-vaccinated individuals.^{(22, 23)} In contrast, for OPV, we assume that pre-existing immunity effectively reduces the probability of take by multiplying the average per-dose take rates (Table I) with the relative susceptibility to reinfection of the immunity state compared to fully susceptible individuals.^{(25, 26)}

Table I:

Constant Inputs Specific for the Pakistan and Afghanistan Model

Model input	Best estimate	Notes and sources
Number of subpopulations	4	Two each in Pakistan and Afghanistan (see Methods)
Size of under-vaccinated relative subpopulations relative to total population		Judgment, informed by size of border provinces and Pashtun populations (42, 43)
- Pakistan	0.05
- Afghanistan	0.10
Number of age groups	11	0–2; 3–11 months; 1; 2; 3; 4; 5–9; 10–14; 15–24a;25–39a; ≥ 40 years
Number of mixing age groups	3	0–4;5–14; ≥ 15 years, consistent with most modeled situations(19, 26)
Proportion of contacts reserved for individuals within the same mixing age group	0.35	Measure of strength of preferential mixing between age groups; value similar to other high-risk settings(26, 27)
Average R0 for WPV		Ratios by serotype based on generic model inputs, values between previously assumed values for northwest Nigeria (7.5 for serotype 1 WPV) and northern India (13)(26, 27)
- Serotype 1	11
- Serotype 2	9.9
- Serotype 3	8.25
R0 amplitude	0.15	Based on judgment and calibration within ranges used for other populations (10, 20, 26) to match incidence pattern
R0 peak day (day of year)		Broadly consistent with typical precipitation patterns and non-polio enterovirus isolation rates in both countries and with range of values used in a global model(10); calibrated to match incidence patterns
- Pakistan	180 (Jun. 30)
- Afghanistan	240 (Aug. 29)
Proportion of transmissions via oropharyngeal route	0.3	Value used for high R0 developing country settings(10, 26, 34)
Average per-dose take rates (serotype 1;2;3)		Value based on review of seroconversion studies (55) and consistent with ranges from a global model (10) and between assumed values for northwest Nigeria and northern India (20, 26)
- tOPV	0.40;0.60;0.52
- mOPV	0.52;0.60;0.52
- bOPV	0.48;NA;0.48
- IPV	0.63;0.63;0.63
Time of IPV introduction in RI		(56, 57)
- Pakistan	Aug. 20, 2015
- Afghanistan	Sep. 30, 2015
Time of switch from tOPV to bOPV	Apr. 30, 2016	(58)
Demographics	Time series	Surviving birth rates and age-specific mortality rates over time computed(26) from UN-estimated medium variant annual number of surviving infants and population(37) in each age group and country using existing methods(26)

Abbreviations: bOPV, bivalent oral poliovirus vaccine; IPV, inactivated poliovirus vaccine; mOPV, monovalent oral poliovirus vaccine; NA, not applicable; R₀, basic reproduction number; RI, routine immunization; tOPV, trivalent oral poliovirus vaccine; UN, United Nations; WPV, wild poliovirus

Notes:

^aDenotes age groups whose immunity profiles determine the fraction of newborns born with maternal antibodies.⁽²⁶⁾

The main calibration effort focused on finding appropriate time-varying model inputs related to the impact of RI (Appendix A3) and SIAs (Appendix A4). Briefly, we estimated RI coverage with 3 or more doses of approximately 75% in the Pakistan general population and 60% in the Afghanistan general population in recent years, with relative RI coverage of approximately 30% in the under-vaccinated subpopulations of each country. For SIAs, we accounted for repeated missed probabilities, 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.”^{(27, p. 4)} We fitted the true coverage and repeated missed probabilities for the general population to minimize the difference between modeled and reported 0-dose proportions among non-polio AFP cases, where available (Appendix A4). For recent SIAs, this resulted in estimated true coverage of 70–95% in the two general populations, with relative SIA coverage of 20–25% assumed for the two under-vaccinated subpopulations compared to their corresponding general population. We report the modeled 0-dose proportions in all four subpopulations against the non-polio AFP data to compare the calibrated SIA assumptions against the limited available data about the cumulative impact of SIAs on dose histories (Appendix A4). To arrive at an approximate equilibrium when RI vaccination starts, we started the model in 1950, ran it for 20 years without any vaccination or seasonality but following historic demographic estimates,⁽³⁷⁾ started seasonality in 1970, and start vaccination in 1980.

