P11.11.B VALIDITY RELATED EVIDENCE FOR COGSTATE TASKS IN PATIENTS WITH BRAIN METASTASES

Le résumé fourni par la source

Abstract BACKGROUND Digital approaches to measuring and monitoring cognitive function (CF) are of interest as screening measures to enhance clinical care and clinical research. Validity related evidence for such measures is required prior to adopting these in practice and trials. Four investigator-initiated clinical trials evaluated CF in patients with brain metastases (BM) with both clinically validated CF tests and experimental computerized CogState (CS) tasks. Construct validity of the CS tasks was evaluated. MATERIAL AND METHODS 140 pts with BM who completed both CF testing and computerized CS tasks prior to treatment were included. Within sample mean/SD were used to convert raw to standardized test scores. The following domains and tests were evaluated: Processing Speed (CS Detection, TMTA, TMTB, Coding), Attention (CS Identification, Digit Span, Coding, TMTA, TMTB), Working Memory (CS One Back, Digit Span), and Learning and Memory (CS One Card Learning and International Shopping List, HVLT-R). Spearman correlations were computed between CF tests and CS tasks, with significance set at the 0.01 level. Impairment was classified as a z-score ≤-1.5 and concordance rates were examined between HVLT-R and ISL. RESULTS On average, pts were 58.3 years old(SD 12.9), completed 14.7 years of education(SD 3.0), 48% female, and 84% right-handed. Spearman correlation coefficients between CS tasks and related CF tests were weak to moderate for processing speed (rs=0.39-0.55) and attention (rs=0.34-0.56), weak for working memory (rs=0.28), and weak to strong for learning and memory (rs=0.36-0.69). Spearman correlations were stronger, while intraclass correlation coefficients were moderate, between the ISLT and HVLT-R TR (rs=0.69 and ICC=0.58) and ISLR and HVLT-R DR (rs=0.69 and ICC=0.58). Concordance rates of impairment for the most highly correlated and parallel measures of learning and memory were poor with no to minimal correspondence based on Cohen kappa coefficients (ISLT and HVLT-R TR k=0.11; ISLR and HVLT-R DR k=0.29). CONCLUSION Spearman correlations between CS tasks and clinically validated CF tests in similar domains were weak with the exception of highly parallel measures of verbal learning and memory. ICC were less robust and Cohen kappa coefficients was negligible for these measures of verbal learning and memory. The substantial lack of agreement in identifying abnormal episodic memory performance is concerning in terms of utilizing the CS ISL task as a surrogate for the HVLT-R. Further associations between clinical variables and these measures of CF will help to determine ability to assess clinically relevant differences due to disease burden. Longitudinal evaluations will further shed light on sensitivity to treatment and disease effects.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.