Rotavirus is the leading cause of severe pediatric gastroenteritis worldwide, and although rotavirus-associated gastroenteritis results in few childhood deaths in the United States, the disease burden is high; prior to the introduction of the pentavalent vaccine, approximately 410,000 physician visits, 70,000 hospitalizations, and 272,000 emergency department (ED) visits each year were attributed to rotavirus, resulting in a total annual cost of almost $1 billion.1–3
RotaTeq (Merck & Co Inc, Whitehouse Station, NJ), an oral pentavalent rotavirus vaccine (RV5), was first licensed in the United States in February 2006 and was subsequently recommended for routine use by the Advisory Committee on Immunization Practices.3 The vaccine contains 5 live reassortant rotaviruses isolated from human and bovine hosts and is indicated for the prevention of rotavirus gastroenteritis caused by serotypes G1–G4, which accounts for approximately 90% of all rotavirus gastroenteritis in the United States.4,5 According to the US product label, the RV5 series consists of 3 ready-to-use liquid doses to be administered starting at 6–12 weeks of age, with subsequent doses administered at 4- to 10-week intervals, and the third dose given by 32 weeks of age.6 This schedule corresponds to standard pediatric visits.
An earlier rhesus-human rotavirus vaccine, RotaShield (Wyeth-Lederle, Radnor, PA), was voluntarily withdrawn from the market in 1999 after it was associated with intussusception (IS),7 a condition that increases during the first year of life, peaking at age 5 to 6 months.8,9 Although an association between RV5 and IS (or other adverse events) was not found in large-scale, prelicensure clinical trials,10–12 this study was conducted as a postlicensure commitment to monitor the short-term safety profile of RV5 in a large number of infants under conditions of routine use.
MATERIALS AND METHODS
This prospective observational cohort study was conducted within a proprietary research database built from electronically captured health insurance claims in a large geographically diverse health plan in the United States. The patients are fully insured for physician, hospital, and prescription drug services. The database is compliant with patient privacy according to the Health Insurance Portability and Accountability Act. The New England Institutional Review Board Privacy Board granted a waiver of authorization for patient medical record abstraction. No data were collected directly from parents or infants.
On a quarterly basis during 2006 and 2007, we identified infants vaccinated with RV5 based on insurance claims bearing the Current Procedural Terminology (CPT) code specific for RV5 (90680) or with the National Drug Code for RV5. We required insurance enrollment within 1 week of birth (to improve ascertainment of vaccine exposure) and age less than 10 months at first vaccine dose (allows for 60 days of follow-up before 1 year of age). The date of the RV5 vaccination was designated as the index date. Each subsequent dose of RV5 set a new, separate index date. We captured claims from birth through 1 year of age. The follow-up was potentially through March 31, 2009, and since 97% of infants received their third RV5 dose by age 32 weeks,13,14 this follow-up allowed for late second and third vaccinations, 60 days follow-up, and any lag for insurance claims.
The study incorporated 2 comparison cohorts: a cohort of infants vaccinated with diphtheria-tetanus-acellular pertussis (DTaP) contemporaneously with the infants receiving RV5 (concurrent DTaP infants), and an historical cohort of infants who received DTaP in 2001–2005, prior to availability of RV5. Applying the same eligibility criteria, the concurrent DTaP infants were identified on a quarterly basis using CPT codes for DTaP vaccination. To address the concern that in the first year of availability infants receiving a first dose of RV5 might be older than those receiving a first dose of DTaP, a potential source of confounding, within each 2006 quarterly cohort, we age-matched (1:2) RV5-exposed infants to the concurrent DTaP infants using a “greedy” matching algorithm with a variable caliper.15 If a concurrent DTaP infant had a subsequent claim for RV5, we censored follow-up time of the infant as of the date of the RV5 vaccination, entered the infant into the RV5 cohort, and selected a replacement concurrent DTaP infant from the pool of eligible comparators. The rapid uptake of RV5 limited the number of infants receiving DTaP and not RV5, and we included all infants first vaccinated with RV5 and DTaP in 2007.
The historical DTaP cohort was created using similar eligibility criteria and follow-up windows.9 We randomly selected 20,000 infants who received a first dose of DTaP within each of the 5 years of the study and followed them through the earlier of 1 year of age, disenrollment, or June 30, 2006.
