Salk inactivated poliovirus vaccine (IPV) became available in the late 1950s and Sabin oral poliovirus vaccines (OPV) in the 1960s. Challenge studies indicate that IPV-induced immunity significantly reduces oral-oral poliovirus transmission but reduces participation in fecal-oral transmission much less than OPV-induced immunity[1, 2]. Israel introduced Salk IPV in early 1957 using mass campaigns (supplemental immunization activities [SIAs]), followed by introduction routine immunization (RI) in 1958. This led to a transition from hundreds of paralytic cases caused by wild polioviruses (WPVs) each year during 1950-1956 to a period of limited transmission with episodic outbreaks.[3] Israel introduced oral poliovirus vaccine (OPV) beginning with monovalent OPV types 1 and 2 (mOPV1 and mOPV2) in SIAs in 1961 and type 3 (mOPV3) in 1962. Israel stopped mOPV SIAs when it switched from IPV to trivalent OPV (tOPV) for RI starting in 1963 (3 doses, 6 weeks apart and a booster dose 6 months later).[4,5] Continued polio cases among the non-Jewish population during 1974-1979 led Israel to conduct annual national 2-week (May-June) SIAs during 1979-1982 with mOPV1 targeting children under age 3 years, including all non-Jews and Jews living near non-Jewish settlements.[4,6]

While most of the country continued to use tOPV for RI without any SIAs after 1982, two subdistricts (Hadera and Ramla) used enhanced-potency IPV for RI at 2, 4, and 12–16 months of age from 1982 to 1988.[6] Annual mOPV1 SIAs for non-Jewish children in Hadera and all children in Ramla in 1982, 1983, and 1984 targeted children under the age of 3 years at the time of the SIA.[6] In 1988, a type 1 WPV outbreak (WPV1) occurred, cases occurring predominantly in Hadera, after an importation most likely from northern Egypt in 1987.[7,8] To control the outbreak, Hadera (and Ramla) conducted an outbreak response SIA with tOPV in September 1988 and Israel conducted a national SIA in October 1988.

Similar to other countries that successfully eliminated WPVs, in 1990 Israel switched to a sequential IPV-OPV RI schedule to avoid possible cases of vaccine-associated paralytic polio (VAPP). The schedule included 3 IPV doses followed by 3 tOPV doses administered by age 12 months, with tOPV booster doses at ages 6 and 16 years.[7] In 2005, 3 years after the World Health Organization (WHO) certified the European Region (including Israel) as polio-free,[9] Israel implemented an IPV-only RI schedule, with doses given at age 2, 4, 6, and 12-16 months and a booster dose at age 7 years.

Environmental surveillance began in northern Israel during the 1987-1988 outbreak and expanded nationally to cover 30-40% of the population with routine sewage sampling in multiple locations in 1989.[10] Israel reported no WPV-associated polio cases after 1989; however, the environmental surveillance system isolated WPV1 in 1995 and type 2 vaccine-derived polioviruses (VDPV) on multiple occasions since 1998 from two sampling locations.[7] On April 7, 2013, WPV1 isolated in environmental samples taken from Beer Sheva and Rahat led to a retrospective analysis of earlier samples, which suggested introduction of WPV1 in early February 2013.[10-12] Repeated environmental sampling revealed evidence of WPV1 transmission based on positive samples collected from multiple sites in the South over extended periods of time, but limited transmission in the rest of Israel [10]. In response to the detected WPV1, Israel conducted a national IPV catch-up SIA campaign targeting all children under age 6 years with incomplete vaccination, combined with a booster dose for adults aged >18 years in Rahat (a South district town at the epicenter [11]). Israel then conducted a national bivalent types 1 and 3 OPV (bOPV) SIA during August-October 2013 for children born since 2005,[11] and added 2 bOPV doses given at ages 6 and 18 months to the national RI schedule in January 2014.[13]