To assess the potential impact of IPV use on WPV1 eradication and cVDPV2 risk management, we consider various scenarios (first column of Table II). The base case assumes introduction of IPV co-administered with the third non-birth tOPV RI dose in mid-2015 (Table I) and inclusion of IPV in some SIAs between November 2014 and September 2017, which prioritized high-risk areas (Appendix A4). We retrospectively consider hypothetical IPV scenarios that assume no IPV use in RI (No IPV RI), exclude the use of IPV SIAs (No IPV SIAs), or exclude both (No IPV RI or SIAs). Furthermore, we consider prospective scenarios that add one SIA targeting all children 0–4 years of age in both subpopulations from January 1–5, 2018 using either bOPV (Add bOPV-only SIA), mOPV2 (Add mOPV2-only SIA), IPV co-administered with bOPV (Add bOPV+IPV SIA; equivalent to IPV-only for serotype 2), or IPV co-administered with mOPV2 (Add mOPV2+IPV SIA; equivalent to IPV-only for serotype 1). For the added SIA, we consider either average SIA quality similar to other SIAs targeting the under-vaccinated subpopulations in 2017 (i.e., true coverage of 0.20 in Pakistan and 0.14 in Afghanistan and repeated missed probabilities of 0.83 in Pakistan and 0.87 in Afghanistan) or assume that substantially improved access allows above-average SIA quality (i.e., coverage of 50% of all targeted children, regardless of whether they received vaccine during prior SIAs). These scenarios assume no constraints on the vaccine supply from the stockpile or other resources required to conduct these SIAs. However, for all IPV scenarios, we report the impact on the approximate number of IPV doses required (appendices A2–3), polio incidence, and population immunity to transmission of all three serotypes.

Table II:

Impact of Various IPV Strategies on Estimated Number of IPV Doses Required and Polio Cases Prevented in Pakistan and Afghanistan, 2014–2018

Base case resultsa	Approximate # IPV doses required, in millions		Polio casesb
	Under-vaccinated subpopulations	General populations	WPV1	cVDPV2
- 2014–2017	0.48 (RI) + 5.3 (SIAs)	18 (RI) + 16 (SIAs)	480;400	36;32
- 2018	0.20 (RI) + 0 (SIAs)	7.6 (RI) + 0 (SIAs)	37;18	<1;2,900
Retrospective scenarios	Difference in IPV doses compared to the base case, in millions (2014–2017)		Polio cases prevented compared to the base case (2014–2017)c
	Under-vaccinated subpopulations	General populations	WPV1	cVDPV2
No IPV RI	−0.48	−18	−9.8	− 1.1
No IPV SIAs	−5.3	−16	−20	− 0.89
No IPV RI or SIAs	−5.7	−34	−33 (−41;−3.8)d	− 2.2 (−2.6;−0.56)d
Prospective scenarios	Difference in IPV doses compared to the base case, in millions (2018)		Polio cases prevented compared to the base case (2018)e
	Under-vaccinated subpopulations	General populations	WPV1	cVDPV2
Add bOPV-only SIA
- Average qualityf			6.8	0
- Above-average qualityg	0	0	13	0
Add bOPV+IPV SIAh
- Average qualityf			8.0	980
- Above-average qualityg	4.2	67	13	1,000
Add mOPV2-only SIA
- Average qualityf			0	2,400
- Above-average qualityg	0	0	0	2,600
Add mOPV2+IPV SIAi
- Average qualityf			4.9	2,600
- Above-average qualityg	4.2	67	11	2,800

