Quinn, Helen E. PhD, MAE*†; Wood, Nicholas J. MB BS, DCH, MPH*†‡; Cannings, Kathryn L. MPH*; Dey, Aditi PhD, MPH*†; Wang, Han MStats*; Menzies, Robert I. PhD, MPH*†; Moberley, Sarah PhD, MPH§; Reid, Su MPH¶; McIntyre, Peter B. MB BS, PhD*†‡; Macartney, Kristine K. MB BS, BMedSci, MD*†‡
In 2009, the World Health Organization recommended that infants worldwide should be vaccinated against rotavirus, the most common cause of severe diarrheal disease in young children.1 This recommendation followed a review of vaccine efficacy, effectiveness and safety data from a number of countries.2,3 The 2 currently registered rotavirus vaccines, a monovalent human vaccine, Rotarix (RV1; GlaxoSmithKline Biologicals) and the pentavalent reassortant vaccine, RotaTeq (RV5; CSL Biotherapies/Merck & Co), had an excellent safety profile in placebo-controlled clinical studies that enrolled >130,000 infants.2,3 However, because of the withdrawal of the tetravalent human-rhesus reassortant vaccine (Rotashield, Wyeth) in the United States in 1999 because of a 37-fold increased risk of intussusception (IS) after dose 1,4,5 a number of countries implemented postmarketing surveillance for IS after vaccination.
Recent published studies from Australia and the United States (for RV1 and RV5),6–9 Brazil and Mexico (for RV1)10 and the United States (for RV5),11 suggest low but variable risk of IS (up to 6 additional cases per 100,000 vaccinated infants in Australia)7 after rotavirus vaccine. Based on emerging evidence, in 2013, the World Health Organization recommended that countries should conduct surveillance for IS when introducing rotavirus vaccines.12 Published studies have predominantly used active surveillance for IS, with clinical review of putative cases to ascertain whether they meet published criteria for IS.13 For such a rare condition, this process needs to occur across many sites and is resource intensive. The use of large healthcare databases to identify cases of IS and their relationship to vaccination is an alternative means to generate data for such a risk analysis that is both faster and less resource intensive. Such databases have previously been used without clinical review of cases, to investigate IS in the United States,14,15 as well as a range of other adverse events after immunization with different vaccines. However, for IS, these studies have not allowed comparison of the clinical characteristics and severity of IS cases that occur within the nominated risk window after vaccination with cases outside the risk window or cases in nonvaccine exposed infants. It is important to understand whether infants with IS temporally related to RV vaccination have a more severe or complicated course.
Using a defined population-based cohort, our first aim was to evaluate the proportion of cases coded as IS in routinely collected large healthcare databases with cases confirmed as Brighton level 1 IS after detailed clinical review; this aimed to inform the validity of using large healthcare database information without clinical review to assess IS risk after RV vaccination. Our second aim was to compare the severity of clinical outcomes of confirmed IS associated with recent receipt of a vaccine dose with IS not temporally associated with vaccination.
IS Case Ascertainment
This retrospective cohort study was conducted in New South Wales, Australia (population 7.1 million, annual birth cohort 94,000) where RV1 is used in a 2-dose course, with the first dose recommended at between 6 and ≤14 weeks of age and the second dose by ≤28 weeks of age.16 Vaccine coverage of 85% for 2 doses by 12 months of age was achieved rapidly after program commencement in July 2007 and adherence to upper age limits for dose administration is high.17,18 New South Wales is a geographically large state containing only 3 tertiary pediatric centers where treatment of IS (enema reduction or surgery) is routinely available.
Hospitalization data from the NSW Admitted Patient Data Collection were used to identify episodes of IS. All hospitalizations coded as K56.1 (IS), using the International Statistical Classification of Diseases and Related Health Problems 10th revision Australian modification (ICD-10AM), in any diagnostic field, were included in the analysis. All emergency department (ED) presentations from the NSW Emergency Department Data Collection coded as K56.1 were also identified to ensure comprehensive ascertainment.
