Impact of Supplementary Immunization Activities using Novel Oral Polio Vaccine Type 2 during a Large outbreak of Circulating Vaccine-Derived Poliovirus in Nigeria

Abstract

Background

Novel oral poliovirus vaccine (OPV) type 2 (nOPV2) has been made available for outbreak response under an emergency use listing authorization based on supportive clinical trial data. Since 2021 more than 350 million doses of nOPV2 were used for control of a large outbreak of circulating vaccine-derived poliovirus type 2 (cVDPV2) in Nigeria.

Methods

Using a bayesian time-series susceptible-infectious-recovered model, we evaluate the field effectiveness of nOPV2 immunization campaigns in Nigeria compared with campaigns using monovalent OPV type 2 (mOPV2).

Results

We found that both nOPV2 and mOPV2 campaigns were highly effective in reducing transmission of cVDPV2, on average reducing the susceptible population by 42% (95% confidence interval, 28–54%) and 38% (20–51%) per campaign, respectively, which were indistinguishable from each other in this analysis (relative effect, 1.1 [.7–1.9]). Impact was found to vary across areas and between immunization campaigns.

Conclusions

These results are consistent with the comparable individual immunogenicity of nOPV2 and mOPV2 found in clinical trials but also suggest that outbreak response campaigns may have small impacts in some areas requiring more campaigns than are suggested in current outbreak response procedures.

Using a bayesian time-series susceptible-infectious-recovered model, we evaluated novel oral poliovirus vaccine type 2 (OPV2) in Nigeria and found the impact on disease transmission comparable to that previously achieved with monovalent OPV2.

Of the 3 types of poliovirus, wild poliovirus (WPV) type 2 (WPV2) was the first to be eradicated, last detected in natural circulation in 1999 in India. However, the type 2 component of the live attenuated Sabin oral poliovirus vaccine (OPV) readily loses its primary attenuation, and in settings of persistently low immunity can transmit from person to person, causing outbreaks of circulating vaccine-derived poliovirus type 2 (cVDPV2). The lack of indigenous WPV2, periodic cVDPV2 outbreaks, and rare adverse events such as vaccine-associated paralytic poliomyelitis associated with OPV type 2 (OPV2) prompted the withdrawal of type 2–containing OPV from routine use in 2016 [1]. However, cVDPV2 outbreaks have continued, requiring additional type 2–containing OPV to control, typically in a monovalent presentation (monovalent OPV2 [mOPV2]). The se of mOPV2 has unfortunately resulted in additional cVDPV2 outbreaks in certain settings, facilitated in part by lower type 2 poliovirus mucosal immunity since 2016, due to OPV2 withdrawal [2]. This cycle of outbreak response and generation of new cVDPV2 outbreaks has prompted the development and emergency use listing of a novel OPV2 (nOPV2), designed to be more genetically stable and therefore less neurovirulent. It is now being used to respond to cVDPV2 outbreaks and is anticipated to reduce the rate of new emergence of cVDPV2 [3–6].

While the adequacy of the humoral and intestinal immunogenicity of nOPV2 is supported by clinical trial data [7–9], its effectiveness in the field when used in supplementary immunization activities (SIAs) for stopping transmission is not yet established. Importantly, the efficacy of OPVs is often lower than that in found in clinical trials [10–12], which has recently been observed for nOPV2 [13]. For type 2 vaccines, their relatively good performance in the field may in part explain the early eradication of WPV2 in 1999, while WPV type 1 (WPV1) is still endemic. Relative to types 1 and 3, the higher primary immunogenicity of the type 2 component of trivalent OPV (tOPV) and higher rates of Sabin 2 infection among contacts of vaccinees both contribute to this [14,15]. Some evidence suggests that nOPV2 is not shed as long as Sabin OPV2 [16] and thus may result in less passive immunization. In addition, since nOPV2 is unable to gain fitness through reversion as easily as Sabin OPV2, it may have a decreased ability to transmit. Furthermore, a recent study found that immunogenicity was reduced when nOPV2 was coadministered with bivalent OPV, suggesting that nOPV2 may be more susceptible to interference from other enteroviruses [17]. Thus, the ability of nOPV2 to stop outbreaks requires additional assessment.

