Sofia Korsavva, Filimena Borisova Valkova, Ignacio Calderon Perez, Alexandros Meletiadis
Neuropsychiatric Investigation - 2025;63(1):1-16
Objective: Anorexia nervosa (AN) affects millions of people worldwide and treatments vary widely with no set treatment guidelines. A network meta-analysis compared and contrasted treatments for adults with AN that reported body mass index (BMI) or weight changes. Methods: Embase, Medline, CINAHL, PubMed, PsycINFO, Cochrane, and clinicaltrials.gov were searched, from January 1974 to March 2024, for articles reporting weight or BMI results pre/posttreatment for anorexia treatments. Outcomes, using pooled-weighted-standard mean effect sizes and subgroup analyses per-intervention-type, were analyzed using a random effects model with bayesian and frequentist statistics. Artificial neural nets were used to predict response to future treatments. About 63 studies were included out of the 650 reviewed articles. Results: The random effects model calculated a pooled-weighted-effect size for 4366 patients of 1.43 (1.00-1.86) 95%CI for surgically invasive neuromodulation (deep brain stimulation (DBS) and Capsulotomy Surgery), 0.65, 95%CI (0.45-0.86) for pharmacological interventions, 0.10 (0.23-0.42) for non-invasive neuromodulation (repetitive transcranial magnetic stimulation, transcranial direct current stimulation, electroshock treatment), -0.02 (-0.32-0.29) for psychotherapeutic interventions, and -0.09 (-0.47-0.29) for compulsory diets. Between-study heterogeneity was tau2 = 0.08. Conclusion: Subgroup analysis suggests that capsulotomy, DBS, olanzapine, and cyproheptadine may result in higher BMI/weight differences and longer posttreatment weight maintenance. Meta-regression with neural nets indicates that the mode of action of interventions, resulting in adverse drug reactions for some patients with specific pharmacogenetic profiles, has a higher chance of affecting treatment outcomes. These findings suggest that pharmacogenomic testing and precision medicine need to be explored further for AN patients. Study registration: PROSPERO:CRD42023421545
