Updated characterization of poliovirus transmission in Pakistan and Afghanistan and the impacts of different outbreak response vaccine options

Abstract Background Pakistan and Afghanistan remain the only reservoirs of wild poliovirus transmission. Prior modeling suggested that before the COVID-19 pandemic, plans to stop the transmission of serotype 1 wild poliovirus (WPV1) and persistent serotype 2 circulating vaccine-derived poliovirus (cVDPV2) did not appear on track to succeed. Methods We updated an existing poliovirus transmission and Sabin-strain oral poliovirus vaccine (OPV) evolution model for Pakistan and Afghanistan to characterize the impacts of immunization disruptions and restrictions on human interactions (i.e., population mixing) due to the COVID-19 pandemic. We also consider different options for responding to outbreaks and for preventive supplementary immunization activities (SIAs). Results The modeling suggests that with some resumption of activities in the fall of 2020 to respond to cVDPV2 outbreaks and full resumption on January 1, 2021 of all polio immunization activities to pre-COVID-19 levels, Pakistan and Afghanistan would remain off-track for stopping all transmission through 2023 without improvements in quality. Conclusions Using trivalent OPV (tOPV) for SIAs instead of serotype 2 monovalent OPV (mOPV2) offers substantial benefits for ending the transmission of both WPV1 and cVDPV2, because tOPV increases population immunity for both serotypes 1 and 2 while requiring fewer SIA rounds, when effectively delivered in transmission areas.

The GPEI prioritized the introduction of inactivated poliovirus vaccine (IPV) into Pakistan and Afghanistan in 2015 with the expectation that adding IPV would help to accelerate WPV1 eradication [5]. Delays in eradication of WPV1 necessitated the release of a 2019-2023 GPEI Strategic Plan [7], and challenges associated with outbreaks of serotype 2 circulating vaccine-derived polioviruses (cVDPV2) led to the release of an addendum to the plan in 2020 [8]. In early 2020, the GPEI anticipated widespread availability of a serotype 2 novel OPV (nOPV2) and that nOPV2 would completely replace mOPV2 to respond to cVDPV2 outbreaks by February 2021 [8]. As of March 2021, field trials with nOPV2 have just begun, and its properties remain uncertain. Because of the co-circulation of WPV1 and cVDPV2, trivalent OPV (tOPV) became the preferred vaccine for SIAs in Pakistan and Afghanistan beginning in late 2020.
Modeling the complex history of the use of different formulations of oral poliovirus vaccine (OPV) and different OPV and IPV vaccination strategies demonstrated the importance of substantially increasing OPV coverage of supplementary immunization activities (SIAs) in under-vaccinated subpopulations above that obtained previously in order to stop WPV1 transmission [15]. Other studies showed: (1) the importance of using proactive strategies to increase population immunity [16]; (2) the nature of the different types of surveillance information with respect to characterizing confidence about the absence of transmission as a function of time with no cases or environmental detections reported [17]; and (3) the tradeoffs of some characteristics of the poliovirus surveillance system in Pakistan and Afghanistan [18]. One statistical analysis reported relatively little role of A c c e p t e d M a n u s c r i p t 4 acute flaccid paralysis (AFP) surveillance data on resource allocation decisions for Pakistan [19].
Several other statistical analyses characterized the sensitivity and role of environmental surveillance in Pakistan and Afghanistan [20][21][22].
Prior to the coronavirus disease 2019 (COVID-19) pandemic, building on prior dynamic transmission modeling [15,16], we updated the transmission model to reflect the actual poliovirus vaccine use and epidemiology through early 2020 [23,24]. These analyses demonstrated that efforts to stop WPV1 poliovirus transmission in Pakistan and Afghanistan remained off track. Other studies highlight a number of important root causes related to the chronic failure to vaccinate in Pakistan and Afghanistan, access issues in areas where SIAs are 'banned' in Afghanistan, and significant issues with vaccine acceptance in Pakistan [2,3]. In 2019, Pakistan reported 22 cVDPV2 cases and this transmission continued in 2020 with a reported 135 cases in Pakistan and 305 cases in Afghanistan [25]. A review of the cVDPV2 epidemiology [26] highlighted the unknown source of the 2019 cVDPV2 emergence in Pakistan. Updated global modeling of cVDPV2 transmission and risks [27] emphasized that the widespread cVDPV2 transmission in 2019 and early 2020 increased the chances of needing to globally restart use of serotype 2 OPV (OPV2) in routine immunization (RI) in OPVusing countries [28]. Extensive efforts to monitor the quality of SIAs continue to show gaps, and poliovirus surveillance data, which now includes the detection of polioviruses in environmental samples, provide further evidence that Pakistan and Afghanistan are not on track to interrupt transmission [2,3].
As of the end of 2020, we recognized the need for another global model update due to the reduced social interactions, population mixing, and polio vaccine coverage caused by national responses to the COVID-19 pandemic [29]. This motivated us to also update our modeling of Pakistan and Afghanistan to consider the impacts of disruptions caused by the COVID-19 pandemic.

