Inter-rater reliability of CTCAE assessments with or without EORTC patient-reported outcome data in a mixed cancer population: a multinational, open-label, randomised controlled trial

BACKGROUND: The Common Terminology Criteria for Adverse Events (CTCAE) is the standard for provider-based rating of adverse events in oncology, but its reliability for symptomatic adverse events is low. Patient-reported outcome (PRO) data might improve the inter-rater reliability of CTCAE assessments. We investigated whether providers' access to PRO data enhances the inter-rater reliability of CTCAE ratings and the detection of symptomatic adverse events.

METHODS: This multinational, open-label, randomised controlled trial was done in 11 hospitals in ten countries. Adults (aged ≥18 years) with any cancer diagnosis receiving chemotherapy, immunotherapy, or radiotherapy of curative or palliative intent were eligible. Patients were randomly assigned (1:1) to the intervention group (in which providers saw patients' PRO data as they were doing CTCAE assessments) or the control group (in which providers did not have access to PRO data) within each centre through a minimisation algorithm and a weighted cutoff; no additional stratification factors were applied. Patients completed the European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 and 16 additional items from the EORTC Item Library, covering 17 symptomatic adverse events. Two independent providers (oncologists or trained nurses) did CTCAE ratings. The primary endpoint was the inter-rater reliability of CTCAE ratings, expressed as intraclass correlation coefficients (ICCs), for all patients with complete assessments. The trial was registered at ClinicalTrials.gov (NCT04066868) and is now closed.

FINDINGS: Between Feb 10, 2020, and Dec 6, 2024, a total of 1067 patients were enrolled (538 in the control group and 529 in the intervention group), with full data from 512 patients in the control group and 501 in the intervention group included in the analysis. Median age was 59 years (IQR 46-68); 558 (55·1%) of 1013 were female and 455 (44·9%) were male. Most had good functional status (403 [79%] of 512 had ECOG 0-1 in the control group and 371 [74%] of 501 in the intervention group). The most common cancer types were haematological (control group: 137 [27%] of 512; intervention group: 131 [26%] of 501), breast (control group: 118 [23%]; intervention group: 107 [21%]), and gastrointestinal (control group: 98 [19%]; intervention group: 89 [18%]). Inter-rater reliability was significantly higher in the intervention group for 13 of 17 symptomatic adverse events, with the largest differences for memory impairment (ICC 0·176; p<0·0001), irritability (0·161; p<0·0001), concentration impairment (0·157; p<0·0001), depression (0·126; p=0·0012), and anxiety (0·109; p=0·0018). There was no significant difference for pain, rash, and peripheral sensory neuropathy and the reliability was significantly higher in the control group than the intervention group for diarrhoea (ICC -0·066; p=0·013).

INTERPRETATION: Availability of PRO data to providers for CTCAE rating improves the consistency of provider-based symptomatic adverse event detection in patients with cancer. PRO data supports reliable assessment of treatment-related toxicity, highlighting the value of integrating PRO data into clinical evaluations within cancer trials.

FUNDING: EORTC Quality of Life Group.