Prior modeling results suggested that as the world approaches the eradication of WPVs and countries increasingly shift from OPV to IPV, national health leaders will need to (1) recognize the heterogeneity in population immunity might create pockets of susceptible individuals and (2) develop immunization strategies to protect the entire population.[14] Because of the relatively lower population immunity to transmission induced by IPV compared with OPV, the observation of WPV1 transmission in Israel without detected polio cases confirms the possibility for live polioviruses (LPVs) to circulate in the context of high IPV-only coverage (in the absence of WPVs).[15] This paper models population immunity for Israel to provide insights about potential LPV transmission in a country using IPV for national RI and the development of immunization strategies to prevent transmission following LPV importations.

We collected the best available information about RI and SIAs histories (summarized in Table 1). Table 2 shows the inputs we used in our differential equation based model with developed country settings and assumptions related to demographics, poliovirus transmissibility and seasonality.[1,2,16,17] We divided the population of Israel into 13 age groups, with three preferentially mixing age groups (i.e., 0–4, 5–14, and 15 or more years). Considering the different immunization schedules in the 1980s and the epidemiology of the 1988 WPV1 outbreak, we divided the Israeli population into two preferentially mixing regions for Hadera and the rest of the country before 1994. From 1994, when population immunity in both regions became similar, we changed the geographic division of the population to the South district and the rest of Israel to capture differences in hygiene and crowding conditions. Each region also includes a preferentially mixing subpopulation that represents historically undervaccinated non-Jews for which Israel separately reports polio incidence and vaccination coverage.[3-7,18,19] Thus, the model includes four subpopulations representing Jews and non-Jews in different parts of the country before 1994 (i.e., Hadera and the rest of Israel) and from 1994 (i.e., South district and the rest of Israel).

We assumed an average basic reproductive number (R₀ ) for WPV1 of 5 for the entire period, with seasonal variation (Table 2). However, after changing the subpopulation in 1994, we assumed a higher average basic reproductive number for WPV1 (R₀ = 6) and a lower proportion of oropharyngeal transmission (p^oro =0.6) in the non-Jewish subpopulation in the South district based on conditions more conducive to transmission (i.e., relatively poor hygiene, crowding).

Our model considers eight immunity states: fully susceptible individuals with no successful vaccination or LPV infection; maternally immune infants born to mothers with any active, recent, or historic immunity; individuals who acquired immunity from either one or ≥2 prior LPV infections; individuals with 1, 2, or ≥3 successful IPV doses; and individuals with some combination of both ≥1 LPV infection and ≥1 IPV dose.[1,2,16,17] We assumed that LPV infection can lead to paralytic poliomyelitis only in fully susceptible individuals (including infants born with maternal immunity who lose protective maternal antibody levels); whereas individuals in all other immunity states remain completely and permanently protected from paralytic poliomyelitis. Individuals in any immunity state can get reinfected and participate asymptomatically in poliovirus transmission to varying degrees (depending on immunity state).[1,2,16,17]

Our model includes fecal-oral and oropharyngeal transmission routes, 20 stages of OPV reversion, and five stages of immunity waning. Similar to prior work,[17] we characterized population immunity to poliovirus transmission by computing the effective immune proportion adjusted for age- and sub-population-mixing (EIPM), with an effective immunity threshold EIP*=(1- 1/R₀), above which infections eventually die out.[1,2,16,17] To simulate die-out of a given virus strain in the model, we set the force-of-infection for that strain to zero whenever the infectiousness-weighted prevalence of infectious individuals falls below a certain “transmission threshold” calibrated based on the experience with poliovirus die-out in diverse situations.[17] The model characterizes infections, and we estimated cases of paralysis as a function of the paralysis to infection ratios for first infections, which vary by serotypes for WPVs, VDPVs, and VAPP.[1,2,16,17]

