A novel frailty index for stratifying risk in pituitary adenoma resection

Introduction: Pituitary adenoma (PA) resection is a common skull-base procedure, with patient selection key in clinical decision-making. Frailty is a useful risk stratification tool to predict postoperative outcomes. Multiple frailty assessment tools have been proposed, including the modified 11-item frailty index (mFI-11), and retrospective risk analysis index (RAI-C). However, a direct comparison between metrics has not been conducted. Here, we compare performance of multiple common frailty indices in PA resection, and also developed a novel preoperative index that outperformed all others studied.

Methods: A 12-year single-center retrospective review was conducted of all patients who received transsphenoidal pituitary adenoma resection. Statistical analysis was performed using the SciPy Statistics Python library. Frailty metrics were compared by measures of univariate association with surgical outcomes, and ROC-AUC values. For feature selection, a univariate association matrix was created between potential features and key outcome measures; linear regression, chi-squared analysis, Fisher’s exact test, and two-tailed T-test were conducted according to variable type. Feature consistency was assessed by the number of outcome variables with a nonrandom association probability above a set threshold, and feature specificity by the average nonrandom association probability among all outcome measures. The 6 top-performing features per category were retained, yielding 7 unique features; one was replaced with an average-performing feature to prevent overfitting. The features were then summed as the index.

Results: 433 patients were studied. The novel metric outperformed all others for multiple post-operative complication measures, including meningitis (p=0.023), hemorrhage (p=0.0056), and vision loss (p=0.0012). Additionally, linear regression analysis demonstrated it was the best predictor of key outcome measures, including mRS both at discharge (p < 0.0001, r=0.331) and 90 days (p < 0.0001, r=0.219). Finally, it was the only metric with a statistically-significant mean p-value across all variables studied (p=0.032).

Conclusion : Using institutional data, this study identifies a novel frailty metric as outperforming other indices in patients who underwent transsphenoidal pituitary adenoma resection. Further validation on a multi-center database is necessary to determine if this index can be applied to stratify patient risk more broadly.