Self-controlled risk interval study of rotavirus vaccine safety: Findings and implications
International Journal of Risk & Safety in Medicine, vol. 35, no. 3, pp. 207-215, 2024 DOI: 10.3233/JRS-230049
Original Article Drug Safety
Self-controlled Risk Interval Study of Rotavirus Vaccine Safety: Findings and Implications
Jacob Puliyel MD MRCP MPhil
Visiting Professor
International Institute of Health Management Research
Delhi India
Orchid Number
0000-0002-6303-000X
Puliyel@gmail.com
Phone +91 9811725193
Brian S. Hooker, PhD
Emeritus Professor of Biology
Simpson University
Redding, California, USA
Chief Scientific Officer
Children's Health Defence
New Jersey, USA
drbrianhooker@gmail.com
Phone 509-366-2269
Short Title: Re-evaluation ‘Rotavac’ Safety Using SCRI
Address for correspondence
Jacob Puliyel, 37 Panchavatti, Cochin 689019 India puliyel@gmail.com
?
Abstract
BACKGROUND
The self-controlled case series (SCCS) is often used to monitor vaccine safety. The evaluation of intussusception after the rotavirus vaccine is complicated because the baseline rate varies with age. Time-varying baseline risk adjustments with data from unexposed cohorts are utilised. Self-controlled risk interval (SCRI), with a shorter observation period, can also mitigate the problem by studying a control period close to the risk period.
An Indian rotavirus vaccine was studied using SCCS. The risk of intussusception in the high-risk windows (21 days after vaccination) was comparable to the background risk.
OBJECTIVE
To re-analyse data of an existing SCCS study using alternate statistical methods to examine vaccine safety.
METHODS
We examined the mean age of intussusception in the vaccinated and the unvaccinated.
We did an SCRI analysis of the surveillance data from the SCCS study, limiting the observation period to 180 days.
We analysed the time-to-intussusception from the last vaccination.
Finally, we performed an SCCS analysis, excluding unvaccinated cases from the analysis.
RESULTS
We found the mean age of intussusception was significantly lower in the vaccinated (205 days) compared to the unvaccinated (223 days) (p-value 0.0026). The Incident Risk Ratio (IRR) on SCRI analysis was 1.62 (95% CI 1.07-2.44). There were significantly more intussusceptions in the first 30 days after vaccination compared to the next 30-day window. (92 vs 63 p-value = 0.009). We found that excluding unvaccinated infants from the SCCS analysis demonstrated significantly increased risk for the risk period 1-21 days after the 3rd dose (2.47 (CI 1.70-3.59). The risks of intussusception were missed in traditional SCCS analysis using unvaccinated infants as controls.
CONCLUSION
Traditional risk adjustments using data from unexposed cohorts in SCCS may not be appropriate for investigating the risk of intussusception where vaccination lowers the mean age of intussusception.
Key Message
Age-dependent risk adjustments in SCCS, utilising data from unexposed cohorts are not appropriate for detecting intussusception-risk with rotavirus vaccine because vaccination lowers the mean age of intussusception.
Declarations
Funding: No funding was received for this research.
Conflicts of Interest Statement: The authors have no conflicts of interest to declare.
Availability of data and material: The study data is being supplied in a supplementary file
Ethics / IRB Statement This study uses the anonymised sentinel surveillance data as part of the public record with a public authority. The study was originally approved by the institutional ethics committees of surveillance hospitals.
Contributions of authors. JP and BSH contributed to the conception and design of this study; JP drafted the manuscript. Both authors read and approved the final manuscript.
Self-controlled Risk Interval Study of Rotavirus Vaccine Safety: Findings and Implications
Background
Randomised control trials (RCTs) done before the licensure of vaccines are often not powered to detect rare adverse events. Post-licensure monitoring of safety is crucial for this. Databases of people with their immunization histories and medical outcomes may be used to look for adverse events among the vaccinated compared to the unvaccinated. However, vaccine coverage is often high and matched unvaccinated controls may be difficult to find.
The self-controlled case series (SCCS) can detect the temporal association between vaccination and adverse events [1]. The incidence of the adverse event within a period where the risk is assumed to be high is compared with the incidence during other periods. Individuals act as their own controls for all fixed covariates (which do not vary with time) in this epidemiological analysis. The method performs less well where covariates change with time like the age of young children. Time-varying baseline risk adjustments have to be made in such instances, often using data from unexposed cohorts. Carlin describes how adjustments can be made for age [2]. SCCS has been used in several studies looking at intussusceptions after the deployment of rotavirus vaccines [3-9}.
