We performed a systematic review of studies that investigated the effect of abnormalities of the tumour suppressor gene p53 upon prognosis in patients with colorectal cancer. The methods used to assess p53 status were immunohistochemistry (IHC), indicating abnormal accumulation of p53, and sequence analysis, indicating presence of p53 mutations (mut). We identified 168 reports, with 241 comparisons of relevant end points and survival data on 18 766 patients. We found evidence of both publication bias and heterogeneity of results. Our analysis was hampered by variability in both the assessment of p53 status and the reporting of results. We used a trim and fill method to correct for publication bias and minimised heterogeneity by using well-defined clinical subgroups for the assessment of outcomes. Overall, patients with abnormal p53 were at increased risk of death: relative risk (RR) with IHC 1.32 (95% confidence interval (c.i.) 1.23-1.42) and with mutation analysis 1.31 (95% c.i. 1.19-1.45). The adverse impact of abnormal p53 was greater in patients with lower baseline risk of dying: good prognosis RR (mut) 1.63 (95% c.i. 1.40-1.90) and poor prognosis RR (mut) 1.04 (95% c.i. 0.91-1.19). We found no effect of abnormal p53 on outcome in patients treated with chemotherapy. Abnormal p53 was associated with failure of response to radiotherapy in patients with rectal cancer: RR (mut) 1.49 (95% c.i. 1.25-1.77).