Aims: Equivocal results of association between metformin and cancer-specific survival need more investigation. We tested the hypothesis that adherence to the drug had a lower cancer-specific mortality in a homogeneous population (i.e. regular users). Methods: The Australian Cancer database was linked to the Pharmaceutical Benefits Scheme data and the National Death Index. Adherence to metformin was calculated by proportion of days covered. Cox regression models with time-varying covariates were used to estimate multivariable-adjusted cause-specific hazard ratios (HRs) and 95% confidence intervals (CI) for the association of adherence to metformin and cancer-specific mortality. Results: Between 2003 and 2013, three separate cohorts of 6717, 3121, and 1854 female patients were identified with newly diagnosed breast, colorectal, or endometrial cancer. The 1-year adherence was similar at baseline in three cohorts, on average 75%. Each 10% increase in 1-year adherence to metformin reduced cancer-specific mortality among women with breast cancer (adjusted HR = 0.95; 95% CI, 0.93–0.97), colorectal cancer (adjusted HR = 0.94; 95% CI, 0.91–0.96), or endometrial cancer (adjusted HR = 0.95; 95% CI, 0.90–0.99). The inverse associations remained unchanged in most subgroup analyses. Conclusions: Among metformin users, adherence to this drug is inversely associated with reduced cancer-specific mortality. If confirmed, metformin could be considered as an adjuvant treatment.