Methods
We updated our deterministic, differential equation-based (DEB) poliovirus transmission and OPV evolution Pakistan and Afghanistan model [15][16][17][18]24] to include the epidemiological experience through 2020 and to account for disruptions that occurred due to COVID-19 [29]. Briefly, the model divides the population into eight immunity states for fully susceptible, maternally immune, and six partially immune states following live poliovirus infections (WPV and/or OPV), A c c e p t e d M a n u s c r i p t 5 and/or successful IPV vaccination [23,24]. The model describes waning of immunity as a five-stage process, infection as a process with two latent and four infectious stages for both fecal-oral and oraloral transmission, and OPV evolution as a 20-stage process (i.e., stage 0 for fully attenuated Sabin strains to stage 19 for fully reverted cVDPV strains) [23,24].
The model divides the populations of each country into a general population and an undervaccinated subpopulation. The under-vaccinated population represents a conceptual construct characterized primarily by historically low vaccination levels, rather than geography [15][16][17][18]. The model aggregates all under-vaccinated communities from different parts of Pakistan and Afghanistan and including mobile populations [15][16][17][18]. As previously estimated, we assume the under-vaccinated subpopulation represents 5% of the total population of Pakistan and 10% of the total population of Afghanistan [15][16][17][18]. The top of Table 1 summarizes general inputs for Pakistan and Afghanistan related to the model population structure, poliovirus transmission, and vaccination that remained constant in all model runs [15][16][17][18]. Consistent with historical epidemiological evidence, the model assumes a constant seasonal variation pattern for each country, although recent epidemiological data do not show the same temporal pattern. The bottom of Table 1 summarizes model inputs used to the characterize the COVID-19 disruptions (i.e., beginning and end time of restrictions, change in average R 0 and RI coverage) consistent with prior global modeling [29]. Based on the model behavior in relation to cVDPV2 transmission in Pakistan, and its spread to Afghanistan, we assumed that both under-vaccinated subpopulations remain isolated relative to the general population such that they continue to preferentially mix with themselves and each other (Table 2a). We intensify the isolation from 2017 due to temporary border closures [15][16][17][18], which complicated (but did not stop) the process of crossing the border (Table 2b). Starting on March 20, 2020, we assume a second intensification of isolation due to the restrictions related to COVID-19 pandemic (Table 2c) followed by a return to the pre-COVID-19 levels after relaxation of the restrictions that we assumed would occur on January 1, 2021 [29].
We updated the RI coverage information based on the recent estimates of the national coverage with 3 doses of diphtheria-tetanus-pertussis vaccine (DTP3) by country [30], and recent

Demographic and Health Study (DHS) point estimates of poliovirus vaccine birth dose coverage
A c c e p t e d M a n u s c r i p t 6 (POL0) and coverage for 1, 2, and 3 poliovirus vaccine doses (POL1, POL2, and POL3) [31], Figure 1 shows the updated assumed RI coverage by dose over time. We also updated the SIA history from 2017 through 2020, including gaps in the SIA schedule during the COVID-19 pandemic disruption.  [24]. In response to the outbreak, Pakistan started using mOPV2 rounds from November 2019. We assume that some of the children in mobile populations that received mOPV2 in Pakistan cross the border into Afghanistan. We capture this in the model by introducing mOPV2 at the time of outbreak response SIAs in areas surrounding the border (with an assumed 1-day delay relative to the start day of the related mOPV2 SIA) to a small number of children ages 3 months to 4 years residing in Afghanistan. We assume that this spillover of mOPV2 continues until COVID-19 related restrictions led to more isolation between the two countries. For these introductions, we assumed a frequency of 2 fully susceptible children per age group per day received the mOPV2 prior to entering Afghanistan for the duration of the related mOPV2 SIA.
Given the substantial and widespread transmission of both WPV1 and cVDPV2 in Pakistan and Afghanistan in 2020, the GPEI and countries plan to use trivalent OPV (tOPV, containing all three OPV serotypes) for some SIAs instead of separate SIAs using serotype 2 monovalent OPV (mOPV2) and bivalent OPV (bOPV, containing serotypes 1 and 3 OPV). Table 3 shows the base case (BC) vaccination schedule assumed for both countries from October 2020 to December 2023 based on information available in October 2020, which includes the use of two tOPV rounds (October 2020 and January 2021) in both countries according to plans at that time, followed by the assumed use of bOPV for SIAs (and assuming the same estimates for true coverage and repeatedly missed and substitutes mOPV2 for the same bOPV rounds as the "tOPV use" scenario beginning in the second half of 2021, and (iv) "mOPV2 use," which substitutes all tOPV rounds with mOPV2 until mid-2021 and substitutes the same bOPV rounds as the "tOPV use" scenario with mOPV2 beginning in the second half of 2021.  serotype 1 and (b) serotype 2. The WPV1 paralytic incidence of "tOPV use 2021 only" and "tOPV use" for Pakistan closely approximate the base case, whereas the incidence of "tOPV and mOPV2 use" is considerably higher, and much higher still for "mOPV2 use." Figure 4 shows a similar, but less to the base case, with the worst outcome observed for the "mOPV2 use" scenario (i.e., tOPV never reintroduced into those countries).