Cases were eligible for inclusion if the hospital admission or ED presentation date was between July 1, 2007, and June 30, 2010, and they were aged <12 months at presentation. Medical records, including laboratory and radiologic reports, were reviewed and data extracted using a standardized data collection form. Each case was then classified using the Brighton collaboration criteria (Table 1).13
Number and date of RV doses were confirmed from the Australian Childhood Immunisation Register. The Australian Childhood Immunisation Register records immunizations provided to children aged <7 years enrolled in the Australian universal health insurance scheme, Medicare, and constitutes a nearly complete population register (approximately 99% of the birth cohort).19 Greater than 97% of vaccinations are reported to the Australian Childhood Immunisation Register.20
Self-controlled Case Series Analysis
The self-controlled case series (SCCS) method was developed as a statistical approach suited to investigation of associations between vaccination and adverse events.21 The association between rotavirus vaccination and IS was examined using a modified Poisson regression model, which calculates a relative incidence (RI) by comparing the frequency of IS in predefined risk periods after vaccination, with the frequency of IS during “not at risk” periods derived by including both vaccinated and unvaccinated subject data.22 Cases were excluded from SCCS analysis where no vaccination record was available, or a second dose of RV1 was recorded with missing data for the first dose. The analysis assessed risk in 2 exposure periods, chosen on the basis of previously published studies: 1–7 and 8–21 days postvaccination.7 The model included an adjustment for month of age to account for the changing background rate of IS across the age period of observation, as done in previous studies.7 SAS for Windows version 9.2 was used for data analysis.
The clinical case review was commissioned and approved by the NSW Ministry of Health as an inquiry into a matter affecting the health of the public under section 71(1) of the NSW Public Health Act 1991. The Paediatric Active Enhanced Disease Surveillance network, which provided clinical data for cases treated in the ED, has approval for surveillance from the Children’s Hospital at Westmead Human Research Ethics Committee (No. 2007/009).
Chart Review of IS-coded Episodes in the Vaccination Period
Using available variables in the hospitalization records, interhospital transfers (on the same or next calendar day) coded as IS in the same individual were identified. After combining these records, 229 ICD-coded records were reduced to 175 unique episodes in the 3-year period (Fig. 1). In 2 cases, no patient medical record could be found, leaving 173 ICD-coded hospitalized IS episodes. Four individuals had an additional admission for IS, separated from the first episode by at least 72 hours (12, 71, 100 and 159 days, respectively); these were all kept in the analysis. In the ED database, there were 171 ICD-coded episodes of IS, of which 29 were not identified as being in the same child on the same date in hospitalization records (Fig. 1). Only 13 of the 29 episodes (those that occurred at the Children’s Hospital at Westmead) could be reviewed because of lack of access ED clinical records across the state. A small proportion of New South Wales infants (n = 7) vaccinated with RV5 in other Australian states were also excluded from the analysis. Thus, a total of 179 unique ICD-coded episodes were available for chart review and classification, with 110 (61%) meeting the Brighton level 1 criteria. Those meeting the Brighton level 2–5 criteria, including those without IS (Brighton level 5), some of whom had a clear alternate diagnoses, were considered to be “non-confirmed IS”.
Self-controlled Case Series
Of the 110 confirmed Brighton level 1 IS episodes, 102 (93%) met the eligibility criteria for inclusion in the SCCS analysis (Table 2). There was an increased risk of IS in the 1–7 days after vaccination with the first RV1 dose, with a RI of 11.1 (95% CI: 2.6–48.0). There was also an increased RI of IS in the 1–21 day period post dose 1 of vaccine [RI: 5.5 (95% CI: 1.7–17.8)]. The point estimates for risk post dose 2 suggested an increased relative incidence, but this was not statistically significant.
When unique IS-coded episodes obtained from the hospitalization and ED databases (without consideration for chart review results) were analyzed using the SCCS method (164 of 179; 92%), a lesser, but still increased relative incidence of IS in the 1–7 and 1–21 day periods after vaccination with the first dose of RV1 remained (Table 2). The risk of IS in the 1–7 and 1–21 day periods after the second dose of RV1 were unchanged but just reached statistical significance (Table 2).
