726 patients) had hyperthyroidism, and 35 (4.8%) more may have had hypothyroidism. Even if all these patients represent subclinical thyroid disease, it is important to rec¬ ognize these patients. Patients with hyperthyroidism, even subclinical forms, are at a greater risk for recurrent atrial fibrillation, osteopenia, and iodine-induced clinical hy¬ perthyroidism. Patients treated during subclinical hypo¬ thyroidism show improved symptom and psychometric testing scores.3 Patients with atrial fibrillation should also undergo testing for hyperlipidemia and coronary artery dis¬ ease, conditions that overlap in the need to be tested for thyroid status. It is unfortunate that proper follow-up re¬ ports are not available on the tested subjects.During the course of the study, serum thyrotropin level testing replaced free thyroxine analysis, a strategy that is supported by the literature4 but is more expen¬ sive. It is also more accurate, however, particularly in the hospitalized sick population, such as the one under study. Even using the data in the study by Krahn et al,1 it cost them $ 1760 to find each new patient with hyperthyroid¬ ism, a figure far lower than other acceptable costeffectiveness assessments,5,6 in which costs as high as $26 130 per quality-adjusted life-year (1994 dollars) are considered worthwhile investments.To answer the question of what test strategies are effective to assess patients with recent-onset atrial fibril¬ lation requires better exclusion and inclusion criteria and better data analysis than that supplied by Krahn et al.1The letter by Schreiber calls into question the methodology and presentation of our data on the results ofthyroid screening in 726 patients with recent-onset atrial fibrillation.1 This article attempted to determine the yield of routine screening in this population with a discussion of the resultant utility and cost implications. The first concern revolves around patient selection for the Canandian Registry of Atrial Fibrillation study.1 Indeed, patients with transient atrial fibrillation surrounding cardiac surgery or other procedures warrant attention, but the purpose of this database was to establish the natural history of atrial fibrillation with current therapy in a patient population at significant risk for recurrent atrial fibrillation. These data do not necessarily generalize to all patients with atrial fibrillation, the management of which must be tailored to the individual clinical presentation. Second, Table 2 suggests that 18 patients had previous hyperthyroidism (6 taking thyroxine) and 44 patients had previous hypothyroidism (37 taking thyroxine). The text discusses those patients with abnormal serum thyrotropin levels and their preceding clinical thyroid history but does not discuss the patients with previous thy¬ roid disease and biochemical euthyroidism on enrollment.Third, the follow-up in patients with abnormal results of the thyroid screening was arranged at the dis¬ cretion of the patient's physician and as such was not routine in the sense that pat...