Development and validation of predictive models for 6-month gastrostomy timing in amyotrophic lateral sclerosis

Background Dysphagia is common in amyotrophic lateral sclerosis (ALS), contributing to malnutrition and accelerated disease progression. Although early nutritional intervention is recommended, the optimal timing for percutaneous endoscopic gastrostomy (PEG) placement remains uncertain. This study aimed to develop and validate simple prediction models, accessible via an online calculator, to identify ALS patients likely to require PEG within 6 months.Methods We conducted a retrospective cohort study including ALS patients followed at three Italian reference centres between February 2018 and October 2023. Predictors of PEG placement within 6 months were identified using univariate and multivariable binary logistic regression models. Prediction models were developed following Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines and underwent both internal and external validation.Results In the development cohort (n=263; median age 63.8 years), 138 patients (52.5%) underwent PEG within 6 months. Three models were developed: the Anamnestic Prediction Model, based on age, onset site and non-invasive ventilation (NIV), showed fair predictive performance. The Anamnestic and Functional Prediction Model, incorporating age, bulbar subscore of Amyotrophic Lateral Sclerosis Functional Rating Scale—revised (ALSFRS-r) and forced vital capacity (%), demonstrated strong predictive performance (Brier score: 0.1230), excellent discrimination (concordance index (c-index) 0.91) and good calibration (Hosmer-Lemeshow p=0.59). The Anamnestic and Nutritional Prediction Model, including age, onset site, NIV, body mass index and weight loss, showed good predictive performance (Brier score: 0.1719), discrimination (c-index 0.81) and calibration (Hosmer-Lemeshow p=0.48). These findings were confirmed in an external validation cohort of 116 ALS patients.Conclusions The prediction models provide accurate, easily implementable tools to predict PEG need within 6 months, enabling timely nutritional interventions that may improve outcomes and care quality in ALS.