Intussusceptions are known to be age-dependent [10] with low incidence in the first 6 months of life when the vaccine is administered. The incidence of naturally occurring intussusception (intussusception in the unvaccinated) is higher in the second 6 months of life. Age-adjusted SCCS, using unvaccinated infants in the control group, have been used under such circumstances. However, if the act of vaccination lowers the mean age of intussusception, it may be inappropriate to use unexposed babies as controls.
We reanalyse data from a published SCCS study [3] of rotavirus vaccination which had reported that the risk of intussusception 1-21 days after any dose was not higher than the background risk. We use four alternative analyses to examine if the mean age of intussusception in the vaccinated is different from the unvaccinated and if the conclusions drawn on SCCS analysis are valid.
Material and Methods
A SCCS study of an Indian live attenuated rotavirus vaccine (116E) (G9P[11]) Rotavac (Bharat Biotech) has been reported [3]. Rotavirus vaccination status was ascertained from the vaccination record. Among the 589 infants with intussusception selected for the SCCS, 212 had not received the vaccine and 377 received at least one dose of the vaccine. 33 had received only one dose of the vaccine, 55 received 2 doses and 289 received all three doses. Although contraindicated, 22 children received another dose of rotavirus vaccine after intussusception. There were 4 cases of intussusception in the risk window of 21 days after the first dose of the vaccine, 19 cases after the second dose and 37 cases in the 21-day risk period after the third dose. We were provided with the anonymised raw data on all 589 cases of Brighton criteria Level 1 intussusception (radiological or surgical diagnosis of intussusception). 377 of them had received the study rotavirus vaccine in the sentinel surveillance. This data was provided as part of the public record available with a public authority. The data was transferred to an Excel sheet by the authors and is being provided as a supplementary file.
Analysis 1. In our preliminary analysis, we examined the age at which intussusception developed in the 377 vaccinated babies compared to the 212 who had not received the rotavirus vaccine. The mean difference and 95% confidence interval of the difference were calculated.
Analysis 2. In the second analysis, we used the self-controlled risk interval (SCRI) design [11]. In this analysis, to minimise confounding by age, a shortened control period close to the risk period is studied. We selected to limit our analysis to 6 months because intussusceptions peak around 7 months in the unvaccinated [10].
The first 21 days after any dose of the vaccine was the high-risk period studied. The control period was the period of observation from 29 days to 180 days, excluding the high-risk period. SCRI analysis has been used in the study of rotavirus in the USA previously [12]. The data was analysed using the statistical software package STATA developed by StataCorp LLC (4905 Lakeway Drive College Station, Texas).
Analysis 3. Vaccination is contraindicated after intussusception, so intussusceptions can be attributed to the last vaccination or otherwise to the background risk of intussusception. In the third analysis, we examined the time-to-intussusception from the last dose of vaccine received. We compared the frequency of intussusception in 2 time windows, 0 to 30 days and 31 to 60 days after the last rotavirus vaccination. We did this analysis to detect a temporal association with vaccination. We assumed that if intussusceptions were unrelated to vaccination there would be no temporal association to vaccination.
Analysis 4. If vaccination reduces the mean age of intussusception significantly it would be inappropriate to use unvaccinated infants as controls in SCCS analysis. We performed an SCCS analysis including only the 377 vaccinated infants. The observation period was from day 30 to day 365. The high-risk period was 21 days after any vaccination dose. The risk period i.e. 1-7 days, 8-21 days, and 1-21 days after each dose of vaccine was studied. The period of observation outside the high-risk period was considered the control period. We considered age groups as less than and greater than 6 months separately in the Poisson regression model, because intussusceptions are generally expected beyond 6 months [10]. The relative incidences (RIs) were calculated on the statistical software STATA
We looked at 95% confidence intervals and p-values to study statistical significance. A p-value less than 0.05 was considered statistically significant.
Results
Age of Intussusception
The mean age of intussusception was 205 days in the 377 vaccinated babies compared to 223 days in the 212 unvaccinated. This was statistically significant (p-value 0.002). The results are tabulated in Table 1. Intussusception was seen at a younger age in the vaccinated cohort than the unvaccinated.
SCRI Analysis
In the SCRI analysis, we limited the period of observation to age 6 months (180 days) starting at 29 days. 149 individuals developed intussusception during the observation period. There were 59 cases in the high-risk window and 90 cases in the low-risk window. The Incidence Risk Ratio was 1.62 (95% CI 1.07-2.44). The data is tabulated in Table 2
Time to Intussusception from Vaccination
Figure 1 is a histogram showing the number of intussusception cases against time elapsed from the last rotavirus vaccination, shown at 30-day intervals. The histogram shows that the incidence of cases tapers off with time. Table 3 compares the number of cases of intussusception in the first two time frames of 30 days. 92 cases of intussusception occurred within 30 days of the last dose of the rotavirus vaccine whereas there were only 63 cases in the second interval between 31 and 60 days (p=0.009).
