ObjectiveThe aim of this study was to construct a composite scoring system to predict the probability of placebo response in adolescents with Major Depressive Disorder (MDD). MethodParticipants of the current study were 151 adolescents (aged 12–17 years) who were randomized to the placebo arm (placebo transdermal patches) of a randomized controlled trial (RCT) comparing the selegiline transdermal patch with placebo (DelBello et al., 2014). The primary outcome of response was defined as a CGI-I score of 1 or 2 (very much or much improved) at week 12 (study-end) or exit. As a first step, a multiple logistic mixed model was used to estimate the odds of placebo response from each predictor in the model, including age, CDRS-R total at baseline (depressive symptom severity), history of recurrent depression (yes vs. no), sex (female vs. male), and race (non-Caucasian vs. Caucasian). On the basis of the initial logistic mixed model analysis, we then constructed an Adolescent Placebo Impact Composite Score (APICS) that became the sole predictor in a re-specified Bayesian logistic regression model to estimate the probability of placebo response. Finally, the AUC for the APICS was tested against a nominal area of 0.50 to evaluate how well the APICS discriminated placebo response status. ResultsAmong the 151 adolescents, with a mean age of 14.6 years (SD = 1.6) and a mean baseline CDRS-R total of 60.6 (SD = 12.1), 68.2% were females, 50.3% was Caucasian, and 39.7% had a history of recurrent depression. Placebo response rate was 58.3%. Based on the logistic mixed model, the re-specified equation with the highest discriminatory ability to estimate the probability of placebo response was APICS = age + (0.32 × CDRS-R Total at baseline) + (−2.85 × if female) + (−5.50 × if history of recurrent depression) + (−5.85 × if non-Caucasian). The AUC for this model was 0.59 (p = .049). Within a Bayesian decision-theoretic framework, in 95.5% of the time, the 10,000 posterior Monte Carlo samples suggested that as APICS decreased the probability of placebo response increased. The observed APICS and related probability of responding to placebo in this adolescent sample ranged from 14.1 = 74.1% (in placebo responders) to 39.1 = 41.8% (in placebo non-responders). ConclusionThe APICS model estimates the probability of placebo response in adolescents with MDD with a modest degree of accuracy (AUC = 0.59) and with a reasonable degree of positive predictive value (74.5%), and is based on five previously identified patient characteristics of placebo response from prior meta-analytic studies (Bridge et al., 2009; Cohen et al., 2010) of randomized placebo-controlled trials of antidepressants in youth with MDD. Calculation of the APICS should be restricted to the range of the adolescent ages (12–17 years) and CDRS-R total scores (17–113); thus, the APICS can assume possible calculated values and related probability of responding to placebo ranging from about 3 (84%) to 53 (25%). The APICS Bayesian logistic model, based on a given aggregate patient profile, has a range of predicted probabilities of placebo response that is fairly wide, albeit truncated, but potentially meaningful for predicting the probability of placebo response among adolescent youth with MDD. The ability of the APICS to objectify the probability of placebo response in adolescents with MDD could be accounted for in the clinical research design of the trial itself and perhaps aid clinicians in treatment strategy for youth who are more likely to experience placebo response.