The predictive value of multidimensional frailty instruments for postoperative adverse outcomes in cancer patients: a systematic review and meta-analysis

PURPOSE: To systematically review the predictive performance of Tilburg Frailty Indicator (TFI), Groningen Frailty Indicator (GFI), and Edmonton Frailty Scale (EFS) for adverse outcomes including postoperative complications, unplanned readmission, 30-day mortality, prolonged length of stay among cancer patients. MATERIALS AND METHODS: A comprehensive search was conducted across English and Chinese databases until October 2, 2025. Risk of bias was assessed using Quality Assessment of Diagnostic Accuracy Studies 2 tool. Predictive performance was evaluated by pooling sensitivity, specificity, and summary receiver operating characteristic curves. RESULTS: There were 21 studies (4,435 individuals) that were included for the meta-analysis. TFI demonstrated pooled sensitivities of 0.64 (for postoperative complications), 0.77 (for unplanned readmission), and 0.50 (for prolonged hospital stay), with corresponding specificities of 0.67, 0.54, and 0.64 respectively. The areas under the curve (AUC) were 0.70, 0.71, and 0.60. GFI demonstrated sensitivities of 0.59, 0.65, and 0.55 for complications, 30-day mortality, and functional decline, with specificity of 0.73, 0.63, and 0.77, and the AUC of 0.71, 0.68, and 0.74. EFS had sensitivity 0.39, specificity 0.87, and AUC 0.57 for complications. Subgroup analysis revealed that TFI had reasonable predictive value for adverse outcomes with sensitivity 0.61-0.81 and specificity 0.60-0.68 among most gynecological and gastrointestinal cancer subgroups. In most subgroups, GFI showed higher specificity (0.64-0.84) relative to sensitivity (0.43-0.68). CONCLUSION: TFI and GFI demonstrated moderate predictive validity for adverse outcomes, whereas EFS exhibited poor predictive performance. These findings highlight the necessity for caution interpretation of frailty assessments in clinical practice and underscore the importance of further validating these tools within diverse oncological contexts.