Potential cases of IS and Kawasaki disease (KD) were identified by searching cohort members’ health insurance claims for qualifying diagnosis or procedure codes from hospitals or EDs: IS (International Classification of Disease, 9th revision [ICD-9] diagnosis code 560.0, 543.9, or CPT 74283) and KD (ICD-9 code 446.1, or 414.11, an exploratory diagnosis possibly related to KD). We sought medical records for all potential cases. Two independent, 3-member adjudication committees, one for each outcome and blinded to vaccinations, reviewed the abstracted medical records and determined whether the case met the study criteria. Claims-based cases for which a medical record could not be obtained were considered to be noncases in the analyses. Potential deaths were identified, and associated medical records were reviewed in an effort to capture study outcomes that might not have a corresponding health insurance claim.
The case definition for IS required radiographic or surgical confirmation, or evidence of IS at autopsy, corresponding to the level one definite criteria of the Brighton Collaboration.16 For KD, the case definition was based on a US Centers for Disease Control surveillance case definition that included the following criteria: rash, cervical lymphadenopathy, bilateral conjunctival injection, oral mucosal changes, peripheral extremity changes, and coronary artery abnormalities.17 For the confirmed IS cases, we sought vaccination histories from medical records to confirm that the vaccine received (DTaP or RV5) matched that identified on the basis of health insurance claims.
The expected number of IS cases relative to person-time exposed to RV5 was plotted against a monitoring boundary that assumed a background incidence rate of 1 per 2000 (or 0.5 per 1000) infant-years and was adjusted for the sequential nature of the study with an overall alpha equal to 0.05 (one-sided) at the end of the study. The monitoring boundary curve is consistent with previously developed sequential monitoring methods for monitoring the number of observed cases of IS based on the accrued people and person-time.18 For any amount of person-time, there is an expected number of cases, and a boundary above which an observed number of cases would be significantly above background. A similar curve was plotted for KD assuming a background rate of 1 per 5000 (or 0.2 per 1000) person-years.
Additionally, general safety was assessed by monitoring any health insurance claims event resulting in inpatient hospitalization or ED visits occurring in the time windows day 0, days 1–7, days 8–14, and days 0–30 after any RV5 dose relative to historical and self-controls using previously established methods.19 The results of these screenings were tabulated using prespecified statistical and clinical criteria to identify any unexpected health event. An independent safety monitoring committee (SMC) comprised of pediatric vaccine specialists, and a biostatistician reviewed these tables as well as all study outcome results and the updated monitoring boundaries each quarter.
Infants were characterized by age, gender, geographic region, and calendar year of first vaccination. Within different windows of time following each dose (days 0–30, 0–21, 0–60, and 31–60 for self-controls), we determined available person-time. We censored person-time of follow-up in a window on the occurrence of a subsequent dose of a relevant vaccination, occurrence of a study outcome, disenrollment, or the infant reaching 1 year of age. The initial claims-based event date was replaced by the actual event date for chart-confirmed cases if they were different.
We calculated incidence rates of chart-confirmed events (events divided by person-time), 95% confidence intervals (CIs), and associated relative risks (RRs). We compared observed cases with prespecified monitoring boundaries developed to have 80% power to identify a RR of 2.5. All data management and determination of person-time and incidence rate estimation were conducted in SAS/STAT software20; RRs and exact 95% CIs were determined using STATA.21 The one-sided non-mid-p exact probabilities were derived in SAS.20,22
We identified 85,150 infants vaccinated with RV5 (>210,000 doses) in the 2-year accrual period, contributing 17,433 person-years of follow-up within 30 days and 32,799 person-years within 60 days (Table 1).Infants with a first vaccination of RV5 in 2006 (n = 18,807) were matched to 30,464 concurrent DTaP controls on date of birth and vaccine dose. We matched 93% of the RV5 recipients on exact date of birth; for the remaining infants, we matched within 7 days. All eligible DTaP recipients in 2007 (n = 32,153) were included in the comparison cohort. The 62,617 concurrent DTaP controls contributed 12,339 person-years of follow-up within 30 days and 23,211 person-years within 60 days. We censored follow-up of 3160 concurrent DTaP infants on the occurrence of a subsequent claim for RV5 vaccination. The infants in the 2 cohorts were comparable with respect to age at first dose and fraction receiving subsequent doses. Even without matching on age in 2007, the difference in mean age at first dose between the 2 cohorts was less than 1 week. The proportion of infants receiving a first RV5 vaccine dose at the schedule-recommended age (6–12 weeks) increased from 49% in quarter 1 (2006) to 92% in quarter 4 (2007).