The first and largest use of nOPV2 to date has been in Nigeria, where mOPV2 was also used extensively, allowing for a detailed comparison. Nigeria is of particular interest, since it has been a focus for cVDPV2 transmission historically [18] and the recent outbreaks in which nOPV2 was used in response were far larger than previous outbreaks in which mOPV2 was used (see Figure 1). However, inferring the impact of vaccination activities and comparing vaccines of different types is not straightforward. Differing background immunity, transmissibility, outbreak response speed, and coverage will all lead to variation in outbreak behavior that can be difficult to separate from the effects of the vaccine used.

Figure 1.

Circulating vaccine-derived poliovirus type 2 (cVDPV2) cases and supplementary immunization activities (SIAs), from 2016 to June 2022. Top, cVDPV2 cases (bars), environmental surveillance detections (rug plot), and SIAs as percentage of the population targeted. Bottom, Geographic distribution of cVDPV2 cases for years where there were cases. Abbreviations: IPV, inactivated poliovirus vaccine; mOPV2, monovalent OPV type 2; nOPV2, novel OPV type 2.

In the current study, we used a time-series susceptible-infectious-recovered (TSIR) model to evaluate a series of cVDPV2 outbreaks that occurred in Nigeria since 2016. In our model we found that outbreak response with nOPV2 had an impact in Nigeria comparable to that of previous responses with mOPV2, despite qualitatively different outbreak behavior, and we found that delayed vaccination response likely played a critical role in the larger recent outbreaks. This work supports continued use of nOPV2 for outbreak response but also highlights the critical role of timely and high-quality vaccination response.

METHODS

Data

cVDPV2 isolates from acute flaccid paralysis and environmental surveillance data were obtained from the Polio Information System maintained by the World Health Organization and mapped to the first administrative unit (states in Nigeria) to produce time series of cVDPV2 cases and environmental samples. Immunization activities relevant to cVDPV2 were provided by Nigeria's Polio Emergency Operations Center and mapped to the same geographic areas. We considered vaccination campaigns containing nOPV2, mOPV2, tOPV, and inactivated poliovirus vaccine (IPV). No tOPV SIAs occurred during the period considered for analysis. We did not distinguish between fractional-dose (intradermal) IPV administration and full-dose intramuscular administration, since both are believed to provide comparable type 2 boosting [19]. Data reported before 11 October 2022 were included, and the analysis was limited to surveillance data with paralysis onset or environmental sample collection between 1 June 2016 and 31 July 2022. We included 30 states and the Federal Capital Territory in the analysis, removing 6 states where there were no cVDPV2 detections during the period considered.

Initial immunity estimates at the time of OPV withdrawal (May 2016) are obtained from a cohort model described elsewhere [20], using an 80% SIA coverage assumption. State-level population estimates were obtained from WorldPop population rasters [21] and extracted to the state boundaries in the World Health Organization polio geodatabase.

Model

The impact of SIAs was estimated using a TSIR model, forms of which have been used elsewhere to estimate impacts of interventions from disease surveillance data [22,23]. Our implementation is described in more detail in the Supplementary Materials. Briefly, the model estimates the number of infected and susceptible children in each state in Nigeria in discrete 1-month time steps. The impacts of SIAs are quantified by the (log) proportional reduction in remaining susceptible children following an SIA and thus the proportional reduction in infections and expected cases in the subsequent month. Fractional SIAs are permitted for SIAs covering only part of a province. We assume a 1-month lag time between the conducting of an SIA and the reduction in the susceptible population. The impacts of SIAs are allowed to vary between SIAs with the quality of implementation and other factors using random effects. We use a 1-month generation time, which is consistent with other models fit to WPV1 transmission in Nigeria estimating a mean infectious period of 27 days with a 1–5-day incubation period [24], derived from studies of shedding duration [25].

Cases are observed according to a Poisson distribution related to the number of infections, using a case-to-infection ratio of 1:2000. The case-to-infection ratio assumption is somewhat relaxed by including an inflation factor on the number of infections as they are removed from the susceptible population. This flexibility can also capture impacts of other factors, such as a subset of the population participating in transmission or a reduction in the proportion of infections that result in paralysis due to IPV coverage. We also include an estimate of susceptibility based on SIAs alone, in which the impacts of infection are removed. Environmental surveillance samples are related to infections via a logit link with sensitivity varying by site. The model was fit using the Stan software version 2.21.0.

