Most intravascular ultrasound (IVUS) studies of arterial remodeling in native coronary arteries reported a remodeling index obtained at a single time point. We analyzed serial IVUS examinations, including the vessel cross-sectional area changes (remodeling behavior), of 60 hemodynamically nonstenotic left main lesions (baseline vs 18.4 +/- 9.4 months follow-up). Lumen reduction resulted from vessel reduction (sometimes despite plaque + media decrease), plaque + media increase (with or without vessel increase), or both. The percent annual changes in lumen area correlated strongly with changes in vessel (r = 0.84), but not with changes in plaque + media area. Plaques were classified as group A lesions, reflecting positive remodeling behavior (vessel changes >0), or group B lesions, reflecting negative (or intermediate) remodeling behavior (vessel changes <==0). Both groups did not differ significantly in demographics, laboratory data, and medications. Group A lesions (n = 40) more often showed plaque + media increase than group B lesions (32 of 40 [80%] vs 9 of 20 [45%]; p = 0.02). Group A lesions had, on average, mild annual lumen increase despite mild plaque + media increase, i.e, overcompensation of remodeling for plaque + media increase (vessel increase greater than plaque + media area increase, 19 of 40 [47%]). Conversely, group B lesions (n = 20) showed a significant lumen area reduction (-2.8 +/- 2.6 mm(2)/year) as a result of a decrease in vessel area only. Thus, serial long-term reduction of lumen size may result from vessel shrinkage (sometimes despite plaque decrease), plaque increase (with or without vessel increase), or both; overall, only the remodeling behavior has a significant relation to lumen changes. More than 30% of lesions show a negative remodeling behavior, which shows no relation to patient characteristics or initial plaque burden.