Analysis of a multicenter clinical trial on pressure ulcer development by a marginal approach in the Cox model

Background: We exemplify the use of a marginal approach with proportional h azards model when failure times are correlated. Methods: The marginal distribution for each failure time is formulated by t he Cox proportional hazards model, while the dependance structure is unspec ified However, a correct variance-covariance estimate of the regression coe fficients that takes into account the intra-group correlation is proposed, The program MULCOX2 which implements this statistical methodology is used t o assess the effect of a nutritional supplementation intervention on pressu re ulcer development in critically ill older patients from a multicentric t rial (involving 19 wards). We compare the results obtained with those of th e usual Cox regression. Results: The naive approach yields much smaller standard error estimates of the regression parameters than the robust approach. Conclusion: In our example, the results obtained with the marginal approach do not modify the conclusions: a nutritional supplementation intervention tend to decrease significantly the formation of pressure ulcers, However in others situations, ignoring the intra-cluster dependance could lead to inv alid statistical inference, The variability of the estimated effects by MUL COX2 can be quire sensitive to the number of clusters in the sample and to the clusters size.