Efficacy of intracytoplasmic sperm injection in women with non-male factor infertility: A systematic review and meta-analysis.

LY and FL substantially contributed to conception and design. RZ and QW contributed to data acquisition, analysis, and interpretation. LY and FL drafted the manuscript for important content. LY critically revised the manuscript for important intellectual content. All authors gave their final approval of the manuscript for publication.

Database searches initially identified 1838 records; the other 10 were from other sources. After removing duplicates, we assessed 1796 records by screening the titles and abstracts. Among these, 1725 records were excluded. After reading the full texts of the remaining 71 studies, 28 were excluded for male‐factor infertility, 14 for non‐ICSI, four for non‐IVF, two for the missing original data, and the others were excluded for no related outcomes reporting. Finally, 18 randomized studies were included in the quantitative synthesis. 13 , 14 , 15 , 16 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 The flow chart of study selection is summarized in Figure  1 . Preferred reporting items for systematic reviews and meta‐analyses (PRISMA) flow chart of study selection. Detailed information on the included studies is outlined in Table  1 . A total of 18 RCTs, with 3249 cycles and 30 994 sibling oocytes, were included in this review. Two studies 13 , 36 were conducted in multiple centers, whereas the others were single‐center RCTs. The average age of females ranged from 29 to 41. Among the 18 RCTs, 11 14 , 15 , 16 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 used sibling oocytes as the unit of randomization, and the others 13 , 35 , 36 , 37 , 38 , 39 , 40 randomly allocated patients to ICSI or IVF groups. The etiology of non‐male factor infertility varied among trials; four enrolled patients with tubal factor, 33 , 35 , 37 , 40 four with unexplained cause, 27 , 28 , 30 , 38 one with polycystic ovary syndrome, 29 one with low ovarian response, 39 six with multiple causes, 13 , 16 , 31 , 32 , 34 , 36 and the others not reported. 14 , 15 For controlled ovarian stimulation protocols, 14 studies used GnRH agonist long, 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 and four applied GnRH antagonist protocols. 13 , 14 , 15 , 16 Within the 18 studies, 12 13 , 14 , 15 , 16 , 27 , 28 , 32 , 34 , 35 , 36 , 38 , 39 followed WHO's sperm evaluation criteria, while the other six 29 , 30 , 31 , 33 , 37 , 40 combined it with Kruger's strict criteria. Four studies 14 , 30 , 34 , 36 used discontinuous gradient and eight 27 , 28 , 29 , 31 , 32 , 35 , 37 , 40 used swim‐up for sperm preparation. Two studies 13 , 33 used either swim‐up or discontinuous gradient. The remaining four 15 , 16 , 38 , 39 studies did not mention any specific techniques of sperm preparation. Only six 13 , 14 , 27 , 31 , 33 , 36 studies gave detailed results of insemination concentrations for IVF, and the data was similar. The evaluation methods of sperm samples, sperm preparation techniques, and insemination concentrations for IVF were basically consistent (Table  S3 ). Characteristics of included studies. IVF:138 ICSI:160 IVF: 58 ICSI: 58 IVF: 736 ICSI: 748 IVF: 35.6 ± 5.3 ICSI: 35.5 ± 4.8 1994.07 −1994.09 IVF: 136 ICSI: 144 Unexplained (90), endometriosis (10) IVF:551 ICSI:589 1996.01 −1997.04 IVF: 52 ICSI: 52 IVF: 234 ICSI: 229 IVF: 36.7 ± 0.6 ICSI: 35.3 ± 0.6 1996.09 −1998.08 IVF:334 ICSI:328 IVF: 33.1 ± 4.4 ICSI: 34.3 ± 5.2 1994 −1997 IVF:536 ICSI:1072 Previous ovarian surgery ICSI/IVF attempts IVF: 38 ICSI: 38 IVF: 308 ICSI: 345 IVF: 34 ± 3.9 ICSI: 34 ± 3.8 Previous FR12 IU/L; ≥3 IVF cycles IVF: 219 ICSI: 204 IVF: 2065 ICSI: 2153 IVF: 30.9 ± 4.1 ICSI: 31.6 ± 3.2 IVF: 45 ICSI: 44 IVF: 350 ICSI: 291 IVF: 33 ICSI: 32.7 Tubal (56), Unexplained (38), endometriosis (6) IVF: 187 ICSI: 188 1997.07 −2002.04 IVF:754 ICSI:907 1998.01 −2000.01 Female aged>38 years; baseline FSH>12 mIU/ml; 42 years; history of TFF; <6 COCs retrieved; PGT cycles IVF: 64 ICSI: 74 IVF:1306 ICSI:1331 2018.11 −2019.04 IVF:283 ICSI:285 2018.03 −2019.08 Tubal (24), Unexplained (35.5), diminished ovarian reserve (25), ovulation disorder (12), endometriosis (3.5) IVM cycles; frozen semen; previous FR ≤25%; >2 cycles of IVF/ICSI IVF: 532 ICSI: 532 IVF: 5852 ICSI: 5852 IVF: 32.6 ± 4.7 ICSI: 32.7 ± 4.6 2018.03 −2019.12 IVF:258 ICSI:257 Abbreviations: BMI, body mass index; COC, cumulus‐oocyte complexes; COS, controlled ovarian stimulation; CPR, clinical pregnancy rate; FR, fertilization rate; FSH, follicle‐stimulating hormone; GnRH, gonadotropin‐releasing hormone; GnRH‐a, GnRH agonist; GnRH‐A, GnRH antagonist; ICSI, intracytoplasmic sperm injection; IR, implantation rate; IVF, in vitro fertilization; IVM, in vitro maturation; LBR, live birth rate; MR, miscarriage rate; NA, not available; PGT, preimplantation genetic testing; RCT, randomized controlled trial; TFF, total fertilization failure; UAE, United Arab Emirates; UK, United Kingdom. The risk of bias is shown in the risk of bias graph and risk of bias summary (Figure  S1 ). We rated nine trials 13 , 14 , 15 , 27 , 29 , 31 , 36 , 37 , 39 as low risk of random sequence generation and four at low risk of allocation concealment as they reported explicit random method and allocation concealment. Four studies 13 , 16 , 29 , 37 reported explicit information about the blinding of outcome assessors and were judged at low risk of detection bias. However, it might be impossible to blind operators when using sibling oocytes as the unit of randomization. Three studies 13 , 39 , 40 that allocated patients explicitly mentioned no blinding of performance. We finally judged 12 studies 13 , 15 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 39 , 40 at high risk of performance bias. The remaining trials were considered at unclear risk of bias due to insufficient information. All trials were free from attrition bias and other biases. We judged two studies 33 , 34 at unclear risk of selective reporting due to the missing outcome definition. Three trials 13 , 37 , 38 reported LBR, and no significant difference was found between ICSI and IVF (RR = 1.11, 95% CI: 0.94–1.30, I 2  = 0%, 1190 cycles) (Figure  2 ). The proportion of cycles until given births were 34.2% (204/596) in the ICSI group vs 30.8% (183/594) in the IVF group. Forest plots of live birth rate and fertilization rate in non‐male factor fertility. Intracytoplasmic sperm injection (ICSI) vs in vitro fertilization (IVF). (A) live birth rate, (B) fertilization rate per oocyte retrieved, and (C) fertilization rate per oocyte inseminated/injected. The effects of ICSI on FR per oocyte retrieved were identified in all included studies. 13 , 14 , 15 , 16 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 Overall, there was no significant difference in FR per oocyte retrieved (RR = 1.05, 95% CI: 0.98–1.13, I 2  = 92%, 30 994 oocytes retrieved) (Figure  2 ). Across all subgroup analyses, two factors (unit of randomization and female age) differed significantly (Table  S4 ). We further revealed that the unit of randomization ( p  = 0.019) rather than female age ( p  = 0.118) was the factor that influenced the estimated effects significantly from multivariate meta‐regression analysis (Table  S5 ). Regarding the unit of randomization based on patients and sibling oocytes, the estimated effect from studies of randomization based on patients might be more reliable (RR = 0.95, 95% CI: 0.86–1.05, I 2  = 93%). In addition, we also analyzed FR per oocyte inseminated/injected (Figure  2 ). We found that FR per oocyte inseminated/injected was significantly higher in the ICSI group compared to the IVF group (RR = 1.14, 95% CI: 1.08–1.20, I 2  = 69%, 15 716 oocytes inseminated/injected). Subgroup analyses showed no differences between the two groups (Table  S4 ). A total of 16 studies, comprising 3119 cycles, evaluated the effects of ICSI on TFF. 13 , 14 , 15 , 16 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 36 , 37 , 38 , 39 , 40 The incidence of TFF was significantly lower in the ICSI group (3.1%) compared to the IVF group (9.7%) (RR = 0.26, 95% CI: 0.13–0.50, I 2  = 58%) (Figure  3 ). However, the benefit was observed only in studies that randomized sibling oocytes (RR = 0.14, 95% CI: 0.07–0.30, I 2  = 14%) and in women under 35 years old (RR = 0.20, 95% CI: 0.09–0.43, I 2  = 60%). Multivariate meta‐regression analyses identified that the unit of randomization ( p  = 0.022) was associated with the decreased TFF and the potential source of heterogeneity (Tables  S4 and S5 ). Forest plots of total fertilization failure in non‐male factor infertility. Intracytoplasmic sperm injection (ICSI) vs in vitro fertilization (IVF). Six trials 13 , 27 , 36 , 37 , 38 , 39 reported IR in both ICSI and IVF groups. The IR with ICSI was similar to that with IVF (RR = 0.99, 95% CI: 0.86–1.14, I 2  = 23%) (Figure  S2 ). No subgroup effect was found (Table  S6 ). Seven trials, 13 , 27 , 36 , 37 , 38 , 39 , 40 including 1922 embryo transfer cycles, provided information on CPR. The overall effect estimate did not favor either of the two groups (RR = 1.01, 95% CI: 0.90–1.14, I 2  = 0%) (Figure  S2 ). No subgroup effect was found (Table  S6 ). Four trials 13 , 27 , 39 , 40 with 532 embryo transfer cycles were included in this analysis. The pooled estimates showed that ICSI did not decrease the MR (RR = 1.03, 95% CI: 0.72–1.47, I 2  = 0%) compared to the conventional IVF (Figure  S2 ). In addition, no subgroup effect was found (Table  S6 ). Four trials 31 , 34 , 37 , 38 assessed outcomes concerning the good quality of embryos. There was no difference between ICSI and IVF (RR = 1.02, 95% CI: 0.86–1.20, I 2  = 66%) (Figure  S2 ). In the subgroup analysis by the unit of randomization, a trend toward a higher rate of good‐quality embryos was observed in the ICSI group in trials randomizing sibling oocytes. However, no considerable subgroup difference was detected (See Table  S6 ). Four trials 13 , 14 , 15 , 36 met the criterion for studies at low risk of bias (for which both selection and attrition bias were rated as low risk). In addition, the results from these trials were consistent with overall summary effect estimates for all outcomes (Table  S7 ). Funnel plots were used to detect publication bias for primary outcomes, except LBR, due to the limited number of studies ( n  < 10). The symmetrical appearance of funnel plots indicated no evidence of publication bias for FR and TFF (Figure  S3 ). The pooled estimates from studies randomizing patients were further assessed using the GRADE approach (Table  2 ). Overall, the certainty of primary outcomes was moderate, indicating that more research is still needed to define the effects of ICSI. The certainty of the evidence assessment. Live birth 342 per 1000 (290–400) 34 (−18 to 92) RR 1.11 (0.94–1.30) 1190 cycles (3 RCTs) ⊕ ⊕ ⊕㊀ Moderate b Fertilization per oocyte retrieved 557 per 1000 (504–615) −29 (−82 to 29) RR 0.95 (0.86–1.05) 20 161 oocytes 7 RCTs ⊕ ⊕ ⊕㊀ Moderate c Fertilization per oocyte inseminated/injected 721 per 1000 (708–741) 77 (64–97) RR 1.12 (1.10–1.15) 13 932 oocytes 4 RCTs ⊕ ⊕ ⊕⊕ High 47 per 1000 (31–69) −18 (−34 to 4) RR 0.72 (0.48–1.06) 1805 cycles 6 RCTs ⊕ ⊕ ⊕㊀ Moderate b Implantation 261 per 1000 (227–301) −3 (−37 to 37) RR 0.99 (0.86–1.14) 3649 embryos 6 RCTs ⊕ ⊕ ⊕⊕ High 366 per 1000 (326–413) 4 (−36 to 51) RR 1.01 (0.90–1.14) 1922 cycles 7 RCTs ⊕ ⊕ ⊕⊕ High Miscarriage 156 per 1000 (104–234) 9 (−43 to 87) RR 1.06 (0.71–1.59) 532 cycles 4 RCTs ⊕ ⊕ ⊕㊀ Moderate b 414 per 1000 (340–506) −46 (−120 to 46) RR 0.90 (0.74–1.10) 990 embryos 2 RCTs ⊕ ⊕ ㊀ ㊀ Low d Abbreviations: CI, confidence interval; ICSI, intracytoplasmic sperm injection; IVF, in vitro fertilization; RR, risk ratio. The risk in the intervention group (and the 95% confidence interval) is based on the risk in the control group and the relative effect of the intervention. Downgraded one step due to imprecision of results as shown in wide confidence intervals. Downgraded one step due to the presence of moderate heterogeneity. Downgraded two steps due to imprecision of results as shown in wide confidence intervals and high risk of bias.