Because of limited information about immigrant immunity status,[20] the model generally assumes the same population immunity profile for immigrants as in the Israeli population (Table 2).[17] However, the population of Israel increased rapidly in the 1950s because of major immigration of Jewish people, which probably led to net inflow of unvaccinated families. Therefore, up until 1965, we assumed 10% of immigrants under age 15 years that would enter as immunes according to the existing population immunity profile instead entered the population as fully susceptible individuals. Moreover, because of persistent WPV circulation in neighboring countries, during 1962–1982 we assumed annual effective introductions of WPV1 to all Israeli subpopulations at the threshold at which transmission can occur.[17] We continued effective WPV introductions on a biannual basis during 1983-1987 only into non-Jewish subpopulations, because we assumed they disproportionally interacted with other populations with persistent WPV circulation during that time (e.g., Egypt). We introduced WPV1 into the non-Jewish population of the South district at the beginning of 1995 as the importation first detected in Gaza and again at the beginning of 2013 as the importation that originated from Pakistan.[11, 12]

Depending on the RI schedule in a given period, we generally assumed administration of RI doses when a child aged from the 0 to 1-month into the 2-month age group, from the 3-month into the 4 to 5-month age group, from the 4 to 5-month into the 6 to 11-month age group, and from the 6 to 11-month into the 1-year age group. However, in Hadera during 1982-1988, we assumed no IPV dose administered when a child aged from the 4 to 5-month into the 6 to 11-month age group. Consistent with prior work,[21,22] we modeled the cumulative effect of IPV doses administered at the same time as OPV doses as an effective boost by 1 IPV dose. Also, we characterized each SIA by its coverage and its probability of repeatedly missing individuals targeted in the preceding SIA round (Table 1).[23]

We collected information about vaccination coverage[3-7, 10,11] and used national WHO-UNICEF estimates of the third dose of polio vaccine coverage for Jewish subpopulations during 1980-1993 (Figure 1).[19] Our model assumes low relative coverage for non-Jewish subpopulations compared to Jewish subpopulations starting from the early years of vaccination, which we increased up to 1 gradually during 1979-1993. We used Israel Ministry of Health data as the basis for RI coverage estimates for Jewish and non-Jewish subpopulations for the South district and the rest of the country from 1994 forward (Figure 1). Moreover, because of cross-border population movements between Israel and the West Bank and Gaza, we accounted for the possibility of secondary OPV spread to the Israeli population from all SIAs in the West Bank and Gaza (Table 3). During these SIAs, we assumed effective OPV coverage of 0.01 for non-Jewish and 0.001 for Jewish subpopulations.

For the reference case vaccination strategy, we assumed the RI schedule included 4 IPV doses by age 12 months and 1 booster dose by age 7 years during 2005-2013. We included the IPV catch-up SIA conducted during July-August 2013 and the bOPV SIA conducted from August to October of 2013, as well as the addition of 2 bOPV doses to the RI schedule starting in January 2014. Including the 2 OPV doses given in Rahat, we assumed a SIA duration of 50 days for the reference case (i.e., the SIA reached 90% of all individuals vaccinated within 50 days.[10])

We compared the reference case with several alternatives, including the counterfactual scenario of maintenance of the sequential IPV/OPV schedule after 2005 and no response (Table 4). We considered the option of typical outbreak response 45 days after onset of the first paralytic case, with detection of the case occurring 10 days after infection,[17] to simulate a typical outbreak response in the absence of environmental surveillance. Because speed matters during a response,[24] we considered the impact of a more rapid response by shifting the SIA round to start in July (i.e., SIA 1 month earlier), or by assuming an option with the bOPV SIA conducted over only 4 days (i.e., SIA duration 4 days), or both (i.e., SIA lasting 4 days, 1 month earlier). Finally, we considered a scenario of 20% higher absolute SIA coverage in the rest of Israel (excluding the South district) (i.e., better SIA coverage).

