1. Identifying health and frailty in multiple myeloma : a clinical and biochemical approach to improve outcome prediction
- Author
-
Carter-Brzezinski, Luke
- Subjects
biomarkers ,machine learning ,frailty ,multiple myeloma - Abstract
Fitness and frailty can have a significant impact on the capability of a patient to tolerate cancer treatment. However, identifying these characteristics can be challenging in the setting of cancer where a range of physical and psychological factors may affect perceived fitness, and where cancer-related health deficits may be reversible. An objective tool to distinguish effects of frailty from those of disease could significantly aid patient management and treatment selection. This research investigated a real-world population of patients with myeloma seeking to define how relative impacts of different health impacts might be detected and quantified. Consecutive patients undergoing therapy for newly diagnosed or relapsed multiple myeloma at a tertiary centre were recruited over a three-year period (n=91). Data comprised standard of care tests of myeloma or general health, together with serum samples for quantitative protein analysis using SWATH mass spectrometry. Additional quality of life (QoL) assessments was performed using standard validated tools. Data sets were tested as appropriate using standard testing, survival analysis, exploratory statistical analysis, and supervised machine learning to look for features primarily reflecting fitness frailty or myeloma disease. The patient group had characteristics that were comparable with published cohorts. Initial use of Principal Component Analysis (PCA) suggested that three clinically distinct groups could be identified: those with high myeloma disease burden, those with myeloma and a high frailty burden and those with myeloma but with minimal adverse features; the groups had distinctive patterns of survival with a significantly inferior outcome in our novel high frailty burden group. Although these groups have potential clinical utility, large clinical datasets and PCA are not well suited to prospective patient classification in a routine clinical setting. Therefore, the serum mass spectrometry was performed on patient serum, and the datasets obtained were explored to identify candidate serum protein biomarkers that could prospectively identify the groups. Supervised analysis of serum proteins using Random Forest was performed on a representative group of patients to identify potential biomarkers. This was refined to identify a candidate set using standard significance approaches ANOVA, T-test and ROC analysis as well as looking for potential biological differences between proteins in the different subsets. The QoL results were used to 10 examine the relationship between frailty, myeloma and QoL. The EORTC MY-20 questionnaire was found to be a suitable and acceptable tool. Analysis of the results of QoL data demonstrated a deterioration of QoL as the number of lines of therapy increased and was related to the overall survival in our cohort. It was also shown that our novel frail cohort of patients had an inferior self-reported quality of life. In summary, this study has identified clinically important subsets of patients that can be identified by standard of case testing and serum protein analysis. The findings of this research should require validation in prospective analyses but have the potential for changing treatment selection and management of patients with multiple myeloma.
- Published
- 2023