Table 2.

Summary of Published Studies Using Nonequivalent Exposures

Citation	Nonequivalent Outcome	Nonequivalent Exposure	Primary Study Endpoint	Exposure or Treatment	Primary Study Effect Measure	Nonequivalent Outcome Effect	Nonequivalent Exposure Effect	Proposed Source of Bias	Causal Criteria
Zaadstra, Mult Scler. 2008	N/A	Broken arm, concussion, tonsilectomy	Multiple sclerosis	Childhood infections (rubella, chicken pox, mono, measles, mumps)	Odds ratios from logistic regression	N/A	Patients with MS had higher rates of concussion and tonsillectomy	Recall bias	Unstated
Dusetzina, Breast Cancer Res Treat. 2013	N/A	Aromatase inhibitors	Co-prescribing of antidepressants and endocrine therapy	Tamoxifen	Risk ratios from binomial regression using a difference-in differences approach	N/A	Greater decline in strong inhibitor antidepressant use among tamoxifen users than aromatase inhibitor users	Unobserved confounding, temporal changes in antidepressant use	No biological plausibility
Rasmussen, JAMA. 2007	Cancer-related hospital admissions	Calcium channel blockers	Long-term mortality post-MI	Adherence to statins and beta-blockers	Hazard ratio from Cox model	No increase in cancer-related admissions by statin or beta-blocker adherence level	Adherence to calcium channel blockers was not associated with mortality endpoints, as expected	Healthy user bias	No biological plausibility