Slight decreases in cumulative live birth rates when using frozen-thawed semen samples for ICSI with donor oocytes do not mean we should stop freezing

Abstract Study question Does the use of autologous frozen semen samples for ICSI with donor oocytes make any difference in reproductive outcomes compared to using fresh samples? Summary answer Although it seems slightly detrimental in several intermediate outcomes, the advantages of using frozen samples in cycle logistics outweigh this potential decrease in results. What is known already Sperm freezing has become a part of day-to-day practice in infertility clinics due to its many logistic benefits. However, there is no clear consensus on the impact of freezing-thawing semen samples prior to ICSI on cycle outcomes. Moreover, most studies express results according to classical outcomes of reproductive success (i.e. pregnancy rate or live birth rate (LBR) per embryo transfer (ET)). Cumulative LBR (CLBR) are a more realistic approach to expressing the probability of achieving a livebirth when assessing the effect of a technique, since they compute the contribution of all used oocytes and embryos from the same cohort. Study design, size, duration Retrospective multicenter observational cohort study involving 45,120 ICSI procedures with donor oocytes (34,847 with fresh semen samples and 10,269 with frozen samples) from January 2008 to November 2021. The fresh sample group included real-world data from the clinical records of these patients and cycles from a cohort of 387,903 inseminated oocytes and 75,116 embryos transferred in 49,775 ET; whereas the frozen sample group considered 110,983 injected oocytes and 24,334 embryos replaced in 14,953 ET. Participants/materials, setting, methods Pregnancy rates and LBR per ET were computed and then adjusted to female and male patients’ age and BMI, total progressive motile sperm, semen capacitation, transfer before or at blastocyst stage, and use of PGT, using generalized linear models to compute adjusted odds ratios (adjOR) and p-values comparing both groups. CLBR per ET, embryo replaced and used oocyte were plotted in Kaplan-Meier curves adjusted according to donors’ and female patients’ age in Cox regression models. Main results and the role of chance There were statistically significant differences between both groups in terms of biochemical, clinical and ongoing pregnancy rates per ET, and LBR per ET. After the multivariate analysis, these differences remained statistically significant with adjusted p-value <0.001. The adjOR were 1.091 (1.045-1.139), 1.133 (1.086-1.183), 1.115 (1.067-1.165) and 1.168 (1.114-1.223) respectively when considering the frozen sample group as the reference. CLBR was 57.73% (57.05-58.41) after three embryos were transferred and 73.31% (72.44-74.16) after five embryos replaced in the fresh sample group, compared to the 49.65% (48.40-50.88) and 66.58% (64.92-68.16) CLBR in the frozen sample group (p < 0.001). The adjusted hazard ratio for this comparison was 1.146 (1.100-1.195). The CLBR in the fresh sample group was 45.74% (45.10-46.37) and 69.25% (68.46-70.02) after 10 and 14 donor oocytes were used, while the rates for the frozen sample group were 43.87% (42.66-45.05) and 68.83% (67.17-70.40) respectively (p = 0.0002). The adjusted hazard ratio was 1.067 (1.023-1.112), with an adjusted p-value of 0.002. In the fresh sample group, 5.6% of the used oocytes resulted in the couples’ first newborn (267,462 oocytes and 15,153 newborns), whereas in the frozen sample group it was the 5.2% of oocytes (77,284 oocytes and 4,040 newborns), which is consistent with the published literature. Limitations, reasons for caution The retrospective nature of the study subjects data to biases and inaccuracies in their annotation in clinical sources from which they were exported, although the multivariate statistical analyses aimed to control those biases. Considering only donor oocytes also reduces some of the variability since it homogenizes factors concerning oocyte quality. Wider implications of the findings Although the frozen samples group had slightly decreased success rates, the CLBR per used oocyte remained between around two percent lower compared to using fresh samples. Thus, the large sample size and real-world data included show that the subtly lower rates do not outweigh the logistic benefits of sample freezing. Trial registration number Not applicable