Abbreviations: bOPV, bivalent oral poliovirus vaccine; cVDPV2, serotype 2 circulating vaccine-derived poliovirus; IPV, inactivated poliovirus vaccine; mOPV2, serotype 2 monovalent oral poliovirus vaccine; Ref., reference; RI, routine immunization; SIA, supplemental immunization activity; WPV1, serotype 1 wild poliovirus

Notes:

^aSee Table A2 in the Appendix for a list of assumed IPV SIAs conducted until January 2017

^bNumbers separated by semicolons represent results for More isolated and Less isolated base case, respectively, in the absence of any further outbreak response measures in the event that the cVDPV2 outbreak resurges

^cResults for the More isolated mixing scenario (corresponding to Figure 2)

^dRange between parentheses corresponds to the range among the different IPV assumptions considered in Figure 5

^eResults for the Less isolated mixing scenario (corresponding to Figure 3) and assuming no further outbreak response measures following the cVDPV2 outbreak resurgence

^eScenarios add an SIA starting on January 1, 2018 that targets all four subpopulations, assuming similar quality of other SIAs in 2017 (i.e., relative SIA coverage compared to the general population of 0.25 in Pakistan and 0.20 in Afghanistan)

^fScenarios add an SIA starting on January 1 2018 that targets all four subpopulations, assuming that substantially improved access results in 50% coverage, regardless of prior vaccination history, in the two under-vaccinated subpopulations (with no change in the two general populations).

^gEquivalent to IPV-only SIA scenarios for serotype 2

^hEquivalent to IPV-only SIA scenarios for serotype 1

3. RESULTS

Figure 1 compares the reported monthly polio incidence with the model base case for both mixing scenarios (see Appendix A5 for a more detailed discussion of the model results and data). The model estimates substantially higher incidence before AFP surveillance met the standards required to adopt the virological case definition (i.e., about twice higher in Pakistan during 1997–1999 and six times higher in Afghanistan during 1997–2002). However, the model approximates the reported pattern of peaks and troughs during those years in Pakistan, which experienced a nationwide outbreak in 1997.⁽⁴⁴⁾ For Afghanistan, the patterns do not match well before 2000, possibly reflecting poor-quality AFP surveillance. Afghanistan reported only 28 total AFP cases during 1997 (including 19 polio cases), compared to 335 in 2002 and over 1800 each year since 2011 (i.e., a 60-fold increase between 1997 and 2011).⁽³⁹⁾ During subsequent years, the curves in Figure 1 track increasingly closely. Pakistan detected a cVDPV2 in Quetta, Balochistan in late 2016⁽⁴⁵⁾ and conducted a local mOPV2 response in January 2017 and two successively larger mOPV2 SIAs in February and March 2017. In the model, a cVDPV2 similarly emerges after the tOPV-bOPV switch in the Pakistan under-vaccinated subpopulation, and we assume that the mOPV2 SIAs specifically targeted this subpopulation and achieved high quality (i.e., true coverage up to 80%). Whether this controls all serotype 2 live poliovirus transmission depends on the extent to which the under-vaccinated subpopulation in Pakistan mixes with other subpopulations, particularly with the under-vaccinated subpopulation in Afghanistan. If the subpopulations remain highly isolated (More isolated mixing scenario), then we find that cVDPV2 transmission stops and the mOPV2 used does not cause any extensive transmission elsewhere. However, if the subpopulations mix more intensely (Less isolated mixing scenario), then, although the cVDPV2 transmission still stops in mid-2017, the mOPV2-related strains spread to the rest of Pakistan and the Afghanistan under-vaccinated subpopulation. Given that the population immunity to serotype 2 transmission steadily declined since the tOPV-bOPV switch in those populations (see appendix A5), the mOPV2-related viruses in this scenario eventually evolve to a new cVDPV2 outbreak that appears first in the Afghanistan under-vaccinated subpopulations. In the absence of any further outbreak response in the model, the outbreak expands to all four subpopulation and causes almost 3,000 polio cases (Table II).

