Observational research--opportunities and limitations

Abstract

Medical research continues to progress in its ability to identify treatments and characteristics associated with benefits and adverse outcomes. The principal engine for the evaluation of treatment efficacy is the randomized controlled trial (RCT). Due to the cost and other considerations, RCTs cannot address all clinically important decisions. Observational research often is used to address issues not addressed or not addressable by RCTs. This article provides an overview of the benefits and limitations of observational research to serve as a guide to the interpretation of this category of research designs in diabetes investigations. The potential for bias is higher in observational research but there are design and analysis features that can address these concerns although not completely eliminate them. Pharmacoepidemiologic research may provide important information regarding relative safety and effectiveness of diabetes pharmaceuticals. Such research must effectively address the important issue of confounding by indication in order to produce clinically meaningful results. Other methods such as instrumental variable analysis are being employed to enable stronger causal inference but these methods also require fulfillment of several key assumptions that may or may not be realistic. Nearly all clinical decisions involve probabilistic reasoning and confronting uncertainly, so a realistic goal for observational research may not be the high standard set by RCTs but instead the level of certainty needed to influence a diagnostic or treatment decision.

Keywords: Bias; Causal Inference; Confounding; Epidemiology; Pharmacoepidemiology; Randomized Controlled Trial.

Figure 1

Cross-sectional study of Native Americans of the Pima and Papago Indian tribes in Southern Arizona on the associations between the GM haplotype Gm^3;5,13,14, native quantum, and diabetes mellitus prevalence. Panel A displays all participants combined with Native quantum of either 0/8, 4/8 or 8/8 by presence of diabetes mellitus in relation to Gm^3;5,13,14 presence or absence. The overall (crude) odds ratio for the association is shown. Panel B displays all participants from Panel A stratified by Native quantum, demonstrating confounding by Native quantum as judged by the discordance between the crude and stratified or Quantum-adjusted results. Panel C demonstrates that Quantum meets the criterion as a confounding variable due to its negative association with Gm^3;5,13,14 and positive association with diabetes prevalence.

Figure 2

A hypothetical population of 2000 identical persons with type 2 diabetes differing only by renal function as measured by serum creatinine and assigned to either metformin or glipizide based on the serum creatinine level. The active treatment, though, is never dispensed, and instead substituted with a identical placebo. An expected difference in mortality is seen between the two groups given the association between poorer renal function and mortality in the glipizide group. This difference cannot be explained by the effect of the active pharmaceutical (since there was none) and therefore represents an example of confounding by indication.