Monday, May 23, 2011

Why restrict the types of NSAIDs used as active comparators?

I am always surprised when people do not include negative controls when there is a very obvious candidate to be used as such. Consider this article:

Chronic analgesic use with either COT or COX-2 was associated with an increased risk of cardiovascular outcomes. These findings suggest either a selection of high-risk patients to chronic analgesic treatment, coupled with unmeasured or residual confounding, or a potential cardiovascular effect of these medications. Further research is warranted to evaluate causes for this association.


Why did the researchers not use an active comparator that is known to be null (a negative control)? After all, the participants who are tkaing pain medication may be systematically different from those who do not. It is prescription claims data so it is unclear whether or not you can adjust for these kinds of differences.

So why would you not at least look at Naproxen and Ibuprofen users?

Yes, the categories were: Opioids, Rofecoxib, Celecoxib,Valdecoxib, and General population. Covariates were:

We derived variables representing demographics, medical history of angina, coronary heart disease (CHD), congestive heart failure, arrhythmias, ischemic stroke, transient cerebral ischemia, peripheral vascular disease, diabetes, hypertension, hyperlipidemia, hypercholesterolemia, smoking, and obesity; and dispensing of nitrates, anti-platelet agents, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, beta-blockers, calcium channel blockers, and diuretics. Baseline history of chronic diseases of the musculoskeletal system, diseases of the esophagus, hyperthyroidism, medical care required for general ill-defined symptoms and respiratory or chest symptoms, including dyspnea and upper respiratory symptoms, were also included.


Now I have done a paper on the misclassification of ibuprofen and naproxen in claims data but the issue there was sensitivity and not specificity. There is no reason that naproxen or ibuprofen could not be negative controls (or that aspirin could not be a positive control). It would certainly make the unexpected results of this analysis easier to interpret!

No comments:

Post a Comment