Confounding or bias amplification? Clues for the researcher seeking causal inference.

In epidemiology studies, associations between exposure and outcome could be affected by the well-known problem of confounding, but also by amplification of bias due to the presence of unknown or unmeasured confounders and correlation between the exposure of interest and co-occurring exposures. However, determining which of these types of bias – or both! – may exist is not straightforward, and requires knowledge of and assumptions about, the underlying causal structure. An example where both ‘traditional’ confounding and bias amplification may occur, is examination of health effects due to perfluoroalkyl and polyfluoroalkyl substances (PFAS), where correlation between PFAS is usually present, and the sources (and therefore potential confounders) are not always well understood. As an example, several epidemiological studies have reported associations between PFAS biomarker levels and immune endpoints including vaccine response. In one of these studies, correlations between PFAS were moderate to high (range: 0.22 to 0.78), and results were presented for both single PFAS and multi-PFAS models. In some cases, adjusting for other PFAS changed point estimates (e.g., a 37% attenuation of the point estimate of PFDA in multi-PFAS models, compared with the single-PFAS model) while in other cases (e.g., PFOA), results were similar. We performed a simulation exercise with a dataset constructed to mimic the data from this study and examined how point estimates were changed with different causal assumptions. We found that results were sensitive to assumptions about the underlying presence and strength of associations, but that when amplification bias was present the ‘better’ choice is the single PFAS model. If the strength of associations or degree or correlation were changed, however, the balance could shift such that the multi-PFAS model yielded lower bias.    *Disclaimer: The views expressed in this abstract are those of the authors and do not necessarily reflect the views or policies of the U.S. EPA

Impact/Purpose

Confounding is a commonly encountered phenomenon in epidemiology but in practice it is difficult to determine when study results are impacted by confounding alone, or by amplification of bias due to co-occurring exposures with unmeasured confounders (or both). This abstract describes the challenge and some theoretical comparisons using simulated data.

Citation

Christensen, K., M. Leung, E. Radke-Farabaugh, J. Wright, T. Bateson, AND M. Weisskopf. Confounding or bias amplification? Clues for the researcher seeking causal inference. Society for Epidemiologic Research 2024 Annual Meeting, Austin, TX, June 18 - 21, 2024.