5.  Themes
In the process of extracting data from the different studies, we captured some common themes in Table 4 and we identified instances in which the modeling groups provided similar or conflicting insights or recommendations.
Table 4. Summary of themes explored by multiple modeling groups.
Theme KRI IC IDM Other
Outbreak response speed [14, 35, 43, 55] [98]   [162, 166]
Expanded age group SIAs [35] [98] [129]  
Population immunity* [10, 16, 18, 19, 26, 33–40, 43, 46, 47, 49, 52–58, 60, 68–70, 73–77] [83, 84, 87, 89, 92, 93, 100–102, 109, 111] [131, 133, 140–142] [155, 162]
OPV cessation dynamics [26, 38, 39, 49, 68] [100, 111, 113] [134, 143] [153, 165, 166]
Silent transmission on an IPV background and/or delayed detection of transmission due to IPV use [26–28, 39, 41] [97] [133] [162, 164, 166, 170]
Role of IPV after OPV cessation [18–20, 25, 28, 33, 39, 51–60, 64, 65, 68, 69, 73–77, 80] [90, 91, 94, 97, 103, 114, 115, 118–120, 122] [133] [153]
Undetected circulation [27, 46, 70, 71, 75, 76]   [132, 136] [161, 165, 167–169]
Role of IPV in outbreak response SIAs [51, 55, 64, 68]   [133] [150, 164]
Environmental surveillance [43, 46, 55, 67, 71, 73–76, 78] [117] [136, 145, 147] [162, 166]
Vaccine stockpile [21, 23, 53, 65, 77]     [174]
iVDPVs [4, 13, 50, 61, 81] [125]    
Abbreviations: IC, Imperial College; IDM, Institute for Disease Modeling; IPV, inactivated poliovirus vaccine; iVDPVs, immunodeficiency-associated vaccine-derived poliovirus; KRI, Kid Risk, Inc.; OPV, oral poliovirus vaccine; SIAs, supplementary immunization activities.
* As indicated in text, defined differently by the 3 modeling groups: KRI focuses on modeling infection and defines ‘population immunity to transmission’ based on all individuals of all ages integrated over all immunity states in a DEB model as a function of serotype, population-specific inputs, and time, which is a model-based concept that does not vary by paper (see details in [34, 202]). KRI publications earlier than 2013 discussed ‘population immunity’ as the same concept (i.e. over the entire population), but characterized it as an input for some analyses based on data (see e.g. [10]); IC focuses only paralysis (i.e. not infection) and defines ‘population immunity’ including only vaccine-induced immunity (i.e. excluding immunity from maternal antibodies and immunity induced by infection with any live poliovirus via community spread), and varies by paper depending on the data used (e.g. nonpolio AFP data for: serotype 1 only for children <5 years old [83, 84, 87], serotypes 1 and 3 for children <2 years old [89], serotypes 1, 2, and 3 for children <36 months [92], serotype 1 for children <5 years old [101], serotype 2 for children <2 years [100], serotype 2 for children <36 months [111, 113], and serotype 1 for children <36 months [102]; multiple metrics used for regression analyses [93, 109], see individual papers for specific definitions); IDM definition of ‘population immunity’ includes only vaccine-induced immunity (i.e. excluding immunity from maternal antibodies and immunity induced by infection with any live poliovirus via community spread), focuses on paralysis (i.e. not infection), and varies by paper depending on data used (e.g. OPV-induced immunity for nonpolio AFP cases in children <5 years old in a district within a 6-month period [131, 133, 141, 142], children <15 years old [140], dose estimates based on SIAs, see individual papers for specific definitions).

5.1.  Responding quickly to outbreaks
We found consistency in the recommendations made independently from different transmission modeling studies [14, 35, 43, 55, 98, 133, 162, 166] with respect to the importance of rapidly detecting and responding to outbreaks. Multiple studies also recommended that in the event of detection of a transmitting virulent virus (i.e. WPV or cVDPV) after OPV cessation, using OPV for outbreak response offered the best option [14, 35, 52, 55, 68, 133, 134], although its use comes with risks. Specifically, all three modeling groups expected the risks associated with using OPV for outbreak response after OPV cessation would increase as a function of the time since cessation (i.e. as more birth cohorts without exposure to LPVs accumulate). The modeling motivated the creation of mOPV vaccine stockpiles for outbreak response after OPV cessation to ensure sufficient supplies. For the review inclusion time (2000–2019), only KRI applied transmission modeling to questions related to creating, funding, and managing stockpiles of poliovirus vaccines [21, 23, 53, 65, 77], although one economic analysis considered the US stockpile [174].

