The aim of this study was to estimate the effect of the radiologist's interpretation of plain radiographs on the earliness of diagnosis of cancer recurrences. Data consisted of patients who had undergone primary treatment and were resident in the Tampere University Hospital Area in Finland during 1991-1997. Consecutive patients were randomised in a double-reading arm (an oncologist and a radiologist independently interpreting radiographs), and in a single-reading arm (interpretation by an oncologist only; if necessary, a radiologist's clinical report was obtained following a separate request). The time of diagnosis of recurrence and death were estimated by the cumulative probabilities of actuarial method with the Wilcoxon (Gehan) test. There were 869 eligible participants, mostly breast cancer patients (n = 516). In total, 227 recurrences were diagnosed, and of these 55 on plain radiographs, which is 24.2% of the total number of recurrences. There was no statistically significant difference between the arms in the number of recurrences (P = 0.85) or in the time of detecting the recurrence (P = 0.64). Altogether, 225 (25.9%) died from cancer and 38 (4.4%) from other causes. There was no statistically significant difference (P = 0.34) in survival between the two arms during the follow-up to 5 years. Double-reading of plain radiographs does not offer any extra benefit for the detection of recurrences or for patient's survival compared with single-reading.