A decision curve analysis of the clinical usefulness of a two-step frailty assessment strategy in older patients with prostate, breast, colorectal, or lung cancer

• Published in: Journal of clinical oncology Vol. 40; no. 16_suppl; p. 12011
• Main Authors: Gonzalez Serrano, Adolfo, Laurent, Marie, Barnay, Thomas, Martinez-Tapia, Claudia, Audureau, Etienne, Boudou-Rouquette, Pascaline, Aparicio, Thomas, Rollot-Trad, Florence, Soubeyran, Pierre, Bellera, Carine A., Paillaud, Elena, Caillet, Philippe, Canoui-Poitrine, Florence
• Format: Journal Article
• Language: English
• Published: 01-06-2022
• ISSN: 0732-183X; 1527-7755
• DOI: 10.1200/JCO.2022.40.16_suppl.12011

12011
Background: Geriatric Assessment (GA) is recommended to assess the health status and select the most appropriate cancer treatment in older patients. However, GA is resource- and time-consuming. Thus, a two-step approach using frailty screening has been recommended. We aimed to evaluate the usefulness of frailty screening over GA for identifying unfit individuals who need GA and reducing unnecessary GA in fit individuals in a population of older outpatients with cancer. Methods: We analyzed patients age 70 and older with prostate, breast, colorectal, or lung cancer included in the multicenter, prospective ELCAPA cohort study (NCT02884375) between February 2007 and December 2019. All patients had a GA at inclusion. GA was the reference test. We defined unfit patients as those having at least one abnormal score in the following domains: functional status, mobility, comorbidity, cognition, mental health status, nutrition, and polypharmacy. We defined unfit patients according to the G8 and modified G8 scores using the recommended cut-offs (≤ 14 out of 17 points and ≥ 6 out of 35 points, respectively). We calculated each screening tool's sensitivity, specificity, and positive and negative predictive values. We used decision curve analysis to estimate the net benefit (the percentage of patients found to be unfit) of screening over GA. We assessed the avoided unnecessary GAs for each screening tool (reducing unnecessary GA in fit patients without decreasing the number of unfit patients undergoing [necessary] GA). We calculated these estimates across different threshold probabilities corresponding to the value of missing an unfit patient compared to exposing a fit patient to an unnecessary GA. A probability of 0.33 indicated that missing an unfit patient was two times worse than referring a fit patient to an unnecessary GA. A probability of 0.50 indicated that missing an unfit patient was the same as exposing a fit patient to an unnecessary GA. Results: We analyzed 1,648 patients with prostate (15%), breast (52%), colorectal (22%), or lung cancer (11%). The median age was 81 years, 559 patients (34%) had metastatic disease, and 1,428 patients (87%) were unfit. The sensitivity (95% CI) and specificity were 85% (84-87) and 59% (57-61) for the G8 score, and 86% (84-87) and 60% (58-63) for the modified G8 score. With a threshold probability of 0.33, the net benefit was 0.71 for the G8 score, 0.72 for the modified G8 score, and 0.80 for GA. With a threshold probability of 0.50, the net benefit was 0.68 for the G8 score, 0.69 for the modified G8 score, and 0.73 for GA. We did not observe a reduction in unnecessary GA of screening tools over GA. Conclusions: Frailty screening tools showed good diagnostic performances. However, our findings suggest that the GA-for-all strategy provides the higher clinical benefit in older patients with cancer.