Stat Med

Global and partial agreement among several observers.

Year 1998
Rogel A. Boelle PY. Mary JY.
Centre de Bioinformatique, INSERM U444, Universite Paris 7, France.
The advantage of modelling agreement on a categorical scale among observers rather than using summarizing indices is now well established. However, analysis of agreement among more than two observers is essentially based on pairs of observers. We present a global and partial agreement modelling approach derived from the quasi-independence and quasi-symmetry log-linear models. This approach addresses high order interactions in the contingency table rather than two-way interaction in the pairwise agreement approach. Pairwise, global and partial agreement models were applied to the detection by six pathologists of three lesions in biopsy specimens arising from patients suspected to be affected by Crohn's disease. The global and partial agreement approach surpasses the pairwise agreement approach, especially if there is heterogeneity among ratings.

The multi-centre assessment of quality of life: the Interdisciplinary Group for Cancer Care Evaluation (GIVIO) experience in Italy.

Year 1998
Mosconi P. Torri V. Cifani S. Ruggiata R. Meyerowitz BE. Apolone G. Liberati A.
Istituto di Ricerche Farmacologiche Mario Negri, Milano, Italy. mosconi@irfmn.mnegri.it
One of the main issues to be considered in conducting clinical trials concerns the presence of missing data. This aspect is particularly relevant in oncology longitudinal studies, characterized by a long follow-up, and especially in quality of life studies where there is still little knowledge about patients' characteristics that predict loss of data. Since the middle of the 1980s the GIVIO (Interdisciplinary Group for Cancer Care Evaluation) co-operative group has been involved in conducting quality of life assessment studies, also focusing on the development of some strategies aimed at the minimization of missing data. In this paper we report on the results of two trials, which are now completed, concerning the quality of life assessment in a sample of breast and colon cancer patients. In order to cope with the problem of missing data, in both the trials the strategy of follow-up mailing was adopted, which proved to be an effective way to increase the response rate by nearly 50 per cent at each time point.

A comparative analysis of quality of life data from a Southwest Oncology Group randomized trial of advanced colorectal cancer.

Year 1998
Troxel AB.
Columbia University School of Public Health, New York, NY 10032, USA. atroxel@biostat.columbia.edu
Longitudinal quality of life measurements from an advanced-stage cancer clinical trial are analysed using a variety of methods, and the results compared. The methods used require different assumptions about the mechanism that produces the missing data. They include analyses that require the data to be missing completely at random; fixed-effects models and weighted generalized estimating equations, which require missing at random data; and a fully parametric approach where the outcomes and the missingness mechanism are jointly modelled, allowing non-ignorable missing data. The data show evidence of non-random missingness, but a formal test of non-ignorable missing data is not significant.

Parametric likelihoods for multiple non-fatal competing risks and death.

Year 1998
Shen Y. Thall PF.
Department of Biomathematics, UT M.D. Anderson Cancer Center, Houston, Texas 77030, USA.
Clinical trials of fatal diseases often focus on one or more non-fatal events, in addition to survival, both to characterize morbidity and to improve survival estimates. Three statistical complications are that the time to each non-fatal event and subsequent residual survival may be either positively or negatively associated, the times to death with or without an antecedent event often have very different distributions, and death may censor some of the non-fatal event times. Consequently, the overall survival time distribution is a mixture of the distributions corresponding to the possible antecedent non-fatal events. These conditions violate the usual assumptions underlying many statistical methods for analysing multivariate time-to-event data. In this paper, we consider a general parametric model for multiple non-fatal competing risks and death. The model accounts for positive or negative association between the time of each non-fatal event and subsequent survival while accommodating covariates and the usual administrative censoring. Each event time distribution is specified marginally by a three-parameter generalized odds rate model, and the time of each non-fatal event and subsequent residual survival are combined under a bivariate generalized von Morgenstern distribution. The approach is illustrated by application to two data sets from clinical trials in colon cancer and acute leukaemia.