Posts tagged Meta-analysis
As ever, the best piece of writing (in my humble opinion) from over the weekend comes from the “Bad Science” blog. Ben Goldacre’s mind is hardwired for statistics, and this week he has turned it to the oft touted health benefits of smoking, and in particular, protection against Alzheimers. He discusses a recently published systematic review which – surprise surprise (and I’m sure I’m not spoiling it for anyone here – hey? what? smokings bad! You don’t say) – shows that the risk of alzheimers is increased in smokers. The really interesting part however, if you are into this sort of thing, is the analysis of papers where the researchers were associated with the Tobacco industry. Now that is a surprise I won’t spoil.
A recent paper in Chest has suggested that the staple treatment for COPD may not be the knees of the bees after all. Inhaled corticosteroids (ICS) have been the cornerstone of COPD management for some time, with the aim of reducing inflammation and (hopefully) cutting exacerbation rate. The TORCH study successfully did away with thoughts of ICS improving survival (ahh, the agony of p=0.052) but at least we have the principle of symptom control to cling to don’t we? Well, don’t we?
The clever bods from Chandigarh, India have taken 11 large prospective, double-blind, randomized, placebo-controlled trials and passed it through their Metaregression-ometer and come up with the conclusion that “The benefits of ICS in preventing COPD exacerbations seems to be overstated”. The collected statistics from the 8,164 patients pooled from seminal papers including ISOLDE, TRISTAN and TORCH showed that “we did observe a modest benefit of ICS in preventing COPD exacerbations in patients with FEV 1 < 50%. However, this finding was not further substantiated on a metaregression analysis wherein we did not observe a significant effect of baseline values of FEV 1 (percentage predicted) on the benefits from ICS compared with placebo. Thus, ICS is likely to have only modest benefits in preventing COPD exacerbations, if at all, and should be judiciously used in patients with COPD keeping in mind the risk-benefit ratio.”
I’ll admit my utter ignorance of the nuances of metaregression analysis and its potential flaws when used like this, but I would hope that the use of words such as “likely” and “modest” would hint that the analysis has scope to be challenged by people with more statistical savvy than my good self. Mind you, it doesn’t really matter does it, what with the veritable panacea of COPD drugs at our disposal…………(right, anyone got shares in Roflumilast?).