Background: Baclofen is a GABA-B receptor agonist currently used in the treatment of spasticity. In recent years, baclofen has been used to reduce craving, voluntary alcohol intake and withdrawal syndrome of alcoholic patients. To date, there are no data available to estimate the relationship between baclofen exposure and the variation of craving. The first objective of this study was to investigate the variation of craving as a function of exposure, and the second was to explore the possible existence of baclofen responders and nonresponders. Methods: Sixty-seven outpatients, 43 men/24 women (weight: 73 kg [42 to 128]; age: 49 years old [29 to 68]) followed in the addictology unit, were studied during 3 months after treatment initiation. Baclofen was administered by oral route. Therapeutic drug monitoring enabled the measurement of plasma concentrations. Craving level was assessed by the Obsessive-Compulsive Drinking Scale (OCDS). A population pharmacokinetic (PK)-pharmacodynamic analysis of the OCDS variation following baclofen administration was performed. Demographic data, biological data, and tobacco consumption were tested for their influence on the parameters. Results: Data were modeled with a 1-compartment model with first-order absorption and elimination. PK analysis confirms the results of our previous study. An E max model best-described the exposure-OCDS relationship. None of the covariates tested was able to improve the fit or decrease intersubject variability. However, 2 subpopulations were defined for the exposure corresponding to half the maximal effect (BE50). The proportion of patients being classified as responders was 38% (95% confidence interval [CI] 7 to 76), the maximal decrease in OCDS (E max) was 72% (95% CI 25 to 85), and the BE50 was 12.6 (95% CI 0.02 to 74.3) or 4,390 (95% CI 20.4 to 31,800) h mg/l for responders and nonresponders, respectively. Conclusions: We defined the relationship between baclofen exposure and craving in patients with alcohol use disorder. Baclofen treatment decreased craving in all patients. However, we drew up the hypothesis of 2 subpopulations of patients differentiated by their speed of response. A wide interindividual variability in response was depicted, making it currently impossible to predict which group a patient will belong to.