National Immunization Campaigns With Oral Polio Vaccine May Reduce All-cause Mortality: An Analysis of 13 Years of Demographic Surveillance Data From an Urban African Area

Abstract

Background

Between 2002 and 2014, Guinea-Bissau had 17 national campaigns with oral polio vaccine (OPV) as well as campaigns with vitamin A supplementation (VAS), measles vaccine (MV), and H1N1 influenza vaccine. We examined the impact of these campaigns on child survival.

Methods

We examined the mortality rate between 1 day and 3 years of age of all children in the study area. We used Cox models with age as underlying time to calculate adjusted mortality rate ratios (MRRs) between “after-campaign” mortality and “before-campaign” mortality, adjusted for temporal change in mortality and stratified for season at risk.

Results

Mortality was lower after OPV-only campaigns than before, with an MRR for after-campaign vs before-campaign being 0.75 (95% confidence interval [CI], .67–.85). Other campaigns did not have similar effects, the MRR being 1.22 (95% CI, 1.04–1.44) for OPV + VAS campaigns, 1.39 (95% CI, 1.20–1.61) for VAS-only campaigns, 1.32 (95% CI, 1.09–1.60) for MV + VAS campaigns, and 1.13 (95% CI, .86–1.49) for the H1N1 campaign. Thus, all other campaigns differed significantly from the effect of OPV-only campaigns. Effects did not differ for trivalent, bivalent, or monovalent strains of OPV. With each additional campaign of OPV only, the mortality rate declined further (MRR, 0.86 [95% CI, .81–.92] per campaign). With follow-up to 3 years of age, the number needed to treat to save 1 life with the OPV-only campaign was 50 neonates.

Conclusions

OPV campaigns can have a much larger effect on child survival than otherwise assumed. Stopping OPV campaigns in low-income countries as part of the endgame for polio infection may increase child mortality.

Epidemiological and immunological studies have shown that vaccines have nonspecific effects (NSEs); that is, they not only prevent the vaccine-targeted disease but also reduce mortality from other infections. Vaccines apparently train the immune system in ways that reduce or enhance susceptibility to unrelated infections [1, 2]. The World Health Organization (WHO) recently reviewed the potential NSE of BCG, diphtheria-tetanus-pertussis (DTP), and measles (MV) vaccines [3, 4]. The review concluded that BCG and MV were associated with large reductions (45%) in overall mortality. Prevention of tuberculosis and measles infection did not explain these effects and the results thus suggested that vaccines have beneficial NSEs. The WHO’s Strategic Advisory Group of Experts on Immunization (SAGE) recommended further research into the NSEs of vaccines.

All live attenuated vaccines examined so far, including BCG, MV, and oral polio vaccine (OPV), have beneficial NSEs, also in randomized controlled trials (RCTs) [5–8]. The RCTs have found a 26%–38% reduction in mortality not ascribed to the vaccine-targeted diseases. Low-income countries have experienced numerous eradication campaigns with OPV and MV in the last 15–20 years. Removing vaccines after eradication could have negative consequences by depriving children of beneficial NSEs. Throughout this period, the Bandim Health Project in Guinea-Bissau has maintained a health and demographic surveillance system (HDSS) in an urban area with around 100 000 inhabitants. We therefore examined how OPV campaigns affected mortality among children <3 years of age. There has been no polio infection in Bissau since 1999 [9], so if campaigns reduced mortality, it would be evidence of NSEs of OPV. Since other campaigns could have NSEs, we also examined the campaigns with vitamin A supplementation (VAS), MV, and H1N1 vaccines.

METHODS

Study Setting

The Bandim Health Project has worked in Guinea-Bissau since 1978. The HDSS involves monthly registration of new pregnancies and deliveries through home visits. It also conducts daily registration of routine vaccinations at health centers, and visits to children <3 years of age every 3 months to register growth, morbidity, vaccinations, survival, and risk factors for child survival [5–8]. Details of study children are provided in Supplementary Table 1.