We obtained medical records for 21 of 22 (95%) of the potential IS cases identified in the 60 days following all doses among recipients of RV5, and for 16 of 17 (94%) among concurrent DTaP controls. For KD, we obtained 12 of 13 (92%) medical records for recipients of RV5 and 4 of 4 (100%) for the concurrent DTaP controls.
Six of the 14 (43%) potential IS cases with medical records in the 0–30-day follow-up window and 11 of 21 (52%) in the 0–60-day window were adjudicated as cases. The corresponding confirmation among the concurrent DTaP controls was 5 of 9 (56%) potential cases in the 0–30-day window, and 6 of 16 (38%) in the 0–60-day window. For the historical DTaP cohort, 5 of 11 potential cases (45%) in the 30-day follow-up window were confirmed, as were 13 of the 28 potential cases (46%) in the 0–60-day follow-up window.9 Potential cases that were not confirmed were primarily those for whom the claim reflected a visit for evaluation of the condition (a “rule-out” diagnosis). The percentage of cases positively adjudicated as KD was 25% in both the RV5 and concurrent DTaP control cohorts.
None of the RRs at any dose in any follow-up window indicated an increased risk of IS (Table 2).The RR of IS after any dose in the 0–30-day follow-up window among recipients of RV5 compared with concurrent DTaP infants was 0.8 (95% CI: 0.22–3.52); the RR compared with the 2004–2005 historical DTaP controls was similar.
Although the chart abstraction process obtained almost all of the medical records sought for IS (95% and 94% for the RV5 and concurrent DTaP control cohorts respectively), we conducted a sensitivity analysis that assumed the medical records that could not be obtained (one for RV5 and one for DTaP) were positively adjudicated. The results of this analysis did not differ appreciably (RR = 1.0, 95% CI: 0.27–3.96) from the results based on confirmed cases.
For the outcome of KD, 1 case was confirmed within 30 days in an RV5 recipient, 2 cases in days 31–60 (self-controls), and 1 case in the concurrent DTaP cohort in the 0–30-day window. The RR for KD among RV5 recipients compared with concurrent DTaP controls was 0.7 (95% CI: 0.01–55.56; any dose, 0–30-day follow-up window) and 0.4 (95% CI: 0.01–8.47; any dose compared with self-controls). The single chart that could not be obtained was for an RV5 self-control; the result was similar with this case included.
We identified 31 infants in the RV5 and concurrent DTaP cohorts who had an inpatient discharge status indicating death or claims indicative of resuscitative efforts within the follow-up windows and obtained 28 (90%) of the requested records. The records for 2 recipients of RV5 and 1 concurrent DTaP infant were not obtained. The occurrence of death was balanced across the 2 study cohorts with 15 chart-confirmed deaths in the 0–60-day follow-up window among 32,800 person-years of follow-up among infants vaccinated with RV5 and 10 confirmed cases in 23,213 person-years of follow-up in the concurrent DTaP control cohort. None of the information available in the abstracted charts indicated the deaths were related to IS or KD.
The 6 confirmed cases of IS among recipients of RV5 were plotted against the number of subjects vaccinated with RV5 periodically during the study (Fig. 1). With 17,433 person-years accrued at the conclusion of the study, 8 or 9 cases of IS were expected, and 18 cases would have been required to cross the monitoring boundary. Even the addition of the single case for which the medical record was not obtained to the 6 confirmed cases meant that the total number of cases was considerably below the monitoring boundary, so that incomplete access to medical records did not account for being within the monitoring boundary. For KD, approximately 10 cases would have been required to cross the monitoring boundary, and there was one case in the 0–30 days following RV5.
The study was powered to assess risks in the 30-day period after any dose. An additional an ad hoc exploratory analysis was conducted to assess the risk of IS within 7 days after vaccination. We compared rates of IS among recipients of RV5 and DTaP in the 1–7-day follow-up window after any dose, where there were 4 cases of IS among recipients of RV5 and 1 case among recipients of DTaP (RR = 2.8, 95% CI: 0.3–139.5, P = 0.31). The broad CI around this RR reflects the low power, although the finding is consistent with the overall study conclusion of no association between RV5 vaccination and IS. Using data from validated vaccination histories, the 4 cases in the RV5 recipients were distributed as follows: 1 case of IS occurred after dose 1, and 3 cases occurred after dose 2. For the DTaP concurrent controls, 1 case occurred after Dose 3.
