To determine the accuracy of using nitroglycerine as a 'test of treatment' in the diagnosis of cardiac chest pain we undertook a systematic review of studies of diagnostic accuracy. Databases searched included PubMed, Cochrane Database, Google Scholar, Science Citation Index, EMBASE and manual searching of bibliographies of known primary and review articles. Studies were included if sublingual nitroglycerine was the index test, its effect on the patient's pain score was recorded and the reference test was performed on at least 80% of patients. The data from the five papers were used to form 2x2 contingency tables. Five eligible studies were found, all in the acute setting (although one paper collected its data in the follow-up setting, all patients had acute presentations). The sensitivity ranged from 35% to 92% and the specificity from 12% to 63%. However, in all but one paper the Youden indices were close to zero suggesting that the response to nitroglycerine is not useful as a diagnostic test. The combined sensitivity was 0.52 (95% CI 0.48 to 0.56) and combined specificity was 0.49 (95% CI 0.46 to 0.52). The diagnostic OR from the combined studies was 1.2 (95% CI 0.97 to 1.5), which is not significantly different from 1. In the acute setting, nitroglycerine is not a reliable test of treatment for use in the diagnosis of coronary artery disease. However, further studies are needed to determine the diagnostic accuracy of nitroglycerine for recurrent exertional chest pain.