Figure 1:

Base case results from the differential equation-based model for Pakistan and Afghanistan compared to polio incidence reported through the acute flaccid paralysis surveillance system, 1997–2017. We allocated cases of unknown serotype (prior to adoption of the virological case definition) to serotype 1 wild poliovirus (WPV1) or serotype 3 wild poliovirus (WPV3) based on the ratio of available WPV1 to WPV3 cases in the same year.⁽³⁹⁾

Figure 2 shows how past IPV use affects the model-estimated WPV1 and cVDPV2 behavior, assuming the More isolated mixing scenario. Removing all IPV used in RI since mid-2015, all IPV used in SIAs since late-2014, or both initially yields a small effect on the incidence. Over time, the cumulative effect becomes more considerable, although the expected increase in polio cases during 2014–2017 without the use of approximately 40 million IPV doses remains below 35 cases for all scenarios (Table II). Looking beyond 2017, we found that because the base case with the More isolated mixing scenario barely controls the post-switch cVDPV2 outbreak, removal of IPV in RI and/or SIA makes enough of a difference to allow the cVDPV2 to continue to circulate.

Figure 2:

Comparison of expected polio incidence in both countries combined for different retrospective scenarios of inactivated poliovirus vaccine (IPV) use in routine immunization (RI) or supplemental immunization activities (SIAs), assuming the More isolated mixing scenario (see appendix A5 for the results with the Less isolated mixing scenario).

Figure 3 explores prospective scenarios involving different vaccine choices. It focuses on the Less isolated mixing scenario because with no continuing cVDPV2 outbreak in the More isolated mixing scenario, using serotype 2-containing OPV would represent an unnecessary risk. A failure to use IPV in RI from 2018 moderately increases WPV1 incidence (Figure 3a). Adding one SIA targeting both under-vaccinated subpopulations in January 2018 with average quality moderately reduces WPV1 incidence, but does not interrupt transmission by the end of 2018 regardless of vaccine choice. In contrast, with WPV1 apparently very close to the threshold, adding one above-average quality SIA that includes both bOPV and IPV interrupts transmission at the end of 2018. While an SIA of the same quality but with only bOPV fails to interrupt WPV1 transmission, additional model runs shows that if we increase the bOPV SIA quality from 50% to 60% coverage (regardless of vaccination status during prior SIAs), it interrupts WPV1 transmission during the same month as the bOPV+IPV SIA with 50% coverage. Thus, we find that the addition of IPV to a bOPV SIA exerts roughly the same effect as a 10% increase in bOPV coverage at a cost of 4.2 million IPV doses in the two under-vaccinated subpopulations (and another 67 million IPV doses if we include the IPV used in both general populations). Figure 3a further shows that adding an SIA that includes IPV but no serotype 1-containing OPV (i.e., Add mOPV2+IPV SIA) results in much less impact on WPV1 incidence than adding an SIA that includes serotype 1-containing OPV. Table II shows the associated numbers of IPV doses and prevented cases for all prospective scenarios from Figure 3. Figure 3a includes only results with the Less isolated mixing scenario (see appendix A5 for the full results, which show similar WPV1 behavior for the More isolated scenario, except that even the added above-average quality bOPV+IPV SIA does not interrupt WPV1 transmission). We emphasize that the projections in Figure 3a assume no change in SIA quality between 2017 and 2018 and adopt the tentative SIA plans. If the quality improves and/or the number of SIAs increases or more specifically targets the under-vaccinated subpopulations than we can assess bases on current plans, then WPV1 elimination may occur by the end of 2018 for all the scenarios in Figure 3a.

Figure 3:

Comparison of expected polio incidence in Pakistan and Afghanistan combined, assuming the Less isolated mixing scenario (see appendix A5 for the results with the Less isolated mixing scenario) for different prospective scenarios that remove inactivated poliovirus vaccine (IPV) use from routine immunization (RI) or that add one supplemental immunization activity (SIA) that targets both under-vaccinated subpopulations in January, 2018 with bivalent oral poliovirus vaccine (bOPV), serotype 2 monovalent oral poliovirus vaccine (mOPV2), and/or IPV, with or without substantially improved access.

