Investigating the correlation structure of quadrivariate udder infection times through hierarchical Archimedean copulas.

The correlation structure imposed on multivariate time to event data is often of a simple nature, such as in the shared frailty model where pairwise correlations between event times in a cluster are all the same. In modeling the infection times of the four udder quarters clustered within the cow, more complex correlation structures are possibly required, and if so, such more complex correlation structures give more insight in the infection process. In this article, we will choose a marginal approach to study more complex correlation structures, therefore leaving the modeling of marginal distributions unaffected by the association parameters. The dependency of failure times will be induced through copula functions. The methods are shown for (mixtures of) the Clayton copula, but can be generalized to mixtures of Archimedean copulas for which the nesting conditions are met (McNeil in J Stat Comput Simul 6:567-581, 2008; Hofert in Comput Stat Data Anal 55:57-70, 2011).

Outcome of Epithelial Ovarian Cancer: Time for Strategy Trials to Resolve the Problem of Optimal Timing of Surgery.

INTRODUCTION: The standard treatment of ovarian cancer is the combination of debulking surgery and chemotherapy. There is an ongoing discussion on which treatment is best: primary debulking surgery (PDS) or neoadjuvant chemotherapy with interval debulking (NACT-IDS). Even a large randomized trial has not settled this issue. We examined whether comparing a specified treatment protocol would not be a more logical approach to answer this type of discussions. METHODS: A retrospective study of 142 consecutively treated patients according to a fixed protocol between 2000 and 2012 was conducted. Disease-free survival and overall survival were calculated by univariate and multivariate analyses for the whole group and for advanced stages separately. Specific differences between PDS and NACT-IDS were studied. Comparison of results from large databases was made. RESULTS: Disease-free survival and overall 5-year survival for the whole group were 35% and 50%. For the advanced stages, disease-free survival and overall 5-year survival were 14% and 36%, with a median disease-free and overall survival of 16 and 44 months. Of the 98 women with advanced ovarian carcinoma, 54% of operable patients underwent PDS and 44% underwent NACT-IDS. More patients in the PDS group were optimally (<1 cm) debulked: 80% vs 71%. There was no significant difference in survival between PDS or NACT-IDS. Optimally debulked patients had a significant better overall survival in multivariate analysis with a hazard ratio of 2.1. DISCUSSION: Outcome of treatment according to a fixed protocol with a mixture of PDS and NACT-IDS was similar to results from large databases. We hypothesize that comparison of a specific strategy may yield more useful results than awaiting the perfect randomized trial.