Predicting patient-reported symptom clusters in prostate cancer patients: A machine learning approach
Rammant, E. Deman, E. Poppe, L. Bultijnck, R. Dirix, P. De Meerleer, G. Haustermans, K. Van Hecke, A. Azria, D. Chang-Claude, J.
Rammant, E.
Deman, E.
Poppe, L.
Bultijnck, R.
Dirix, P.
De Meerleer, G.
Haustermans, K.
Van Hecke, A.
Azria, D.
Chang-Claude, J.
Citations
Altmetric:
Abstract
Introduction & Objectives: Prostate cancer (PC) is the most common urological cancer in the world, with patients suffering from multiple co occurring symptoms (=symptom clusters (SC)). Identifying SC is important to anticipate on other symptoms within a SC and to uncover possibly
overlooked symptoms. Also, supportive care interventions should aim to target multiple symptoms within a SC by addressing 1 or 2 symptoms
and therefore alleviating the severity of other symptoms within that SC. This way, greater gains in a patients’ health-related quality of life (HRQoL)
and more efficient patient care can be achieved. The aim of this study is to identify (1) SC and their changes over time in PC patients receiving
radiotherapy (RT), (2) the impact of SC on HRQoL, and (3) demographic, clinical and, treatment-related predictors of SC.
Materials & Methods: Data were used from REQUITE: an international prospective cohort study including PC patients receiving RT (26 hospitals,
8 countries). SC were identified based on patient-reported outcomes collected before RT(T1), end of RT(T2), month 12(T3), and month 24 after
RT(T4) with the EORTC QLQ-C30 and pelvic symptom questionnaire. A combination of machine learning techniques were used to identify SC at
different timepoints, to assess the impact of SC on HRQoL and to predict the SC, resp.: Hierarchical agglomerative clustering, multivariate linear
regression and random forest regression. A first part of the dataset was used to develop the prediction model and a second part to validate the
model for unseen data.
Results: Data from 1538, 1490, 1322, and 1219 PC patients were analysed at T1, T2, T3 and T4, respectively. Three SC were identified at T1:
SC1 (gastro-intestinal symptoms), SC2 (fatigue, urinary symptoms, emotional and cognitive functioning), and SC3 (pain, physical, role, and social
functioning). At T2, changes in SC were seen: SC1 (gastro-intestinal symptoms), SC2 (fatigue, urinary problems, insomnia), SC3 (social and role
functioning), and SC4 (pain, bowel problems, physical, emotional and cognitive functioning). At T3, SC returned to the 3 T1 SC and remained more
or less stable at T4 (‘fatigue’ left SC2 and clustered together with ‘dyspnoea’ (SC4)). SC including ‘fatigue’ or ‘urinary symptoms’ had the highest frequencies across time-points. At T1, T3 and T4, cluster 2 and 3 (35-45% explained variance) had the strongest impact on the patients’ overall
HRQoL. At T2, cluster 4 (52%) had the strongest impact. Planned RT target volume, PSA at prediagnostic biopsy, age and alcohol consumption
were the best predictors of SC2 at T2 and SC3 and SC4 at T4.
Conclusions: Several SC were identified in PC patients receiving RT. Although SC including fatigue and urinary symptoms were most common
across time-points, the ‘pain, bowel problems, physical, emotional and cognitive functioning’ SC at T2 had the strongest impact on HRQoL. The
predictors can be used to tailor future interventions.
Authors
Rammant, E.
Deman, E.
Poppe, L.
Bultijnck, R.
Dirix, P.
De Meerleer, G.
Haustermans, K.
Van Hecke, A.
Azria, D.
Chang-Claude, J.
Choudhury, Ananya
De Ruysscher, D.
Lambrecht, M.
Rosenstein, B. S.
Seibold, P.
Sperk, E.
Symonds, R. P.
Valdagni, R.
Vega, A.
Webb, A.
West, Catharine M L
Veldeman, L.
Fonteyne, V.
Van Hoecke, S.
Deman, E.
Poppe, L.
Bultijnck, R.
Dirix, P.
De Meerleer, G.
Haustermans, K.
Van Hecke, A.
Azria, D.
Chang-Claude, J.
Choudhury, Ananya
De Ruysscher, D.
Lambrecht, M.
Rosenstein, B. S.
Seibold, P.
Sperk, E.
Symonds, R. P.
Valdagni, R.
Vega, A.
Webb, A.
West, Catharine M L
Veldeman, L.
Fonteyne, V.
Van Hoecke, S.
Description
Date
2022
Publisher
Collections
Keywords
Type
Meetings and Proceedings
Citation
Rammant E, Deman E, Poppe L, Bultijnck R, Dirix P, De Meerleer G, et al. Predicting patient-reported symptom clusters in prostate cancer patients: A machine learning approach. European Urology. 2022 Feb;81:S1687-S8. PubMed PMID: WOS:000812320401536.