Statistical analysis

The target sample size was 700 infants, with an enrolment target of 888. We assumed 90% of infants would achieve a type 2 immune response with two doses of mOPV2 in the standard 4 week group.¹⁶ We estimated a sample size of 175 participants per arm to have 90% power to detect non-inferiority in the type 2 immune response induced by two mOPV2 doses given at 1 week or 2 week intervals versus the standard 4 week interval with a continuity corrected Z test with pooled variance. This assumes a non-inferiority margin of 10% and p=0·05 (one-tailed). The enrolment target was increased to 222, assuming 10% of enrolled infants would have baseline titres too high to detect an immune response (seroconversion or boosting) and 10% attrition. The type 2 immune response in the 1 week (11 weeks of age) or 2 week groups (12 weeks of age) was deemed non-inferior to that in the standard 4 week study group (14 weeks of age) if the lower bound of the 90% CI of the differences in immune responses, calculated with a two-sided Wald test, was greater than −10%. We chose a priori to set the type 1 error at 5%; therefore, we used a 90% two-sided CI. We selected non-inferiority to compare short and standard interval schedules because the potential benefit of a faster increase in population immunity achieved with a short-interval schedule would offset a reduced, but not programmatically meaningful decline in type 2 immune response. The margin of non-inferiority was based on a probable and acceptable public health decline in immune response in lieu of quicker programmatic action. Distributions of antibody titres were compared by use of the Kruskal-Wallis test. Antibody titres were plotted as reverse cumulative distribution curves, which were constructed by representing on the vertical axis the proportion of participants with antibody titres equal to or greater than that represented on the horizontal axis.

We assumed 98% of infants would achieve a type 2 immune response with two doses of mOPV2 and one IPV dose (4 week plus IPV group). No previous studies have assessed type 2 immunogenicity of simultaneous administration of mOPV2 and IPV. A sample size of 175 infants in the 4 week plus IPV group (enrolment target of 222) would provide 80% power through use of a two-sided test with a significance of p=0·05 to compare with the 4 week group with continuity corrected Z test with pooled variance. We used Fisher’s exact test to compare the proportion of participants showing type 2 immune response in the 4 week group and 4 week plus IPV group after receiving two doses (14 weeks of age) or one dose (10 weeks of age) of mOPV2. We also used the Kruskal-Wallis test to compare distributions of antibody titre.

Several post-hoc analyses were done to further investigate initial findings. We assessed the benefit of an additional dose of mOPV2 by comparing type 2 immune response and titres 4 weeks after the first and second dose of mOPV2 (10 weeks and 14 weeks of age, respectively) within participants in the 4 week group. McNemar’s test was used for within-participant comparisons of proportions and the signed rank test for antibody titre distributions. We also assessed whether IPV co-administration affected type 2 immune response in a subset of participants who were seronegative at baseline by comparing immune response and titres 4 weeks after one and two doses of mOPV2 (10 weeks and 14 weeks of age, respectively) in the 4 week group and 4 week plus IPV group. Fisher’s exact test was used for comparisons of proportions of patients between study groups and the Kruskal-Wallis test for comparison of distributions of antibody titres. Results of post-hoc analyses were assessed by use of a Bonferroni corrected significance level of less than 0·01; not all post-hoc analyses are presented. We did not apply Bonferroni correction for a-priori hypotheses.

Baseline analyses included infants who completed all study visits and provided samples (intention to treat). Participants with adjusted baseline titres too high to detect a four times increase in a poliovirus serotype were excluded from further analysis for that serotype (modified intention to treat). Except for baseline findings, results are presented for infants who completed the study per modified intention to treat; results from per-protocol analyses were similar (not reported here). All analyses were done with SAS version 9.3¹⁷ and graphics were created with R version 3.3.3.

This trial is registered at ClinicalTrials.gov, number {"type":"clinical-trial","attrs":{"text":"NCT02643368","term_id":"NCT02643368"}}NCT02643368, and is closed to accrual.