In this systematic review, we evaluated the effects of ICSI on reproductive outcomes in couples with non‐male factor infertility. Overall, ICSI reduced total fertilization failure (TFF) and increased FR per oocyte inseminated/injected, but did not improve the LBR, FR per oocyte retrieved, or other outcomes compared to IVF. Although our subgroup analyses suggested ICSI could lead to a higher FR per oocyte retrieved and lower TFF when using sibling oocytes as the randomization unit, when addressing studies based on patients randomly allocated, no benefit was found, which might be more reliable. Further, no evidence suggested that other predefined factors, including female age, study year, and controlled ovarian stimulation protocols, could influence the overall estimates. Equally allocating sibling oocytes to different groups is a widely used study design to explore differences between insemination methods in assisted reproductive technique. In our meta‐analysis, about two‐thirds of studies applied this design, randomizing sibling oocytes rather than patients. The design allowed each patient to serve as their own control, reducing confounders such as maternal age, ovarian stimulation protocol, oocyte/sperm quality, and laboratory conditions. 9 It was also believed that it could avoid TFF since oocytes from the same patient were used for different insemination methods to obtain embryos. In this meta‐analysis, we found ICSI had a significant advantage over IVF in studies that randomly allocated sibling oocytes, including lower TFF, higher FR per oocyte retrieved, and possibly more high‐quality embryos. In contrast, studies randomizing patients found no substantial benefit in these outcomes. This difference might be related to bias from the study design randomizing sibling oocytes. The number of oocytes randomly allocated to each group was small, about 5 to 20, limited by the obtained oocytes from a single patient. Randomly assigning so few oocytes could compromise the randomization, consequently failing to equalize the difference in oocytes across groups. 41 , 42 Summarizing data from each patient in a study further accumulated the imbalanced baseline of oocytes. This is due to multiple randomizations based on small samples, which increases the risk of bias in these studies. Given the possible methodological biases, we believed studies randomizing sibling oocytes could not reflect rigorous statistical analysis results and may exaggerate effect sizes. In addition, it was unlikely to obtain reliable data on clinical outcomes from studies allocating sibling oocytes because mixed embryos from ICSI and IVF might be transferred to ensure the quality of transfer. 9 , 10 , 43 However, studies randomizing patients effectively avoided these shortcomings and objectively reflected the difference between ICSI and IVF on clinical outcomes. Seven RCTs randomizing patients were strictly included in our meta‐analysis to compare the efficacy of ICSI and IVF for non‐male factor infertility. The FR directly reflects the effects of insemination methods. In our study, ICSI showed no difference in FR per oocyte retrieved compared to IVF. However, it was much higher than IVF when considering FR per oocyte inseminated/injected. We believed that the difference in FR might be due to the following reasons. In both ICSI and IVF, a fertilization check was performed 16–18 h after insemination. However, only mature (metaphase II) oocytes were injected and potentially fertilized in the ICSI group, while metaphase I oocytes had no chance of fertilization. In the conventional IVF group, both metaphase II and some metaphase I oocytes had the chance to fertilize naturally. This was because some metaphase I oocytes would mature to metaphase II in vitro during this period, increasing the overall FR of the IVF group. 27 , 37 This might explain why the FR per oocyte retrieved was similar in both groups, but the FR per oocyte inseminated/injected was higher in the ICSI group compared to the conventional IVF group. In terms of the other primary outcome of TFF, a huge emotional and financial blow to patients and the main reason ICSI is widely used in non‐male factor infertility. Many studies have concluded that ICSI could reduce the risk of TFF 36 , 44 , 45 ; however, we only observed a trend of TFF reduction with ICSI. Specifically, for numbers needed to treat, an additional approximately 50 cycles of ICSI were required to avoid one occurrence of TFF after IVF. This ratio argues against the routine use of ICSI for TFF reduction. Compared to surrogate outcomes, such as FR and TFF, LBR is recommended as the ultimate goal of clinical treatment in reproductive research. 