Empirical identification and validation of eating disorder phenotypes in a multisite clinical sample.

Identified problems with the classification system of eating disorders (EDs), including its imperfect application to clinical samples, challenge its validity and limit its utility. The present study aimed to empirically identify and validate ED phenotypes in a multisite clinical sample using latent profile analysis (LPA). ED symptom data collected from 687 individuals were included in LPA. Identified latent profiles (LPs) were compared on clinical validators. Five LPs were identified: LP1 (n = 178), objective bingeing and multiple purging methods; LP2 (n = 172), objective bingeing without purging; LP3 (n = 130), objective bingeing and vomiting; LP4 (n = 108), low/normal weight and excessive exercise; LP5 (n = 99), low/normal weight and absence of ED symptoms. Validation analyses demonstrated the most extreme psychopathology/medical morbidity in LP1 and the least in LP5. LP1 and LP3 were most likely to report medication treatment for EDs. Identified LPs imperfectly resembled diagnostic and statistical manual of mental disorders-IV-TR EDs. Multiple purging methods and the absence of ED cognitions marked differences in severity across groups, whereas low weight did not. Clinical differences in psychopathology, medical morbidity, and treatment utilization validated groups. Future research should examine longitudinal stability of empirically-derived phenotypes and incremental validity of alternative classification schema to inform DSM-V.