To clarify the theory of the pattern-mixture models with adjusting imputed values and achieve its model analysis in the clinical trial of acupuncture with longitudinal data. Methods:The three-stage methods were used to construct the pattern-mixture model with non-random data missing mechanism, and the imputed values were merged by the multiple imputation approach. All parameter estimation and test analyses were completed by the SAS 9.4 software. Results:Data of 384 （384/1075, 35%） patients were extracted from the original database, a total of 15 cases were missing （15/25, 60%）. The curative effect of electroacupuncture were stable at different correction levels. Conclusion:The combination of pattern-mixture model with adjusting imputed values and multiple imputation provides a reliable methodological support for the sensitivity analysis of the multiple imputation approach, and provides a new idea for the clinical study of longitudinal data with non-random missing mechanism.