Clinical Characteristics of Non-confirmed IS Episodes
Non-confirmed IS episodes (n = 69) had a significantly shorter length of stay than confirmed episodes (P < 0.001). Although many of infants with a Brighton level 2 episode (88%) had an ultrasound performed, fewer cases classified as Brighton level 3–5 had an ultrasound (Table 3). A high proportion of infants with a Brighton level 2 episode [12 of 14 (86%)] had an initial ultrasound suggestive of IS (often before referral; Table 3); however, IS was not confirmed on further investigation (usually a contrast enema). Of the 53 episodes that met criteria for Brighton level 3–5, 5 (9%) had a clear alternative final diagnosis by ultrasound [urachal cyst (1), volvulus (1), hepatoblastoma (1), gallstones (1) and mid small bowel obstruction (1)].
Infants with IS-coded episodes that were non-confirmed presented to 35 different hospitals initially, with 70% (n = 48) requiring transfer to 1 of the 3 tertiary pediatric centers in New South Wales.
Clinical Characteristics of Brighton Level 1 Confirmed IS Episodes
Brighton level 1 IS episodes were twice as likely to occur in males (75 vs. 35, ratio = 2.1:1) and the median length of stay for confirmed episodes was 1 day (range 1–12 days). Ninety-two episodes (84%) occurred in infants who had received at least 1 dose of RV1, with 18 occurring within 21 days of RV1 (6 within 7 days; Fig. 2). There were 5 episodes after dose 1 and 13 after dose 2. The 18 vaccine proximate episodes, defined as RV1 within 21 days of IS, occurred at a significantly younger age than those in infants where RV1 was given 21 days or more previously (4.8 months vs. 7.4 months; P < 0.001). Of 23 infants with confirmed IS after their first vaccine dose, 12 (52%) went on to receive dose 2; none had a recurrence of IS documented within the study period. Of 4 infants with recurrent IS, 3 of the original episodes were confirmed as Brighton level 1, as were 2 of the 4 recurrent episodes. For both infants with recurrent Brighton level 1 episodes, both episodes occurred >21 days after dose 2 of RV1.
Almost all confirmed IS cases (94%) were treated in 1 of the 3 pediatric centers, with abdominal ultrasound the first investigation for 93% (n = 102; Table 3). The most common site of IS was ileocolic (46%), followed by colonic (34%), ileoileal (7%) or not stated (13%). Almost all confirmed IS cases (95%) received an air contrast enema, but in 33 of 105 (31%), the IS did not resolve or complications occurred (Table 3). Surgical reduction was required for 35% of confirmed episodes and included 10 bowel resections and 5 perforation repairs, with 1 caecal duplication cyst removal. The median length of stay for those requiring surgery was 4 days. The likelihood of surgery was no different if the case had onset of IS within 21 days of vaccination (39% vs. 34%, P = 0.67). There were no deaths, but 5 cases required admission to an intensive care unit (ICU). Two of these 5 ICU cases had received their dose of rotavirus within the previous 21 days (8 days post dose 1 and 9 days post dose 2, respectively), both were older infants (aged 5.2 and 6.3 months, respectively) who were vaccinated in the first year of the program.
This study presents a comprehensive population-based retrospective analysis of hospitalization and ED episodes ICD-coded as IS for a 3-year period in infants in whom RV1 vaccine coverage was uniformly high. An increased risk of IS after receipt of RV1 vaccine has been found in 2 Australian studies,6,7 both of which included some of the cases (18%6 and 57%7) presented in this review. This study is unique in presenting a population-based review of the clinical characteristics and outcomes of all ICD-coded episodes of IS. As expected, the risk of IS derived using all ICD-coded episodes, as compared with episodes confirmed as IS, was generally lower because of misclassification of cases. This is also the first published study to review the clinical characteristics and outcome of confirmed IS cases in relationship to vaccine exposure.
