We aimed to develop a metformin dosing strategy to optimise efficacy and safety in patients with reduced kidney function. Metformin data from two studies stratified by kidney function were analysed. The relationship between metformin clearance and kidney function estimates was explored using a regression analysis. The maintenance dose range was predicted at different bands of kidney function to achieve an efficacy target of 1 mg/L for steady-state plasma concentrations. The dosing strategy was evaluated using simulations from a published metformin pharmacokinetic model to determine the probability of concentrations exceeding those associated with lactic acidosis risk, i.e. a steady-state average concentration of 3 mg/L and a maximum (peak) concentration of 5 mg/L. A strong relationship between metformin clearance and estimated kidney function using the Cockcroft and Gault (r2 = 0.699), MDRD (r2 = 0.717) and CKD-Epi (r2 = 0.735) equations was found. The probability of exceeding the safety targets for plasma metformin concentration was <5% for most doses and kidney function levels. The lower dose of 500 mg daily was required to maintain concentrations below the safety limits for patients with an eGFR of 15–29 mL/min. Our analysis suggests that a maximum daily dose of 2250, 1700, 1250, 1000, and 500 in patients with normal kidney function, CKD stage 2, 3a, 3b and 4, respectively, will provide a reasonable probability of achieving efficacy and safety. Our results support the cautious of use metformin at appropriate doses in patients with impaired kidney function.