SCCS Excluding the Unexposed
Of the 377 infants who developed intussusception 60 developed it during the high-risk period of 21 days after rotavirus vaccination (any dose). Four developed it after the first dose, 19 after the second dose and 37 after the 3rd dose. The analysis showed a significantly increased risk after the 3rd dose both in the risk window 1-7 days (IRR 3.15 95% CI 1.90-5.22) and the risk window 8-21 days (IRR 2.12: 1.32-3.40). The data are depicted in Table 4. Traditional SCCS (3) using unvaccinated controls failed to detect this risk.
Discussion
Our analysis found that intussusception occurred early in the vaccinated and there were significantly more intussusceptions (1.62-fold increase) within 3 weeks of rotavirus vaccination which was not detected using SCCS analysis [3].
Age of Intussusception.
There was a significant lowering of the mean age intussusception among the vaccinated. If vaccination is merely a coincidental event and it does not disturb the normal incidence of intussusceptions in the population, the mean age of intussusception would be the same in the vaccinated and the unvaccinated. Tate et al have noted that there was an increase in intussusception in the USA among younger (8 to 11-week-old) infants after the introduction of rotavirus vaccines compared to the era before vaccines [13]. We discuss later, how this lowering of the mean age of intussusception may have blunted the safety signal in the SCCS analysis.
SCRI Analysis
Limiting our observation to the first 6 months of life showed that the risk of intussusception with the rotavirus vaccine was significantly higher in the high-risk period. The selection of risk periods is based on intelligent guesswork and it is not uncommon to examine multiple risk periods to cover different possible risk windows [14]. The susceptibility to adverse events need not be highest immediately after vaccination, especially if reactions manifest after a lag period. For vaccines administered repeatedly at monthly intervals, the high-risk period of two doses may overlap. For comparison with the previous study (3) we considered 21 days after a dose of the vaccine as the high-risk window. For the SCRI we selected 21 days after any dose as the high-risk period.
The statistical power of an SCRI analysis is less than that of an SCCS study because of the reduced number of events in the shorter control duration [11]. Despite this lowered power, a statistically significant increased risk of intussusception from the vaccine was seen. This increased risk of intussusception went unnoticed in the published SCCS study where the observation period was between 28 and 365 days [3].
The incidence of naturally occurring (non-vaccine induced) intussusceptions increases from 5 to 6 months and peaks at around 7 months [10]. Even using the SCRI design, the risk of vaccine-induced intussusception would be underestimated because the baseline risk of natural intussusception is comparatively low in the 21-day risk periods, which ends by 4.5 months in most children (- those taking the doses on the schedule advised, complete their vaccinations by 14 weeks), compared to the control period between 5 and 6 months.
Time to Intussusception from Vaccination
We looked at the time that had elapsed between the last dose of vaccine administered and intussusception. We found significantly more intussusceptions in the period proximal to vaccination when we chose 30-day intervals. This analysis, where a high-risk period is compared to a period with a lower risk (further away, temporarily, from the exposure), is easier to perform and more straightforward to understand than the SCCS and SCRI analyses while utilising the same underlying principle. It also points to vaccine-induced intussusception. The analysis includes only vaccinated infants. In this aspect, it is similar to SCRI and different from the traditional SCCS analysis which employs unexposed controls. This type of analysis has been used previously to detect safety signals. Adverse events after herpes zoster vaccination in the first 28 days compared with days 29-56 have been studied. The authors of the herpes study used data mining techniques that are more complex than the analysis we adopted [15].
SCCS Analysis
The previous SCCS [3] had studied the risk windows 1-7 days, 7-21 days and 1-21 days, so we used the same risk periods for our SCCS analysis. In the new analysis, we excluded unvaccinated children from the analysis. Unexposed babies who have intussusception at a later stage in life, are not appropriate cases to consider as controls. We hypothesised that including unvaccinated infants in the original SCCS analysis [3] may have blunted its ability to detect safety signals. Excluding unvaccinated babies from the SCCS analysis, we were able to detect an increased risk of adverse events after the 3rd dose of the vaccine.
