Skeletal muscle density performance for screening frailty in older adults with cancer and the impact of diabetes: The CARE Registry.

introductionSkeletal muscle density (SMD) measurements from imaging scans identify myosteatosis and could screen patients for geriatric assessment. We assessed SMD performance as a screening tool to identify older adults with cancer likely to be frail and who could benefit from in-depth assessment; we compared performance by sex and diabetes status. MATERIALS AND

methodsWe analyzed patients in the Cancer & Aging Resilience Evaluation (CARE) Registry. Frailty and diabetes were captured using a patient-reported geriatric assessment (CARE tool). Frailty was defined using CARE frailty index (CARE-FI) based on principles of deficit accumulation. SMD was calculated from computed tomography scans (L3 vertebrae). Analyses were conducted by sex and diabetes status. Scatterplots and linear regression described crude associations between SMD and frailty score. Classification performance (frail vs. non-frail) was analyzed with (1) area under the receiver operating characteristic curves (AUC) and confidence intervals (CIs); and (2) sensitivity/specificity for sex-specific SMD quartile cut-offs (Q1, median, Q3). Performance was compared between patients with and without diabetes using differences and estimated CIs (2000 bootstrap replicates). We additionally calculated positive and negative likelihood ratios (LR+, LR-).

resultsThe analytic cohort included 872 patients (39% female, median age 68 years, 27% with diabetes) with predominately stage III/IV gastrointestinal cancer; >60% planning to initiate first-line chemotherapy. SMD was negatively associated with frailty score; models were best fit in male patients with diabetes. AUC estimates for female (range: 0.58-0.62) and male (0.58-0.68) patients were low. Q3 cut-offs had high sensitivity (range: 0.76-0.89), but poor specificity (0.25-0.34). Diabetes did not impact estimates for female patients. Male patients with diabetes had greater sensitivity estimates compared to those without (sensitivity differences: 0.23 [0.07, 0.38], 0.08 [-0.07, 0.24], and 0.11 [0.00, 0.22] for Q1, median, Q3, respectively). LR estimates were most notable for male patients with diabetes (LR+ = 2.92, Q1 cut-off; LR- = 0.46, Q3 cut-off). DISCUSSION: Using SMD alone to screen older patients for geriatric assessment requires improvement. High-sensitivity cut-off points could miss 11-24% of patients with frailty, and many non-frail patients may be flagged. Screening with SMD is practical but work is needed to understand clinical andresource impacts of different cut-off points. Future research should evaluate performance with additional clinical data and in subgroups.