The general safety component of the study compared the claims-based occurrence of any health event that might be identified by an ICD-9 code from a hospital or ED among infants vaccinated with RV5 compared with RV5 self-controls and historical DTaP controls. These comparisons of diagnoses based on ICD-9 diagnosis codes entailed over 11,000 comparisons that accounted for site of service, dose and multiple follow-up windows. Tests were conducted using a one-sided α level of 0.025; therefore, approximately 275 of the 11,000 comparisons might be expected to be statistically significant elevated due to chance alone. Of the 11,000 comparisons, 194 were found to have statistically significant differences in rates (132 elevated and 62 lowered). The SMC discussed the specific diagnoses with significant findings to assess possible associations with RV5 based on pattern of elevation, clinical judgment, knowledge of the scientific literature, biologic plausibility, and clinical significance. The SMC found that none of the specific diagnoses or patterns of diagnoses warranted additional follow-up for safety purposes.
This study included 85,150 infants vaccinated with at least one dose of RV5 and 62,617 vaccinated with at least one dose of DTaP from 2006 and 2007. The follow-up of these identified no significant increased risk of IS or KD. At no point in this study did the number of confirmed cases of IS or KD cross the prespecified monitoring boundaries for these outcomes based on the number of RV5 recipients. The broad-based general safety evaluation for RV5 did not identify any other outcomes that might represent safety concerns. The finding of no significant increased risk of IS in the 7 days after any dose should be interpreted with caution due to the small sample size available for this analysis.
Approximately half of the potential IS cases were confirmed. The sensitivity analysis conducted to assess the effect of incomplete medical record abstraction supported the conclusion of no association between vaccination with RV5 and risk of IS or KD in the 0–30-day follow-up window. We also retrieved vaccination histories of confirmed IS cases to confirm the actual vaccine received (DTaP or RV5) based on health insurance claims. There were some differences related to the number of vaccine doses received from these 2 data sources, but the actual vaccine received (DTaP vs. RV5) was completely concordant, providing assurance that this study's overall null finding was not an artifact of incorrect vaccine coding in the insurance claims.
These results, including the historical component of the study, establish the epidemiology of IS among an insured infant population.9 The estimates of the incidence rates of IS and their associated CIs for any dose in the 0–30-day follow-up window among all 3 of the cohorts are slightly lower but still consistent with the rate assumed in developing the monitoring boundary (0.5 cases per 1000 person-years), and are similar to the background rate of IS reported by the US Vaccine Safety Datalink,23 suggesting that these results may be generalizable to other populations. The results also demonstrate that while the incidence of IS peaks among infants at 5–6 months of age, the condition is diagnosed and confirmed in younger infants. A description of the epidemiology of KD also emerges from these results, with the incidence observed among the RV5 and concurrent DTaP control cohorts being consistent with that assumed in developing the monitoring boundary (0.2 cases per 1000 person-years), but the small number of confirmed cases that occurred during the follow-up of these cohorts does not provide much precision. These results are also consistent with published data from a Vaccine Safety Datalink analysis of more than 207,000 doses administered between May 2006 and May 2008; the study noted no elevation in risk for IS, seizures, meningitis/encephalitis, myocarditis, Gram-negative sepsis, gastrointestinal bleeding, or KD.24 More recently, in the largest study of RV5 safety conducted to date, the Vaccine Safety Datalink completed an updated analysis of a total of 839,364 RV5 doses compared with 405,785 doses among concurrent controls receiving vaccinations other than RV5; this analysis continued to suggest that receipt of RV5 is not associated with an increased risk for IS 1–30 days after any dose (RR = 0.79, 95% CI: 0.43–1.50), 1–7 days after any dose (RR = 0.79, 95% CI: 0.18–3.92), or 1–7 days after dose 1 (RR = 0.65, 95% CI: 0.03–38.04).25
The authors would like to acknowledge the insightful review of the members of the Safety Monitoring Committee: Dr. Georges Peter (Chair), Professor Emeritus Brown Medical School; Dean Blumberg, University of California Davis; Dr. Lisa LaVange, University North Carolina Chapel Hill; Dr. Gary Marshall, University of Louisville School of Medicine; Mark Sawyer, University California San Diego; and Dr. David Yost, United States Public Health Service, Arizona. We similarly recognize the contributions of the members of the Intussusception Adjudication Committee: Dr. Daniel Doody (Chair), Massachusetts General Hospital; David Bates, Children's Hospital, Columbus, OH; Dr. Denesh Chitkara, University North Carolina Chapel Hill; and Dr. Alejandro Flores Sandoval, New England Medical Center, Boston, MA; and the Kawasaki disease Adjudication Committee: Dr. Anne Rowley (Chair), Northwestern University, Dr. Jane Newburger, Children's Hospital, Boston, MA; and Dr. Jane Burns, University California San Diego.
