Negative controls are increasingly used to detect and correct for bias in observational studies, but their effectiveness relies on sharing the same confounding structures as the primary exposure-outcome relationship (U-comparability). Despite their growing use, limited research exists evaluating whether the assumption is met in applied studies. This dissertation addresses this gap through three interconnected aims. Aim 1 systematically reviewed sixteen pharmacoepidemiologic studies employing negative control analyses. The review found that while researchers often reported covariate distributions within negative control analyses, they rarely compared confounding structures directly across primary and negative control analyses. When comparisons were made, methods varied considerably, with no standardized approach for assessing U-comparability. Aim 2 developed and evaluated covariate ranking similarity as a diagnostic tool for assessing U-comparability of negative control exposures (NCEs). This approach compares the rankings of measured covariate effects on the primary exposure and candidate NCEs using rank correlation. Through Monte Carlo simulation, the method was tested under varying conditions of unmeasured confounding strength, direct effects of measured covariates on primary exposure/candidate NCEs not mediated through unmeasured confounders, and measured covariate-unmeasured confounder correlations. Results demonstrated that U-comparable NCEs showed strongly positive rank correlations (Spearman coefficient up to 0.811) that increased with confounding strength. The method performed best when direct covariate effects on exposures were minimized and measured covariate-unmeasured confounder correlations had high variance. Aim 3 applied the covariate ranking similarity approach in a real-world pharmacoepidemiologic study examining associations between oral bisphosphonate (BP) adherence and cardiovascular (CV) events. Using NCEs as proxies for unmeasured confounders, we empirically evaluated multiple candidate negative control outcomes (NCOs) for their U-comparability. Pneumonia and certain preventive services demonstrated strong confounding similarity with the primary outcome as NCOs, while accidents, transfusions, as well as influenza vaccination failed to show comparable associations with NCEs. Limiting bias detection to the empirically validated U-comparable NCOs produced more consistent conclusions on confounding due to overall health/functional status and health-seeking tendency in the CV protective effect of oral BP adherence. This dissertation provides a systematic framework for empirically validating U-comparability in negative control analyses, potentially improving confidence in detecting unmeasured confounding in observational studies.