Testicular sperm aspiration enhances fertilization rates in MMAF patients undergoing ICSI: a retrospective cohort study

Abstract

Multiple morphological abnormalities of the sperm flagella (MMAF) compromise intracytoplasmic sperm injection (ICSI) outcomes. Testicular sperm aspiration (TESA) may be beneficial, but its impact on early embryonic development and pregnancy in MMAF remains uncertain. This study evaluated the effect of TESA on fertilization, cleavage, on-time Day-3 embryo development, and clinical pregnancy and explored modification by female age and MMAF genotype. In this retrospective cohort, 196 MMAF patients undergoing ICSI (January 2019–December 2024) without female-factor infertility were analyzed: TESA group (n = 65) versus non-TESA group (n = 131). Only fresh embryo transfer cycles were included. Outcomes were fertilization, cleavage, on-time Day-3 embryo development, and pregnancy. Multivariable regression was complemented by inverse probability of treatment weighting, propensity score matching, and stratified analyses by female age and genotype. Robustness was examined using bootstrap resampling and outlier-sensitivity analyses. Notably, TESA was associated with higher fertilization rates (adjusted coefficient: +5.63%, P = 0.022; inverse probability of treatment weighting average treatment effect [IPTW-ATE]: +5.9%, P = 0.022; Cohen’s d: 0.371). Cleavage gains were smaller (IPTW-ATE: +3.8%, P = 0.013; Cohen’s d: 0.246) and method-sensitive. No significant differences were found for on-time Day-3 development or pregnancy; power was limited for these endpoints. TESA mostly improved fertilization in women aged 30–35 years (IPTW-ATE: +9.5%, P = 0.017) and cleavage in genotype-negative patients (IPTW-ATE: +10.1%, P = 0.012). Propensity models showed good balance (area under the receiver operating characteristic curve: 0.638, and standardized differences <0.1); power was adequate for fertilization (0.825). Overall, TESA is associated with higher fertilization rates in MMAF-ICSI, with modest and method-sensitive evidence for cleavage, supporting selective use. Prospective, multi-center validation is needed. Limited downstream effects suggest a role primarily at early stages.

INTRODUCTION

Multiple morphological abnormalities of the sperm flagella (MMAF) represent a severe form of asthenozoospermia characterized by absent, short, bent, coiled, or irregular-caliber flagella, often leading to profound male infertility.1,2 This condition disrupts sperm motility and morphology, with underlying genetic etiologies frequently involving mutations in genes such as dynein axonemal heavy chain 1 (DNAH1), dynein axonemal heavy chain 5 (DNAH5), cilia- and flagella-associated protein 43 (CFAP43), and others that encode proteins essential for axonemal structure and ciliary function.2,3,4,5,6 Due to the functional limitations of sperm in MMAF, conventional in vitro fertilization (IVF) is often ineffective, rendering intracytoplasmic sperm injection (ICSI) a clinically viable alternative. Research indicates that clinical pregnancies can still be achieved through ICSI, even with completely immobile sperm.6,7,8,9

However, reproductive outcomes in MMAF patients remain suboptimal, with reported fertilization rates ranging from 54.3% to 80.1%6,10,11 and cleavage rates often below 82.1%,10,11 largely attributed to potential sperm DNA damage, mitochondrial dysfunction, or flagellar remnants interfering with pronuclear formation.1 Traditional sperm sources like ejaculated semen may exacerbate these issues in MMAF, as epididymal transit could introduce additional oxidative stress or maturational defects.12

Testicular sperm aspiration (TESA), a minimally invasive technique to retrieve spermatozoa directly from testicular tissue, offers a promising alternative. Although testicular sperm are typically immature, they often exhibit superior DNA integrity and reduced exposure to epididymal stressors compared to ejaculated sperm.13,14 Preliminary studies suggest TESA could enhance ICSI success in non-obstructive azoospermia or severe oligoasthenoteratozoospermia, with some reports indicating improved fertilization and embryo quality.15,16 For MMAF specifically, case series have hinted at benefits, but evidence is anecdotal, limited by small samples, and confounded by heterogeneous patient factors. Critically, few studies have rigorously assessed TESA’s effects while controlling for confounders, or explored potential effect modifiers that could guide clinical practice.