Figure 2 displays the annual paralytic polio incidence in Israel caused by all WPVs in the model compared with the total reported for all three serotypes during 1950-2015 (broken into three periods to accommodate the dramatic changes in the y-axis scales as immunization reduced incidence). Overall, our model shows a similar pattern of “low” and “high” years during 1950-1965 (Figure 2a), although the model estimates fewer cases than reported for the 7 years before initial IPV vaccination (1950-1956). The lower model estimates compared to the reported cases reflects uncertainty about reported cases, potential inclusion of non-polio cases in early monitoring, and/or the extensive immigration of infected or susceptible individuals, especially from Eastern European countries, which we only partially considered in the model by disproportional inflow of fully susceptible individuals. The model also produces good correlation with the observed incidence during 1965-1980 (Figures 2a and 2b), when high population immunity resulted from the increase in RI coverage and four annual mOPV1 SIAs conducted during 1979-1982. Figure 2b shows the 1988 outbreak with WPV1 importation in 1987 and relatively quick elimination after the immunization response. The sequential IPV/OPV schedule adopted in 1990 maintained population immunity high enough to avert widespread circulation despite assumed periodic importations of WPV1 and VPDV2 (i.e., no reported cases after 1988 in Figures 2b and 2c). Our model mimics the behaviour of the 1995 imported WPV1 resulting in 1-month of low-level circulation followed by die-out.

Figure 3 shows the age- and mixing-adjusted effective immune proportion (i.e., the EIPM or overall population immunity for all of Israel) and the seasonally varying threshold EIP* (i.e., the threshold above which transmission dies out) for all three serotypes. Comparison of the curves in Figure 3a shows that the RI schedule with OPV significantly increased overall population immunity to transmission over the threshold until we observed a considerable drop during 1982-1988 that corresponded to the adoption of the IPV-only schedule in Hadera. The population immunity decreased because of the lower effect of IPV vaccination on transmission compared with OPV and the reduced number of doses (i.e., 2 instead of 3 doses by age 6 months) and lack of secondary OPV spread associated with the IPV-only schedule in Hadera.[12] The increase observed in the late 1980s occurs due to the outbreak in Hadera, outbreak tOPV response SIAs including a national SIA, and subsequent use of a sequential IPV/OPV schedule nationally. After the 2005 change in the RI schedule to all-IPV, Figure 3a shows that population immunity for type 1 started to decrease and over time dropped significantly below the EIP*. However, it dropped relatively more slowly than during 1982-1988 because of better coverage and more doses in the schedule. For the 2013 introduction of WPV1 the model estimates 360 days of circulation with 85% of infections in the South and 0.8 paralytic cases assuming a paralysis-to-infection ratio of 1/200[16,17] (see Figure 2c). After the IPV catch-up campaigns, bOPV SIA response, and the addition of 2 bOPV doses to the RI schedule in 2014, population immunity for serotype 1 for the reference case returned to high levels similar to those observed before the 2005 change in the RI schedule (Figure 3a).

Figures 3b-e show the results of the reference case and the options for maintaining population immunity by responding to the signal of silent circulation of WPV1 from environmental surveillance during 2013–2014. Figure 3b shows the population immunity at the beginning of 2013 significantly below the EIP* (see run up in Figure 3a), and Figure 3c shows the estimated number of daily new infections (including in partially infectible individuals who benefit from immunity to paralysis because of prior infection or vaccination with IPV).

Table 4 displays the estimated numbers of infections and cases associated with the different options. Maintaining the sequential IPV/OPV schedule prevents the circulation of the imported WPV or other LPVs. The model estimates 18 cases in 2014 with no response to interrupt transmission of the imported WPV or four cases if outbreak response started 45 days after onset of the first paralytic case. The results also suggest that when circulation occurs, earlier and more rapid response result in shorter times required to interrupt transmission (Figure 3c).