Relevance of the Findings
The rotavirus vaccine (116E) (G9P[11]) Rotavac (Bharat Biotech) was licenced after a small randomised control trial (RCT) involving 6719 infants (4532 received the vaccine and 2187 were controls) [16]. RCT investigators calculated that there was an excess of 11 cases of intussusception per 10,000 among the vaccinated, compared to controls although they did not show statistical significance in the small sample studied [17]. This is 5–10 times higher than the risk of intussusception with the RotaShield vaccine (which had to be withdrawn from the market). The SCCS [3] was supposed to put to rest questions about the safety of this vaccine. Our analysis suggests that there is a significant risk of intussusception which went unnoticed in the SCCS analysis.
Recognition of the risk of intussusception with vaccination is particularly important in countries like India where the vaccine may be administered to babies born to illiterate and impoverished parents in remote places, who do not have access to professional radiologists and paediatric surgeons, to identify and treat this complication. Parents need to be made aware of this risk of the vaccine and alert so that they do not mistake the symptoms of intussusception – the passage of blood and mucus - for the more familiar infective dysentery.
Limitations of the Study
There are several limitations to our study. First, we did not acquire the data ourselves and so we were dependent on others for the accuracy of their ascertainment methods. Furthermore, a large cohort of eligible children were excluded (12%) from the data provided because vaccination status could not be confirmed. This along with the relatively small sample studied resulted in wide confidence intervals and reduced precision and power. However our use of multiple methods all point to a risk of intussusception missed in the original SCCS study. The generally consistent results obtained from two complementary designs (SCRI and our modified SCCS) are a notable strength of the study and add to the robustness of the conclusion of this analysis.
Conclusion
Age-adjusted SCCS analysis is used when the baseline risk varies with age. Unvaccinated children can be included in the analysis to provide data on the baseline risk at different ages. However, it is inappropriate to include unvaccinated children in the SCCS analysis, if the act of vaccination alters the mean age at which intussusception occurs. This shift in the age of intussusception was seen in our study and has also been reported by others [13]. SCCS must include only vaccinated babies if vaccination alters the age-related baseline risk. A SCRI analysis in infants under 6 months can also be used for rotavirus evaluations as the vaccination is completed by 6 months when the risk of naturally occurring intussusception is relatively low.
A meta-analysis of SCCS studies of intussusception caused by rotavirus showed a relative risk of 5.71 (95%CI 4.50-7.25) in the 7 days after the first dose of the vaccine [18]. It may be useful to re-examine the raw data in these studies so the real relative risk of vaccine-induced intussusception can be re-estimated on meta-analysis.
References
1. Baker MA, Lieu TA, Li L, et al. A vaccine study design selection framework for the postlicensure rapid immunization safety monitoring program. Am J Epidemiol. 2015;181(8):608-618. doi:10.1093/aje/kwu322
2. Carlin JB, Macartney KK, Lee KJ, et al. Intussusception risk and disease prevention associated with rotavirus vaccines in Australia's National Immunization Program. Clin Infect Dis. 2013;57(10):1427-1434. doi:10.1093/cid/cit520.
3. Reddy SN, Nair NP, Tate JE, et al. Intussusception after Rotavirus Vaccine Introduction in India. N Engl J Med. 2020;383(20):1932-1940. doi:10.1056/NEJMoa2002276
4. Stowe J, Andrews N, Ladhani S, Miller E. The risk of intussusception following monovalent rotavirus vaccination in England: A self-controlled case-series evaluation. Vaccine. 2016;34(32):3684-3689. doi:10.1016/j.vaccine.2016.04.050
5. Yung CF, Chan SP, Soh S, Tan A, Thoon KC. Intussusception and Monovalent Rotavirus Vaccination in Singapore: Self-Controlled Case Series and Risk-Benefit Study. J Pediatr. 2015;167(1):163-8.e1. doi:10.1016/j.jpeds.2015.03.038
