Relative risk reduction and absolute risk reductionBMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b3626 (Published 16 September 2009) Cite this as: BMJ 2009;339:b3626
All rapid responses
I am aware that this rapid response is a tad late in the day to expect a decent response from Professor Talluri, but I would appreciate it if you can bring this to the attention of Prof Talluri or the current contributor of statistical articles in Endgame, Dr Philip Sedgwick.
I hope to illustrate possibly any method of big pharma using recent statistical reporting trends in manipulating perception of results which is presented in line with current Evidence-based Medicine reporting style
Trial A compared a drug with a placebo for the prevention of stroke, the incidence of stroke at 1 year was 3% in the group receiving drug A (treatment group) and 4% in the group receiving placebo (control group); P=0.01.
Relative risk reduction (RRR) is 25% and Absolute risk reduction (ARR) is 1%.
Hence the motivation behind recent trends to promote ARR as a better gauge of the clinical significance in the effect of the drug.
Trial B compared another drug B with a placebo for the prevention of stroke in a different study population, the incidence of stroke at 1 year was 6% in the group receiving drug B (treatment group) and 8% in the group receiving placebo (control group); P=0.01.
Relative risk reduction (RRR) is still 25% and Absolute risk reduction (ARR) is 2%.
Should drug B be considered better than drug A on the basis of a larger ARR (assuming all other factors between the 2 drugs are identical)?
I believe not, since the population cohort is different since that baseline stroke rate (placebo arm) for Trial B is double that of A.
However the big pharma is using the latest fashion of using ARR as a solo direct comparison of drug effects between different competing drugs to promote their products without discussing the study population characteristics. I believe the uninitiated and in some cases, authors of review articles and meta-analysis are also falling into the trap using ARR mainly to determine which drug produce more clinically significant effect.
If my analysis is correct and this trend is real, it represent another disturbing phenomenon in modern medicine that we can no longer rely on the self-appointed guardians producing meta-analysis and reviews
"quis custodiet ipsos custodes?"
Satires of Juvenal, Satire 6.346–348
Competing interests: No competing interests