Bias amplification of unobserved confounding in pharmacoepidemiological studies using indication-based sampling

Key points

-Restricting a pharmacoepidemiological study to those with drug indications can result in a net increase in bias through bias amplification.

-It is, generally, less biased to use regression adjustment than to restrict the sample to those with a particular indication.

-The amount of bias amplification depends largely on the relationship and strength of the relationship between factors.

-Even if confounding by indication exists, it is often not warranted to restrict analyses to those with a particular drug indication.
Bias amplification should be considered in the list of biases and other considerations during the design phase of any pharmacoepidemiological investigation using observational data.



Abstract

Purpose
Estimating causal effects in observational pharmacoepidemiology is a challenging task, as it is often plagued by confounding by indication. Restricting the sample to those with an indication for drug use is a commonly performed procedure; indication-based sampling ensures that the exposed and unexposed are exchangeable on the indication - limiting the potential for confounding by indication. However, indication-based sampling has received little scrutiny, despite the hazards of exposure-related covariate control.

Methods
Using simulations of varying levels of confounding and applied examples we describe bias amplification under indication-based sampling.

Results
We demonstrate that indication-based sampling in the presence of unobserved confounding can give rise to bias amplification, a self-inflicted phenomenon where one inflates pre-existing bias through inappropriate covariate control. Additionally, we show that indication-based sampling generally leads to a greater net bias than alternative approaches, such as regression adjustment. Finally, we expand on how bias amplification should be reasoned about when distinct clinically relevant effects on the outcome among those with an indication exist (effect-heterogeneity).

Conclusion
We conclude that studies using indication-based sampling should have robust justification - and that it should by no means be considered unbiased to adopt such approaches. As such, we suggest that future observational studies stay wary of bias amplification when considering drug indications.




Plain language summary

To understand the benefits and harms of drug use epidemiologists often rely on observational data – where the world is observed as it naturally occurs. In doing so, epidemiologists can study a wide range of phenomena. But this is not straightforward, as it is not always clear why some individuals use a drug and some do not. The difference between drug users and non-users may obscure the results, sometimes resulting in an incorrect conclusion. One commonly used approach to resolve this complexity is to study individuals who share a similar indication for drug use, and assume that this makes individuals more comparable. However, in this study, we describe that doing so is not always as beneficial as it might seem. In fact, we show that restricting the study to individuals with a particular drug indication – as is commonly done in pharmacoepidemiology – can hamper rather than aid epidemiologists in their pursuit. Using causal graphs, mathematical simulations and real-world examples we showcase how this bias appears and what epidemiologists can do to resolve it. We conclude our study with a recommendation that epidemiologists should consider this form of bias in all pharmacoepidemiological investigations.