Purpose: To quantitatively evaluate the complementary roles of two risk-benefit estimation methods by applying them to the data comparing low molecular weight heparin (LMWH) to regular heparin for DVT prevention following trauma published by Geerts et al (NEJM 1996).

Methods: Method 1: the Lynd & O'Brien (J. Clin. Epi 2004) approach was used to plot 1) the incremental probability of a DVT prevented against the incremental probability of an additional bleed expressed as an incremental risk-benefit ratio ; 2) the risk-benefit acceptability curve, a plot of the probability that LMWH is net-beneficial relative to heparin at any risk-benefit acceptability threshold (µ). Method 2: the Sutton et al (J. Clin. Epi 2005) Bayesian method for calculating net clinical benefit (benefit - harm) was also applied to the same data. Uniform priors for DVT and bleeding risk were assumed for both LMWH and heparin. Trial data from the Geerts paper were combined with the priors to estimate posterior distributions of DVT and bleeding risk. Both approaches use preference measures to convert from risk units to benefit units. They are µ for Method 1, and outcome ratio (OR =1/µ) for Method 2. µ = 1 when 1 bleed = 1 DVT. µ = 1/4 when 4 DVTs = 1 bleed.

Results: Method 1 shows that the probability that LMWH is net-beneficial relative to heparin is 96.5% when µ = 1; the probability drops to 69.5% when µ = 1/4. For Method 2 when mean OR =1 and variance =1/4, net clinical benefit is -0.106 (106 less DVT events per 1000 patients on LMWH versus H). The probability that net clinical benefit is in favor of LMWH is 95.4%. Assuming a mean OR =4 and variance = 4, net clinical benefit is only -.0361 and the probability that net clinical benefit is in favor of LMWH drops to 67.7%.

Conclusion: Method 1 provides the decision maker a functional relationship between different values of µ (preference) and the probability that LMWH provides a net benefit. Method 2 on the other hand allows the decision maker to frame the uncertainty around the preference measure OR (1/µ) and how it affects the probability that LMWH provides a net benefit. Methods 1 and 2 can be used together for better decision making.