AbstractObjectivesTo investigate the variation of endometrial responsiveness between cycles within the same women undergoing assisted reproduction.MethodsThe sonographic endometrial thickness in ovarian stimulation cycles was compared with that of subsequent natural cycles. One hundred and thirty‐six ovarian stimulation cycles of in‐vitro fertilization and embryo transfer were evaluated. Women who did not conceive in in‐vitro fertilization cycles were subsequently seen in natural cycles (n = 97) or the next in‐vitro fertilization cycle (n = 39). Based on a receiver–operating characteristics (ROC) curve using endometrial thickness to predict pregnancy, the first in‐vitro fertilization cycles were classified according to the endometrial thickness as optimal (>8 mm) in 98 cycles, or suboptimal (≤8 mm) in 29 cycles. Similarly, spontaneous cycles were classified as suboptimal (≤7 mm) in 28 cycles and optimal (>7 mm) in 69 cycles.ResultsThe pregnancy rates were significantly lower (P < 0.05; Fisher's Exact test) in the suboptimal group in both the in‐vitro fertilization and frozen embryo transfer cycles. There was a strong correlation (r 2 = 0.745) and a significant difference (P < 0.001; Wilcoxon signed rank sum test) between the endometrial thickness of stimulation and natural cycles.ConclusionIt is possible to predict the occurrence of optimal or suboptimal endometrial response in natural cycles of women, after evaluation in stimulated cycles, with a high degree of reliability. Risk of implantation failure can be identified before subsequent treatment cycles and adjuvant therapeutic strategies may be planned to improve the endometrial response before embryo transfer. Copyright © 2002 ISUOG