Chart review showed that 38% of ICD-coded episodes did not meet the Brighton level 1 criteria for diagnosis of IS, in most cases because no definitive diagnostic or therapeutic intervention, such as enema or surgery, was performed. This proportion is higher than a US inpatient discharge data study and a Swiss capture-recapture study, which both found that 12% of cases did not meet the Brighton level 1 criteria,23,24 but sits between recent US estimates of 25% from the Vaccine Safety Datalink project25 and 54% from the Mini-Sentinel program.9 Almost all of the confirmed Brighton level 1 episodes (94%) in New South Wales were managed at the 3 tertiary pediatric centers, which accept transfers of cases with suspected IS from smaller hospitals that lack access to specialist services. The large geographic area in our study may have resulted in more episodes being duplicates because of interhospital transfer (24%), compared with Switzerland where 9% of total cases were found to be duplicates because of transfer.23 Of non-confirmed episodes, 23% were classified as Brighton level 2. Some of these cases may have been transient, self-resolving episodes of IS, but may also have been coded based on suspicion or false positive ultrasound results from smaller metropolitan and rural hospitals. Our high rate of interhospital transfer would have provided more time for spontaneous resolution of some of the IS episodes to occur, potentially contributing to a smaller proportion of ICD-coded episodes classified as Brighton level 1 in our study compared with others.23,24 Patterns of health care delivery and specialist referral specific for the management of IS will vary among study settings and must be taken into account when evaluating ICD-coded data, although trends in the same setting will be less liable to variation.
We found a similar risk of IS in the 1–7 day period after dose 1 of RV1 to that in the national Australian study that included confirmed cases from New South Wales reported here, together with RV1 exposed infants from other Australian jurisdictions6,7 and to reports from studies conducted in Mexico and the United States.8,10,26 When we conducted a SCCS analysis including all ICD-coded episodes irrespective of the Brighton classification, the point estimate of risk post dose 1 was lower, as would be expected with non-differential misclassification and is consistent with the expectation that non-confirmed episodes would dilute the risk of IS.27 Recently, statistical methods have been developed that can adjust risk estimates for vaccine safety analyses derived from healthcare databases where misclassification is a potential issue.28 The availability of such methods may make the use of large dataset analyses without chart review a more viable option. This data suggest that estimates of risk for this rare condition derived from database studies in the absence of case verification14,15 are likely to be less precise and could inhibit the detection of a real but low-level increased risk of IS.
Our findings that only 61% of coded IS were confirmed also suggest that historical incidence estimates from Australia, and potentially other countries, that were derived using ICD-coded data only, may have overestimated true populated-based incidence of IS. The incidence of IS in Australia was notably higher in Australia (71–81 cases per 100,000)29,30 than that in some other comparable countries (37–57 cases per 100,000 population).14,15,23,31 Differential levels of misclassification between study settings may account for some differences in incidence reported using such data. Of note, however, recent vaccine-attributable risk estimate of 5–6 additional IS cases for every 100,000 infants vaccinated included adjustment of ICD-coded IS confirmed as Brighton level 1.7,8
Almost all (95%) confirmed episodes in our study were initially treated by air enema. In 31% of cases, enema reduction was unsuccessful and surgery was required; a higher proportion than reported in Switzerland (21%).23 The total proportion of cases requiring surgery in our study (35%) was slightly lower than in the United States for Brighton level 1 cases (37–46%),11,15,24 but higher than the Swiss experience (23%).23 Of cases requiring surgical reduction, 10 of 38 (26%) had bowel resection, lower than previously reported (40%).24 In contrast, more than 85% of infants in Mexico and Brazil were managed with surgery, although bowel resection rates were similar to those from high income countries (16–46%).10,26 There were no deaths in our study; however, 5 infants required admission to an ICU. There was no apparent difference in severity between vaccine proximate and other confirmed IS episodes in our study, as measured by length of stay, success of air enema or need for surgery (Table 3); however, power was limited. Vaccine proximate episodes occurred at a significantly younger age, which was not unexpected given the upper age limits for vaccination.
Of all Brighton level 1 episodes in our study treated at the 1 tertiary referral center (Children’s Hospital at Westmead), 10 (17.9%) were found only in the ED database because treatment occurred without hospital admission. This is much higher than the findings of Cortese et al24 in the United States, whose study showed that 4.5% of confirmed IS cases were not admitted to hospital. Together with other findings in this study, this highlights how understanding variation in patterns of diagnosis, clinical management and referral of IS cases is important in assessing the best methods for surveillance of IS and also for analysis of any vaccine-associated change in IS risk over time.
