Validating an instrument for direct patient reporting of chemotherapy toxicity among South African cancer patients.

Abstract

172 Background: In the USA, use of patient-reported outcome (PRO) instruments for monitoring treatment toxicity has been associated with improved patient quality of life (QoL) and clinical outcomes. No similar PRO tools exist for the African cancer patients. We aimed to validate the Patient Reported Symptoms – South Africa (PRS-SA) survey, a novel PRO tool designed to measure subjective stress and symptoms in South African (SA) cancer patients receiving chemotherapy. Methods: We enrolled patients receiving chemotherapy at the oncology clinic at Charlotte Maxeke Hospital, Johannesburg. At 3 separate visits, participants completed the PRS-SA survey, which allowed reporting of stress on a numerical ranking scale (range 0-10) and severity of 11 common chemotherapy-related symptoms (i.e., pain, fatigue, fever, dyspnea, cough, mucositis, nausea, vomiting, diarrhea, constipation, and neuropathy) using lay language descriptions corresponding to Common Terminology Criteria for Adverse Events grades. They simultaneously completed the EORTC Core Quality of Life Questionnaire (QLQ-C30), the Global Impression of Change (GIC) questionnaire, and the Hospital Anxiety and Depression Scale (HADS). All instruments were offered in English and isiZulu. We constructed a receiver operator characteristics (ROC) curve for stress scale values and a total HADS score ≥15, representing clinical depression/anxiety. We evaluated construct validity for each symptom item by comparing symptom severity to the simultaneous QLQ-C30 summary score (Pearson correlation tests) and ECOG performance status (PS) (Mann-Whitney U tests). We assessed symptom item responsiveness by comparing change in severity across visits to change in QLQ-C30 summary score and comparing standardized mean symptom scores in those reporting negative, no, or positive change on the GIC using a Jockheere-Terpstra trend test. Results: Overall, 196 participants completed instruments at visit 1, 173 at visit 2, and 150 at visit 3. Area under the ROC curve for stress score was 0.76, and a score of 4 had 82% sensitivity and 55% specificity for clinical depression/anxiety. All symptom items showed construct validity by association with QLQ-C30 score (all p-values < 0.0001), and all but mucositis showed validity by association with ECOG PS (highest p = 0.03). All but cough showed responsiveness to change in QLQ-C30 score (highest p = 0.045), and the standardized mean in symptom scores demonstrated a trend across negative, neutral, and positive GIC scores (p = 0.03). Conclusions: In SA cancer patients, the PRS-SA’s stress scale behaves similarly to the distress thermometer in other populations, and the symptom items demonstrated construct validity and responsiveness. Further research should address implementation of the PRS-SA in the clinical setting for monitoring cancer treatment-related toxicity and its impact of patients’ QoL and outcomes.