Sensitivity Analyses

We explore several univariate sensitivity analyses, in particular to the generation time, priors on transmission process variability and participation rates, removal of IPV, seasonality estimates from WPV1, and inclusion of environmental surveillance data. In our base model, SIAs are allowed to have negative effects in order to avoid the implicit assumption that vaccination campaigns are effective, so we also explore sensitivity analyses we constrain the per-SIA reduction in the susceptible population to be between 0 and 1.

RESULTS

Description of the Outbreak Data

cVDPV2 detections and immunization activities are summarized in Table 1, and in more geographic detail in the Supplementary Materials. Outbreaks of cVDPV2 first peaked in 2018, driven by the JIS-1 lineage, and again in 2021, driven by the ZAS-1 lineage, with 34 and 417 cases, respectively. Similar trends can be seen in environmental surveillance.

Table 1.

Circulating Vaccine-Derived Poliovirus Type 2 Detections and Supplementary Immunization Activities by Year

cVDPV2 Detections and SIAs	2016	2017	2018	2019	2020	2021	2022
cVDPV2 detections
Cases, no.	1	0	34	19	7	417	40
Environmental samples, no. (proportion positive)	0 (0)	0 (0)	46 (0.03)	58 (0.03)	5 (0)	303 (0.12)	67 (0.03)
Age, median (IQR), mo	30 (…)	…	24 (18–35)	30 (18–46)	18 (14–27)	26 (19–38)	24 (17–36)
SIAs
IPV, country-equivalentsa	0.02	0.01	0.73	0.24	0.03	…	…
mOPV2. country-equivalentsa	0.39	0.64	1.01	1.5	0.09	0.01	…
nOPV2, country-equivalentsa						2.54	2.58
OPV response time, median (IQR), d	36 (30–43)		26 (9–52)	24 (14–52)	94 (79–116)	25 (12–44)	21 (21–21)

Abbreviations: cVDPV2, circulating vaccine-derived poliovirus type 2; IPV, inactivated poliovirus vaccine; IQR, interquartile range; mOPV2, monovalent OPV type 2; nOPV2, novel OPV type 2; OPV, oral poliovirus; SIAs, supplementary immunization activities.

^aCountry-equivalents were defined as the average number of SIAs conducted, where fractional numbers result from subnational SIAs. Data as of October 2022.

Overall, the median age at paralysis was 26 months (interquartile range, 18–38 months). We found no evidence for differences in the average age at infection across years (analysis of variance P = .38) or by geographic area (analysis of variance P = .95).

In response to these outbreaks, a median of 2 SIAs were conducted per area. In 2021 generally more SIAs were conducted across broader areas and followed by additional rounds in the same areas in 2022. The overall response timeliness (defined as the time between paralysis date or environmental sample collection and initiation of the first OPV SIA in that area, for areas without an SIA in the previous 6 months) was 28 days (interquartile range, 13–50 days), which did not vary substantially between years, except in 2020 where a temporary pause in global polio immunization activities likely played a role. However, as we will see below, despite overall similarity in response timelines throughout the analysis period, there were consequential differences in the speed of response in some subnational areas.

Fit of the Model

Posterior estimates of infections scaled to paralysis incidence, environmental detections, and underlying susceptible population are displayed in Figure 2, aggregated to the national level. State-level estimates of incidence and susceptibility are given in the Supplementary Materials. Overall, the model fits incidence well: estimated infections follow trends of reported cases, and the 95% Poisson prediction intervals cover 98.9% of the reported cases at the state level and 87.8% at the national level. Likewise, for total estimated environmental detections, we find that 93% of observations fall within the 95% Poisson-binomial prediction intervals. Environmental surveillance site sensitivity is described in more detail in the Supplementary Materials. Estimated susceptibility increases after cessation of tOPV use, from a low of 13% in July 2016 to a high of 63% reached in both September 2018 and April 2021, suggesting similar immunity profiles at the start of the 2 major outbreaks. Ignoring the impact of naturally acquired infections resulted in substantially higher susceptibility estimates over 2021–2022, comparable to the minimum susceptibility over the 2018–2019 outbreak.

Figure 2.

Posterior estimates of paralysis incidence (A), positive environmental surveillance (ES) samples (B) and susceptible fraction (C) at the national level. A, B, Black dots represent the observed data where greater than zero; dark bands, 95% posterior mean; and light bands, 95% posterior prediction intervals. Abbreviation: SIAs, supplementary immunization activities.

