Selecting the most informative positive and negative controls for self-controlled case series (SCCS): Rationale, approach, and lessons from studies investigating the safety of COVID-19 vaccines

The self-controlled case series (SCCS) is a study design used in epidemiology to test for variation in outcomes for a group of individuals before and after an exposure or intervention, with each pre-intervention individual serving as a control for themselves in the post-intervention period. The design is particularly useful for studying rare events or outcomes that occur in a time-dependent manner, such as uncommon side-effects of a medical intervention [1–3]. In the Early Pandemic Evaluation and Enhanced Surveillance of COVID-19 (EAVE II) study, we used this SCCS design to evaluate the safety of coronavirus disease 2019 (COVID-19) vaccines in various age groups of the Scottish and UK populations [4–13].

The SCCS design involves using data only from individuals who have the outcome of interest; for this reason, the study population for analysis consists of individuals who have experienced the outcome and, in most cases, the exposure or intervention of interest. However, it is also possible to include individuals with the outcome, but not the exposure of interest, which can further strengthen the study design if the additional information obtained in this way helps answer the research question [1–3]. Controls who are either exposed or not exposed will not contribute to the precision of comparisons between the periods before and after exposure in those who experienced the outcome. However, they can still serve to increase the precision of the different estimates in the period where there was or was not an exposure.

Each study subject can thus serve as their own control; however, this typically requires the effect of the exposure to be transient. In SCCS, this approach eliminates the need for external unexposed control groups. The design takes advantage of within-person comparisons over time. It compares the occurrence of the outcome (incidence rate) during exposed (risk) periods to the occurrence of the outcome (incidence rate) during baseline (control, unexposed) periods within the same individuals. In vaccine safety studies, the exposure (risk) period is a short period of time following vaccination; the baseline (control, unexposed) period is usually a period of time before vaccination, but may also include time after the exposure period has ended, as the effect of the exposure is assumed to be transient. This allows for the calculation of the incidence rate ratio (IRR) [1–3]. We recap these standard features of SCCS before discussing a beneficial addition to the typical SCCS methodology: incorporating positive and negative controls.

Positive controls, where there is known to be a causal relationship between the exposure and outcome [14], and negative controls, where there is no such causal relationship between the exposure and outcome [15], are control comparisons that can be used to help detect the presence of residual confounding. It is important to note that there can be different types of positive or negative controls: these can be exposure-based or outcome-based. In this regard, they play a similar role to positive or negative controls in laboratory assays [16,17].