For serotype 2, if transmission continues, as it does for the Less isolated mixing scenario, then different prospective strategies imply very large differences in the number of cases due to the absence of regular use of serotype 2-containing OPV and no assumed further outbreak response when the cVDPV2 outbreak continues (Figures 3b–c). Adding an SIA in early 2018 while cVDPV2 transmission remains limited slows the outbreak by immunizing children before exposure. Using IPV as the only serotype 2-containing vaccine (i.e., Add bOPV+IPV SIAs) slows down the outbreak considerably but does not prevent the occurrence of a very large outbreak in the second half of 2018 (Figure 3b). In contrast, an added mOPV2-only SIA yields much more impact. Co-administering over 70 million total IPV doses with mOPV2 (including approximately 4 million in the two under-vaccinated subpopulations) provides some additional benefit, although none of the scenarios we considered suffice to stop a rapidly expanding cVDPV2 outbreak with a single SIA. An above-average quality mOPV2-only SIA prevents approximately the same number of cases as an average-quality mOPV2+IPV SIA (Table II). While all of the scenarios in Figure 3a fail to stop serotype 2 transmission, we emphasize that for the More isolated mixing scenario serotype 2 transmission does stop after the mOPV2 outbreak response in early 2017. Thus, the need for any further mOPV2-containing SIA depends strongly on the uncertainty about the degree of isolation of the Pakistan under-vaccinated population that experienced the recent cVDPV2 transmission and mOPV2 response. Moreover, Figure 3 focuses on comparing different vaccine choices in the context of a single added SIA. In reality, if further cVDPV2 detections occur, the full response would involve multiple SIAs⁽⁴⁶⁾ to both stop the cVDPV2 transmission and prevent the mOPV2-related viruses from creating new cVDPV2 outbreaks.⁽³³⁾

Figure 4 shows the population immunity to transmission (i.e., EIPM) in the Pakistan under-vaccinated subpopulation (see Appendix A5 for Afghanistan figures) for selected retrospective and prospective scenarios and assuming the Less isolated mixing scenario. Figure 4a suggests that the WPV1 EIPM hovers around the threshold, with minimal difference from past IPV use, and a notable but temporary benefit from one average-quality SIA in January 2018 targeting all four subpopulations. The added SIA results in clearly higher EIPM if it contains bOPV. The serotype 2 EIPM hovered around the threshold before dropping well below it after the switch in April 2016 (Figure 4b). After the switch, IPV use increasingly affects the serotype 2 EIPM, but from 2017 on the cVDPV2 outbreak and response bring the curves with and without retrospective IPV use back together. An added high-quality SIA with any serotype 2-containing vaccine significantly increases EIPM, with a much greater increase if the SIA includes mOPV2 than if it relies on IPV-alone for serotype 2. Despite the effect on polio incidence from adding IPV to an mOPV2 SIA (Figure 3b and Table II), the incremental effect of IPV on population immunity to serotype 2 transmission remains small. For serotype 3, EIPM also remains close to the threshold, but given WPV3 die-out in 2012 in the model and very low EIPM needed to generate a serotype 3 cVDPV,⁽⁴⁷⁾ short periods below the threshold do not cause any transmission.

Figure 4:

Impact of scenarios of inactivated poliovirus vaccine (IPV) use in routine immunization (RI) or supplemental immunization activities (SIAs) with or without bivalent oral poliovirus vaccine (bOPV) or serotype 2 monovalent oral poliovirus vaccine (mOPV2) on population immunity to transmission in the Pakistan under-vaccinated subpopulation. All results shown assume the Less isolated mixing scenarios and the prospective scenarios that add an SIA assume average quality (see Appendix A5 for the results for the Afghanistan under-vaccinated subpopulation).