46 Our results showed that ICSI did not improve LBRs or other clinical outcomes compared with IVF, consistent with the recently published randomized trial that enrolled more than 1000 couples. 13 In summary, ICSI did not provide significant benefits in laboratory measures or clinical outcomes. Whether ICSI could benefit non‐male factor infertility patients with advanced age is controversial. Some studies have suggested that structural defects, such as thickened zona pellucida in oocytes, might result in fertilization failure in advanced‐age women, and ICSI could improve the outcomes. 6 However, our study showed similar reproductive outcomes between older and younger women and was consistent with a previous meta‐analysis that included retrospective studies which found ICSI did not reduce FR in women over 38 years old. 12 Also, a recent randomized trial that enrolled women over 40 showed that ICSI does not improve FR and good‐quality embryos. 15 We also explored whether the development of ICSI techniques or GnRH analogue protocols, including antagonist or agonist long protocols, affect related outcomes. However, no evidence supports that these factors can influence the results. In addition to the clinical efficacy, we should not ignore the cost and safety of ICSI in decision‐making. A recent study found that ICSI may increase the risk of singleton congenital abnormalities in non‐male factor infertility couples. 47 Compared with IVF, Kissin et al. also claimed that ICSI was associated with a higher risk of autism and congenital malformations in offspring, though the mechanism was unclear. 48 These potential risks might relate to genetic changes, immature germ cell use, links between genetic disorders, and certain infertility forms. 49 Additionally, the cost of ICSI must be considered. ICSI usually increases complexity and cost, requiring extra embryologist resources, materials, and time. 46 For example, ICSI costs at least £500 more than IVF for the assisted reproductive technique in the UK. 50 Our research has its unique strengths. To our knowledge, this is the first meta‐analysis to focus on randomized trials comparing the efficacy of ICSI and IVF on non‐male factor infertility. Compared with previous studies, which have mainly focused on FR and rarely on live births, we focused more on clinical outcomes to make our analysis comprehensive. In addition, as our study was registered in PROSPERO and performed strictly according to the Cochrane Handbook, all steps were faithfully conducted. Furthermore, we performed subgroup analyses to explore the influence of different factors on the outcomes. More importantly, this review first investigated the impact of the randomization unit on outcomes, possibly because previous reviews did not have enough studies to do so. We found that studies randomizing sibling oocytes or patients differed significantly in key outcomes, and we further explored the possible reasons for the differences. The main limitation of this study was that the congenital disabilities and neonatal growth were not analyzed and summarized. Among the included studies, only one trial reported neonates‐related outcomes, and there was no significant difference between the two groups. 