The reproductive success of bovine sperm after sex-sorting: a meta-analysis

Systematic search, selection, and data extraction

The electronic search of Scopus returned 3,136 hits, with an additional 10 identified through the screening of reference lists. Following the removal of duplicates, 860 studies were assessed for eligibility under the inclusion criteria. After reading the titles and abstracts, 75 studies were found to directly compare conventional and sexed sperm in at least one of the fertility measures of interest. Following the inclusion criteria, 72 trials across 45 studies were eligible for inclusion in the meta-analysis (Fig. 1). Table 1 provides the descriptive data of all studies for the meta-analysis highlighting the characteristics and reproductive outcome for each trial.

Non-return rates (NRR 24/60)

The NRR 24/60 was investigated in 6 trials in 3 publications (Fig. 2). The NNR 24/60 was significantly reduced from 70.7% (CI 95: 66.1–75.3 54.7–68.7) to 61.7% (CI 95: 54.7–68.7) (p < 0.001, Cochran–Mantel–Haenszel test) when using sex-sorted sperm for insemination compared to conventional sperm. The rate ratio was 0.87 (CI 95:0.81–0.94) indicating a 13% reduction in the occurrence of a successful early pregnancy 24 – 60 days after AI with sex-sorted sperm as compared to 24–60 days after AI with conventional sperm (p < 0.001, Cochran–Mantel–Haenszel test, Fig. 2). There was a statistically significant heterogeneity between trials (I²: 98%, tau²: 0.01, p < 0.001, chi-square test).

Pregnancy rates

Pregnancy rates were investigated in 67 different trials. The overall pregnancy rates in all cows were significantly reduced from 56.1% to 43.9% when using sex-sorted sperm (p < 0.001, Cochran–Mantel–Haenszel test, Table 2). The rate ratio was 0.77 pointing to a reduction of 23% in pregnancy rates after sex-sorting (CI 95: 0.75–0.8, p < 0.001, Cochran–Mantel–Haenszel test, Table 2, Fig. 3). The heterogeneity was statistically significant (I² = 93%, tau² = 0.01, p < 0.001, chi-square test) confirming the use of a random-effects model and pointing to the necessity of subgroup analyses.

When comparing dairy and beef cows, pregnancy rates were significantly higher (6.7–8.5 percentage points) in beef cows compared to dairy cows irrespective of using conventional or sex-sorted sperm (p = 0.024 and p = 0.009, respectively, t-test, Table 2). Overall, the negative effect of sex-sorting on pregnancy rates was not significantly different in dairy and beef cows (p = 0.450, Cochran–Mantel–Haenszel test, Table 2).

When comparing heifers and cows, both in heifers and cows pregnancy rates were significantly reduced (p < 0.001 and p < 0.001, respectively, Cochran–Mantel–Haenszel test) when using sex-sorted sperm. In heifers, the pregnancy rate was reduced from 60.2% (CI 95: 58.4–62.6) to 48.4% (CI 95: 45.0–51.8, Table 2). The RR was 0.79 (CI 95: 0.75–0.82) pointing to a reduction in pregnancy rates of 21% in heifers. In cows, pregnancy rates were reduced from 46.2% (CI 95: 39.3–53.2) to 34.4% (CI 95: 28.8–39.9), The RR was 0.75 (CI 95: 0.72–0.80) indicating a decrease of pregnancy rates of 25% in cows. The heterogeneity was higher in heifers than in cows (I²: 93%, p < 0.001 in heifers, I²: 81%, p < 0.001 in cows, tau²: 0.01, chi-square test, Table 2). The subgroup comparison between heifers and cows revealed that the negative impact of sperm sexing on pregnancy rate was similar in cows and heifers (p = 0.240, Cochran–Mantel–Haenszel test, Table 2).

The reproductive success of frozen and fresh semen was analysed in 49 and 7 different trials, respectively. Pregnancy rates after sex-sorting were significantly decreased after freezing and thawing (p = 0.008, Cochran–Mantel–Haenszel test). The rate ratio was 0.77 (CI 95: 0.75–0.80) for frozen semen and 0.84 (CI 95: 0.80–0.88) for fresh semen indicating a decrease of 7 percentage points in pregnancy rate ratios when using frozen semen (Table 2). The heterogeneity was high in frozen sperm (I²: 77%, tau²: 0.01, p < 0.001, chi-square test, Table 2) and low in fresh sperm (I²: 12%, tau²: 0.01, p = 0.340, chi-square test, Table 2).

