Many factors can signal risk of relapse, especially in AN patients.
High relapse rates among eating disorders patients, especially those with anorexia nervosa (AN), are still problematic. For example, 40% to 50% of AN patients experience relapse. Those with bulimia nervosa (BN) and binge eating disorder (BED) have relapse rates of about 30%. Up to 40% of patients with other specified eating disorders (OSFED) can relapse.
A team led by Dr. Margaret Sala, Yeshiva University, the Bronx, and colleagues at the University of North Carolina, Chapel Hill, the Karolinska Institut, Stockholm, Yale University, and the University of Louisville recently conducted a meta-analysis of studies of relapse among eating disorders patients through December 2021. Thirty-five studies were selected for further analysis (J Psychiatry Res. 2023. 158: 281).
Relapse rates and factors identified
Dr. Salas and her team identified a number of factors that pointed to the risk of relapse. The authors found that, across eating disorders diagnoses, approximately one-third of individuals experienced relapse. This high rate of relapse is problematic, as relapse can lead to cycles of continuous readmission and discharge from treatment (Child Adolesc Ment Health. 2014. 19:115; Clin Psychol. 2017. 21:143).
The authors advise considering watching other eating disorder characteristics for risk of relapse. The search begins with lower pre- and post-treatment body weights, particularly among patients with AN and BN. More severe eating disorder psychopathology predicted a greater likelihood of response for all eating disorders. However, they note that results from a few studies showed the opposite trend—more severe psychopathology predicted a lower likelihood of relapse (Int J Eat Disord. 1996.19:279). A third factor was post-treatment dietary intake, including energy density, variety in diet, protein levels, and daily caloric intake (Int J Eat Disord. 2012. 45:79).
Age. While a diagnosis of AN binge-purge subtype (AN-BP) and longer illness produced a higher likelihood of relapse among individuals with AN, this was not the case with younger AN patients, who had a shorter duration of illness (Int J Eat Disord. 2007. 40:129).
Comorbidities. Psychiatric comorbidity and worse overall psychosocial/global functioning predict relapse across eating disorders diagnoses (Appetite. 2010. 55: 656). Specific comorbidities associated with a higher likelihood of relapse have included exposure to traumatic events (Psychiatry Res. 2012. 200: 518), obsessive-compulsive symptoms (Psychol Med. 2004. 34:671), depressive symptoms (Appetite. 2010. 55:656), a history of suicide attempts, and postpartum depression (Br J Psychiatry. 1999. 174:135). However, other research has not found a significant relationship between psychiatric comorbidities (such as depression) and likelihood of relapse (Am J Psychiatry. 2002. 159:96).
Leptin levels. Higher leptin levels were a robust predictor of a lower likelihood of relapse among AN patients, but no conclusions could be drawn about how leptin levels affected the course in other types of eating disorders. And, the leptin level only predicted lower odds of relapse when it was assessed at discharge, suggesting that leptin levels at discharge may be a biomarker of AN relapse. Leptin levels may be indicative of fat mass or could reflect the reward status of food restriction (Eur Eat Disord Rev. 2021. 29:634). However, the authors advise that this finding should be interpreted with caution, given the small number of cases included in this study.
How did treatment affect the course?
One factor to consider is the timing when the factor linked to relapse is measured. The timing may influence the strength with which it is related to likelihood of relapse. Eating disorder characteristics assessed immediately after treatment may not robustly predict the likelihood of relapse, because some types of treatment (e.g., inpatient and residential care) are conducted in a controlled environment. Therefore, the ability to abstain from disordered eating during treatment may not predict the ability to do so outside of treatment. For example, some research suggests that whereas indicators of severity of eating disorder pathology assessed before treatment predicted likelihood of relapse, indicators of the severity of eating disorder pathology assessed immediately after treatment did not (Int J Eat Disord. 2015. 48:337). Specific psychiatric comorbidities were predictive of relapse, and only comorbid depression was significantly associated with a higher likelihood of relapse.
Types of eating disorders. Several effects were only significant among samples from individuals with AN and/or BN. Specifically, having a higher body mass index (BMI), and a comorbid psychiatric disorder significantly predicted a higher likelihood of relapse in samples comprised solely of individuals with AN, but not in samples of individuals with other eating disorders (such as BN and mixed eating disorders). More severe psychopathology predicted a higher likelihood of eating disorder relapse. Almost all eating disorder symptoms (except over-exercise) were significantly associated with relapse. Dietary restraint was most strongly associated with relapse, suggesting that although all eating disorder symptoms should be targeted. It may be particularly important to watch for restraint during treatment, and to ensure that it lessens before the patient is discharged.
Finding that AN-BP is associated with a higher likelihood of relapse suggested that AN-BP may reflect a later and more severe stage of illness (Eating and Weight Disorders-Studies on Anorexia, Bulimia and Obesity. 2021. https://doi.org/10.1007/s40519-021-01226-0). Patients with AN-R only have one route to relapse (i.e., weight loss), whereas those with AN-BP have two routes to relapse (i.e., weight loss or binge/purge behaviors) (Psychol Med. 2004. 34:671).
Eating behaviors. Normalized and varied eating behaviors were associated with a lower likelihood of relapse. Notably, most of the studies that analyzed the predictive effect of meal energy density/variety were studies of AN patients. The authors recommend that future approaches to AN treatment and relapse prevention research should examine the efficacy of having a strong emphasis on helping patients learn to eat a larger variation of energy-dense meals.
Patients who are trying to maintain a weight that is too low may be engaging in eating disordered behaviors to do so. Accepting a body-appropriate weight may improve a patient’s prognosis. Thus, having a lower BMI at discharge could be an indication of a higher likelihood of returning to an underweight BMI.
Length of follow-up. In 2005, Richard and colleagues found that whereas patient motivation predicted a higher likelihood of relapse in individuals with AN six months after remission, it also predicted a lower likelihood of relapse two years following remission (Eur Eat Disord Rev. 2005. 13:180). Knowing the time points in which certain factors may be most predictive of relapse and would allow for optimal timing in preventing it.
The authors pointed out a number of limitations in their study. For example, the terms “recovery” and “relapse” were not carefully defined across studies. Analyzing a small number of cases in some studies also made it difficult to calculate or categorize the effects. The small group sizes in some studies also reduced confidence in the precision of the predictors.
Overall, these findings have important implications for helping identify characteristics that may lead to relapse, shedding light on key factors maintaining eating disorders, and identifying areas in greater need of assessment before and during treatment, to minimize relapse.
The authors’ findings also have important implications for guiding treatment development research, including: (1) the need for in-depth assessment and monitoring of several factors predictive of relapse; (2) developing treatments that target these factors; and (3) establishing evidence-based guidelines for the optimal time to discharge individuals with eating disorders from treatment.
—Leah Graves, RDN, LDN, CEDS-S, FAED