Design
Two independent, consecutive cohorts of patients (group 1, n=28; group 2, n=17) underwent all-night polygraphic sleep recordings and other clinical evaluations during the first and fourth weeks of a 1-month inpatient treatment program within a Veteran Affairs Medical Center. They were reevaluated 3 months following discharge to the community. None were treated with disulfiram or other medications during or after hospitalization.

Patients
All subjects were male veterans with primary alcoholism and without significant preexisting, secondary, or comorbid diagnoses such as major medical problems, depression, antisocial personality, or drug addiction.

Outcome Measures
Relapse was defined as any alcohol consumption between discharge from the hospital and 3-month follow-up.

Results
Ten (36%) of 28 patients in group 1 were Relapsers at 3-month follow-up. Relapsers in group 1 showed significantly shorter Rapid Eye Movement (REM) latency, increased Rapid Eye Movement percent (REM%), and increased REM Density during their admission sleep studies compared with Abstainers. To replicate these observations, group 2 was then studied as a validation sample. Six (35%) of 17 patients relapsed. As in group 1, Relapsers had significantly shorter REM latency and increased REM% compared with Abstainers; REM Density was not significantly different in the Relapsers as compared with Abstainers in group 2. Using a principal components analysis based on these three REM sleep measures to determine "REM pressure," three separate discriminant function analyses (DFAs) were calculated: one for each group and one for all patients (n=45) together. The DFA from group 1 correctly classified 22 (78.6%) of the 28 patients in group 1 and 13 (76.5%) of the 17 patients in group 2 as Relapsers or Abstainers. The DFA from group 2 correctly classified 13 (76.5%) of the 17 patients in group 2 and 23 (82.1%) of the 28 patients in group 1. The DFA formed from both groups together correctly classified 36 (80%) of the 45 patients. When the REM sleep measures at hospital admission and discharge were compared, no statistically significant effect of time was observed. Abstinence and relapse were not consistently related to other clinical measures at the time of hospital admission such as age, duration and severity of alcoholism, marital status, employment, hepatic enzyme levels, cognitive performance, or depression ratings.

Conclusion
Short REM latency, increased REM%, and, possibly, increased REM Density at the time of admission to a 1-month inpatient alcohol treatment program predict relapse in nondepressed patients with primary alcoholism by 3 months following hospital discharge.