Figure 5 shows the base case results along with alternative assumptions about the effect of IPV on individual and population immunity (see appendix A6 for details about the varied assumptions). Varying the average per-dose take rate of IPV affects the proportion of IPV recipients who acquire both individual immunity to disease and partial immunity from participation in poliovirus transmission, while the remaining variations affect only the extent to which IPV-immunized individuals can participate in poliovirus transmission. For serotype 1, Figure 5a shows a moderate effect of varying the take rate over a very wide range. However, even with a take rate of only 30%, we find a marked reduction in the incidence curves compared to no IPV use at all. Controlling for the IPV take rate and only varying how IPV affects transmission, we see that the IPV boosting effect accounts for approximately half of the reduction in incidence from the IPV use to date (i.e., the curve No effect IPV-only on fecal-oral and oropharyngeal transmission lies approximately midway between those for the base case and No IPV RI or SIAs). The other half of the reduction comes primarily from the assumed effect of IPV-only and oropharyngeal transmission (i.e., the curve for No effect IPV-only on fecal-oral transmission remains very close to that for the base case). Figure 5b shows similar relative importance of the assumptions for serotype 2, with any scenario that reduces the IPV impact on transmission allowing the cVDPV2 outbreak to resurge in 2018.

Figure 5:

Sensitivity analysis of assumptions related to the effect of IPV on the base case with the More isolated mixing scenario.

4. DISCUSSION

Accumulating experience from the last four polio-endemic countries strongly indicates that the failure to reach under-vaccinated communities represents the main impediment to WPV elimination. Consistent with models by different groups of recently endemic areas and outbreaks,^{(19, 26–29, 48, 49)} explicit characterization of under-vaccinated subpopulations proved essential to reproduce the epidemiological experience in Pakistan and Afghanistan. The introduction of mOPV1 in northern India to address the presumed tOPV failure⁽⁵⁰⁾ did not interrupt WPV1 transmission but allowed a WPV3 resurgence, followed by mOPV3 use and cVDPV2 emergences.⁽²⁸⁾ WPV elimination in northern India ultimately succeeded when vaccination efforts reached previously unidentified missed children. Similarly, interruption of WPVs and cVDPV2s in northwest Nigeria required intense efforts to identify missed communities.⁽³⁰⁾ However, insufficient tOPV intensification before the switch allowed the emergence of a new cVDPV2 in Pakistan, and prolonged access challenges in Borno in northeast Nigeria allowed cVDPV2 transmission to persist through the switch.⁽⁴⁵⁾ As of late 2018, all known ongoing WPV1 transmission in the world occurs in parts of countries with known access issues, unstable security, and/or programmatic failures to vaccinate, most of which also supported post-switch cVDPV2 transmission.

Genetic data suggest that endemic transmission of WPV1 still occurred independently in several foci of un- and under-immunized communities throughout Pakistan and in parts of Afghanistan,⁽⁵¹⁾ which our model captures at a conceptual level as under-vaccinated subpopulations that preferentially interact with each other. Although under-vaccinated communities tend to reside in the areas close to the Pakistan-Afghanistan border, some of them represent highly mobile populations and under-vaccinated communities exists in other parts of these countries as well, with surveillance regularly detecting poliovirus transmission in all areas.⁽⁵¹⁾ Therefore, eradication efforts should focus on identifying and reaching under-vaccinated communities regardless of geography. At the same time, maintaining high enough population immunity to transmission in the general population remains critical to avoid more widespread transmission. Finishing WPV1 eradication and controlling any further cVDPV2 spread will require vaccination of under-vaccinated communities with OPV vaccines that include both serotypes 1 and 2. Although tOPV represents the best vaccine to simultaneously respond to WPV1 and cVDPV2,^{(52, 53)} the emergency stockpile does not include tOPV. Consequently, efforts must focus on alternating mOPV2 and bOPV SIAs or addressing the logistics of similar vials to manage mOPV2-bOPV co-administration.