6. Escolano S, Hill C, Tubert-Bitter P. Intussusception risk after RotaTeq vaccination: evaluation from worldwide spontaneous reporting data using a self-controlled case series approach. Vaccine. 2015;33(8):1017-1020. doi:10.1016/j.vaccine.2015.01.005
7. Groome MJ, Tate JE, Arnold M, et al. Evaluation of Intussusception After Oral Monovalent Rotavirus Vaccination in South Africa. Clin Infect Dis. 2020;70(8):1606-1612. doi:10.1093/cid/ciz431
8. Leino T, Ollgren J, Strömberg N, Elonsalo U. Evaluation of the Intussusception Risk after Pentavalent Rotavirus Vaccination in Finnish Infants. PLoS One. 2016;11(3):e0144812. Published 2016 Mar 7. doi:10.1371/journal.pone.0144812
9. Maglione MA, Gidengil C, Das L, et al. Safety of Vaccines Used for Routine Immunization in the United States. Evid Rep Technol Assess (Full Rep). 2014;(215):1-740. doi:10.23970/AHRQEPCERTA215
10. Tate JE, Simonsen L, Viboud C, et al. Trends in intussusception hospitalizations among US infants, 1993-2004: implications for monitoring the safety of the new rotavirus vaccination program. Pediatrics. 2008;121(5):e1125-e1132. doi:10.1542/peds.2007-1590
11. Li R, Stewart B, Weintraub E. Evaluating efficiency and statistical power of self-controlled case series and self-controlled risk interval designs in vaccine safety. J Biopharm Stat. 2016;26(4):686-693. doi:10.1080/10543406.2015.1052819
12. Yih WK, Lieu TA, Kulldorff M, et al. Intussusception risk after rotavirus vaccination in U.S. infants. N Engl J Med. 2014;370(6):503-512. doi:10.1056/NEJMoa1303164
13. Tate JE, Yen C, Steiner CA, Cortese MM, Parashar UD. Intussusception Rates Before and After the Introduction of Rotavirus Vaccine. Pediatrics. 2016;138(3):e20161082. doi:10.1542/peds.2016-1082
14. Whitaker HJ, Farrington CP, Spiessens B, Musonda P. Tutorial in biostatistics: the self-controlled case series method. Stat Med. 2006;25(10):1768-1797. doi:10.1002/sim.2302
15. Yih WK, Kulldorff M, Dashevsky I, Maro JC. Sequential Data-Mining for Adverse Events After Recombinant Herpes Zoster Vaccination Using the Tree-Based Scan Statistic. Am J Epidemiol. 2023;192(2):276-282. doi:10.1093/aje/kwac176
16. John J, Kawade A, Rongsen-Chandola T, et al. Active surveillance for intussusception in a phase III efficacy trial of an oral monovalent rotavirus vaccine in India. Vaccine. 2014;32 Suppl 1:A104-A109. doi:10.1016/j.vaccine.2014.03.036
17. Bajaj J, Puliyel JM. Intussusception risk with 116E rotavirus vaccine in Vellore, South India. Vaccine. 2016;34(4):403. doi:10.1016/j.vaccine.2015.03.007
18. Koch J, Harder T, von Kries R, Wichmann O. Risk of Intussusception After Rotavirus Vaccination. Dtsch Arztebl Int. 2017;114(15):255-262. doi:10.3238/arztebl.2017.0255
?
Table 1: Age of intussusception Vaccinated vs. Unvaccinated
Vaccinated Unvaccinated
Mean Age at Intussusception 205.36 days 223.25 days
Standard Deviation 65.8 74.12
Mean Difference -17.89
Standard Error 5.915
95% CI -29.51 to -6.27
p value 0.0026
?
Table 2: Risk of intussusception after vaccination (SCRI analysis)
Number of cases in the risk period 1-21 days after any vaccine Number of control cases in the low-risk period during the observation from 28 days to 180 days IRR (95% CI)
(adjustment for age; grouped in intervals of 30 days)
59
90
1.62 (95% CI 1.07-2.44)
?
Table 3: Cases of intussusception within the first 30 days of vaccination compared to cases in the next 30 days
Within 30 days of vaccination Between 31 and 60 days from vaccination
Cases 92 63
Non-cases 285 314
Odds Ratio 1.61 (95% CI: 1.12-2.30) p-value=0.009
(Pearson Chi-Square without the Yate's correction)
?
Table 4: The relative incidence of intussusception in vaccinated babies in the risk periods 0-7 days and 8-21 days after each dose of the rotavirus vaccination by self-controlled case series method
Doses of rotavirus vaccine Risk Period (days) No. of cases in risk period IRR (95% CI)
Dose 1 1-7 days 2 0.26 (0.06-1.06)
8-21 days 2 0.15 (0.03-0.61)
1-21 days 4 0.19 (0.07-0.51)
Dose 2 1-7 days 4 0.72 (0.29-1.79)
8-21 days 15 1.25 (0.72-2.17)
1-21 days 19 1.06 (0.65-1.72
Dose 3 1-7 days 15 3.15 (1.90-5.22)
8-21 days 22 2.12(1.32-3.40)
1-21 days 37 2.47 (1.70-3.59)
?
Figure 1
?
Legend for Figure
Figure 1: Cases of intussusception against time since last rotavirus vaccination