5.2.  SIAs with expanded age groups
All three modeling groups gave similar recommendations to the GPEI partners based on the application of transmission models in response to questions about the potential benefits of using expanded age group as the target for SIAs [35, 98, 129]. Notably, although the populations modeled by the groups differed, the primary conclusions of the application of transmission models to the question of expanding the target age ranges for OPV SIAs emphasized the importance of reaching susceptible children (typically the younger ones and those in undervaccinated subpopulations) as quickly as possible. Some of the modeling groups also highlighted the substantially lower cases (and costs) associated with performing pSIAs to prevent the need for oSIAs [34, 35, 159].

5.3.  Population immunity
All of the modeling groups recognize the need for high population immunity to achieve and maintain polio eradication. However, one of the most notable sources of conflicting recommendations from the three modeling groups comes from the use of different definitions for population immunity. As shown in Table 4, all three modeling groups used the term ‘population immunity’ in numerous 2000–2019 publications. The KRI papers that mention population immunity use a dynamic transmission model that focuses on the characterization of the transmission of infections based on the understanding that eradication requires achieving and maintaining the end of all LPV transmission (i.e. permanent prevention of infection). As such, KRI defines ‘population immunity to transmission’ for each serotype as dynamic measure of the overall immunity by serotype of all individuals in a population, including partial immunity for those with prior vaccination or infection who can become (re)infected and participate in transmission due to the nature or waning of their immunity. In contrast, statistical and epidemiological models developed by IC defined population immunity differently, even from paper to paper depending on the research question and data used, see note at the bottom of Table 4, which indicates the serotype-specific definitions applied in some papers. The IC concept of population immunity focuses on vaccine coverage and prevention of paralysis (instead of infection). While this narrower concept of population immunity provides an indication of susceptibility to transmission in an important part of the population (i.e. young children) and can characterize variability in relatively small geographic areas (e.g. districts), it excludes the (i) the immunity of young children induced by exposure to WPVs, secondary spread of OPV-related viruses, and cVDPVs, (ii) serotype-specific immunity in some instances (particularly when countries use mOPV or bOPV), (iii) differences in the nature of immunity induced by OPV and IPV, and/or (iv) the potential role of older children and adults in transmission. The IDM papers that discuss population immunity also focus on vaccine coverage in young children. In the review, we noted two other modeling studies that mentioned population immunity [155, 162]. Although not captured by the review, a study of the impact of SIAs in the Democratic Republic of the Congo also estimated population immunity and emphasized the importance of achieving and maintaining high population immunity [231].

5.4.  OPV2 cessation dynamics
All three modeling groups provided recommendations to the GPEI related to OPV2 cessation. KRI integrated modeling [18] helped to support the GPEI establishment of a 2008 global agreement to stop OPV use after WPV eradication [232], and to do so with globally coordinated OPV cessation and with the contingency of mOPV vaccine stockpiles for outbreak response [21]. Despite delays in achieving WPV eradication, later integrated analyses reaffirmed this strategy [51, 54], while also emphasizing the need to carefully manage the risks associated with OPV cessation and to ensure sufficient OPV vaccine supplies [52, 53]. In preparation for OPV cessation, KRI applied DEB modeling to explore OPV cessation dynamics and recommended that the GPEI partners increase population immunity to transmission for serotype 2 to stop any existing cVDPV2s and prevent the creation of future cVDVP2s prior to globally coordinating OPV2 cessation by intensifying tOPV pSIAs [26, 38, 39, 49, 52–54, 68]. IC used an SC model to explore theoretical concepts related to OPV cessation dynamics [100]. When first presented to the GPEI partners, this modeling initially did not consider the seeding of OPV2 from routine immunization in all tOPV-using countries, which led IC to recommend caution about tOPV pSIAs and contrasted with the recommendations from KRI [38, 39]. However, in its published results, IC considered tOPV use in routine immunization, and supported the strategy of ‘focused tOPV SIAs before OPV2 withdrawal in areas at risk of VDPV2 emergence and in sufficient number to raise population immunity above the threshold permitting VDPV2 circulation’ [100]. A separate statistical analysis by IC supported the GPEI decision to globally coordinate OPV2 cessation in 2016 based on its assessment and expectations about population immunity for Nigeria and Pakistan [111]. After OPV2 cessation, IC and IDM performed statistical analyses that reported that the high population immunity achieved in most areas helped with the prevention of cVDPV2s [113, 143], while also noting problem areas. KRI and IDM characterized the expected increasing vulnerability of populations to transmission of serotype 2 LPVs as a function of time after OPV2 cessation and the risks posed by reintroductions of LPVs from multiple potential sources, including the risks of using mOPV2 use for outbreak response [56–58, 60, 133]. After OPV2 cessation, in a review of lessons learned KRI emphasized the importance of reaching under-vaccinated subpopulations [69], characterized the probabilities of potentially needing to restart OPV2 vaccine production and use on a large scale [77], and discussed the complex vaccine choices and logistics of managing vaccine supplies [80]. Several studies by others also explored the dynamics of OPV cessation and the risks of reestablished transmission [153, 170].