In agreement with reported mixed evidence about the benefit of past IPV use in endemic areas,⁽²⁴⁾ this study finds a moderate effect of IPV use to date in Pakistan and Afghanistan on population immunity to WPV1 or cVDPV2 transmission compared to reliance on OPV-only through 2017. The benefit comes primarily from the assumed effect of IPV boosting on intestinal immunity observed in clinical trials^{(22, 23)} but further traces to our assumption that oropharyngeal transmission plays some role in overall poliovirus transmission and that IPV-only can effectively limit oropharyngeal transmission.^{(15, 34)} Due to the latter mechanism, in the absence of regular use of serotype 2-containing OPV, IPV-alone provides a noticeable effect on population immunity to cVDPV2 transmission because the comparator does not involve any serotype 2-containing vaccine. However, the impact of IPV-alone does not suffice to control the outbreak and remains much smaller than the impact that the same number of mOPV2 doses would achieve. The findings about the moderate incremental benefit of IPV added to mOPV2 show that in the context of inadequate OPV use, IPV can provide benefits. This contrasts with the observation of a very limited incremental effect of adding IPV to an aggressive mOPV2 response that by itself suffices to control a cVDPV2 outbreak.⁽²⁵⁾ In an ideal world without IPV supply or financial constraints, the incremental benefit of IPV may justify its use in addition to OPV, but unfortunately IPV supply remains highly limited. Numerous countries did not receive IPV for RI in 2017, and catching up these populations with IPV should remain a high priority. Newborn cohorts that do not receive any serotype 2-containing vaccine represent a significant concern because of an increasing risk of importing cVDPV2s or outbreaks due to iVDPV2s within their borders. Prior studies suggested a growing cumulative effect of IPV in RI over time compared to no use of any homotypic OPV after OPV cessation,⁽⁴⁷⁾ more effect of IPV in settings with less intense fecal-oral transmission,^{(19, 26, 32)} a higher expected iVDPV prevalence in those settings,⁽⁵⁴⁾ and a 3-fold increase in the risk of uncontrolled outbreaks leading to an OPV restart if all countries that used OPV-only in 2013 fail to introduce at least one IPV RI dose.⁽¹⁰⁾ Our results suggest that Pakistan, and by extension Nigeria and possibly Afghanistan all need to use mOPV2 to control cVDPV2 outbreaks. Instead of using IPV in SIAs for a relatively small incremental benefit, the GPEI should focus on using mOPV2 aggressively to rapidly stop any future detections of cVDPV2 transmission before it goes out of control. These countries should reserve the available IPV for use in RI to reach as many children born since the switch as possible in these and other countries because these children otherwise would receive no serotype 2-containing vaccine, as required and planned as part of the switch strategy.⁽⁷⁾

