Article
Identifying Predictors for Low Health-Related Quality of Life in Patients with Myelodysplastic Syndromes – Research Protocol
Search Medline for
Authors
Published: | March 6, 2018 |
---|
Outline
Text
Background and objectives: The myelodysplastic syndromes (MDS) refer to a heterogeneous cluster of clonal hematopoietic stem cell disorders with an increased risk for leukemic transformation, diagnosed mainly among the older population. Beside its nature, the disease burden is increased by the common comorbidities and the difficulty of choosing between treatment strategies. Therefore, an accurate risk assessment in the process of MDS management requires the input from the patients’ physical, mental, spiritual, emotional, and social well-being. Although this approach might be underestimated in the clinical practice, the importance of the health-related quality of life (HRQoL) as an independent predictor of overall survival and for treatment assessment has been well documented. Most of this research was conducted as part of effectiveness and safety trials, overlooking the continuous assessment of HRQoL. The studies also reported several limitations, such as the small number of included patients and their diversity, relatively short follow-up period and limited patient data. The extensive data of the European Myelodysplastic Syndrome (EUMDS) Registry provide an opportunity for additional analyses. For example, the inclusion of time-varying predictors in the assessment of HRQoL and prediction of HRQoL at multiple time points measured by the EQ-5D every six months allows for a dynamic prediction. In this segment of the research, we aim to identify disease-specific and patient-related predictors of low HRQoL among MDS patients. The resulting findings will improve the personalized treatment approach for MDS patients by focusing on the important predictors and guiding the therapy towards each risk group.
Methods: In consultancy with the MDS clinical experts, we will use the receiver operating characteristics (ROC) curve as a binary classifier of the dependent variable, with variations of the cut-off point being considered for sensitivity analyses. Potential predictors of HRQoL will be identified using univariate variable screening with a less strict p-value (p<0.15). Final predictors with 95% confidence intervals will be determined using multivariate logistic regression analyses and stepwise variable selection. Two-way interactions and multicollinearity will be assessed through the regression analyses. Classification-and-regression-tree (CART) analysis will be used to detect subgroups based on higher level interactions, which would potentially have the most benefit out of the identified predictors.