13 At the same time, there was a lack of relevant data on thawing cycles, which should be explored in the future. In addition, high heterogeneity was observed in FR, which might be related to different sample sizes, study protocols, and participant characteristics. Finally, we could not analyze whether various etiologies affected the outcomes due to limited available data. To accurately and comprehensively clarify the role of ICSI in non‐male factor infertility, future research should pay more attention to the endpoint and clinical outcomes, such as LBR and neonatal outcomes, rather than surrogate outcomes, such as FR and high‐quality embryo rate. In addition, our research showed that ICSI may have a slight benefit in reducing TFF, requiring about 50 additional cycles to avoid one case of TFF. However, we need to consider the cost of ICSI and patient preferences more in this case. Therefore, future studies should pay more attention to the cost‐effectiveness of ICSI and focus more on patient‐centered outcomes. In addition, whether the efficacy of ICSI differed among various causes of non‐male factor infertility is still unknown. Therefore, it is necessary to explore the role of ICSI in a specific etiology, such as PCOS, or provide detailed data about different types of patients in a large mixed randomized trial in the future. Furthermore, we found that studies randomly allocating sibling oocytes may not offer reliable reproductive outcomes due to possible bias; future trials should adopt the study design assigning patients to ICSI or IVF while limiting the risk of bias as much as possible to provide more reliable and higher quality evidence.

Compared with IVF, ICSI leads to no difference in reproductive outcomes compared to IVF in patients with non‐male factor infertility. Considering the cost and safety of this technique, our results may not support the routine use of ICSI in patients with non‐male factor infertility. In the future, high‐quality RCTs with large sample sizes will be needed to confirm our results. More attention should be focused on the cost‐effectiveness of ICSI and its impact on clinical and neonatal outcomes.

Intracytoplasmic sperm injection (ICSI) has been routinely performed as an assisted reproductive technique in severe male infertility since its first use in the 1990s. 1 However, the percentage of patients with non‐male factor infertility using ICSI has increased worldwide over the past few decades. 2 In the United States, the use of ICSI for non‐male factor infertility rose from 15% to 67% during 1996–2012. 3 Moreover, the data showed a 1.2% annual increase in the overall use of ICSI cycles from European countries in 2016, with a more significant rise in non‐male factor infertility than in male infertility. 4 , 5 The prevailing rationale for applying ICSI to couples with non‐male factor infertility is to prevent total fertilization failure (TFF) and raise the fertilization rate (FR), thus increasing the likelihood of pregnancy. For instance, some studies suggested that structural defects, such as thickened zona pellucida in oocytes, might result in fertilization failure in advanced‐age women, and ICSI could improve the outcomes. 6 Additionally, ICSI was suggested as an effective treatment for unexplained infertility in previous research, especially in improving FR and clinical pregnancy rate (CPR). 7 , 8 Although a large amount of research has explored the efficacy of ICSI, there remains debate over whether ICSI can effectively improve reproductive outcomes in non‐male factor infertility. Additionally, evidence from existing systematic reviews suggested conflicting results of ICSI in non‐male factor infertility couples. Johnson et al. analyzed 11 studies involving 901 couples and found that ICSI increased FR and decreased the risk of TFF in couples with unexplained non‐male infertility. 9 In contrast, other reviews suggested that ICSI offers no obvious advantage regarding FR and CPR. 10 , 11 , 12 Further, the applicability of the evidence is also limited by the methodological quality of the reviews, for instance, pooling estimates from randomized and observational studies, introducing selection bias and reducing credibility, and relying on laboratory surrogate outcomes rather than patient important outcomes to draw conclusions. 9 , 10 , 11 , 12 Recently, several new randomized controlled trials (RCTs) are accessible and yielded more reliable ICSI results. 13 , 14 , 15 , 16 Therefore, it is necessary to conduct a new systematic review and meta‐analysis of RCTs to avoid the above limitations in existing reviews to provide robust evidence clarifying the efficacy of ICSI in non‐male factor infertility couples.

