To monitor the safety of rotavirus vaccines in use in national immunization programs, intussusception surveillance was established at sentinel hospitals in seven countries that were early adopters of RV1 in sub-Saharan Africa (Ethiopia, Ghana, Kenya, Malawi, Tanzania, Zambia, and Zimbabwe). Countries joined the network on a rolling basis following introduction of RV1 into their routine childhood immunization programs. Surveillance began in the first country in February 2012 and ended in all countries in December 2016. Participating countries used a common surveillance protocol to allow for pooling of data across sites and countries. Site investigators conducted active surveillance to identify cases of intussusception at major pediatric hospitals located in large urban areas of the participating countries. Children <12 months of age meeting the Brighton Collaboration criteria for level 1 of diagnostic certainty for intussusception were enrolled, regardless of RV1 vaccination status. Level 1 of diagnostic certainty requires confirmation of intussusception during surgery or by specific radiological findings if reduction occurred by enema or at autopsy.¹⁹ Limited clinical and sociodemographic data were collected from interviews with parents and by review of the child’s medical record. Onset of intussusception was defined as the date of first reported symptoms by the parent or guardian. Vaccination status and dates were obtained from vaccination cards brought by the parent or guardian to the hospital or, if the vaccine card was unavailable, by visiting the child’s home or clinic where the child was vaccinated. For most cases, a photocopy or photograph of the vaccine card or clinic record was made for future reference and for confirmation of vaccination status. In all participating countries, two doses of RV1 were recommended to be given at the first two Expanded Program on Immunization (EPI) visits at 6 and 10 weeks of age along with the other EPI vaccines including oral polio. Additional doses of rotavirus vaccine were contraindicated if a child experienced an episode of intussusception prior to completion of vaccine series. For intussusception case-patients who were diagnosed prior to 8 months of age, efforts were made to re-contact their families when the child reached 8 months of age to determine if the child had received any additional doses of rotavirus vaccine or had a recurrent episode of intussusception and to assess the vital status of the child.

A total of 1060 children <12 months of age with intussusception were enrolled in surveillance from 29 sentinel hospitals located in 7 countries. (Table 1) Of these, 239 children were excluded from the analysis because they were <28 days or >245 days at time of intussusception symptom onset and an additional 104 children were excluded because their vaccination status could not be confirmed, resulting in 717 children included in the analysis. Ghana, the first country in the network to introduce rotavirus vaccine, contributed the greatest number of cases (258; 36%) to the analysis.

The median age of children included in the analysis was 25 weeks with few cases of intussusception detected in very young infants 4–11 weeks of age (Figure 1); 61% (436/717) were male. Among those for whom information was available, only 2% (10/664) of children had never received any breastmilk prior to intussusception onset, 68% (438/644) lived in a household where at least one person was employed, 73% (455/627) lived in a household that had electricity at least part of the time, and 80% (522/650) live in a household with a mobile phone. The median duration between symptom onset and admission to a surveillance facility was 3 days (interquartile range (IQR): 1–4 days) with 78% (420/537) of children first seeking care at another facility. The majority of the children were treated surgically (87%; 615/704) with 57% (353/615) of these children requiring resection of the bowel; 13% (89/704) were treated by enema. Overall, 12% (80/681) of intussusception cases died.

Vaccination coverage was high and receipt was timely. (Figure 1) Ten percent of the intussusception cases included in the analysis were unvaccinated, 6% received only one dose of RV1, and 84% received two doses of RV1. The median age at dose 1 was 6 weeks (IQR: 6–7 weeks) and at dose 2 was 11 weeks (IQR: 10–12 weeks). Despite the contraindication, 5 children received at least one dose of RV1 following intussusception. Of the 445 (69%) surviving children re-contacted at 8 months of age, 14 (3%) children had experienced a second episode of intussusception and 7 (2%) children died after hospital discharge and before reaching 8 months of age.

No clustering of cases occurred in any of the risk windows (1–7 days, 8–21 days, or 1–21 days) following receipt of either dose of RV1. (Figure 2) One case was observed in the 1–7 days and 6 cases in the 8–21 days post dose 1. Five cases and 16 cases were observed in the 1–7 days and 8–21 days, respectively, post dose 2. (Table 2) No elevated risk of intussusception was detected in the 1–7 days post RV1 dose 1 (relative incidence (RI), 0.25; 95% confidence interval (CI), <0.001–1.16) or in the 1–7 days post RV1 dose 2 (1–7 days: RI, 0.76; 95% CI, 0.17–1.70). Similarly, no elevated risk was detected in the 8–21 or 1–21 days post dose 1 or post dose 2. (Table 2)

