No abstract

No abstract

6589 Background: Establishing the Minimal Clinically Important Difference (MCID) is essential for interpreting the clinical relevance of patient reported outcomes. This is the first study to date to determine the MCID of FACT-Cog, a 37-item validated subjective neuropsychological instrument designed to evaluate cancer patients’ perceived cognitive deterioration on their quality of life. Methods: This prospective, observational study involved 220 breast cancer patients who have completed FACT-Cog and EORTC-QLQ-C30 at two time points: baseline and at least 3 months following chemotherapy. The MCID was computed using 3 approaches: 1) an anchor-based approach utilized the validated EORTC-QLQ-C30-Cognitive Functioning scale (CF) as the anchor for patients who showed a minimal deterioration on the CF (defined as a one-step deterioration on the CF scale); 2) a Receiver Operating Characteristic (ROC) curve was used to identify an optimal MCID cut-off point for deterioration; 3) a distribution-based approach utilized the 0.33 SD, 0.5 SD and one standard error of measurement (SEM) of the total FACT-Cog score (148 points) to estimate the MCID. Results: There was moderate correlation between the mean change scores of FACT-Cog and CF (rp= 0.43, p<0.001). The CF-anchored MCID was 9.6 points (95% CI 4.4 - 14.8). MCID derived from the ROC method was 7.5 points (AUC: 0.75; sensitivity: 75.6%; specificity: 68.8%). Using the distribution-based approach, MCID corresponding to effect sizes of 0.33SD to 0.5SD of the total FACT-Cog score ranged from 6.9 – 10.3 points and one-SEM criterion resulted in a MCID estimate of 10.6 points. Combining results from all approaches, the MCID identified for FACT-Cog ranged from 6.9 – 10.6 points (4.7% to 7.2% of total score). Conclusions: A 6.9 to 10.6 points reduction of the FACT-Cog score corresponds to the smallest clinically-relevant perceived cognitive deterioration. These estimates are important as they can facilitate the interpretation of patient-reported cognitive changes and sample size estimation in future studies.