In his talk at the Duke-NUS Graduate Medical School Singapore, Dr. Marco Huesch cautioned investigators on selection biases that can color comparative effectiveness research, while he advocated for valid approaches to control for them. Dr. Huesch’s talk was based on a study he recently published, where he gauged the sensitivity of observational long-term coronary stent outcomes results to unmeasured confounding in a US state from 2004-2008.
About his talk:
Understanding the comparative effectiveness of treatments in the community requires valid approaches to control for selection biases such as confounding by indication. This
can arise if characteristics of a patient that affect the real-world choice of treatment also affect the outcomes of interest.
For example, drug-eluting coronary stents (DES) were shown to be more effective compared with the older bare metal stents (BMS) in initial randomized clinical trials and in large observational studies with propensity score adjustment. Yet if patients who receive BMS are sicker in ways not captured by observed confounders, then propensity score approaches may fail. The resulting selection bias may distort the comparative effectiveness of DES. For example, physicians may have assessed the patient’s ability to afford or to comply with subsequent antiplatelet therapy in ways that were unobserved by the analyst. BMS may then be used in patients independently more prone to adverse outcomes.
In the study illustrated during the lecture, we gauged the sensitivity of observational long-term coronary stent outcomes results to unmeasured confounding in one US state in 2004-2008. We exploit a technique originating in work by Cornfield and colleagues (1954) on the causal relationship between smoking and lung cancer, and further described by Greenland (1996). We also conduct an IV analysis using observed physician preferences as an instrument for stent type.
The tools described will be of broad interest to investigators conducting observational studies of the real-world effectiveness of treatments in settings in which unmeasured confounding cannot be ruled out. Such settings are increasingly common in comparative effectiveness research.
Information about the talk can be found here: http://www.imcb.a-star.edu.sg/bseminars/20130108b.pdf
If you have access to Wiley, you can read the article using the link below.