Unlike previous data from high and upper-middle income countries, we did not find an increased risk of intussusception associated with RV1 in low and low-middle income, sub-Saharan African countries. We hypothesize several possible explanations for this difference in risk by setting. First, although the exact mechanism is not known, intussusception is possibly related to intestinal replication of the orally administered, live vaccine rotavirus strain. Because oral rotavirus vaccines are less efficacious and shedding of vaccine virus, a potential marker of vaccine replication, is less frequently detected in low-income countries compared with high- and middle-income countries^15,17, rotavirus vaccines might also be associated with a lower intussusception risk in low-income countries. Second, rotavirus vaccine is co-administered with oral polio vaccine in low-income countries and the first dose of oral polio vaccine, which is associated with the greatest replication of the vaccine poliovirus, has been shown to decrease the immunogenicity of the first dose of RV1 when co-administered.²³ This phenomenon was hypothesized as a potential reason why no increased risk of intussusception was seen after the first dose of RV1 in Brazil, the other country where no such association was observed post dose 1.¹⁰ However, a low level association between RV1 and intussusception was observed after the second dose in Brazil.¹⁰ Third, the two doses of RV1 were administered at a young age (6 and 10 weeks) in African countries compared with older age schedules (generally 2 and 4 months) in high and middle come countries. Because intussusception is uncommon in the first 2 months of life, RV1 administration at these young ages might not be associated with intussusception in African countries, especially if the causes of intussusception are different in younger infants compared with older infants. Finally, other factors that may play a role in the risk of intussusception in younger infants and that are different in these low-income African countries compared with high- and middle-income countries (e.g. diet, breastfeeding practices, microbiome, or levels of maternal antibodies) might also partially explain the differences in the risk of intussusception following RV1 vaccination via unknown mechanisms.

Given the higher background rate of intussusception in older children, the initial WHO recommendations issued in 2009 specified that the rotavirus vaccine series be initiated by 15 weeks of age to avoid amplifying any potential vaccine-associated risk of intussuception.²⁴ However, recognizing that many children could be excluded from receiving rotavirus vaccine under these age restrictions, especially in some low-income, high burden countries where delays in vaccination are more common, these age restrictions were reviewed. A modeling study using vaccine efficacy data from clinical trials and the available data on risk of intussusception following vaccination from post-licensure evaluations in middle- and high-income countries found that removing these age restrictions would avert 154 rotavirus deaths for every intussusception death related to the vaccine in low-income countries.²⁵ Based on these data, in January 2013, WHO recommended removal of age restrictions for rotavirus vaccines to improve vaccine coverage. The timely administration of RV1 in the early vaccine introducing countries included in this analysis, with 3% of children receiving RV1 after 15 weeks of age, may limit the generalizability to countries with significant delays in vaccine administration. If the dynamics of RV1 replication and immune response are different (e.g. greater) in older children, the risk of intussusception could be different.

Our study had some limitations. First, the lack of suspicion of intussusception and delays in seeking health care could have resulted in some children dying from intussusception prior to reaching a surveillance facility. While this would have reduced the number of cases identified, these delays would likely have been independent of vaccination status and therefore would not likely have biased our results. However, if intussusception cases that did not present for treatment were more likely be vaccinated late, then our results may not be generalizable to those children. The age distribution of our cases was similar to that seen in the United States and other countries with good access to care^14,26 with few cases occurring in the first 12 weeks of life indicating that we were not selectively missing cases in younger children during the period when vaccine doses are given. Exact reasons for lower rates of intussusception in younger children are not well understood but may be related to the decline of maternal antibodies to pathogens associated with intussusception or to age-related milestones in intestinal lymphoid tissue maturation. If we did selectively miss cases among younger children, then we may have underestimated the true risk of intussusception following vaccination. Second, we used date of first reported symptom by the parent as the date of intussusception onset rather than date of hospital admission as has been used in many of the previous analyses. We chose to use the symptom onset date because there was a median of 3 days between symptom onset and admission to the surveillance facility. If admission date was used as the marker for the date that intussusception occurred, we were concerned that some children would be excluded from the 1–7 day risk window. If we used date of hospital admission as the onset date for the analysis or a longer risk window, we also did not see an increased risk of intussusception. Third, children frequently travelled long distances to reach the surveillance facilities and often did not bring their vaccine cards with them. This necessitated great effort by surveillance staff to find ways to confirm the vaccination status of the enrolled children. Overall, we were able to confirm the vaccination status of 87% of age-eligible children and there was no difference in the age distribution of children for whom we confirmed the vaccination status and those for whom we did not. Finally, the presentation and treatment of intussusception in the countries this evaluation differs from that observed in other regions of the world in terms of delays in presentation, surgery rates, and mortality.¹⁴ While better understanding these data are important for better treatment and outcomes for intussusception patients, such analyses were outside the scope of the current evaluation.

The self-controlled case-series approach provided an efficient method that can be applied in resource-limited settings to evaluate the risk of intussusception following rotavirus vaccination. As intussusception is a rare adverse event and a large sample size is required to assess the risk of intussusception following rotavirus vaccination, such evaluations may not be feasible and practical in all countries introducing rotavirus vaccine. As this vaccine or other rotavirus vaccines in the pipeline are introduced into other regions such as Asia where evaluations of rotavirus vaccine and intussusception have not been performed and given the regional differences in the epidemiology of intussusception, similar evaluations will be important to assess the regional benefits and risks of rotavirus vaccination. In the 29 African countries that had introduced rotavirus vaccine into their national immunization program by the end of 2014, ~135,000 rotavirus hospitalizations and 21,000 rotavirus deaths were estimated to be prevented in 2016.²⁷ Given these large health benefits, the lack of increased risk of intussusception following RV1 administration in our study is reassuring.