We examined the effects of replacing dietary fat with carbohydrates on high-density-lipoprotein (HDL) subclasses as measured by nondenaturing polyacrylamide-gradient-gel electrophoresis. One hundred five men received a 6-wk low-fat diet (24% of total energy) and a 6-wk high-fat diet (46% of energy) in a crossover design. Absorbency of protein stain was measured within five HDL subclasses: HDL3c (7.2-7.8 nm), HDL3b (7.8-8.2 nm), HDL3a (8.2-8.8 nm), HDL2a (8.8-9.7 nm), and HDL2b (9.7-12 nm). The low-density-lipoprotein-(LDL) subclass pattern was determined by gradient-gel electrophoresis, with pattern B men defined as having an LDL-predominant peak diameter < or = 25.5 nm and an LDL distribution skewed toward larger size particles. On the high-fat diet, 18 men exhibited LDL-subclass pattern B and 87 men exhibited the alternative LDL pattern A. Twelve men had the apolipoprotein (apo) epsilon 2 allele. Replacing dietary fat with carbohydrates 1) significantly decreased HDL3a, HDL2a, and HDL2b; 2) reduced HDL2b significantly more in pattern A than in pattern B men; and 3) increased plasma HDL3b concentrations significantly more in those men with the epsilon 2 allele. Our results suggest that unfavorable HDL changes were significantly more likely to occur in men who had LDL-subclass pattern A or the apo epsilon allele than in men who had pattern B or lacked the epsilon 2 allele.