Estimated Epidemiological Parameters and Average SIA Impacts

The basic reproduction number among 6–36-month-olds was estimated to be 2.7 overall (95% confidence interval [CI], 2.45–2.95) but varied substantially between areas of the country, being generally higher in northern Nigeria than in the south (Figure 3). Our model included an inflation factor on infections to allow more flexible relationship between reported cases and their impact on the susceptible population. We estimated this infection inflation factor to be 5.1 (95% CI, 3.81–6.50), suggesting that the case-to-infection ratio is smaller than 1:2000 and/or that a smaller subset of the population (perhaps 20%) drives transmission.

Figure 3.

Estimated supplementary immunization activity (SIA) effectiveness (right) and estimated basic reproduction number (R₀) (left), by state.

Likewise, we found that SIA impact, which we indexed to mOPV2 SIAs as the reference group, varied widely with geography. The average SIA impact was smaller in northwestern states and comparatively larger in most southern states. Extreme examples are Kebbi and Zamfara states, where the per-SIA impact is estimated to be close to zero (average impacts of 0.01 and 0.05, respectively). This is apparently because there was an increase in reported cases in those states after the first 2 rounds of SIAs had been conducted (data not shown). The average SIA effectiveness appeared to be correlated negatively with the estimated reproduction number (ρ = −0.2), with notable exceptions to this trend, such as Kano State, which appeared to have both high reproduction number and high average SIA effectiveness.

Relative Effectiveness of nOPV2 SIAs

Impacts of SIAs using nOPV2 reduced the estimated susceptible population by 42% with each SIA (95% CI, 28%–54%), compared with 38% for Sabin OPV2 (20%–51%); these were not distinguishable from each other in this analysis (relative effect 1.1 [95% CI, .7–1.9]). It is notable though, that these results do not rule out potentially meaningful differences: the lower limit of the 95% credible interval allows 30% less per-SIA impact for nOPV2 compared with mOPV2.

Estimates of IPV SIA effect were lower than for nOPV2 and mOPV2 and estimated with a high degree of uncertainty (Table 2). This is likely, since IPV was used less frequently and in short succession with OPV, such that any effect it had could not be separated from the impact of the OPV immunization activities.

Table 2.

Estimated Effectiveness of Supplementary Immunization Activities by Type of Vaccine

Vaccine	Effectiveness (95% CI)	Effectiveness Relative to Sabin OPV2
Sabin OPV2	0.38 (.2–.51)	…
Novel OPV2	0.42 (.28–.54)	1.1 (.7–1.9)
IPV	0.24 (−.37–.59)	0.63 (−.94–1.77)

Abbreviations: CI, confidence interval; IPV, inactivated poliovirus vaccine; OPV2, oral poliovirus type 2.

Sensitivity Analyses

The results of the sensitivity analyses on estimated SIA impacts for novel and Sabin OPV are displayed in Figure 4. None of the sensitivity analyses change the overall conclusions of comparability of nOPV2 and mOPV2. The point estimates of SIA impact are sensitive to assumptions and constraints of the model, and all overlap with the base model estimates. Decreasing the generation time has the largest attenuation of effects but also makes nOPV2 look relatively more favorable than mOPV2. Allowing for a more diffuse prior on the transmission process results in the largest increase in the effect estimates and results in mOPV2 having a larger point estimate than nOPV2.

Figure 4.

Sensitivity analyses. Left, Estimated supplementary immunization activity (SIA) impact for novel oral poliovirus (OPV) type 2 (nOPV2) and monovalent OPV type 2 (mOPV2). Right, Relative impact of nOPV2 compared with mOPV2. Abbreviations: ES, environmental surveillance; IPV, inactivated poliovirus vaccine.

DISCUSSION

In the current study, we developed a TSIR model of successive cVDPV2 outbreaks in Nigeria to estimate impacts of SIAs using nOPV2 relative to mOPV2, based on the timing of cVDPV2 detections relative to SIAs. Despite relatively small outbreaks during a period of mOPV2 use (2016–2020) and much larger outbreaks during a period of nOPV2 use (2021–2022), we found that nOPV2 campaigns had effectiveness comparable to that of Sabin OPV2 campaigns. This complements earlier clinical trial data showing that nOPV2 and mOPV2 were similarly immunogenic in individuals and, given the lower trends of new emergences due to nOPV2 use, supports its continued use as the vaccine of choice for cVDPV2 outbreak response.

