Rasch model-based testing of the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire–Chemotherapy-Induced Peripheral Neuropathy (QLQ-CIPN20) using Alliance for Clinical Trials in Oncology (Alliance) A151408 study data

Abstract

Purpose To test the psychometric properties of the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire–Chemotherapy-Induced Peripheral Neuropathy (QLQ-CIPN20) using Rasch-based methods. Methods A secondary data analysis was performed using pooled QLQ-CIPN20 data from patients (N = 1008) who had participated in any of four multi-site chemotherapy-induced peripheral neuropathy (CIPN) treatment and prevention trials. QLQ-CIPN20 responses were evaluated using a polytomous Rasch partial credit model. Data were assessed for person-item fit using the chi-square statistic, item scaling based on response proportions, threshold ordering using item characteristic curves and logit threshold locations, differential item response (DIF) (i.e., response bias) using likelihood ratio tests, and unidimensionality using cluster analysis. Results A statistically significant chi-square test indicated poor fit of the observed to the expected responses. More than 70% of the respondents reported a complete absence of six symptoms, reflecting significant floor effects and poor item scaling. Disordered/non-ordinal or narrow response thresholds were found for 11 of the 20 items. Item responses were significantly different by gender (p < 0.0001) and chemotherapy type (p < 0.0001). Cluster analysis findings suggest that the QLQ-CIPN20 is a unidimensional scale due to the absence of item clusters. Conclusions Rasch model testing revealed psychometric weaknesses that could be addressed by revising the QLQ-CIPN20’s problematic items and response options. Alternatively, perhaps the new gold standard CIPN measurement approach in future intervention trials should involve use of only the best items, which would also allow comparisons across previous trials that utilized the QLQ-CIPN20.

Overview and critical revision of clinical assessment tools in chemotherapy‐induced peripheral neurotoxicity

“…A secondary Rasch‐model‐based analysis of the CIPN20 in over 1000 patients evaluated the relationship between neuropathy severity and the probability of scoring a particular item . For most items, analysis of the distribution of responses revealed a high proportion of low severity responses, implying that the scoring range of 1 to 4 was inadequate to detect small changes in symptom severity . The Rasch analysis identified psychometric weakness with 11 out of 20 items displaying disordered or non‐ordinal thresholds.…”

“…Autonomic and hearing items were highlighted as problematic. The authors proposed eliminating or revising a number of items and rescaling the 0 to 4 scale to allow a 0 to 10 response range …”

“…The International collaborative CI‐PeriNomS group has examined the validity and reliability of a number of tools, demonstrating the psychometric properties of TNSc and EORTC CIPN20, highlighting the important issue of training of examiners and the dearth of clinical information regarding neuropathic pain in chemotherapy‐treated patients . Similarly, a number of studies have examined the psychometric properties of the EORTC CIPN20, suggesting changes to improve structural properties . On the basis of these studies, it has been proposed that the sensory, motor, and autonomic subscales are not supported by evidence and that grading as an overall score provides better validity.…”

“…On the basis of these studies, it has been proposed that the sensory, motor, and autonomic subscales are not supported by evidence and that grading as an overall score provides better validity. Furthermore, there are a number of individual questions which have been demonstrated to be unclear to patients and lack appropriate psychometric robustness . Identified psychometric weaknesses warrant consideration to recalibrate the scale for future optimal use, including the use of 0 to 10 scaling …”

In Search of a Gold Standard Patient-Reported Outcome Measure for Use in Chemotherapy- Induced Peripheral Neuropathy Clinical Trials

“…Results of the first 2 studies, briefly described here, informed this study. The first study (article in press) 43 was a secondary data analysis to evaluate the sensitivity, internal consistency reliability, and structural validity of the original QLQ-CIPN20 using merged data from European and North American patients (N = 1155) who had participated in studies focused on CIPN measurement, prevention, or treatment. 26 , 41 , 44 - 46 We found the QLQ-CIPN20 to be sensitive, reliable, and valid; however, items designed to evaluate autonomic neuropathy (dizziness, blurred vision, erectile dysfunction) and hearing loss compromised the instrument’s psychometric properties and were therefore removed.…”