It is notable that 4 of the IS episodes in our study were a second episode in the same individual, although none occurred within 21 days of a vaccine dose. Three of the 4 original episodes were confirmed as Brighton level 1; as were 2 of the 4 repeat episodes. Previous data suggests that approximately 9–14% of infants experience a recurrence of IS.32,33 Although now a contraindication to vaccination, at the time of this study, a prior episode of IS was not considered a contraindication to vaccination.34 What precise risk repeat vaccination poses in a child who has previously had an IS episode is hard to quantify.
In this study, we reviewed the epidemiology and clinical features of IS cases recorded in a defined population of infants with high coverage of RV1 and evaluated this data to assess the risk of IS. We conclude that IS episodes confirmed as Brighton level 1 should ideally be used in a vaccine risk analysis to provide the most accurate measures of RI. However, this is resource intensive and not always feasible, where no alternative exists, all unique ICD-coded episodes could be used for analysis, but this may underestimate risk or fail to detect a low absolute level of risk. Understanding patterns of care for infants with IS, including referral patterns, diagnosis, treatment and outcomes is important to inform such analysis and the overall risk-benefit assessment of vaccination. Our data suggest that the morbidity of IS in recently vaccinated infants does not differ to that in nonvaccine exposed infants; however, this should be confirmed in larger studies and in diverse populations.
We wish to acknowledge all Public Health Unit staff who conducted chart reviews and data extraction for this study and Ms Sue Campbell-Lloyd (Manager, Immunisation Branch, NSW Ministry of Health) for help in initiating the study. We also thank coinvestigators involved in the Australian national Paediatric Active Enhanced Disease Surveillance network (Peter Richmond, Christopher Blyth, Helen Marshall, Michael Gold, Nigel Crawford, Jenny Royle, Jim Buttery, Elizabeth Elliott, Robert Booy and Yvonne Zurynski) for their cooperation in allowing data from Paediatric Active Enhanced Disease Surveillance to be included in this study.
1. World Health Organization Meeting of the immunization strategic advisory group of experts, April 2009—conclusions and recommendations. Wkly Epidemiol Rec. 2009; 84:220–236
2. Ruiz-Palacios GM, Pérez-Schael I, Velázquez FR, et al. Human Rotavirus Vaccine Study Group Safety and efficacy of an attenuated vaccine against severe rotavirus gastroenteritis. N Engl J Med. 2006; 354:11–22
3. Vesikari T, Matson DO, Dennehy P, et al. Rotavirus Efficacy and Safety Trial (REST) Study Team Safety and efficacy of a pentavalent human-bovine (WC3) reassortant rotavirus vaccine. N Engl J Med. 2006; 354:23–33
4. Kramarz P, France EK, Destefano F, et al. Population-based study of rotavirus vaccination and intussusception. Pediatr Infect Dis J. 2001; 20:410–416
5. Murphy TV, Gargiullo PM, Massoudi MS, et al. Rotavirus Intussusception Investigation Team Intussusception among infants given an oral rotavirus vaccine. N Engl J Med. 2001; 344:564–572
6. Buttery JP, Danchin MH, Lee KJ, et al. PAEDS/APSU Study Group Intussusception following rotavirus vaccine administration: post-marketing surveillance in the National Immunization Program in Australia. Vaccine. 2011; 29:3061–3066
7. Carlin JB, Macartney K, Lee KJ, et al. Intussusception risk and disease prevention associated with rotavirus vaccines in Australia’s national immunisation program. Clin Infect Dis. 2013; 57:1427–1434
8. Weintraub ES, Baggs J, Duffy J, et al. Risk of intussusception after monovalent rotavirus vaccination. N Engl J Med. 2014; 370:513–519
9. Yih WK, Lieu TA, Kulldorff M, et al. Intussusception risk after rotavirus vaccination in U.S. infants. N Engl J Med. 2014; 370:503–512
10. Patel MM, López-Collada VR, Bulhões MM, et al. Intussusception risk and health benefits of rotavirus vaccination in Mexico and Brazil. N Engl J Med. 2011; 364:2283–2292
