Reliability, validity, and clinically important differences (CIDs) on the NCCN/FACT Bladder Symptom Index (NFBISI-18) among individuals with locally advanced or metastatic urothelial cancer (UC).

Abstract

408 Background: The NFBlSI-18 is a measure of advanced bladder cancer–specific symptoms composed of a total scale and 3 subscales representing physical disease–related symptoms (DRS-P), emotional disease–related symptoms (DRS-E), treatment side effects (TSE), and function/well-being (F/WB). There is evidence for the reliability and content validity of this instrument, but a full psychometric evaluation of the full 18-item format has not been done. In addition, CIDs have not been estimated. Methods: With the exception of test-retest (TRT) analyses, baseline data (n=651) from the JAVELIN Bladder 100 trial (NCT02603432), which compared maintenance treatment with avelumab + best supportive care (BSC) vs BSC alone in patients with unresectable, locally advanced or metastatic UC that did not progress with first-line platinum-containing chemotherapy, were used for this study. Since we focused on baseline, we did not analyze the TSE. We estimated internal consistency reliability (Cronbach coefficient α), tested convergent validity by estimating Spearman ρ correlations with the EQ-5D-5L utility index (UI) and visual analog score (VAS) scales, and estimated known group validity using age (<65, ≥65 years), ECOG performance status rating (PSR), and number of comorbidities/symptoms (1-9, ≥10) as anchors. We estimated TRT reliability using data from treatment cycles 2-3 with intraclass correlation coefficients (ICCs). To estimate and compare CIDs, we calculated differences in means between categories of the known group anchors for which Cohen’s d was >0.2 (ie, at least a small effect). To provide context for the CIDs, we calculated distributional properties of the scales (1/2 standard deviation [SD], 1 standard error of measurement [SEM]). Results: The table shows the reliability, convergent validity, and CID estimates. Reliability estimates often exceeded thresholds for reliability generally considered acceptable. Cohen’s d for NFBlSI-18 scale score differences between known groups ranged between 0.05 and 0.25 (age), 0.35 and 0.6 (ECOG PSR 0 vs 1), and 0.1 and 0.41 (number of comorbidities/symptoms). Conclusions: This analysis demonstrated that the NFBlSI-18 is a reliable and valid instrument to measure symptoms in patients with advanced UC. The CID estimates can help clinicians and researchers to understand what difference in patient symptoms are clinically meaningful, as measured by NFBlSI-18, to inform clinical practice. [Table: see text]