The concept of a very limited effective period for prophylactic antibiotic action was challenged in a prospective, controlled and randomized study featuring a simplified method for assessing the degree of contamination in the operative field during colorectal surgery. The 226 patients were given 1 g metronidazole IV on induction of anaesthesia and 12 h postoperatively. Following standardized saline irrigation of the abdominal cavity, a uriculture dipslide was dipped in the residual fluid just before abdominal closure. The dipslide was incubated for 18 h, and colony counts of coliforms and Staphylococcus aureus were made. The number of colony-forming units (CFU) was classified as 0, 1–4 or ⩾ 5 (n = 123, 33 and 70). The patients with CFU ⩾ 5 were randomized to receive cefuroxime (1·5 g IV/8 h for 2 days) or no more antibiotics. In the patients given only metronidazole, the incidence of postoperative infections rose with the colony counts (5·7, 9·1 and 41·2 per cent with CFU 0, 1–4 and ⩾ 5). Addition of cefuroxime reduced the 41·2 per cent infection rate to 8·3 per cent (P = 0·003). The method thus identified a high-risk group in which a short postoperative course of cefuroxime significantly lowered the rate of sepsis.