Background: Both mesh and suture repair are used for the treatment of umbilical hernias, but for smaller umbilical hernias (diameter 1–4 cm) there is little evidence whether mesh repair would be beneficial. In this study we aimed to investigate whether use of a mesh was better in reducing recurrence compared with suture repair for smaller umbilical hernias. Methods: We did a randomised, double-blind, controlled multicentre trial in 12 hospitals (nine in the Netherlands, two in Germany, and one in Italy). Eligible participants were adults aged at least 18 years with a primary umbilical hernia of diameter 1–4 cm, and were randomly assigned (1:1) intraoperatively to either suture repair or mesh repair. In the first 3 years of the inclusion period, blocked randomisation (of non-specified size) was achieved by an envelope randomisation system; after this time computer-generated randomisation was introduced. Patients, investigators, and analysts were masked to the allocated treatment, and participants were stratified by hernia size (1–2 cm and >2–4 cm). At study initiation, all surgeons were invited to training sessions to ensure they used the same standardised techniques for suture repair or mesh repair. Patients underwent physical examinations at 2 weeks, and 3, 12, and 24–30 months after the operation. The primary outcome was the rate of recurrences of the umbilical hernia after 24 months assessed in the modified intention-to-treat population by physical examination and, in case of any doubt, abdominal ultrasound. This trial is registered with ClinicalTrials.gov, number NCT00789230. Findings: Between June 21, 2006, and April 16, 2014, we randomly assigned 300 patients, 150 to mesh repair and 150 to suture repair. The median follow-up was 25·1 months (IQR 15·5–33·4). After a maximum follow-up of 30 months, there were fewer recurrences in the mesh group than in the suture group (six [4%] in 146 patients vs 17 [12%] in 138 patients; 2-year actuarial estimates of recurrence 3·6% [95% CI 1·4–9·4] vs 11·4% (6·8–18·9); p=0·01, hazard ratio 0·31, 95% CI 0·12–0·80, corresponding to a number needed to treat of 12·8). The most common postoperative complications were seroma (one [<1%] in the suture group vs five [3%] in the mesh group), haematoma (two [1%] vs three [2%]), and wound infection (one [<1%] vs three [2%]). There were no anaesthetic complications or postoperative deaths. Interpretation: This is the first study showing high level evidence for mesh repair in patients with small hernias of diameter 1–4 cm. Hence we suggest mesh repair should be used for operations on all patients with an umbilical hernia of this size. Funding: Department of Surgery, Erasmus University Medical Center, Rotterdam, Netherlands.