To test the hypothesis that "induction" intensive insulin therapy (IIT) needs to be followed by "maintenance therapy" to preserve β-cell function, and to evaluate the impact on β-cell function over 2 years of two approaches to maintenance therapy: intermittent short-term IIT every 3 months vs daily metformin. In this trial, 24 adults with a mean type 2 diabetes mellitus (T2DM) duration of 2.0 ± 1.7 years and glycated haemoglobin (HbA1c) levels 6.4 ± 0.1% (46 ± 1.1mmol/mol) were randomized to 3 weeks of induction IIT (glargine, lispro) followed by either repeat IIT for up to 2 weeks every 3 months or daily metformin. Participants underwent serial assessment of β-cell function using the Insulin Secretion-Sensitivity Index-2 (ISSI-2) on an oral glucose tolerance test every 3 months. The primary outcome of baseline-adjusted ISSI-2 at 2 years was higher in the metformin arm compared with intermittent IIT (245.0 ± 31.7 vs 142.2 ± 18.4; P = .008). Baseline-adjusted HbA1c at 2 years (secondary outcome) was lower in the metformin arm (6.0 ± 0.2% vs 7.3 ± 0.2%; P = .0006) (42 ± 2.2 vs 56 ± 2.2mmol/mol). At study completion, 66.7% of participants randomized to metformin had an HbA1c concentration ≤ 6.0% (≤42mmol/mol), compared with 8.3% of those on intermittent IIT (P = .009). There were no differences in insulin sensitivity. After induction IIT, metformin was superior to intermittent IIT for maintaining β-cell function and glycaemic control over 2 years. The strategy of induction and maintenance therapy to preserve β-cell function warrants exploration in early T2DM.