The effects of an increased sperm dosage (mostly 4 million sperm/straw) were investigated in 16 trials whereas 44 trials used a sperm concentration of 2.5 million or less. Overall, the increase of sperm dosage resulted in significantly higher pregnancy rates both in conventional sperm (p = 0.004, t-test) and in sex-sorted sperm (p < 0.001, t-test, Table 2). The increase in pregnancy rates was 14.1 percentage points (PR: 41.4% and 55.5% for the increased concentration) in sex-sorted sperm and 8.3 percentage points in conventional sperm (PR 55.2% and 63.5%, for the increased concentration). For 2.5 million sperm per straw the RR was 0.76 (CI 95: 0.73–0.79) and for the increased concentration the RR was 0.83 (CI 95: 0.78–0.88) indicating a disproportionately higher degree of improvement of pregnancy rates by 7 percentage points (p = 0.010, Cochran–Mantel–Haenszel test) in sex-sorted sperm compared to conventional sperm when increasing sperm dosage. The use of increased sperm concentration resulted in a decrease of heterogeneity (I²: 79%, tau²: 0.01, p < 0.001 in 2.5 mill/straw and I²: 52%, tau²: 0.01, p = 0.009 in > 2.5 mill/straw sperm concentration, chi-square test, Table 2).

Regarding the method of sex-sorting the Ultra sexing technology proved to achieve a significantly higher pregnancy rate compared to the conventional sexing method (p = 0.047, Mann–Whitney U test). The conventional sexing technology was used in 54 trials, the Ultra sexing method was applied in 13 trials. The RR was 0.82 (CI 95: 0.79–0.86) in the Ultra sexing technology compared to 0.76 (CI 95: 0.73–0.78) in the conventional method indicating a 6 percentage points increase in pregnancy rate ratio when using this technology (p = 0.002, Cochran–Mantel–Haenszel test). Further to that the heterogeneity was higher in the conventional sexing technology (I²: 93%, tau²: 0.01, p < 0.001, chi-square test, Table 2) compared to the Ultra sexing technology (I²: 55%, tau²: 0.01, p = 0.009, chi-square test, Table 2). In line with this, the sex-sorted sperm used in the trials published in the years 2016–2020 achieved significantly higher pregnancy rates compared to the trials published between 1999 and 2015 (p = 0.034, Mann–Whitney U test, Table 2). Further to that, the heterogeneity was significantly higher (p = 0.003, Levene’s test) in the trials published before 2015 (I²: 94%, tau²: 0.01, p < 0.001, chi-square test, Table 2) (in 2015 no respective studies were published) compared to those published after 2015 (I²: 59%, tau²: 0.01, p < 0.001, chi-square test, Table 2). When analysing the combined effects of the use of sex-sorted sperm after freezing in heifers and after 2015 (8 trials), the overall reduction of pregnancy rates was 17.7% (CI 95: 12.3–22.8). The combination of frozen sperm sorted by the SexedUltra™ Technology in heifers (2 trials) resulted in a reduction of pregnancy rates of 13.5% as compared to AI with conventional sperm (CI 95: 7.2–19.9).

When correlating the pregnancy rates with the publication year there was no correlation in the years 1999–2020 (Spearman rho = 0.078, p = 0.532, number of trials: 67). However, there was a correlation in the years 1999–2015 (Spearman rho = -0.305, p = 0.042, number of trials: 45). In the years 2016–2020 there was no correlation between pregnancy rate and publication year (Spearman rho = -0.085, p = 0.707, number of trials: 22). As shown in Fig. 4a, the pregnancy rate ratios show a high variation before 2015 whereas the data after 2016 reveal much less heterogeneity and an increased number of studies near the regression line (Fig. 4a).

