### Article

## A critical assessment of the proportion in favor of treatment as a new effect measure for clinical trials with composite endpoints

### Search Medline for

### Authors

Published: | September 13, 2012 |
---|

### Outline

### Text

**Introduction:** Composite endpoints combine several events of interest within a single variable. They are commonly used if the clinical effect of interest cannot directly be assessed by a single specific outcome. Moreover, using a composite measure instead of single event variable increases the number of expected events and is thus meant to increase the power. However, the interpretation of results can be difficult as the observed effect for the composite does not necessarily reflect the effects for the components. Even if a large positive effect in the composite has been observed, the effects of some components may be of very different magnitude or even point in adverse directions. This is especially a problem if the single endpoints forming a composite are of different clinical relevance.

Common effect measures such as the hazard ratio for composite time-to-event endpoints or the event rate difference for composite binary endpoints give equal weight to all types of events irrespective of their clinical relevance. An effect measure which takes account of the different priorities of the components was proposed by Buyse [1] and Pocock et al. [1], [2]. The idea of the new effect measure, which Buyse [1] referred as the ‘proportion in favor to win’, is as follows: All outcome variables of interest are ordered with respect to their clinical importance. Every patient in the intervention group is compared to every patient in the control group and the patient with the more favorable outcome with respect to the endpoint of primary priority is determined. If no decision can be done the comparison is made based on the endpoint of secondary priority and so on. By this, the effect measure gives higher weight to more important endpoints which is meant to reduce masking effects. Some of the properties of this new effect measure have been studied by Buyse [1]. However, no systematic evaluation of the new effect measure has been performed so far.

**Material and Methods:** We compare the proportion in favor of treatment to standard effect measures such as the hazard ratio or the absolute rate difference. Moreover, we apply it to a variety of data situation resulting from simulations and from real clinical trial examples. We discuss the advantages and problems regarding the new effect measure and give recommendations for its use.

**Results:** In many applications, the proportion in favor of treatment solves the problem of masking by incorporating the priority of the components. However, there also exist situations where masking effects still occur so that the new effect measure provides no gain in interpretability. Moreover, the derivation of distributional assumptions and convergence properties is rather complicated as the pairwise comparisons are stochastically dependent.

**Discussion:** The proportion in favor of treatment is an interesting approach which solves some of the problems commonly met when dealing with composite endpoints. However, there also exist some serious disadvantages regarding interpretability of the effect measure and its statistical properties. Therefore, the decision which effect measure to choose should be done carefully depending on the specific clinical trial application.

### References

- 1.
- Buyse M. Generalized pairwise comparisons of prioritized outcomes in the two-sample problem. Stat Med. 2010;29:3245-57.
- 2.
- Pocock SJ, Ariti CA, Collier TJ, Wang D. The win ratio: a new approach to the analysis of composite endpoints in clinical trials based on clinical priorities. Eur Heart J. 2012;33:176-82.