5.5.  IPV
Numerous studies explored the role of IPV use after OPV cessation [18–20, 25, 28, 33, 39, 51–60, 64, 65, 68, 69, 73–77, 80, 90, 91, 94, 97, 103, 114, 115, 118–120, 122, 133, 150, 153, 164], primarily related to IPV use in routine immunization after WPV eradication. These studies included consideration of the use of IPV in oSIAs, which represents a topic on which the modeling groups offered different recommendations [51, 55, 64, 68, 133, 150, 164]. Notably, KRI does not recommend the use of IPV for oSIAs in OPV-using countries except when homotypic OPV is not available, because adding IPV to oSIAs is not effective and not cost-effective based on its DEB and integrated modeling [64]. In contrast, IC suggests that adding IPV may offer some benefit based on statistical modeling of observational data [106, 114]. The health and economic benefits of using IPV in routine immunization in OPV-using countries differ substantially before and after homotypic OPV cessation. Giving IPV doses sequentially before OPV doses in a national immunization schedule can eliminate VAPP, which is important in high- and upper middle-income countries that achieve high coverage and want to minimize risks associated with vaccine use. In contrast, for countries with relatively lower coverage, IPV may provide some protection from paralysis to the small fraction of children who only receive IPV, but it does not substantially contribute to population immunity to transmission and it may lead to the potential for silent transmission or delayed detection of transmission of LPVs [26–28, 39, 97, 133, 162, 164, 166, 170]. The high cost of IPV also remains an issue, with the relatively high cost of the vaccine and its administration making IPV use not cost-effective. IPV offers an expensive option for post-OPV cessation insurance (i.e. a vaccine that provides protection from paralysis to recipients at a high cost for a virus that is supposed to be gone and does not limit participation in transmission if the virus is not gone or is reintroduced).

5.6.  Undervaccinated subpopulations and ‘weak links’
All of the modeling groups recognized the role of undervaccinated subpopulations in sustaining LPV transmission and recommended focus on these weak links. However, the groups recommended different strategies. Based on the application of its DEB modeling, KRI repeatedly emphasized the need to overcome the failure to vaccinate these subpopulations and to reach all populations with sufficient quantities of tOPV prior to OPV2 cessation, and bOPV after OPV2 cessation to achieve high levels of population immunity to transmission to stop and prevent WPV and cVDPV transmission [36, 38–40, 42, 47, 49, 68, 69, 73, 74]. In contrast, IC emphasized vaccine failure based on its characterization of low OPV efficacy from case–control studies of epidemiological data [82–84, 87–92, 103–108, 114, 115], which led IC to recommend new vaccine tools (e.g. mOPVs, bOPV, IPV) as a way to get around poor programmatic performance. IC and IDM also both focused attention on applying statistical models to characterize population immunity (as they, respectively, defined it for different studies, see note at the bottom of Table 4) and on identifying national and subnational areas that previously performed poorly, for which they recommended temporary shifts or optimization of resources to deal with the failure to vaccinate in some populations [87, 89, 92, 93, 100–102, 109, 111, 131, 140–142]. The differences between the recommendations of the modeling groups with respect to the delays in achieving polio eradication as due to failure to vaccinate vs. due to vaccine failure led to substantially different foci and investments. KRI suggests that chasing better (and often more expensive) tools (e.g. mOPV, IPV) has not helped accelerate global polio eradication, that achieving and maintaining eradication depends on continuing to get enough OPV preventively into susceptible children to stop and prevent the transmission of cVDPVs and/or WPVs (followed by careful and aggressive management of the risks of globally coordinated OPV cessation), and that as of early 2020, the GPEI appears off track [202, 203].

5.7.  Undetected circulation
Building on modeling performed prior to 2000 that supported the certification of the Americas as wild poliovirus free [197–199], multiple studies published in 2000–2019 explored the potential of undetected circulation and confidence about no circulation [27, 46, 70, 71, 75, 76, 132, 136, 161, 167–169]. Generally, the modeling studies to date agreed with respect to their recommendations about undetected circulation and high confidence about no circulation after 3 years with no detected evidence of LPV transmission while conducting high-quality surveillance. Although not captured in the review, modeling of one of the last known reservoirs of WPV3 transmission (i.e. Borno and Yobe, Nigeria) published in 2020 [233, 234] also supported the 2019 decision by the Global Certification Commission to certify the global eradication of indigenous WPV3 [235].

5.8.  Environmental surveillance
As the GPEI expanded its use of environmental surveillance, the modeling groups published increasing numbers of studies that included consideration of the information that environmental surveillance provides [43, 46, 67, 71, 75, 76, 117, 136, 145, 147, 162, 166].

5.9.  Other risks
To date, only KRI considered the risks of iVDPVs [4, 13, 50, 61, 81] and (un)intentional re-introduction risks (e.g. breaches in containment) in its global modeling (see Table 4 for references), although IC recently reviewed the WHO database of known iVDPVs [125].