Study question Does the endometrial preparation protocol (artificial cycle (AC) vs. natural cycle (NC) vs. stimulated cycle (SC)) impact the risk of early pregnancy loss and live birth rate after frozen/thawed embryo transfer (FET)? Summary answer In FET, artificial cycles were significantly associated with a higher pregnancy loss rate and a lower live birth rate compared with stimulated cycle or natural cycle. What is known already To date, there is no consensus on the optimal endometrial preparation in terms of outcomes. Although some studies have reported a higher pregnancy loss rate using AC compared with NC or SC, no significant difference was found concerning the pregnancy rate or live birth rate. Furthermore, no study has compared the three protocols in a large population. Study design, size, and duration A multicenter retrospective cohort study was conducted in nine reproductive health units in France using the same software to record medical files between January 1st, 2012 and December 31st, 2016. FET using endometrial preparation by AC, modified NC or SC were included. The primary outcome was the pregnancy loss rate at 10 weeks of gestation. The sample size required was calculated to detect an increase of 5% in the pregnancy loss rate (21 to 26%), with an alpha risk of 0.5 and a power of 0.8. We calculated that 1126 pregnancies were needed in each group, i.e., 3378 in total. Participants/materials, setting, methods Data were collected by automatic extraction using the same protocol. All consecutive autologous FET cycles were included: 14,421 cycles (AC: n = 8,139; NC: n = 3,126; SC: n = 3,156) corresponding to 3,844 pregnancies (hCG > 100 IU/L) (AC: n = 2,214; NC: n = 812; SC: n = 818). Each center completed an online questionnaire describing its routine practice for FET, particularly the reason for choosing one protocol over another. Main results and the role of chance AC represented 56.5% of FET cycles. Mean age of women was 33.5 (SD ± 4.3) years. The mean number of embryos transferred was 1.5 (±0.5). Groups were comparable, except for history of ovulation disorders (P = 0.01) and prior delivery (P = 0.03), which were significantly higher with AC. Overall, the early pregnancy loss rate was 31.5% (AC: 36.5%; NC: 25.6%; SC: 23.6%). Univariable analysis showed a significant association between early pregnancy loss rate and age > 38 years, history of early pregnancy loss, ovulation disorders, and duration of cryopreservation > 6 months. After adjustment (multivariable regression), the early pregnancy loss rate remained significantly higher in AC vs. NC (odds ratio (OR) 1.63 (95% CI) [1.35-1.97]; P < 0.0001) and in AC vs. SC (OR 1.87 [1.55-2.26]; P < 0.0001). The biochemical pregnancy rate (hCG > 10 and lower than 100 IU/L) was comparable between the three protocols: 10.7% per transfer. Limitations, reasons for caution This study is limited by its retrospective design that generates missing data. Routine practice within centers was heterogeneous. However, luteal phase support and timing of embryo transfer were similar in AC. Univariable analysis showed no difference between centers. Moreover, a large number of parameters were included in the analysis. Wider implications of the findings Our study shows a significant increase in early pregnancy loss when using AC for endometrial preparation before FET. These results suggest either a larger use of NC or SC, or an improvement of AC by individualizing hormone replacement therapy for patients in order to avoid an excess of pregnancy losses. Study funding/competing interests The authors declare no conflicts of interest in relation to this work. G. P.-B. declares consulting fees from Ferring, Gedeon-Richter, Merck KGaA, Theramex, Teva; Speaker’s fees or equivalent from Merck KGaA, Ferring, Gedeon-Richter, Theramex, Teva N.C. declares consulting fees from Ferring, Merck KGaA, Theramex, Teva; Speaker’s fees or equivalent from Merck KGaA, Ferring C.R. declares a research grant from Ferring, Gedeon-Richter; consulting fees from Gedeon-Richter, Merck KGaA; Speaker’s fees or equivalent from Merck KGaA, Ferring, Gedeon-Richter E.M.A. declares Speaker’s fees or equivalent from Merck KGaA, MSD, Ferring, Gedeon-Richter, Theramex, Teva I.C-D. declares Speaker’s fees or equivalent from Merck KGaA, MSD, Ferring, Gedeon-Richter, IBSA N.M. declares a research grant from Merck KGaA, MSD, IBSA; consulting fees from MSD, Ferring, Gedeon-Richter, Merck KGaA; Speaker’s fees or equivalent from Merck KGaA, MSD, Ferring, Gedeon-Richter, Teva, Goodlife, General Electrics Trial registration number Not applicable