Rotarix Vaccine Effectiveness

A case-control design will be utilized to assess the effectiveness of RV1 in preventing rotavirus-associated hospitalizations and Emergency Department (ED) visits using cases and controls from the 2009-2012 rotavirus seasons. Vaccine exposure among cases will be compared to vaccine exposure among controls. There will be one case group and one control group.

Cases will be obtained from children who are enrolled in active surveillance for acute gastroenteritis being conducted at the two initial surveillance sites. These sites include: Cincinnati Children's Hospital Medical Center (2009-2012) and the Medical University of South Carolina (2009-2012). Cases and controls will be identified retrospectively for the 2008-2011 seasons and prospectively through active surveillance for the 2011-2012 season. The active surveillance programs conduct prospective surveillance for hospitalizations and ED visits due to AGE (acute gastroenteritis) in children < 6 years of age. Even though Rotarix was commercially available on January 1, 2008, the actual date that RV1 was initially used varied across sites. The date that Rotarix was initially used will be determined for each site through examination of the data. Children with laboratory-confirmed rotavirus infection (bulk stool sample positive for rotavirus using a rotavirus EIA) will be included as a case if they were born 2 months prior to the initial date of Rotarix availability at each site. The vaccine record for each case will be obtained to determine the vaccine status of the child.

Controls will be children born 2 months prior to the availability of Rotarix at each site who were enrolled in the active surveillance program at either site and who tested negative for rotavirus. Children who have been previously hospitalized for diarrhea, who have a sibling enrolled in the study, or who have recently moved into the study area (and might not have been eligible for the vaccine) will be excluded.

Laboratory Testing

For children enrolled in the active surveillance program bulk stool specimens are obtained within 14 days of the visit for AGE symptoms and are tested using Rotaclone, a commercial enzyme immunoassay (Meridian Bioscience, Inc) at CCHMC. Children with a positive test for rotavirus are eligible to be cases and children with a negative test for rotavirus are eligible for controls. Specimens positive by EIA for rotavirus will have G and P genotyping done at CCHMC.

Retrospective Enrollment of Children Tested for Rotavirus as Part of Routine Care

With the dramatic decline in rotavirus since the introduction of rotavirus vaccines, the number of children enrolled in surveillance has plummeted compared to pre-licensure surveillance. In order to improve our sample size, we will retrospectively invite children who had a rotavirus test done from August 1, 2008 through June 30, 2012 to participate in this study. The laboratory records of rotavirus testing will be reviewed to identify children with a date of birth >August 1, 2008. Once identified, a letter will be sent to the parents/guardians with a brief explanation of the study and we will ask them to notify us within two weeks if they do not wish to be called. Study staff will then contact the parent/guardian to explain the study and to determine whether the child meets eligibility criteria (same as those outlined for enrollment into surveillance). If the child meets eligibility criteria, the parents/guardians will be asked to allow their child to participate, informed consent will be obtained for the child. If agreed, the same data collected on prospectively enrolled children will be collected and permission will be obtained to contact the child's health care provider for the child's immunization records.

Eligibility

The date that Rotarix was initially used will be determined for each site. Children born 2 months prior to the initial date or later will be eligible to be a case or control for this study. If a child is enrolled more than once during the same season, the visit at which the child tested positive for rotavirus will be selected for inclusion; if the child tested negative at all visits, the child's final visit will be selected for inclusion.

Sample Size

Sample size calculations were done for varying degrees of vaccine efficacy and vaccine coverage. With a vaccine uptake of 20%, 68 cases would allow a detection of vaccine effectiveness of 80% assuming a two-sided test with a significance level of 0.05 and 80% power. If the vaccine uptake is 30% only 44 cases would be needed to detect a vaccine effectiveness of 80%.

One must take into consideration that both cases and controls may have incomplete immunization series. If there are a sufficient number of cases, one dose vaccine efficacy will also be examined.

Statistical Analysis

Analyses will be performed using SAS® (Version 9.2 Cary, NC). Demographic and risk factor variables will be examined. Continuous, parametric data will be presented as mean ± standard deviation (SD); continuous, non-parametric data will be presented as median (inter-quartile range). Categorical data will be presented as frequency (percentage) by category. Continuous, parametric data will be analyzed with ANOVA or Student's t-test. Categorical data will be analyzed with chi-square or Fisher's Exact Test as appropriate.

Logistic Regression will be used to estimate Vaccine Effectiveness (VE) and 95% confidence intervals (CIs) from the adjusted odds ratios (aORs) by using the formula VE = (1-aOR) X 100. Cases and controls will be matched on age, enrollment date, and geographic region. Cases will be compared to matched controls in order to evaluate the effectiveness of full (2 doses) and partial (1 dose) vaccination. Initially, univariate conditional logistic regression analyses will be performed to determine which covariates should be included in the multivariable analysis. The explanatory variable of interest is vaccination status. If sample size permits, vaccine effectiveness will be calculated by serotype. Other possible covariates to be included are: date of birth, age at onset, season, insurance status (public/none versus private) and point of care (hospital or ED). Variables with p-values ≤ 0.20 will selected as candidates for multivariable analysis. Multicollinearity among the covariates will be examined by evaluating variance inflation factors. If collinearity is present, only one of the correlated variables will be entered into the model at a time. Analysis will be rerun choosing alternative correlated variables and the models will be compared. The model with the lowest Akaike information criterion (AIC) will be chosen as the final model. Model selection will include stepwise, forward selection, and backwards elimination. The Hosmer-Lemeshow test will be used to determine goodness of fit.