In a last step the impact of herd management on the reproductive success was investigated. When comparing pregnancy rates after insemination during natural estrus (22 trials) or after synchronization (37 trials) pregnancy rates were not significantly different irrespective of the use of conventional or sex-sorted sperm (p = 0.551 and 0.919, respectively, Mann–Whitney U test, Table 2). Thus, the negative impact of sex-sorting on pregnancy rates was similar in inseminations during estrus and after synchronisation (p = 0.160, Cochran–Mantel–Haenszel test). The RR was 0.8 (CI 95: 0.77–0.83) after insemination during estrus and 0.76 (CI 95: 0.73–0.80) after synchronisation. The heterogeneity was substantial both in estrus and after synchronization (I²: 74% and 66%, respectively, tau²: 0.01, p < 0.001, chi-square test, Table 2). Regarding early and late detection of pregnancy, the late detection was significantly more reliable to detect pregnancy both after the use of conventional and sex-sorted sperm (p = 0.031 and p = 0.010, t-test, respectively, Table 2). The impact of sex-sorting on pregnancy rates was not affected by the timepoint of pregnancy detection (p = 0.130, Cochran–Mantel–Haenszel test, Table 2). When comparing rectal palpation and sonography as method for the diagnosis of pregnancy sonography was more reliable to detect pregnancy rates both after insemination with conventional and sex-sorted sperm (p = 0.050 and p = 0.319, Mann–Whitney U test, respectively). The impact of sperm sexing was not associated with the method of pregnancy detection (p = 0.74, Table 2). The heterogeneity was the same in both methods (I²: 76%, tau²: 0.01, p < 0.001, chi-square test, Table 2).

Further to that the effects of the geographical location of the trials on the pregnancy rate were analysed. The 67 studies were performed in 20 countries of 6 regions (Africa: 1; RR 0.84; Asia: 7, RR 0.73 (CI 95: 0.64–0.84); Australia/New Zealand: 1, RR 0.76; Europe:18, RR 0.83 (CI 95: 0.80–0.86); North America: 31, RR 0.74 (CI 95: 0.71–0.78); South America: 9, RR 0.77 (CI 95: 0.71–0.83). There were significant differences related to the geographic location (Cochran–Mantel–Haenszel test., p = 0.01). When applying pairwise comparisons with Bonferroni-Holm correction for multiple testing, Europe had a significantly higher pregnancy rates ratio compared to North America (Europe vs North America: p = 0.003, Europe vs South America: p = 0.400, Europe vs Australia p = 0.480, Europe vs Asia: p = 0.400, Europe vs Africa: p = 0.900). In Europe the pregnancy rate ratio was increased by 9 percentage points compared to North America.

In order to analyse publication bias, funnel plots were calculated for the overall time period of the meta-analysis (1999–2020) as well as for the time periods 1999–2015 and 2016–2020. The funnel plot of the studies published between 1999 and 2020 revealed asymmetry with outliers located right beside the lines marking the 95% confidence limits (Fig. 4b). In the funnel plot of the studies between 1999 and 2015 (Fig. 4c) as well as in the funnel plot of the publications between 2016 and 2020 there was no asymmetry, and most values were within the 95% confidence limits (Fig. 4d). Accordingly, the Begg’s test for analysis of publication bias in the 67 trials between 1999–2020 resulted in p = 0.007 (< 0.05 is an indication for publication bias). When analysing the trials between 1999 and 2015, p was 0.969 (number of trials: 45). In the time period between 2016 and 2020, p was 0.714 (22 trials).

Calving rates

Calving rates were compared in 19 trials. The overall calving rates in all cows were significantly reduced from 54.6% to 41.3% when using sex-sorted sperm (p < 0.001, Cochran–Mantel–Haenszel test, Table 3). The rate ratio was 0.76 indicating a significant decrease of 24% in calving rates after sex-sorting (CI 95: 0.69–0.83, p < 0.001, Cochran–Mantel–Haenszel test, Table 3, Fig. 5). The heterogeneity was statistically significant (I² = 89%, tau² = 0.03, p < 0.001, chi-square test) indicating the necessity of subgroup analyses (Table 3).

The comparison of the reproductive success in dairy (14 trials) and beef cows (5 trials) showed that the calving rates were higher (6.0–6.4 percentage points) in beef cows compared to dairy cows irrespective of using conventional or sex-sorted sperm. In view of the small and unbalanced number of trials this effect was not significant (p = 0.388 and p = 0.301, respectively, t-test, Table 3). Overall, the negative effect of sex-sorting on calving rates were similar in dairy and beef cows (p = 0.970, Cochran–Mantel–Haenszel test Table 3).

