Early weight gain as a predictor of weight restoration in avoidant/restrictive food intake disorder

Previous research has demonstrated that early weight gain in family-based treatment (FBT) is predictive of remission for adolescents with anorexia nervosa (AN). However, no published data has addressed if early weight gain is also predictive of reaching weight restoration (i.e., 95% EBW) in patients with avoidant/restrictive food intake disorder (ARFID). Furthermore, no studies have evaluated the performance of the statistical models used to predict weight restoration at the end of treatment. This study sought to examine whether early weight gain in ARFID is predictive of weight restoration at 20 weeks using ROC analysis. Additionally, this study assessed how accurately the model classified patients and what types of misclassifications occurred.

Participants (n = 130, 57.7% cisgender female 70.0% white) received virtual outpatient FBT. Receiver operating characteristics (ROC) were used to predict successful weight restoration at end of treatment, using early weight gain as the predictor. Twenty weeks was considered as the end of treatment, to align with the definition of end of treatment in FBT clinical trials. ROC analyses demonstrated that gaining at least 6.2 pounds by week 5 of treatment was the strongest predictor of achieving 95% EBW at 20 weeks (AUC = 0.72 [0.63, 0.81]). ROC analyses misclassified 35% of patients; the most common misclassification was predicting that a patient would not achieve 95% EBW when they actually did (61.6%). A logistical regression model, which included the patients' %EBW at admission in addition to early weight gain as a predictor, outperformed the ROC analyses (AUC = 0.90 [0.85, 0.95]) and provided additional context by showing the probability that a patient would succeed.

Taken together, research demonstrates that early weight gain is a useful predictor of 95% EBW at 20 weeks of treatment for patients with ARFID who require weight restoration. Furthermore, results suggest that statistical models need to take into account additional information, such as %EBW at admission, along with early weight gain in order to more accurately predict which patients will reach weight restoration at week 20.