Table 4 also shows the model estimates of WPV1 elimination times relative to the reference case and the number of vaccine doses used. The model uses a threshold of five effective infections per million people (i.e., weighted to account for differential infectiousness of individuals in different immunity states) as a requirement to sustain transmission. Thus, once the number of infections drops below this threshold in all age groups and subpopulations, the force of infection drops to zero and no new infections occur. For the reference case the model estimated onset of the last new infection around January 31, 2014. Infected individuals continue to excrete poliovirus until recovery (e.g., average duration of approximately 30 days for fully susceptible individuals). Continued excretion implies that environmental surveillance may detect positive samples for some time after the onset of the last infection. To account for the possibility of the detection of poliovirus in sewage after the die-out of transmission, Figures 3d and 3e show the absolute number of individuals excreting and the weighted number of individuals excreting (weighted by relative infectiousness, which accounts for a lower contribution from partially infectible individuals).

The model results imply WPV detection by a sensitive surveillance system in early March 2014, consistent with the actual experience. The laboratory collected the last sample found to contain viable WPV1 on April 3, 2014, with prior positive samples collected from the same site on March 6 and March 20, and the last isolation at any other site occurring in a sample collected February 13, 2014. With weekly sampling, the isolation of live virus in successive sewage samples from the same site might represent virus excreted by the same individual(s) rather than virus from new infections due to continued person-to-person transmission (particularly toward the end of the event). Passage and pooling of virus in the sewage system may continue beyond the time of excretion, depending on the many factors, including flow rates, temperature, and concentrations of organic matter.[25,26]

Maintaining high population immunity to poliovirus transmission will protect countries from LPV transmission after an importation.[27] The recent Israeli experience[10-12] suggests that even with high IPV coverage, developed countries may still face the risk for transmission after importation of LPVs.[14,15] Our results for Hadera also demonstrate that drops in population immunity in one sufficiently large region can significantly affect national population immunity and vulnerability to transmission (Figure 3a).

The strategy of not giving bOPV to children without prior IPV doses reflected concern about exposing previously unimmunized children to the risk for VAPP. This choice leaves these children fully susceptible to the uncertain, small (non-zero) risk for exposure to a WPV that causes paralysis at a significantly higher rate than OPV (i.e., >1,000-fold), compared with giving the child OPV, which comes with a certain, small (again non-zero) risk for VAPP. Our model suggests that targeting all children would have increased population immunity enough to stop transmission more than a month earlier. However, this transmission occurred with no reported cases, and concern about VAPP represented an important health communication challenge. Given the unpredictable nature of paralytic cases, we focused on expected cases, but in reality, each infection in a fully susceptible individual either will cause paralysis or not.

Maintaining a sequential schedule in our model appears to keep population immunity above the threshold required to prevent the transmission of imported WPVs. The Israeli experience and these model insights might help to inform decisions about switching to an IPV-only RI schedule. Further studies should continue to explore the dynamics of population immunity and OPV cessation.[21,22,28,29] Our results remain limited and depend on the assumptions we used. For example, our results depend on the assumption that IPV-induced immunity reduces participation in fecal-oral poliovirus transmission much less than immunity induced from a LPV infection, which remains numerically uncertain, but consistent with the data from challenge studies.[1, 2] Also, because die-out is a stochastic process, the model approximates with uncertainty the real events that we cannot observe.

These results underscore the value of response to an early warning signal from environmental surveillance, which prevented at least four expected cases. Our model results suggest advantages in rapid response to WPV circulation rather than the actual more drawn-out campaign. Conducting an OPV campaign comes with logistical and risk communication challenges, when implemented in a country with IPV-only RI and in the absence of cases.[30] The Israeli experience shows the importance of effectively communicating the risks associated with a silently circulating WPV and the risks and benefits of OPV.[30] As concerns about WPV1 circulation subside, Israel will face choices about continuing bOPV in RI, which might range from no change, to cessation, to replacement with tOPV. National health leaders everywhere should appreciate the importance of managing population immunity to transmission even in well-vaccinated populations and that they may need to consider continued use of OPV after introduction of IPV until global coordinated cessation of OPV use.