Cardiac rehabilitation is widely recommended for patients after acute coronary syndrome, and physical exercise features prominently in guidelines published by professional societies.1 Rehabilitation programmes vary widely in content, specialists involved, location and duration, and the exercise component varies widely from health education, through lifestyle advice and physical activity encouragement to supervised training. There is undoubted variation between centres in hours of formal exercise training offered within programmes.

Sandercock and colleagues postulate that underuse of exercise training could explain why cardiac rehabilitation had no significant effect on principal outcome measures—all-cause mortality and cardiovascular morbidity—in recent randomised controlled trials.2 They report the change in a variety of measures (or ‘proxy measures’) of physical fitness between the beginning and end of cardiac rehabilitation programmes in four centres (of six recruited to the study). Patients were predominantly those after elective revascularisation (51%) and acute myocardial infarction (35%), but also included some with chronic stable angina. Start and end values were taken from monitoring measurements recorded by rehabilitation staff.

The discussion raises important issues over relevance and reproducibility of measures used to assess exercise capacity in patients undergoing cardiac rehabilitation. Briefly, treadmill and bicycle ergometer measurements are more objective and give better estimates of symptom-limited physical working capacity, although stages on some protocols may be too widely spaced for older and cardiac patients, and they depend on expensive equipment. Six minute walk and incremental shuttle walk tests are easier to administer, but are ‘proxy measures’, which do not measure physical working capacity directly, and walking distance may be increased without increasing physical working capacity. Translation of these disparate measures to a common standard, the ‘metabolic equivalent of task’ (MET), depends on the conversion formulae used.3 The study includes findings from four different tests, with ‘standardised effect …