Regarding heifers and cows, calving rate ratios were not significantly different when using sex-sorted sperm (p = 0.940, Cochran–Mantel–Haenszel test, Table 3). In heifers (15 trials), the calving rate was significantly decreased from 56.3% (CI 95: 50.1–62.5) to 42.5% (CI 95: 37.1–47.8, p < 0.001, Cochran–Mantel–Haenszel test, Table 3) after sex-sorting. The RR was 0.75 (CI 95: 0.70–0.81, p < 0.001) pointing to a significant reduction in calving rates of 25% in heifers (p < 0.001, Cochran–Mantel–Haenszel test). In cows, calving rates were reduced from 48.4% (CI 95: 17.9–79.0) to 36.9% (CI 95: 8.0–65.7). The high range of the CI 95 and the lack of significance was due to the small number of 4 trials and to the high heterogeneity of the results of these trials. The RR was 0.74 (CI 95: 0.52–1.06) indicating a decrease of calving rates of 26% in cows. The heterogeneity in cows was considerably higher than in heifers (I²: 57%, p = 0.004 in heifers, I²: 81%, p = 0.001 in cows, tau²: 0.01, chi-square test, Table 3).

The effects of frozen and fresh semen on the calving rates were analysed in 13 and 2 trials, respectively. Calving rates after sex-sorting were significantly decreased (p < 0.001) after freezing and thawing. The rate ratio was 0.77 (CI 95: 0.68–0.86) for frozen semen and 0.95 (CI 95: 0.87–1.03) for fresh semen pointing to a reduction of 18 percentage points in calving caused by the freezing of the sex-sorted sperm (p < 0.001, Cochran–Mantel–Haenszel test Table 3). The heterogeneity was high in frozen sperm (I²: 63%, tau²: 0.03, p < 0.001, chi-square test, Table 3) whereas the heterogeneity was low in fresh sperm (I²: 3%, tau² < 0.01, p = 0.31, chi-square test, Table 3).

The effect of a sperm dosage with more than 2.5 million sperm per straw (4 million) was only investigated in 2 trials. The effects of 2.5 million sperm or less on the calving rate were investigated in 16 trials. Overall, there was no significant impact of the sperm dose on the calving rate ratio (p = 0.450, Cochran–Mantel–Haenszel test, Table 3). For 2.5 million sperm (or less) per straw the RR was 0.76 (CI 95: 0.68–0.85) and for the increased concentration > 2.5 million sperm the RR was 0.69 (CI 95: 0.57–0.85, p < 0.001 for ≤ 2.5 million, p = 0.001 for > 2.5 million sperm per straw) (Table 3).

When comparing the effects of sperm sexing on calving rates in the trials published between 1999–2015 and in the trials published between 2016–2020 calving rates increased from 39.9% (CI 95: 33.4–46.4) to 46.5% (CI 95: 29.6–63.5) in sex-sorted sperm and from 53.7% (CI 95: 45.7–61.6) to 58.3% (CI 95: 49.7–66.9) in conventional sperm (Table 3). The RR increased from 0.74 to 0.77 (p = 0.69, Cochran–Mantel–Haenszel test, Table 3) pointing to an improvement of calving rates after sex-sorting of 3 percentage points after 2015 (Table 3). Further to that, the heterogeneity was higher in the trials published before 2015 (I²: 82%, tau²: 0.04, p < 0.001, chi-square test Table 3) compared to those published 2015 or later (I²: 71%, tau²: 0.01, p = 0.020, chi-square test, Table 3).

When correlating the calving rates with the publication year over the whole period (1999–2020) no significant relationship was found (Spearman rho =  − 0.116, p = 0.637, number of trials: 19).

