OBJECTIVES: Machine learning (ML) has the potential to go beyond conventional statistical methods in handling the growing amount of data produced in cancer research, including PROs. To critically examine the current state of including PROs in ML models in cancer research, evaluate the rigor and quality of currently available studies, and propose areas of improvement for future use of ML in the field.

METHODS: PubMed and Web of Science were systematically searched for publications of studies on cancer patients applying ML models with PROs either as features or outcomes. The methodological quality of applied ML models was assessed utilizing an adapted version of the MI-CLAIM (Minimum Information about CLinical Artificial Intelligence Modeling) checklist (according to Smets et al, 2021). The key variables of the checklist are data preparation, model optimization, performance, and examination. Reproducibility and transparency complement the quality criteria.

RESULTS: The literature search yielded 1634 hits, of which 54(3.3%) were eligible. Thirty-seven (68.5%) publications included PROs as a feature, and 35(64.8%) as an outcome, with instruments from the EORTC Quality of Life Group being the most frequently used PRO measures 10(18.5%). Preliminary results of the quality appraisal indicate a potential for improvement, especially in the areas of model examination (with 44(81.5%) publications lacking discussion about the clinical applicability of the developed model) and reproducibility (as 52(96.3%) publications do not provide code to reproduce the model and the results).

CONCLUSIONS: The herein performed critical examination of the status quo of the application of ML in PRO research in oncology allowed to identify areas of improvement for future uses of ML in the field. The presented results will contribute to enabling the potential of new technologies. The final results will be presented at the conference. The systematic review was registered in Prospero (CRD42023405660).