Identification of causal effects in case-control studies

TL;DR

This paper examines when case-control study results can be causally interpreted, focusing on identifying causal effects for various estimands across different sampling schemes and matching strategies.

Contribution

It revisits classical concepts to determine conditions under which case-control studies can yield valid causal inferences for multiple estimands.

Findings

Identifies conditions for causal interpretation in case-control designs

Analyzes effects of matching and sampling schemes on causal estimand identification

Provides a framework for future research on complex case-control settings

Abstract

Case-control designs are an important tool in contrasting the effects of well-defined treatments. In this paper, we reconsider classical concepts, assumptions and principles and explore when the results of case-control studies can be endowed a causal interpretation. Our focus is on identification of target causal quantities, or estimands. We cover various estimands relating to intention-to-treat or per-protocol effects for popular sampling schemes (case-base, survivor, and risk-set sampling), each with and without matching. Our approach may inform future research on different estimands, other variations of the case-control design or settings with additional complexities.