Relapse in Anorexia Nervosa: a Prospective Longitudinal Study

Anorexia nervosa is considered one of the most difficult to treat mental disorders and often treatment outcomes are unsatisfactory. Outpatient treatment is the ideal choice for the treatment of anorexia nervosa, but a large subgroup of patients (50-60%) does not respond to outpatient treatments and need more intensive care, such as hospital rehabilitation. The hospitalization has often a short-term success, as weight is restored to a healthy level, but high percentage of patients relapse during the first year following the discharge. Strategies to limit the post-hospitalization relapse rate are now focused on assessing the effectiveness of different treatments given after discharge. Our group, after evaluating the results of a controlled study, used another strategy. Starting from the observation that the relapse rate is lower in outpatient than in inpatient settings, the investigators hypothesized that the reason of this difference is the nature of inpatient treatment and post-inpatient therapy. For this reason, different procedures and strategies have been implemented during hospitalization to reduce the relapse after discharge.

The primary outcome of interest is the proportion of patients who relapse after discharge from hospitalization. For the calculation of the sample size a 95% confidence interval was established, an expected prevalence of 40% and a desired accuracy of 10%. Based on these premises, it is necessary to have a group of "completers" at 60 weeks of follow-up of 92 patients. To calculate the initial sample needed to generate 92 patients at 60 weeks of follow-up, the investigators considered the attrition rate during treatment (set at 13%) and a subsequent attrition rate at follow-ups (set at 10%). This calculation produces a result of an initial sample of 118 patients.

The frequency and time of relapse will be evaluated using Kaplan-Meier survival analysis. Cox proportional hazards regression analysis will be used to examine the role of predictive variables in predicting time to relapse. Univariate Cox regressions and multivariate Cox regressions will be used to evaluate the potentially predictive variables at baseline, during the treatment process, at the end of therapy and during post-discharge.