Multicollinearity in prognostic factor analyses using the EORTC QLQ‐C30: identification and impact on model selection

Abstract

Clinical and quality of life (QL) variables from an EORTC clinical trial of first line chemotherapy in advanced breast cancer were used in a prognostic factor analysis of survival and response to chemotherapy. For response, different final multivariate models were obtained from forward and backward selection methods, suggesting a disconcerting instability. Quality of life was measured using the EORTC QLQ-C30 questionnaire completed by patients. Subscales on the questionnaire are known to be highly correlated, and therefore it was hypothesized that multicollinearity contributed to model instability. A correlation matrix indicated that global QL was highly correlated with 7 out of 11 variables. In a first attempt to explore multicollinearity, we used global QL as dependent variable in a regression model with other QL subscales as predictors. Afterwards, standard diagnostic tests for multicollinearity were performed. An exploratory principal components analysis and factor analysis of the QL subscales identified at most three important components and indicated that inclusion of global QL made minimal difference to the loadings on each component, suggesting that it is redundant in the model. In a second approach, we advocate a bootstrap technique to assess the stability of the models. Based on these analyses and since global QL exacerbates problems of multicollinearity, we therefore recommend that global QL be excluded from prognostic factor analyses using the QLQ-C30. The prognostic factor analysis was rerun without global QL in the model, and selected the same significant prognostic factors as before.

Quality of Life in 265 Patients with Gastroenteropancreatic or Bronchial Neuroendocrine Tumors Treated with [177Lu-DOTA0,Tyr3]Octreotate

“…This is in line with reports that have indicated that the EORTC QLQ-C30 could be used for predicting survival as well as tumor response (39,40).…”

The prognostic value of health-related quality-of-life data in predicting survival in glioblastoma cancer patients: results from an international randomised phase III EORTC Brain Tumour and Radiation Oncology Groups, and NCIC Clinical Trials Group study

“…The stability of the final model was investigated using a bootstrap re-sampling procedure as proposed by Sauerbrei and Schumacher (1992), applied in the context of HRQOL (Van Steen et al, 2002). This technique generates a number of samples (each of the same size as the original data set), by randomly selecting patients and replacing them before selecting the next patient (i.e.…”

Using patient-reported health status to improve prognostic assessment in patients with acute myeloid leukemia: current challenges and future applications

“…11 For example, while "multicollinearity" is a known challenge in traditional PFA, it becomes even more problematic when PROs are included. 12 Multicollinearity occurs when two or more predictor variables are highly correlated (which is often the case for PROs) thus leading to incorrect model selection and, in any case, making it difficult to disentangle the real influence of each single predictor variable. 11,12 While there is still no gold standard to address this issue, some statistical techniques have been developed to further test the stability of the final multivariate predictive model and to obtain insight into the real value of a single factor being an independent prognostic variable.…”

“…12 Multicollinearity occurs when two or more predictor variables are highly correlated (which is often the case for PROs) thus leading to incorrect model selection and, in any case, making it difficult to disentangle the real influence of each single predictor variable. 11,12 While there is still no gold standard to address this issue, some statistical techniques have been developed to further test the stability of the final multivariate predictive model and to obtain insight into the real value of a single factor being an independent prognostic variable. In the context of QoL studies, Van Steen et al 12 have extensively illustrated a bootstrap model averaging technique which was later successfully used in several methodologically sound studies of patients with solid tumors.…”

“…11,12 While there is still no gold standard to address this issue, some statistical techniques have been developed to further test the stability of the final multivariate predictive model and to obtain insight into the real value of a single factor being an independent prognostic variable. In the context of QoL studies, Van Steen et al 12 have extensively illustrated a bootstrap model averaging technique which was later successfully used in several methodologically sound studies of patients with solid tumors. 13,14 Also, a crucial aspect that could be considered is the importance of an a priori selection of specific PRO scales to be included in the analyses.…”