Aim: The high placebo response in acne clinical trials has slowed down the progress of new drug development. Therefore, this study will use a model-based meta-analysis method to quantitatively analyze the placebo response and explore its influencing factors. Methods: This study included 21,627 subjects from acne-related randomized double-blind placebo-controlled trials. Pharmacodynamic models were developed for four endpoints: total lesion counts absolute change, noninflammatory lesion counts absolute change, inflammatory lesion counts absolute change, and treatment success rate defined by the investigator global assessment scale. Covariate modeling and subgroup analyses were performed to explore influencing factors. Placebo response under different conditions was simulated based on the final model. Results:. Model simulation results showed that for the typical population, after 12 weeks of placebo intervention, their total lesion counts, noninflammatory lesion counts and inflammatory lesion counts were reduced by 19.05 (95% confidence interval: 16.54-21.52), 10.71 (9.18-12.17) and 8.58 (7.55-9.58) from baseline, respectively; the treatment success rate was 11.83% (9.41%-14.35%). Baseline lesion counts, female proportion, Caucasians proportion, sample size, dosage form, type of control drug, skincare program, number of centers, funding source, study locations, and publication year could all influence placebo response. We also used our model to synthesize placebo arms for two single-arm studies, demonstrating the efficacy superiority of the drugs. Conclusion: This study quantitatively described the time-response relationship of placebo in acne clinical trials, revealing the influence of baseline characteristics and trial design characteristics, which constructed a framework for synthesizing placebo controls and provide a methodological tool for clinical trial design.

Pharmacometrics, as the core discipline of model-informed drug development paradigm, quantitatively characterizes drug exposure-response dynamics using mathematical models. However, its traditional methodology, represented by nonlinear mixed-effects modeling, is inherently resource-intensive and often struggles to fully capture complex biological variability. The rapid advancement of artificial intelligence technologies offers a transformative data-driven paradigm, providing superior efficiency and predictive accuracy. Yet, the artificial intelligence approaches suffer from the “black box” constraint. This review detailed the deep, complementary integration of artificial intelligence and pharmacometrics, demonstrating how this synergy is necessary to satisfy the complex, triple demands of accuracy, efficiency, and regulatory explainability. We systematically explored innovative applications across the entire pharmacometrics workflow, including model optimization, covariate screening, virtual population generation, etc. In each section, we also presented specific application cases or workflows. Furthermore, we outlined current challenges in the implementation of artificial intelligence and proposed a vision for interdisciplinary and cross-institutional collaboration.