Depressive symptoms and weight change in inpatients with anorexia nervosa: a cross-lagged panel model

Adrian Meule, David R. Kolar, Thomas Gärtner, Bernhard Osen, Elisabeth Rauh, Silke Naab, Ulrich Voderholzer

TL;DR

In a massive study of over 3,000 inpatients with anorexia nervosa (AN), researchers found that comorbid depression does not hinder weight restoration during specialized treatment. While the study identified statistically significant bidirectional links between depressive symptoms and BMI change, the actual impact was so small that it is likely clinically negligible.

What Problem Does This Paper Address?

Anorexia nervosa is a life-threatening condition where the primary clinical goal is weight restoration. However, a high percentage of patients also suffer from comorbid depression. This leads to a long-standing clinical concern: does depression act as a "brake" on recovery, making it harder for patients to gain weight? Conversely, does gaining weight actually help alleviate depressive symptoms, or are they independent processes?

Previous research on this topic has been inconsistent, often hampered by small sample sizes or a focus on only one direction of the relationship (e.g., how depression affects weight, but not vice versa). This paper seeks to clarify these dynamics using a large-scale, longitudinal dataset.

Methods

The researchers employed a cross-lagged panel model (CLPM) using structural equation modeling. This statistical approach is designed to estimate the directional influence of two variables over time. Specifically, they looked at:

  1. Autoregressive paths: How well admission BMI predicts discharge BMI (and the same for depression).
  2. Cross-lagged paths: How well admission depression predicts change in BMI, and how well admission BMI predicts change in depression.

By using this model, the authors could control for the patients' starting points and determine if one variable truly "predicts" the trajectory of the other.

Datasets

The study analyzed clinical records from 3,011 inpatients treated at four Schoen Clinics in Germany between 2015 and 2020.

  • Demographics: 96% female; 72.8% adults and 27.2% adolescents.
  • Criteria: All patients had a diagnosis of AN (full or atypical) and an admission BMI < 18.5 kg/m².
  • Measures: Body Mass Index (BMI) for weight and the Patient Health Questionnaire–9 (PHQ–9) for depressive symptom severity.

Key Insights and Ideas

  • The "Length of Stay" Mediator: Surprisingly, higher depressive symptoms at admission initially appeared to predict larger weight gains. However, further analysis revealed this was a "length of stay" effect: more depressed patients tended to stay in the hospital longer, and more time in treatment naturally led to more weight gain.
  • Independence at the Start: At the time of admission, there was no correlation between a patient's BMI and their level of depression. A patient could be severely underweight without being severely depressed, or vice versa.
  • Statistical vs. Clinical Significance: This paper serves as a masterclass in interpreting "big data." Because the sample size was so large ($N > 3,000$), the researchers found statistically significant results ($p < .005$) for almost every path. However, they argue these effects are clinically irrelevant. For example, a massive difference in starting depression scores only predicted a 0.17 kg/m² difference in discharge BMI—a change too small to matter in a real-world clinical setting.

Key Results

  • Overall Improvement: Treatment was highly effective. BMI significantly increased ($15.0 \rightarrow 17.4$) and depressive symptoms significantly decreased ($15.0 \rightarrow 8.95$ on the PHQ-9), both with large effect sizes ($d > 1.0$).
  • The Cross-Lagged Effect: Higher BMI at admission predicted slightly smaller decreases in depression ($\beta = .09$), and higher depression at admission predicted slightly larger weight gains ($\beta = .04$).
  • The Bottom Line: Specialized inpatient treatment for AN works for weight restoration regardless of how depressed a patient is at the start.

Limitations and Future Work

  • Self-Reporting: Depressive symptoms were measured via the PHQ-9, which is subject to recall bias or social desirability, rather than structured clinical interviews.
  • Geographic Specificity: The study was conducted in Germany, where inpatient stays are typically longer (average 81 days) than in many other countries. Results might differ in shorter-term treatment settings.
  • Model Evolution: The authors acknowledge that standard CLPMs have been criticized for failing to separate within-person changes from between-person differences. They suggest future studies use at least three assessment points to apply a Random Intercept Cross-Lagged Panel Model (RI-CLPM) for more granular insights.

Summary updated

Abstract

Anorexia nervosa (AN) is marked by a high rate of comorbid depression, which raises the question whether depressive symptoms may adversely affect treatment outcome. Thus, we examined whether depressive symptoms at admission would predict weight change from admission to discharge in a large sample of inpatients with AN. In addition, we also explored the reverse direction, that is, whether body mass index (BMI) at admission would predict changes in depressive symptoms. A sample of 3011 adolescents and adults with AN (4% male) who received inpatient treatment at four Schoen Clinics was analyzed. Depressive symptoms were measured with the Patient Health Questionnaire–9. BMI significantly increased and depressive symptoms significantly decreased from admission to discharge. BMI and depressive symptoms were unrelated at admission. Higher BMI at admission predicted smaller decreases in depressive symptoms and higher depressive symptoms at admission predicted larger weight gain. The latter effect, however, was mediated by longer length of stay. Results indicate that depressive symptoms do not adversely affect weight gain during inpatient treatment in persons with AN. Instead, higher BMI at admission is predictive of smaller improvements in depressive symptoms but this effect seems to be negligible in terms of clinical relevance.
BibTeX
@article{meule2023depressive,
  author = {Adrian Meule and David R. Kolar and Thomas Gärtner and Bernhard Osen and Elisabeth Rauh and Silke Naab and Ulrich Voderholzer},
  title = {{Depressive symptoms and weight change in inpatients with anorexia nervosa: a cross-lagged panel model}},
  journal = {Journal of Psychosomatic Research},
  year = {2023},
  doi = {10.1016/j.jpsychores.2023.111391},
}
Figure