1. Tucker AW, Haddix AC, Bresee JS, et al.. Cost-effectiveness analysis of a rotavirus immunization program in the United States. JAMA. 1998;279:1371–1376.
2. Widdowson MA, Meltzer MI, Zhang X, et al.. Cost-effectiveness and potential impact of rotavirus vaccination in the United States. Pediatrics. 2007;119:684–697.
3. Parashar UD, Alexander JP, Glass RI. Prevention of rotavirus gastroenteritis among infants and children. Recommendations of the Advisory Committee on Immunization Practices (ACIP). MMWR Recomm Rep. 2006;55(RR-12): 1–13.
4. Griffin DD, Kirkwood CD, Parashar UD, et al.. Surveillance of rotavirus strains in the United States: identification of unusual strains. The National Rotavirus Strain Surveillance System collaborating laboratories. J Clin Microbiol. 2000;38:2784–2787.
5. Abdel-Haq NM, Thomas RA, Asmar BI, et al.. Increased prevalence of G1P genotype among children with rotavirus-associated gastroenteritis in metropolitan Detroit. J Clin Microbiol. 2003;41:2680–2682.
6. RotaTeq [package insert]. Whitehouse Station, NJ: Merck and Co, Inc; 2007.
7. Murphy TV, Gargiullo PM, Massoudi MS, et al.. Intussusception among infants given an oral rotavirus vaccine. N Engl J Med. 2001;344:564–572.
8. Chang HGH, Smith PF, Ackelsberg DL, et al.. Intussusception, rotavirus diarrhea, and rotavirus vaccine use among children in New York State. Pediatrics. 2001;108:54–60.
9. Eng PM, Mast TC, Loughlin J, et al.. Incidence of intussusception among infants in a large insured population of the United States. In: 24th International Conference on Pharmacoepidemiology and Therapeutic Risk Management; August 2008; Copenhagen, Denmark.
10. Vesikari T, Matson DO, Dennehy P, et al.. Safety and effectiveness of a pentavalent human-bovine (WC3) reassortant rotavirus vaccine. N Engl J Med. 2006;354:23–33.
11. CDC. Postmarketing monitoring of intussusception after RotaTeq vaccination—United States, February 1, 2006—February 15, 2007. MMWR. 2007;56:218–222.
12. Haber P, Patel M, Hector IS, et al.. Post-licensure monitoring of intussusception after RotaTeq vaccination in the United States, February 1, 2006—September 25, 2007. Pediatrics. 2008;121:1206–1212.
13. Loughlin J, Wang FT, El Khoury A, et al.. Adherence to the recommended pentavalent rotavirus vaccine administration schedule in the United States: 2006–2007 [Letter]. Ped Infect Dis J. 2009;28:667–668.
14. Parashar UD, Glass RI. Rotavirus vaccines–early success, remaining questions. N Engl J Med. 2009;360:1063–1065.
17. Gibbons RV, Parashar UD, Holman RC, et al.. An evaluation of hospitalizations for Kawasaki syndrome in Georgia. Arch Pediatr Adolesc Med. 2002;156:492–496.
18. Jennison C, Turnbull BW. Group Sequential Methods With Applications to Clinical Trials. Boca Raton, FL: Chapman and Hall/CRC; 2000:171–187, 235–244.
19. Davis RL, Black S, Shinefield H, et al.. Vaccine. 2004;22:536–543.
20. SAS/STAT Software, version 9.1.3. SAS Institute Inc, Cary, NC.
21. STATA Software, release 6. Stata Press. College Station, TX.
22. Brown LD, Cai TT, DasGupta A. Interval estimation for a binomial proportion. Statistica Sci. 2001;16:101–117.
23. Haber P, Patel M, Izurieta HS, et al.. Postlicensure monitoring of intussusception after RotaTeq vaccination in the United States, February 1, 2006, to September 25, 2007. Pediatrics. 2008;121:1206–1212.
24. Belongia EA, Irving SA, Shui IM, et al.. Real-time surveillance to assess risk of intussusception and other adverse events after pentavalent, bovine-derived rotavirus vaccine. Pediatr Infect Dis J. 2010;29:1–5.
25. Baggs J, Haber P. Continued surveillance for intussusception (IS) following RotaTeq in VAERS and VSD. Presented at: Advisory Committee on Immunization Practices (ACIP); October 28, 2010; Atlanta, GA.