Under-reporting of subjective symptoms and its prognostic value. A pooled analysis of 12 cancer clinical trials.

Importance: It has been reported that oncologists tend to under-report subjective symptoms during cancer treatment. Whether and how much under-reporting may affect patients’ outcome is unknown. Objective: To describe under-reporting rate of 6 symptoms (anorexia, nausea, vomiting, constipation, diarrhea and fatigue) and explore its association with overall survival (OS). A secondary aim is to test the association with OS of side-effects directly reported by the patients. Design: Post-hoc analysis on data pooled from 12 prospective randomized trials, promoted by National Cancer Institute of Naples (Italy), enrolling patients between 2002 and 2019, with primary analyses already published. Occurrence and grade of selected side-effects reported by physicians were compared with corresponding symptoms reported by patients in quality-of-life (QoL) questionnaires. Setting: Italian public oncologic centres (plus centers from Canada, Belgium and Germany in two trials). Participants: 3792 patients with advanced lung, ovarian, pancreatic, breast or colorectal cancer were pooled; 2603 (68.6%) were eligible having at least one toxicity assessment reported by physicians and at least one self-reported QoL questionnaire before the first planned restaging of disease. Main outcomes and measures: Under-reporting was defined as the rate of cases where the physician reported grade 0 while the patient reported variable severity of the same side-effect/symptom. OS was the outcome for primary and secondary analyses. Prognostic value was tested in a multivariable model stratified by trial, including age, gender and performance status as confounders. For OS analyses, a landmark threshold was defined excluding patients who died or were censored before the threshold. Results: Concordance between physicians’ and patients’ reporting was low with Cohen’s k coefficients ranging from 0.03 (fatigue) to 0.33 (vomiting). Under-reporting ranged from 52.7% (nausea) to 80.5% (anorexia). Under-reporting was not associated with OS, hazard ratios of death ranging from 0.82 (vomiting) to 1.09 (nausea), 95% confidence intervals being not statistically significant. Anorexia, vomiting and fatigue reported by the patients were significantly associated with OS. Conclusions and Relevance: Under-reporting of symptoms possibly representing side-effects of treatment is frequent, but it does not affect OS. Whether it does worsen patients’ QOL remains to be established. Patients’ reported symptoms should be used for prognostic evaluation.