We previously examined how campaigns affected mortality within 7 RCTs conducted in 2002–2014 [10]. Five RCTs recruited at birth and followed them through infancy, while 2 enrolled older children and followed them to 3 years. In the present analysis, we have included all children aged 1 day to 3 years of age to see whether this provided similar results. We included all HDSS children. Hence, children were included in the analysis before they entered an RCT and after the end of the RCT. However, we also examined whether results were similar in trial and nontrial participants.

Vaccination Program and Campaigns

The routine program in Guinea-Bissau was BCG and OPV at birth. DTP and OPV were given at 6, 10, and 14 weeks and MV at 9 months of age. On average, the 46 219 children followed to 12 months of age had received 3.2 doses of OPV (interquartile range [IQR], 3–4). In 2008, DTP was replaced with pentavalent vaccine (DTP + hepatitis B vaccine + Haemophilus influenzae type B vaccine) and yellow fever vaccine was given with MV.

All vaccination and micronutrient campaigns have been listed in Supplementary Table 2. We distinguished between campaigns with OPV only, OPV + VAS, VAS only, MV + VAS, and H1N1 campaigns [10]. Since mebendazole was always administered with VAS, we have not estimated a separate effect of mebendazole campaigns.

Seventeen national OPV campaigns were conducted between 2002 and 2014. Until 2005, the OPV campaigns were organized with 2 doses given with an interval of 1 month; OPV was administered with VAS in the second round. No other vaccine or medication was administered at the same time. From 2010 the pattern changed; OPV was distributed together with VAS and mebendazole, often with a 6-month interval. In addition to trivalent OPV, monovalent and bivalent OPV were also used in this period. OPV was administered to all children <5 years of age, whereas VAS was only administered after 6 months and mebendazole after 12 months of age (Supplementary Table 2). Hence, all OPV campaigns were OPV-only campaigns for children <6 months of age even if VAS was administered to older children during the campaign. In recent years, coverage has been >90% when we assessed participation (Supplementary Table 2). The 29 973 children followed to 36 months of age had received on average 7.2 (IQR, 6–8) doses of routine and campaign OPV.

Guinea-Bissau had MV campaigns in 2006, 2009, and 2012. In 2006, MV was given starting from 6 months but only after 9 months of age in the 2009 and 2012 campaigns. All MV campaigns included VAS and mebendazole distribution, that is, MV + VAS campaigns. In 2010, H1N1 influenza vaccine was given to children aged 6 months to 5 years.

Statistical Methods

We did not estimate the campaign effect comparing participating and nonparticipating children. As coverage was high, we estimated an intention-to-treat-effect comparing mortality rates among children eligible for a campaign (denoted after-campaign mortality), to rates among children, who lived in Bandim before the campaign or were born after the campaign (denoted before-campaign mortality). Children could contribute time to both before-campaign and after-campaign periods.

To compare mortality rates, we calculated an age-adjusted mortality rate ratio (MRR) as a hazard ratio in Cox models with age as underlying time. We calculated the MRR for after-OPV campaign vs before-OPV campaign mortality. We inspected the proportionality assumption of the Cox models visually. NSEs often differ by sex [5, 6]; we therefore present results separately for females and males. We used data from all 17 OPV campaigns in the 2002–2014 period; OPV campaigns before January 2002 counted as prior exposure to OPV. In the 9 campaigns that also provided VAS for older children, children <6 months of age during the campaign were counted in the after-OPV-only group whereas older children were counted in the after-OPV + VAS group. Stata version 16 software was used in the statistical analyses.

Multivariable Analysis

The main methodological issue with splitting time before and after the campaigns is whether the overall temporal change in mortality or seasonal differences in mortality were allocated unequally to before-campaign and after-campaign periods. We used the data from all 13 years to construct 1 large data set to control for year and season at risk (dry: December–May; rainy: June–November). Subsequent tests using Schoenfeld residuals showed that these effects were age-dependent. We therefore stratified by season, and modeled separate year-trends (year × age group) in the age groups 0–5, 6–11, 12–23, and 24–35 months of age (finer age intervals gave almost identical results) [10].