In a last step the impact of herd management on calving rates was investigated. When comparing calving rates after insemination during natural estrus (10 trials) and after synchronization (9 trials) calving rates were not significantly different irrespective of the use of conventional or sex-sorted sperm (p = 0.183 and 0.261, respectively, t-test, Table 3). The negative impact of sex-sorting on calving rates was similar after inseminations during estrus or after synchronisation (p = 0.990, Cochran–Mantel–Haenszel test). Consequently, the RR was the same (0.75) after insemination during estrus and after synchronisation (CI: 0.66–0.86 and 0.69–0.84, respectively, p < 0.001, Cochran–Mantel–Haenszel test). The heterogeneity was reduced after synchronization (I²: 93% for insemination during estrus and 38%, for insemination after synchronisation, tau²: 0.03 and 0,01, respectively, p < 0.001, Table 3). Publication bias was not detected in the 19 trials analysed in the meta-analysis (Begg’s test, p = 0.087).

A summary of the impact of sperm sexing on reproductive success is provided in Fig. 6. The comparison of NRR 24/60, pregnancy rates and calving rates after insemination with sex-sorted and conventional sperm showed that all these rates were significantly reduced after sex-sorting (Fig. 6). Overall, sex-sorting of sperm resulted in a 13% (9 percentage points) decrease of the NRR, a 23% (12.3 percentage points) decrease of pregnancy rate and a 24% (13.3 percentage points) decrease of calving rate (Fig. 6).

Abortion rates

Abortion rates after insemination with conventional and sex-sorted sperm were compared in 18 trials. The abortion rates were similar when using sex-sorted and conventional sperm for insemination (8.8% (CI 95: 5.0–12.7) and 9.3% (CI 95: 4.4–14.2), respectively, p = 0.62, Cochran–Mantel–Haenszel test, Fig. 7). The odds ratio was 1.08 (CI 95: 0.8–1.45) indicating that the likelihood of the occurrence of an abortion was similar when using sex-sorted and conventional sperm. The heterogeneity was moderate (I²: 49%, tau²: 0.13, p = 0.020, chi-square test, Fig. 7). When correlating the abortion odds ratio with the publication year there was no correlation in the years 1999–2020 (Spearman rho =  − 0.010, p = 0.970). Publication bias was not present in the 18 trials analysed in the meta-analysis (Begg’s test, p = 0.787).

Stillbirth rates

Stillbirth rates were investigated in 12 trials with 10 trials discriminating between stillbirths of male and female calves. Stillbirth rates were 6.9% (CI 95: 4.7–9.1) when using sex-sorted sperm and 6.8% (CI 95: 4.2–9.5) when using conventional sperm for insemination. The odds ratio was 1.00 (CI 95: 0.82–1.20, p = 0.960, Cochran–Mantel–Haenszel test) indicating that the likelihood of the occurrence of a stillbirth was the same irrespective of using sex-sorted or conventional sperm. The heterogeneity was substantial (I²: 71%, tau²: 0.05, p < 0.001, Fig. 8a). When correlating the abortion odds ratio with the publication year there was no correlation in the years 1999–2020 (Spearman rho =  − 0.163, p = 0.612). Regarding the overall stillbirth rate publication bias was not present (Begg’s test, p = 0.784).

When discriminating between the stillbirth rates of male and female calves, the stillbirth rate of male calves was significantly increased from 10.2% (CI 95: 6.0–14.3) to 16.5% (CI 95: 7.0–26, p = 0.003, Cochran–Mantel–Haenszel test) when using sex-sorted sperm for insemination. The odds ratio for stillbirth in male calves was 1.46 (CI 95: 1.14–1,86, p = 0.003, Cochran–Mantel–Haenszel test) indicating that it is significantly more likely to experience a stillbirth in male calves after insemination with X bearing sex-sorted sperm (Fig. 8b). The stillbirth rates of female calves were similar (sex-sorted: 6.6% (CI 95: 3.6–9.5, conventional: 5.9% (2.4–9.3). The odds ratio was 1.03 (CI 95: 0.84–1.27, p = 0.77). Irrespective of the use of sex-sorted or conventional sperm the rate of stillbirths in male calves was significantly higher than in female calves (sex-sorted: p = 0.029, conventional: p = 0.043, t-test). The heterogeneity of stillbirth in female calves was higher than in male calves (female: I²: 87%, tau²: 0.04, p < 0.001, male: I²: 65%, tau²: 0.05, p = 0.002, chi-square test, Fig. 8b).