This systematic review was conducted according to the Cochrane Handbook for systematic reviews of interventions 17 and has been registered in PROSPERO with the number CRD42023427004 on May 16, 2023. 18 We also followed the preferred reporting items for systematic reviews and meta‐analyses (PRIMSA) checklist to report this study (Table  S1 ). 17 A comprehensive search of Medline (via PubMed), Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) was performed from inception to March 2023. Only English publications were included. We also manually checked the references of conference proceedings and identified studies or websites of the clinical trial registry to obtain additional relevant data. It was not limited by publication status or sample size. The details of the search terms in PubMed are shown in Table  S2 . Studies were included based on the following criteria: (1) RCTs; (2) participants were diagnosed with infertility with non‐male factors regardless of etiology, and the male's sperm count and motility were assessed according to the World Health Organization (WHO) criteria; (3) the intervention group used ICSI, and the control group used conventional IVF; and (4) primary outcomes included: LBR, defined as the number of live births/the number of fresh embryos transferred; FR per oocyte retrieved or per inseminated/injected, defined as the number of two pronuclei zygotes observed/the number of oocytes retrieved or the number of two pronuclei zygotes observed/the number of oocytes inseminated/injected; and TFF, defined as the number of total fertilization failure cycles/number of ovulation cycles. Secondary outcomes included implantation rate (IR), CPR, miscarriage rate (MR), and good‐quality embryo rate. Two authors (R.Z. and F.L.) first independently reviewed the titles and abstracts of the records, then screened the full texts of all studies deemed potentially relevant. Finally, any disagreements were resolved through discussion with another author (L.Y.). Two authors (L.Y. and Q.W.) independently conducted the data extraction which was cross‐checked. Any discrepancies were discussed with a third author (X.Z.). The following information was extracted: first author, country, publication year, study period, study design, the unit of randomization, etiology, exclusion criteria, methods for evaluating sperm samples, sperm preparation techniques, insemination concentrations for IVF, female age, number of patients/cycles, number of oocytes retrieved, controlled ovarian stimulation protocol, and primary outcomes as reported. Two authors (L.Y. and Q.W.) independently assessed potential methodology bias in each included trial, and differences were resolved through discussion with a third author (X.Z.). We used the Cochrane Risk‐of‐Bias tool to assess the risk of bias in RCTs. 19 The evaluation grades of each domain were “low,” “unclear,” and “high.” We used the grading of recommendations assessment, development and evaluation (GRADE) approach to assess each outcome's certainty, 20 which was classified as high, moderate, low, or very low using GRADEpro software. 21 Review Manage program version (RevMan) 5.4 (Cochrane Collaboration, Oxford, UK) was used for meta‐analysis. Considering the clinical heterogeneity of these studies due to different study protocols and participant characteristics, a random‐effects model was used to pool overall estimates. We calculated the risk ratio (RR) with a 95% confidence interval (CI) for dichotomous outcomes. 22 The I‐squared ( I 2 ) was used to reflect the heterogeneity of data, and substantial heterogeneity existed when I 2  > 50%. 23 , 24 We also assessed the effect of each predefined factor on primary and secondary outcomes by subgroup analysis to explore potential heterogeneity. The factors included unit of randomization (sibling oocytes or patients), female age (above or below 35 years old), study year (before or after 2002), and controlled ovarian stimulation protocols (gonadotropin‐releasing hormone [GnRH] agonist or antagonist). We considered P  < 0.05 as a subgroup modification. 23 When more than one predefined factor was substantial in one outcome, we used multivariate meta‐regression analysis to confirm the subgroup effect. 25 Sensitivity analysis was used to test the robustness of the results. Studies with only low risk of bias (both selection and attrition bias were rated as low risk) were included for sensitivity analysis. Funnel plots were used to assess potential publication bias if more than 10 studies were included in a meta‐analysis. 26

Appendix S1 Click here for additional data file.