The combined overall effects of OPV only, OPV + VAS, VAS only, MV + VAS, and H1N1 were obtained from a single model including the variables OPV only, OPV + VAS, VAS only, MV + VAS, H1N1, year × age group, and stratified by season. The sex-specific effects were obtained by applying the model separately for boys and girls. Test for interaction between sex and OPV only, OPV + VAS, VAS only, MV + VAS, and H1N1 were obtained from a single model including OPV only × sex, OPV + VAS × sex, VAS only × sex, MV + VAS × sex, H1N1 × sex, sex, year × age group, and stratified by season.

This analysis estimated the effect of campaign-type relative to the other campaigns. Whenever a new campaign was held, the current “last campaign” status was updated to the type of campaign the child was eligible for. However, we also estimated the MRR for the most recent campaign in relation to children who had received no campaign.

Number Needed to Treat

The number needed to treat (NNT) is calculated as 1 / [S_-OPV(t)^MRR − S_-OPV(t)] where S_-OPV(t) is the Kaplan-Meier estimate in the before-OPV campaign group and MRR is the mortality rate ratio for OPV-only campaign obtained from the multivariable analysis [11].

RESULTS

With 2834 deaths and 100 594 person-years (PY), the present study was considerably larger than the previous analysis of RCTs, which had 1244 deaths and 33 822 PY [10]; 1078 deaths from the RCTs were from the study area. Since the RCTs mainly recruited infants, mortality was higher in the previous analysis (36.8/1000 PY) than in the study of all children <3 years of age (28.2/1000 PY). There was a marked decline in under-3 mortality over the study period, from 117 per 1000 in 2002 to 40 per 1000 in 2014 (first-day deaths were excluded) (Figure 1).

OPV-Only Campaigns

Mortality was reduced after OPV-only campaigns, the adjusted MRR for after-campaign vs before-campaign being 0.75 (95% confidence interval [CI], .67–.85) (Table 1). Effects were similar for infants and older children (Table 1), and for boys (MRR, 0.73 [95% CI, .62–.86]) and girls (MRR, 0.77 [95% CI, .65–.92]) (Supplementary Table 3). When we censored children taking part in the RCTs from the time of enrollment, the MRR for OPV-only campaign was 0.71 (95% CI, .60–.83) (Table 2).

We examined whether repeated doses of OPV-only campaign increased the beneficial effect. In the combined analysis adjusted for other campaigns, season, and calendar year, each additional OPV-only campaign reduced mortality by 14% (MRR, 0.86 [95% CI, .81–.92]) (Table 3). Nothing similar was seen for the other campaigns.

When we only looked at the most recent campaign compared to having had no campaign yet, OPV-only campaign was associated with an MRR of 0.85 [95% CI, .73–.98) (Table 4). Monovalent, bivalent, or trivalent OPV strains did not differ significantly (Supplementary Table 4).

Though OPV + VAS campaigns were associated with a negative MRR (Table 1), the effect of all OPV campaigns, both OPV only and OPV + VAS, was clearly beneficial, the MRR being 0.81 (95% CI, .73–.91) (Supplementary Table 5). Boosting with any OPV campaign was also beneficial (MRR, 0.92 [95% CI, .87–.97]) (data not shown).

Other Campaigns

Compared with OPV only, other campaigns had negative effects on overall survival (Table 1) and their effects differed significantly from the effect of OPV-only campaigns. The pattern was similar when we looked at the most recent campaign vs no campaign (Table 4). The negative effects of other campaigns were similar for girls and boys, except that OPV + VAS was associated with a significantly stronger negative effect for girls than for boys (P = .006, test of interaction) (Supplementary Table 3).

Number Needed to Treat

With follow-up to 3 years of age, it was necessary to give OPV-only campaign to 50 neonates to save 1 life.

DISCUSSION

Though there was no polio in Bissau, OPV-only campaigns were associated with a marked reduction in mortality for children <3 years of age. Effects increased with additional doses of OPV. The number of children who should receive the OPV-only campaign to save 1 life was very low (NNT = 50). Other campaigns did not have similar beneficial effects.

Strengths and Weaknesses

The study area experienced a 66% reduction in mortality for children <3 years of age in the period from 2002 to 2014. It is reassuring that we found the same effect of campaign OPV in this much larger study as in the previous study of children enrolled in RCTs [10]. There was no strong difference in the OPV-only effect for the children who took part in the RCTs conducted during this period and those who did not. The RCT analysis [10] mainly included infants. Having older children up to 3 years of age included in the present analysis supports that OPV may indeed have been a major driver in the mortality decline in the study area.

