A person-centred and data-driven approach to phenotyping anorexia nervosa

Anorexia nervosa (AN) is a severe psychiatric disorder (ED) with high mortality, marked functional impairment, and substantial phenotypical heterogeneity. Despite extensive research, treatment outcomes remain poor, and it is unclear why some individuals improve while others follow more persistent and severe courses. Existing diagnostic subtypes show limited value in predicting illness trajectory or treatment response. This study examined the distinct AN phenotypes which emerged within a treatment-seeking sample and considered how these differ in their response to standard treatment.

Using data from the TRIANGLE study, latent profile analysis was used to identify phenotypic subgroups of adult patients with AN or atypical AN admitted to hospital for intensive care (n = 382), based on a variety of anthropometric and clinical variables. Following profile allocation, separate linear mixed model analyses (n = 370) examined differences between groups and over time (18 months) in depression, anxiety, and stress symptoms, work and social impairment, body mass index (BMI), and ED psychopathology.

A four-profile solution best fit the data. Profiles differed in illness duration and symptom severity. One profile with long illness duration (mean ± standard deviation = 9 ± 7 years) and high symptom severity, a long-duration group (21 ± 11 years) with moderate severity, and two short-duration groups (4 ± 3 and 5 ± 3 years), one with high and one with comparatively lower severity, respectively. The high-severity, shorter-duration profile showed significant improvement across all variables over the 18-month follow-up period, whereas the highest-severity, longer-duration profile showed significant improvement on all variables except ED psychopathology. Where improvements were observed in both high-severity profiles, the magnitude of change was generally greater in the longer-duration profile, except for BMI, where it was equal and work and social impairment, where the shorter-duration profile showed greater improvement. In contrast, the lower-severity profiles showed significant improvements only in BMI and work and social impairment, which were smaller in magnitude than those observed in the high-severity profiles.

Subtyping and treatment planning for AN must recognise that prognosis is shaped by multiple interacting factors rather than any single indicator and incorporate psychological, social, and functional complexity.