EDITOR—Connolly's view that Kalra et al's findings clearly show that the results of anticoagulation studies can be replicated in general medical practice may be an overinterpretation of the data. 1 2 The 167 patients in the study may not have been a representative sample of high risk patients found in general practice. Patients were recruited from secondary care medical clinics, not general practice, and bias in the study group could have been introduced by selection for hospital referral. Furthermore, bias could have occurred by exclusion of those high risk patients who were already taking warfarin, who may have been considered by their general practitioner to have been at a lower risk of haemorrrhagic complications or a greater risk of stroke. Knowledge of the outcome of the 76 patients already taking warfarin, who presumably were being managed by the same anticoagulation service as trial patients, is also required.

The most striking finding from Kalra et al's study …