Uses and misuses of statistics: the case of strontium ranelate and the number needed to treat.

In the last 15 years, several pharmacological agents for the prevention of fractures have been developed and commercialized. Most of them showed to be effective in reducing fracture risk. The enhanced availability of drugs to prevent fractures has generated a fierce competition among pharmaceutical companies to conquer a share of the potential market, often with claims of superiority of a drug over another without direct comparisons. The definitive way to compare different treatments would require randomized head to head trials. These trials are expensive, need large samples and are unlikely to be ever performed. Therefore, it has become a common practice to compare pharmacological agents through observational studies on administrative databases or by the indirect comparison of the results of individual randomised-controlled trials (RCT) and their meta-analyses. These studies may produce evidence of clinical value, complementary to that given by RCT. However, without a proper and complete analysis, they may result in a biased picture of effectiveness and be completely misleading. In this article, we critically disclose how such competition may produce biased and misleading picture of evidence, by reviewing the significance of the number needed to treat, absolute risk reduction and relative risk reduction in relation to vertebral fractures prevention with available drugs.