There are several reasons why this observation is unlikely to be due to confounding or a general beneficial intervention effect. We included all children and controlled for age, season, and temporal change in the analysis. We previously conducted a simulation study of campaigns in the period 2002–2014 and found that the expected effect of randomly distributed campaigns was an MRR of 1.00 (95% CI, .99–1.01) for after-campaign vs before-campaign mortality [10]. There was no change in nutritional status measured by mid-upper-arm circumference, which could explain the mortality decline. There was no major change in tuberculosis or human immunodeficiency virus incidence. The decline in malaria started before 2002 due to distribution of bed nets, and in 2010–2011 there was a malaria epidemic [12] at the same time as mortality declined (Figure 1). It also seems highly unlikely that the OPV campaign effect should be due to increased attention to health when other campaigns did not show similar beneficial effects.

Previous studies of live vaccines, including MV, BCG, and smallpox vaccine, have suggested that the beneficial effects were strongest for females. That was not the case for OPV; in this as well as the previous study of OPV campaigns [10], the MRR was slightly better for males. Since we measured overall changes also influenced by other campaigns, it is not surprising that the difference was not statistically significant. In the RCT of OPV at birth, the beneficial effect was also largest for males [6].

Since there has been no polio in Guinea-Bissau since the 1990s [9], the timing of OPV campaigns was mainly determined by outbreaks of polio in other parts of West Africa, mainly Nigeria. There was no OPV campaign in 2003 to the beginning of 2004 and from 2006 to 2009; mortality did not decline in 2003–2004 or in 2007–2008. The decline in 2006 may have been related to the OPV campaign in December 2005.

Consistency or Contradiction With Previous Observations on OPV

Though there are several studies of changes in child mortality when routine MV or MV campaigns were introduced in low-income countries [13], there has been virtually no study of the introduction of OPV or the effect of OPV campaigns. When the first OPV campaigns were conducted in Guinea-Bissau in the spring of 1998, we compared child survival for participants and nonparticipants [14]. OPV had significant beneficial effects for the youngest children, but that could have been due to uncontrolled confounding when comparing participants and nonparticipants.

Subsequently, we conducted 2 RCTs of OPV at birth (OPV0) in Bissau; both suggested 32% lower infant mortality for children who received OPV0, with 1 RCT being large enough to be significant in its own right [6, 15]. In RCTs of OPV vs inactivated polio vaccine (IPV), OPV has been associated with less otitis media in Finland [16] and fewer bacterial diarrheas in Bangladesh [17]. No study has shown IPV as having a better effect for health than OPV. Several natural experiments and register-based studies have also suggested beneficial NSEs of OPV in both low- and high-income countries [18, 19].

In 1 observational study from 2004 covering several periods where OPV was missing in the country, OPV0 was associated with increased mortality compared to not having received OPV0 (no OPV0) for males [20]. However, this was a spurious observation due to unequal exposure to OPV campaigns for children who had received OPV0 and no OPV0, respectively [19]. In a reanalysis of the 2004 study, the age-adjusted mortality rate of study participants was 67% lower after the OPV campaigns than before the campaigns. In the OPV0 group, only 22% of follow-up time was “after” the OPV campaigns whereas 55% of the time in the no-OPV0 group was postcampaign. Hence, censoring for OPV campaigns in the original study removed male excess mortality [19]. Clearly, if the effect of OPV campaigns is not taken into consideration, comparison of other groups may change dramatically.

In the present study, the beneficial effect increased significantly with subsequent doses of OPV. Consistent with this, we have for all live vaccines found that boosting enhanced the beneficial NSEs [21].

Surprisingly, campaign OPV + VAS was associated with a negative effect (Table 1). However, VAS-only campaign had a strong negative effect and after-campaign OPV + VAS periods were often compared with before periods with OPV only. Compared with periods where the children had not yet received campaigns, campaign OPV + VAS was not associated with higher mortality (Table 4).