We found that impact of vaccination varied substantially by geography, which may be reflect variable coverage campaigns, population-specific effectiveness, or other ecological factors, which we are not able to distinguish between in our model without additional assumptions. While we do not explicitly estimate rates of secondary immunization, the persistence of cVDPV2 through multiple SIA suggests that it is insufficient to overcome issues in coverage and efficacy. Thus, this work adds an important addition to clinical trial results by suggesting that repeated nOPV2 immunization activities may be required to stop cVDPV2 outbreaks, a result that has also been seen for other OPVs [11,12,26]. This has relevance for outbreak response protocols, where more vaccination rounds may be necessary in areas where poliovirus has been difficult to control in the past, and more emphasis placed on quality of campaigns where that is known to be an issue.

Because we estimate that the impact of nOPV2 SIAs is comparable to that of mOPV2 SIAs, reasons for the large cVDPV2 outbreak in 2021 are thus unlikely to be due to vaccine choice. One hypothesis for the discrepancy in the size of outbreaks is the size of the susceptible population, as the 2021 outbreak occurred further after cessation of routine OPV2 use. However, the outbreak response to the earlier cVDPV2 outbreaks in 2018–2020 likely made it such that population immunity during 2021 should not have been notably lower than in 2018, which is also suggested by both our estimates of immunity and the age distribution of cases. However, a key difference between the 2 outbreaks was the size of the outbreak when the responses occurred: in 2018 we estimate that the number of infections at the time of the first mOPV2 SIAs was 12 600 (95% CI, 9400–17 100), compared with approximately 10 times that in areas at their first nOPV2 use (118 000 [102 000–137 000]). This high force of infection likely resulted in significant onward transmission after multiple SIAs in 2021.

The reasons for the rapid accumulation of cases and infections in 2021 are not immediately clear, but some evidence suggests that response times played a role. In most areas, the vaccination response was slightly faster relative to detections in 2021 than in 2018 (median, 25 days from paralysis onset or environmental sample collection to response in 2021 vs 26 days in 2018). However, the responses in 2021 occurred further along in the outbreak in some key high-transmissibility northern states. Notably, Jigawa State had the most cases over the period considered (n = 87). There was a gap of 107 days between the onset of the first case in 2021 and outbreak response in Jigawa, compared with 25 days in 2018. Likewise in Kano State, the 2021 outbreak response occurred 66 days after the first case, whereas responses in 2018 were based on environmental surveillance alone, and no cVDPV2 cases occurred. A counterexample to this trend is Kebbi State, where the outbreak response in 2021 preceded the first case yet still saw the accumulation of many cases (n = 79), whereas in 2018 no cases occurred in Kebbi State. In the model this is explained by an ineffective response specific to that state, but other virological, immunological, or ecological factors may play a role, because the estimated ineffective response is specific to the state and has a small impact on overall estimated nOPV2 effectiveness.

Overall, this work suggests that SIAs using nOPV2 were comparably effective to those using mOPV2, and it points to outbreak response fundamentals, particularly the speed of responses, as key drivers in overall outbreak behavior. Furthermore, individual SIAs may have small impacts, and thus many may be required in addition to continued efforts to improve impact coverage, particularly in high-transmissibility settings where low-impact SIAs appear more likely to occur. The favorable safety and genetic stability of nOPV2 may enable more expansive and more frequent SIAs, which may be necessary to stop cVDPV2 outbreaks [4].

The work had several notable limitations. Owing to the observational nature of the study, unadjusted confounding may be present—for instance, if campaign quality varied systematically between nOPV2 SIAs and mOPV2 SIAs. Because we used a relatively simple SIR-type framework, we did not explicitly account for the impacts of population movement, variable risk and transmissibility among subpopulations, or secondary spread of vaccine. Likewise, we had to make assumptions about generation time and discretization of the time scale, though variation in these assumptions had no impact on the overall model conclusions.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. A. V. developed the statistical methodology, conducted the analysis, and prepared initial drafts of the manuscript. H. L. participated in study conceptualization and reviewed the methodology and manuscript. F. S., U. S. A., and C. K. oversaw field operations and data collection, provided interpretation of data and results, and reviewed the manuscript. T. E. managed and curated data used in the analysis. A. S. B. and S. O. participated in study conceptualization and overall guidance for research.

Financial support. This work was supported by the Bill & Melinda Gates Foundation.

Data availability. The detailed disease surveillance data on which this research is based are not available outside partners of the Global Polio Eradication Initiative.