There is a need for large scale, pragmatic trials of complementary and alternative medicine in a primary care setting to answer questions about the 'real world' effectiveness of such methods. Randomisation and treatment in such trials should be prospective, but retrospective recruitment is possible, especially in chronic conditions. This involves contacting patients who have previously consulted their GP (identified through GP database searches) rather than encouraging GPs to refer patients as they present. We describe a prospective randomised trial of acupuncture for chronic headache, currently underway, as an example of retrospective recruitment. A pilot study of recruitment and recruitment modelling was undertaken. The target sample size was 300, 26% (n = 36) of general practices approached took part in the study, 32 completed the required database search. Practices used diagnostic term searches, prescription searches, or a combination of both. On average 1.7% of the total practice populations were identified as headache consulters, letters were sent to 4128 patients. 12% of identified patients were randomised (n = 401). Retrospective recruitment methods are feasible for CAM trials in chronic, stable conditions modelling can provide accurate data for planning such studies. Retrospective recruitment can be more efficient than prospective, further research is required on the generalisibility of results from populations recruited in this way.