To estimate causal effects, analysts performing observational studies in health settings utilize several strategies to mitigate bias due to confounding by indication. There are two broad classes of approaches for these purposes: use of confounders and instrumental variables (IVs). Because such approaches are largely characterized by untestable assumptions, analysts must operate under an indefinite paradigm that these methods will work imperfectly. In this tutorial, we formalize a set of general principles and heuristics for estimating causal effects in the two approaches when the assumptions are potentially violated. This crucially requires reframing the process of observational studies as hypothesizing potential scenarios where the estimates from one approach are less inconsistent than the other. While most of our discussion of methodology centers around the linear setting, we touch upon complexities in non-linear settings and flexible procedures such as target minimum loss-based estimation and double machine learning. To demonstrate the application of our principles, we investigate the use of donepezil off-label for mild cognitive impairment. We compare and contrast results from confounder and IV methods, traditional and flexible, within our analysis and to a similar observational study and clinical trial.

Background

Exposure to endocrine disrupting chemicals such as bisphenol A (BPA) is primarily from the diet through canned foods. Characterizing dietary exposures can be conducted through biomonitoring and dietary surveys; however, these methods can be time-consuming and challenging to implement.

Methods

We developed a novel dietary exposure risk questionnaire to evaluate BPA exposure and compared these results to 24-hr dietary recall data from participants (n = 404) of the Diet Intervention Examining The Factors Interacting with Treatment Success (DIETFITS) study, a dietary clinical trial, to validate questionnaire responses. High BPA exposure foods were identified from the dietary recalls and used to estimate BPA exposure. Linear regression models estimated the association between exposure to BPA and questionnaire responses. A composite risk score was developed to summarize questionnaire responses.

Results

In questionnaire data, 65% of participants ate canned food every week. A composite exposure score validated that the dietary exposure risk questionnaire captured increasing BPA exposure. In the linear regression models, utilizing questionnaire responses vs. 24-hr dietary recall data, participants eating canned foods 1-2 times/week (vs. never) consumed 0.78 more servings (p < 0.001) of high BPA exposure foods, and those eating canned foods 3+ times/week (vs. never) consumed 0.89 more servings (p = 0.013) of high BPA exposure foods. Participants eating 3+ packaged items/day (vs. never) consumed 62.65 more total grams of high BPA exposure food (p = 0.036).

Conclusions

Dietary exposure risk questionnaires may provide an efficient alternative approach to 24-hour dietary recalls to quantify dietary BPA exposure with low participant burden.

Trial registration

The trial was prospectively registered at clinicaltrials.gov as NCT01826591 on April 8, 2013.