Aedating 3 0
Therefore, the conditional Poisson regression model can be implemented using a Cox proportional hazards framework, thus simplifying the analysis of SCCS data using standard statistical packages.In addition, Heinze and Schemper (19) have demonstrated that inference based on the profile penalized likelihood is preferable to the Wald test statistic for the Firth correction method.One main drawback of this approach is that it depends on the existence of the MLE.Often when a sample size is small, data separation occurs if no events are observed in one of the 2 groups defined by a dichotomous covariate (or no events are observed in either the risk period or the control period(s) in the SCCS design), and no MLE is produced.In this study, we used simulations to evaluate 2 bias correction approaches—the Firth penalized maximum likelihood method and Cordeiro and Mc Cullagh's bias reduction after maximum likelihood estimation—with small sample sizes in studies using the SCCS design.The simulations showed that the bias under the SCCS design with a small number of cases can be large and is also sensitive to a short risk period.Several bias correction methods have been studied in generalized linear models for matched case-control studies (11–13), but none of these correction methods have been evaluated in the SCCS study design.
However, limitations still exist when the risk period in the SCCS design is short relative to the entire observation period.
IRRs compare the incidence rate of adverse events in the risk period with the rate in the control period(s).
The IRRs are estimated using conditional Poisson regression models, conditioning on the number of events and the exposure history experienced by each individual during a predetermined observation period. The MLEs of α and β can be obtained using standard statistical software such as STATA (17) and SAS (18).
The procedure used to generate the data set followed the simulation structure described by Glanz et al. Briefly, we first created a vaccinated cohort population of 100,000 under the combination of the parameters as specified below.
Then the cases were simulated using the probability model represents person-time during the interval.In all simulations, the observation period was set to 365 days for all individuals.