Discussion
Significant setbacks in Pakistan's polio eradication program began during 2018, with a sharp increase in WPV1 cases and the emergence of cVDPV2 cases in 2019 [2]. In Afghanistan, widespread bans on house-to-house vaccination in conflict areas since April 2018 resulted in increasing numbers of WPV1 cases, and the spread of the cVDPV2 Pakistan outbreak to Afghanistan in 2020 [3]. The outbreak cVDPV2 viruses spread beyond the outbreak areas and the use of mOPV2 in both Pakistan and Afghanistan led to the detection of some new cVDPV2 emergences in late 2020. Ongoing transmission of WPV1 and cVDPV2 in Pakistan and Afghanistan suggest the need for strategies that will improve the quality of campaigns and vaccine coverage. In this analysis, simply switching SIAs back to tOPV in Pakistan and Afghanistan instead of alternately using mOPV2 and bOPV substantially improves population immunity to transmission for serotype 1, but not enough to stop WPV1 transmission. Our assumptions may prove optimistic with respect to the coverage that SIAs will achieve since we assumed resumption of SIAs at pre-COVID-19 coverage levels starting in January 1, 2021. If programmatic activities do not recover this quickly or as much, then the expected cases will increase. We suggest future modeling efforts should continue to monitor the actual performance of polio immunization activities in Pakistan and Afghanistan. We did not consider the use of IPV for outbreak response since prior studies suggest its inferior impact (compared with OPV) with respect to both effectiveness and cost-effectiveness, and the SAGE recently recommended no IPV use for outbreak response.
A c c e p t e d M a n u s c r i p t 9 The insights from this analysis are limited by our assumptions, the model structure, available information, and uncertainty about our reconstruction of the immunization histories and assumptions about future actions [23]. Notably, this modeling assumes unlimited supplies of vaccine, but real supplies of all OPV2 containing vaccines remain limited. For all of the modeled scenarios, we assume the same total number of OPV doses used, and only vary the vaccine choice related to the formulation of the OPV. Using this approach yields comparisons of scenarios that imply the same vaccine administration costs and similar vaccine purchase costs, such that we can focus only on the implications of vaccine choices for outbreak response on population immunity to transmission. We sought to explore the consequences of mobile under-vaccinated populations by allowing a small amount of secondary spread of OPV used for outbreak response in Pakistan to enter Afghanistan, but whether and how much of this occurs remains uncertain due to limited information. Our assumption of homogeneous mixing within subpopulations can imply rapid transmission across relatively large groups of individuals, although we include heterogeneous mixing between the subpopulations and using mixing matrices that limit the transmission to some degree. With the resumption of tOPV use in Pakistan and Afghanistan and substantial transmission of cVDPV2 in Africa, resuming the use of tOPV in areas that need OPV2 in response to cVDPV2 transmission would likely offer the most cost-effective strategy to keep population immunity to all 3 serotypes higher. Given the challenges associated with delivering vaccine in the wake of COVID-19, this may help best to limit the transmission of cVDPV2 and help to prevent some cVDPV1 and cVDPV3 emergences in areas that otherwise would not receive bOPV SIAs. This analysis underscores the consequences of low SIA quality and reinforces the key findings of prior modeling, that Pakistan and Afghanistan will need to increase SIA quality to stop poliovirus transmission. Using tOPV will offer a tool to provide protection for all three serotypes with fewer total SIA rounds, but it will not compensate for continued poor performance or low immunization coverage.   M a n u s c r i p t 16