Immunological Mechanisms

Other live vaccines, including BCG, smallpox vaccine, and live pertussis vaccine have been found to induce innate immune training effects that reduce susceptibility to unrelated infections in both animals and humans [1, 2, 22, 23]. BCG reprograms monocytes through epigenetic changes to a more proinflammatory response; in animal models this response reduces mortality from challenge to unrelated infections [2], and in human experiments randomization to BCG reduced the viral load after subsequent challenge with yellow fever vaccine [23]. Earlier studies have explained the beneficial effects of OPV as related to interferon production [24]. Newer studies of innate immune training after OPV vaccination have unfortunately not been done and may be difficult because of WHO restrictions on storing polio and OPV samples.

CONCLUSIONS

The world has experienced an unprecedented decline in child mortality in recent years; in Bissau under-3 mortality declined to one-third (Figure 1). The decline has apparently been much stronger in the 2000s than in the 1990s [25]; for example, in Guinea-Bissau there was no change in mortality in the 1990s but the Millennium Development Goal 4 (MDG4) of reducing child mortality by two-thirds was reached between 2002 and 2014 in the study area. The present study suggests that the repeated campaigns with OPV may have played an important part in reaching MDG4 (Supplementary Table 2 and Figure 1).

As part of the endgame for polio eradication, it is planned to phase out OPV over the next few years and gradually introduce IPV in the vaccination schedule in low-income countries. Trivalent OPV was stopped globally in April 2016 by SAGE. No study has shown that IPV has beneficial NSEs [16, 17], and in a meta-analysis of trials in which children had been randomized to IPV as a comparator vaccine, we found that IPV-vaccinated girls had significantly higher mortality than IPV-vaccinated boys [26].

Studies of the introduction of DTP and OPV have also shown that the effect of DTP was worse when DTP was given alone [27]. OPV apparently reduced the negative effect of DTP [26]. In the future there may be no “attenuation” of DTP by OPV. Taken together, the evidence worryingly suggests that replacing OPV with IPV could lead to an increase in mortality in low-income countries.

The potential benefits of OPV should be examined before OPV is stopped globally, either phased out or replaced by IPV. First, it should be possible to conduct cluster-randomized trials of the overall effect of OPV campaigns on child survival. Second, OPV0 is often not used and since we have found OPV0 to be associated with better infant survival, particularly when given in the first days of life [6], it should be possible to conduct RCTs of OPV0 to confirm this finding. Third, trials should compare the impact of the OPV schedule to an IPV-based schedule on survival [16, 17]. Fourth, due to the beneficial NSEs of OPV, testing whether OPV induces partial protection against coronavirus disease 2019 has been suggested [24].

There is increasing evidence that removing a vaccine once the disease is eliminated may not be the best guide to public health [28, 29]. All the studies conducted after the eradication of smallpox and subsequent removal of smallpox vaccine have suggested that smallpox vaccination had beneficial immune training effects unrelated to prevention of smallpox infection [30–32]. Hopefully, we are not about to commit a similar error with the eradication of polio and removal of OPV.

Supplementary Data

Supplementary materials are available at Clinical 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. A. and S. N. conducted the statistical analyses. A. B. F. organized data collection in connection with the campaigns. A. B. F., A. R., C. S. B., and P. A. supervised the individual trials and maintained the demographic surveillance system. The first draft was written by P. A.; all authors contributed to the final version of the manuscript. P. A. and A. A. act as guarantors of the study.

Acknowledgments. The Bandim Health Project’s health and demographic surveillance site was established in 1978 at the request of the Guinean Ministry of Health. All intervention studies are presented for approval to the Guinean National Ethical Committee and for consultative approval to the Danish Central Ethical Committee.

Disclaimer. No funding body had any role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Financial support. The work on nonspecific effects of vaccines has been supported by the Danish Council for Development Research, Ministry of Foreign Affairs, Denmark (grant number 104.Dan.8.f.); the Novo Nordisk Foundation and European Union FP7 support for OPTIMUNISE (grant number Health-F-3-2011-261375). C. S. B. held a starting grant from the European Research Council (ERC-2009-StG-243149). Research Centre for Vitamins and Vaccines is supported by a grant from the Danish National Research Foundation (grant number DNRF108). P. A. held a research professorship grant from the Novo Nordisk Foundation.

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.

References