We studied the gastrointestinal side effects of three formulations of naproxen in 18 healthy male volunteers. In a Latin-square design crossover study, the subjects received 500 mg naproxen twice daily for 7 days as plain tablets, enteric-coated tablets, or enteric-coated granules in capsules. The 51Cr-EDTA absorption test was performed before and at the end of each drug period, to evaluate changes in the distal gut. The test dose was instilled distally in the duodenum to prevent lesions in the stomach from interfering with the evaluation. Upper endoscopy was performed at the same intervals, scoring changes in the middle and distal duodenum separately from findings in the stomach and duodenal bulb. The nature and severity of adverse effects were recorded for each treatment period. Non-parametric methods were used for statistical evaluation. All drugs induced a significant increase in 51Cr-EDTA absorption, but we did not detect any difference between the three formulations. All formulations were associated with a significant increase in all the endoscopic findings monitored. Enteric-coated tablets induced significantly less lesions than enteric-coated granules in the stomach and duodenal bulb, and an advantage over plain tablets was indicated. No difference was seen in the middle and distal duodenum. The proximal endoscopic scores were not correlated to those found in the middle and distal duodenum. Evaluation of the small and large bowel should probably be included in clinical studies of NSAIDs, but our findings suggest that the importance of transfer of mucosal lesions to the distal gut by enteric coating may have been overemphasized.