The model suggests that the degree of mixing of the Pakistan under-vaccinated subpopulations that used mOPV2 in early 2017 with under-vaccinated subpopulations elsewhere in Pakistan and Afghanistan will determine whether these populations may experience cVDPV2 transmission as a result of the outbreak or mOPV2 used to respond to it. Preventing any further serotype 2 transmission would in turn require multiple SIAs that include mOPV2 and target a wider geographical area. Previous work suggested that if these mOPV2 SIAs suffice to stop the cVDPV2 outbreak, then they also suffice to prevent transmission of mOPV2-related viruses in the outbreak-affected population.⁽³³⁾ However, this does not preclude that the mOPV2-related viruses create new cVDPV2 outbreaks in surrounding areas that did not use mOPV2, which again depends on how well the areas that use mOPV2 mix with surrounding areas of low population immunity to serotype 2 transmission. Using IPV only would likely not control a continued cVDPV2 outbreak in Pakistan and Afghanistan, but the currently limited mOPV2 stockpile and the risk that mOPV2 could create new cVDPV2 outbreaks elsewhere compromise the ability to conduct a geographically expansive mOPV2 response. Although we hope that the absence of any cVDPV2 detections since July 2017 means that transmission stopped, it remains prudent to prepare to swiftly and strategically respond with mOPV2 in the event of further detections in an attempt to limit the amount of mOPV2 needed to stop transmission. Specifically, as long as the general population immunity to serotype 2 transmission remains high enough to prevent mOPV2-related viruses from establishing transmission and evolving to cVDPV2s, the outbreak response (in the event of continued cVDPV2 detections) should focus on substantially raising population to serotype 2 transmission both in the communities experiencing cVDPV2 transmission and in connected communities suspected to be under-vaccinated with tOPV prior to the switch. While most countries currently still sustain high enough population immunity to serotype 2 transmission to prevent continued transmission of any introduced mOPV2-related viruses,^{(11, 31, 52)} unfortunately this study suggests that the general populations of Pakistan and Afghanistan may not sustain such high population immunity to serotype 2 transmission for much longer. Exactly how long remains uncertain and depends on the quality of tOPV SIAs and RI prior to the tOPV-bOPV switch, the impact of subsequent IPV use on population immunity to serotype 2 transmission, and the kinetics of waning of intestinal immunity. In contrast to this uncertainty, an insufficiently aggressive response following any further cVDPV2 detections would almost certainly allow the cVDPV2 to spread and thus create more mOPV2 demand at a future time of lower global population immunity to serotype 2 transmission and higher risks of creating new cVDPV2 outbreaks.^{(31, 52)} Thus, despite the risks of mOPV2 use, a failure to respond aggressively with mOPV2 in the event of any further cVDPV2 detections does not represent a realistic option. Determining the best mOPV2 outbreak response requires further analysis of the trade-offs, the supply considerations, and careful consideration of any information about under-vaccinated communities and mixing patterns in Pakistan and Afghanistan.

Like all models, the DEB model comes with limitations and uncertainties. Although we carefully calibrated the generic model, we cannot preclude that other structures and inputs would reproduce the evidence equally well or better.^{(26, 27)} The DEB model assumes no heterologous immunity associated with bOPV, which if it exists could mitigate the cVDPV2 outbreak. However, the DEB model accounts for the boosting effect of IPV, which sends IPV recipients to the highest state of immunity at a higher rate than any OPV boost due to greater effective “take”.⁽²⁵⁾ Tailoring the model to Pakistan and Afghanistan model further involved significant uncertainty related to the immunization activities conducted over the years, population size and structure, mixing, and transmissibility. Reproducing the reported incidence patterns required very low SIA quality in the under-vaccinated subpopulations and some significant differences between individual SIAs. The four-subpopulation model represents a simplification of the complex and changing mixing patterns and heterogeneity,⁽⁵¹⁾ but the close match of the model to the incidence data from the time AFP surveillance reached high quality provides some confidence, as does the ability to verify the cumulative effect of SIAs by comparing modeled to reported zero-dose proportions. Given the large number of SIAs and possible combinations of mixing assumptions, we did not perform a systematic sensitivity analysis. However, we believe that by presenting the results for alternative mixing structures we spanned an informative range of possible outcomes for different vaccination strategies.

LIST OF ABBREVIATIONS

AFP

acute flaccid paralysis

bOPV

bivalent oral poliovirus vaccine

cVDPV(2)

(serotype 2) circulating VDPV

DEB

differential equation-based

EIP*

threshold effective immune proportion

EIPM

mixing-adjusted effective immune proportion

GPEI

Global Polio Eradication Initiative

IPV

inactivated poliovirus vaccine

iVDPV

immunodeficiency-associated VDPV

LPV

live poliovirus

mOPV(1,2,3)

monovalent OPV (serotype 1, 2, or 3, respectively)

NPAFP

non-polio AFP

R₀

Basic reproduction number

RI

routine immunization

SIA

supplemental immunization activity

tOPV

trivalent oral poliovirus vaccine

VDPV

vaccine-derived poliovirus

WPV(1,2,3)

wild poliovirus (serotype 1, 2, or 3, respectively)