Background: The outcome of local excision of early rectal cancer using transanal endoscopic microsurgery (TEM) lacks consensus. Screening has substantially increased the early diagnosis of tumours. Patients need local treatments that are oncologically equivalent to radical surgery but safer and functionally superior. Methods: A national database, collated prospectively from 21 regional centres, detailed TEM treatment in 487 subjects with rectal cancer. Data were used to construct a predictive model of local recurrence after TEM using semiparametric survival analyses. The model was internally validated using measures of calibration and discrimination. Results: Postoperative morbidity and mortality were 14.9 and 1.4 per cent respectively. The Cox regression model predicted local recurrence with a concordance index of 0.76 using age, depth of tumour invasion, tumour diameter, presence of lymphovascular invasion, poor differentiation and conversion to radical surgery after histopathological examination of the TEM specimen. Conclusion: Patient selection for TEM is frequently governed by fitness for radical surgery rather than suitable tumour biology, TEM can produce long-term outcomes similar to those published for radical total mcsorectal excision surgery if applied to a select group of biologically favourable tumours. Conversion to radical surgery based on adverse TEM histopathology appears safe for p T1 and p T2 lesions. © 2009 British Journal of Surgery Society Ltd.