11. Haber P, Patel M, Pan Y, et al. Intussusception after rotavirus vaccines reported to US VAERS, 2006-2012. Pediatrics. 2013; 131:1042–1049
12. World Health Organization Rotavirus vaccines, WHO position paper—January 2013. Wkly Epidemiol Rec. 2013; 5:49–64
13. Bines JE, Kohl KS, Forster J, et al. Brighton Collaboration Intussusception Working Group Acute intussusception in infants and children as an adverse event following immunization: case definition and guidelines of data collection, analysis, and presentation. Vaccine. 2004; 22:569–574
14. Yen C, Tate JE, Steiner CA, et al. Trends in intussusception hospitalizations among US infants before and after implementation of the rotavirus vaccination program, 2000–2009. J Infect Dis. 2012; 206:41–48
15. Zickafoose JS, Benneyworth BD, Riebschleger MP, et al. Hospitalizations for intussusception before and after the reintroduction of rotavirus vaccine in the United States. Arch Pediatr Adolesc Med. 2012; 166:350–355
16. National Health and Medical Research Council The Australian Immunisation Handbook. 2013; Vol 10:ed Canberra Australian Government Department of Health and Ageing
17. Hull BP, Menzies R, Macartney K, et al. Impact of the introduction of rotavirus vaccine on the timeliness of other scheduled vaccines: the Australian experience. Vaccine. 2013; 31:1964–1969
18. Macartney KK, Burgess MA. Rapid impact of rotavirus vaccination in the United States: implications for Australia. Med J Aust. 2009; 191:131–132
19. Hull BP, Deeks SL, McIntyre PB. The Australian Childhood Immunisation Register-A model for universal immunisation registers? Vaccine. 2009; 27:5054–5060
20. Hull B, Lawrence G, MacIntyre CR, et al. Immunisation Coverage: Australia 2001. 2002; Canberra, Australia: Department of Health and Ageing;
21. Farrington P, Pugh S, Colville A, et al. A new method for active surveillance of adverse events from diphtheria/tetanus/pertussis and measles/mumps/rubella vaccines. Lancet. 1995; 345:567–569
22. Whitaker HJ, Farrington CP, Spiessens B, et al. Tutorial in biostatistics: the self-controlled case series method. Stat Med. 2006; 25:1768–1797
23. Buettcher M, Baer G, Bonhoeffer J, et al. Three-year surveillance of intussusception in children in Switzerland. Pediatrics. 2007; 120:473–480
24. Cortese MM, Staat MA, Weinberg GA, et al. Underestimates of intussusception rates among US infants based on inpatient discharge data: implications for monitoring the safety of rotavirus vaccines. J Infect Dis. 2009; 200:(suppl 1)S264–S270
25. Shui IM, Baggs J, Patel M, et al. Risk of intussusception following administration of a pentavalent rotavirus vaccine in US infants. JAMA. 2012; 307:598–604
26. Velázquez FR, Colindres RE, Grajales C, et al. Postmarketing surveillance of intussusception following mass introduction of the attenuated human rotavirus vaccine in Mexico. Pediatr Infect Dis J. 2012; 31:736–744
27. Rothman KJ. Epidemiology—An Introduction. 2002; New York Oxford University Press
28. Xu S, Newcomer S, Nelson J, et al. Signal detection of adverse events with imperfect confirmation rates in vaccine safety studies using self-controlled case series design. Biom J. 2014; 56:513–525
29. Bines JE, Liem NT, Justice FA, et al. Intussusception Study Group Risk factors for intussusception in infants in Vietnam and Australia: adenovirus implicated, but not rotavirus. J Pediatr. 2006; 149:452–460
30. Justice F, Carlin J, Bines J. Changing epidemiology of intussusception in Australia. J Paediatr Child Health. 2005; 41:475–478
31. Patel MM, Haber P, Baggs J, et al. Intussusception and rotavirus vaccination: a review of the available evidence. Expert Rev Vaccines. 2009; 8:1555–1564
32. Daneman A, Alton DJ, Lobo E, et al. Patterns of recurrence of intussusception in children: a 17-year review. Pediatr Radiol. 1998; 28:913–919
33. Justice FA, Nguyen LT, Tran SN, et al. Recurrent intussusception in infants. J Paediatr Child Health. 2011; 47:802–805
34. National Health and Medical Research Council The Australian Immunisation Handbook. 2008; Vol 9th:ed Canberra Australian Government Department of Health and Ageing