The first methadone maintenance treatment clinic in China was opened in March, 2004. By the end of March, 2007, 51,854 heroin drug users had been treated in methadone maintenance treatment clinics and 35,888 patients were currently being treated. It was planned to have 500 MMT clinics operating in China by the end of 2007 and 1,500 clinics by the end of 2008. Retaining patients is a critical challenge of the program. The primary aim of this study was to determine retention rates and identify factors that influence patients' retention in Chinese methadone maintenance program. 1,003 patients were recruited from eight methadone clinics in Guizhou province and were followed. The Cox model was used to identify the factors influencing client retention. Generalized estimating equations (GEE) were used to find factors influencing clients' continued heroin use as measured by random urine tests. The study found that the time weighted average methadone dose was associated with patients' retention. For each one milliliter increased in methadone dose, the hazard of dropping out declined by 1% (HR=0.99, P=0.003). We recommend that at least 50ml/day be used as a maintenance dose for patients treated in methadone clinics in China although higher dose can achieve better retention. Those previously in a detoxification program (HR=1.31, P=0.041) were more likely to drop out of methadone maintenance. Patients who intended to be treated in methadone clinic for life were more likely to stay in methadone maintenance (HR=0.657, P=0.026). Through follow up interviews we also found that methadone maintenance treatment could improve patients' physical health, decrease drug use, improve employ status, improve relationships with family/society, decrease illegal activities, and improve psychological health. This study also found that about 30% of urine samples remained positive for the first 10 months and then decreased fast suggesting that patients need to be treated for at least 10 months to achieve abstinence. Time weighed average dose was negatively associated with urine sample positive rate.