Search

A popular choice when analyzing ordinal data is to consider the cumulative proportional odds model to relate the marginal probabilities of the ordinal outcome to a set of covariates. However, application of this model relies on the condition of identical cumulative odds ratios across the cut-offs of the ordinal outcome; the well-known proportional odds assumption. This paper focuses on the assessment of this assumption while accounting for repeated and missing data. In this respect, we develop a statistical method built on multiple imputation (MI) based on generalized estimating equations that allows to test the proportionality assumption under the missing at random setting. The performance of the proposed method is evaluated for two MI algorithms for incomplete longitudinal ordinal data. The impact of both MI methods is compared with respect to the type I error rate and the power for situations covering various numbers of categories of the ordinal outcome, sample sizes, rates of missingness, well-balanced and skewed data. The comparison of both MI methods with the complete-case analysis is also provided.We illustrate the use of the proposed methods on a quality of life data from a cancer clinical trial.

Purpose Patients with locally advanced gastric cancer benefit from combined pre- and postoperative chemotherapy, although fewer than 50% could receive postoperative chemotherapy. We examined the value of purely preoperative chemotherapy in a phase III trial with strict preoperative staging and surgical resection guidelines. Patients and Methods Patients with locally advanced adenocarcinoma of the stomach or esophagogastric junction (AEG II and III) were randomly assigned to preoperative chemotherapy followed by surgery or to surgery alone. To detect with 80% power an improvement in median survival from 17 months with surgery alone to 24 months with neoadjuvant, 282 events were required. Results This trial was stopped for poor accrual after 144 patients were randomly assigned (72: 72); 52.8% patients had tumors located in the proximal third of the stomach, including AEG type II and III. The International Union Against Cancer RO resection rate was 81.9% after neoadjuvant chemotherapy as compared with 66.7% with surgery alone (P = .036). The surgery-only group had more lymph node metastases than the neoadjuvant group (76.5% v 61.4%; P = .018). Postoperative complications were more frequent in the neoadjuvant arm (27.1% v 16.2%; P = .09). After a median follow-up of 4.4 years and 67 deaths, a survival benefit could not be shown (hazard ratio, 0.84; 95% CI, 0.52 to 1.35; P = .466). Conclusion This trial showed a significantly increased RO resection rate but failed to demonstrate a survival benefit. Possible explanations are low statistical power, a high rate of proximal gastric cancer including AEG and/or a better outcome than expected after radical surgery alone due to the high quality of surgery with resections of regional lymph nodes outside the perigastic area (celiac trunc, hepatic ligament, lymph node at a. lienalis; D2).

Background Although individual studies show that baseline health-related quality of life (HRQOL) is a prognostic factor for survival, contradictory results have been published and very few meta-analyses have confirmed the finding. We examined whether HRQOL scores are associated with survival when pooled across a large sample of patients with different disease sites. Methods We selected 30 randomised controlled trials from the European Organisation for Research and Treatment of Cancer (EORTC) started between 1986 and 2004, which included survival data for 10108 patients with 11 different cancer sites. Patients were eligible for inclusion if they had completed a baseline HRQOL assessment with the EORTC core quality of life questionnaire (QLQ-C30; n=7417). Sociodemographic variables were age (<= 60 vs >60 years) and sex (men vs women), and clinical variables were WHO performance status (0-1 vs 2-3), distant metastases (no vs yes), and cancer site. We assessed prognostic significance of sociodemographic and clinical variables and the 15 QLQ-C30 scales with Cox proportional hazard models. Findings In the stratified multivariate model including sociodemographic, clinical, and HRQOL data, the HRQOL parameters of physical functioning (hazard ratio 0.94, 95% CI 0.92-0-96; p<0.0001), pain (1.04, 1.02-1.06; p<0.0001), and appetite loss (1.05, 1.03-1.06; p<0.0001) provided significant prognostic information in addition to the parameters age (1.17, 1.06-1.28; p=0.0001), sex (0.74, 0.67-0.82; p<0.0001), and distant metastases (1.70, 1.49-1.93; p<0.0001), but not for WHO performance status (1.07, 0.97-1.19; p=0.11). Consideration of the three HRQOL parameters and sociodemographic and clinical data increased the predictive accuracy of prognosis of overall survival by 6% relative to sociodemographic and clinical characteristics alone (C statistic for concordance between predicted and observed overall survival 0.68 for sociodemographic and clinical variables, and 0.72 for sociodemogr

This is one of the few studies that have explored the value of baseline symptoms and health- related quality of life ( HRQOL) in predicting survival in brain cancer patients. Baseline HRQOL scores ( from the EORTC QLQ- C30 and the Brain Cancer Module ( BN 20)) were examined in 490 newly diagnosed glioblastoma cancer patients for the relationship with overall survival by using Cox proportional hazards regression models. Refined techniques as the bootstrap re-sampling procedure and the computation of C- indexes and R-2- coefficients were used to try and validate the model. Classical analysis controlled for major clinical prognostic factors selected cognitive functioning ( P= 0.0001), global health status ( P= 0.0055) and social functioning ( P < 0.0001) as statistically significant prognostic factors of survival. However, several issues question the validity of these findings. C- indexes and R-2- coefficients, which are measures of the predictive ability of the models, did not exhibit major improvements when adding selected or all HRQOL scores to clinical factors. While classical techniques lead to positive results, more refined analyses suggest that baseline HRQOL scores add relatively little to clinical factors to predict survival. These results may have implications for future use of HRQOL as a prognostic factor in cancer patients.