Relative Risk Reduction vs Absolute Risk Reduction
Posted Aug 01 2011 8:30am
Like most pre-med students, I was not a big fan of statistics. Sure, there was a time I could tell you about all the mathematical nuances of Major League Baseball, such as on-base average or slugging percentage, but that was way back during the days of Catfish Hunter , Rollie Fingers & Campy Campaneris . But dry numbers? Yuck, at least for me. However, my lack of attention to statistics has come back to bite me (and most likely, a large proportion of my colleagues in the trenches) as pharmaceutical representatives deluge us daily w/new studies reporting tremendous relative risk reductions in disease-oriented evidence (rather than absolute risk reductions in patient-oriented outcomes that matter).
Case in point: a re-analysis of the Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER - the cardiologists have the best acronyms for their studies) was published in the Journal of the American College of Cardiology demonstrating anywhere from 44% relative risk reduction in cardiovascular events in anyone randomized to rosuvastatin compared to placebo up to 65% reduction in those who achieved LDL <50mg/dL.
Impressive right? Especially considering the short duration of the study, just an average of 2 years for the 17,802 individuals w/o cardiovascular (CV) disease and diabetes but w/CRP >2mg/L and LDL <130mg/dL. In those randomized to placebo, the authors reported 1.18 CV events per 100 person-years, while in those who received rosuvastatin, the authors reported 0.86 CV events per 100 person-years in those who achieved LDL >50mg/dL and 0.44 CV events per 100 person-years in those who achieved LDL <50mg/dL. Therefore, we should be putting this in the water, right?
But look more closely at the units - per 100 person-years. In other words, if you followed 100 persons for one year, just more than 1 would have a cardiovascular event. To be more precise, you'd need to follow 10,000 individuals for a year to observe 118 CV events (or 1,000 individuals for 10 years or 100 individuals for 100 years).
Likewise, you'd have to follow 100 persons on rosuvastatin for one year w/LDL >50mg/dL in order to observe less than one CV event. Or more precisely, you'd need to follow 10,000 individuals for a year to observe 86 CV events (or 1,000 individuals for 10 years or 100 individuals for 100 years).
Same idea for those individuals on rosuvastatin who achieved LDL <50mg/dL. Sure, the relative risk reduction (RRR) is impressive. (1.18-0.86)/1.18 = 27% RRR and (1.18-0.44)/1.18 = 63% RRR, respectively depending upon LDL achieved. This is what we're told. Just look at that tremendous improvement!
But think about how many people you'd need to treat in order to get those outcomes. Number needed to treat (NNT) is the reciprocal of the absolute risk reduction. So NNT = 1/(0.0118-0.0086) = 312.5. In other words, you'd need to treat 312 individuals to prevent 1 cardiovascular event per year. Smaller NNTs are better; larger NNTs are worse since you'd have to treat more patients to get benefit. And given the cost of branded rosuvastatin (which isn't available as a generic), that's a large slice of the healthcare pie going to towards just one disease.
Granted, this is how those in power (read insurance companies & Big Pharma) tend to think - bottom line dollars. You can decide with your own dollars whether or not it makes sense to spend your own money this way. Or perhaps it's better to invest your own hard-earned dollars in better food choices and some daily physical activity. I'm not suggesting that we throw away all statins. However, I do recommend that we think twice before falling for the glossy ads hook, line & sinker. By the way, you have one guess as to who paid for this study (and no, it wasn't the government).