This retrospective cohort study aims to elucidate the true effects of TESA on early embryonic development and pregnancy outcomes in MMAF patients undergoing ICSI. By employing advanced statistical methods – including multivariable regression, inverse probability of treatment weighting (IPTW), and stratified analyses for age and genotype – we sought to mitigate confounding bias and identify effect modifiers. Our findings may help guide clinical decision-making, such as recommending TESA for MMAF patients.

PARTICIPANTS AND METHODS

Study design and participants

A retrospective study was conducted between January 2019 and December 2024 at Reproduction Medical Center of West China Second University Hospital, Sichuan University (Chengdu, China), adhering to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.17 The study was approved by the Ethics Committee of West China Second University Hospital, Sichuan University (Institutional Review Board Approval No. 2017040), which also granted the waiver of individual informed consent. The data supporting the findings of this study are not publicly available but can be obtained from the Ethics Committee upon reasonable written request. We reviewed electronic medical records of male patients diagnosed with MMAF who underwent ICSI between January 2019 and December 2024. MMAF diagnosis was based on semen analysis showing that >5% of sperm exhibit a mosaic of morphological abnormalities presenting absent, short, bent, coiled, and irregular tails,2 confirmed by genetic testing where applicable. Only fresh embryo transfer cycles were included in outcome analyses.

Inclusion criteria were as follows: (1) confirmed MMAF in the male partner; (2) ICSI cycles using either ejaculated or testicular sperm; (3) exclusion of female factor infertility (e.g., normal ovarian reserve, no tubal issues, and anti-Müllerian hormone >1.0 ng ml⁻¹); and (4) complete data on primary outcomes. Exclusion criteria included the following: (1) azoospermia unrelated to MMAF; (2) prior chemotherapy or radiotherapy; (3) concurrent female infertility factors; or (4) incomplete records. A total of 196 eligible cycles were identified and divided into two groups based on sperm source: TESA group (n = 65, testicular sperm obtained via percutaneous aspiration under local anesthesia) and non-TESA group (n = 131, ejaculated sperm). TESA was performed when ejaculated sperm were deemed insufficient or of poor quality by the clinician, per center protocols. Microdissection testicular sperm extraction (micro-TESE) was not routinely considered because all included MMAF cases were non-azoospermic with available ejaculated sperm for ICSI; for the purpose of obtaining limited sperm for ICSI, TESA is less invasive, simpler, and more cost-effective, whereas micro-TESE is primarily indicated for non-obstructive azoospermia requiring extensive sperm retrieval.

TESA

TESA was conducted following the methodology outlined by Esteves and Verza.18 The procedure was administered under local anesthesia, which included spermatic cord block and stabilization of the testis, and involved the use of an 18-gauge needle attached to a 20-ml syringe connected to a syringe holder. This assembly was inserted through the taut scrotal epidermis into the superior pole of the testis. By exerting negative pressure via the syringe holder and executing reciprocal movements of the needle’s tip within the testicular substance in an oblique trajectory, the aim was to fragment seminiferous tubules and assimilate tissue samples from varied loci. Upon aspiration of a diminutive segment of testicular tissue, the needle was gently retracted from the testicular body, with sustained negative pressure. Microsurgical forceps were employed to secure and extract the seminiferous tubules that protruded through the scrotal dermis, thereby easing the specimen’s extraction. Subsequent to the preparatory steps, the specimen was flushed into a receptacle containing 0.5–1.0 ml of warmed sperm medium prior to its conveyance to the laboratory for further analysis.

Ovarian stimulation and embryo development

All participants underwent a treatment regimen involving either a gonadotropin-releasing hormone (GnRH) agonist or antagonist protocol. In the GnRH agonist long protocol, pituitary suppression was initiated during the mid-luteal phase of the preceding cycle utilizing triptorelin acetate at a dosage of 0.1 mg every 48 h (Lot No. TA202103; Ipsen Pharma Biotech, Paris, France). Following verification of pituitary suppression through ultrasonographic examination and assessment of baseline hormonal profiles, the administration of recombinant follicle-stimulating hormone (FSH; Lot No. FSH2106; Merck KGaA, Darmstadt, Germany) commenced at dosages ranging from 150 IU to 225 IU. The initial dosage of gonadotropin was tailored based on the patient’s age and body mass index (BMI). Follicular development was evaluated after a 4-day period of consistent dosage, with adjustments made optimally based on follicular number and size.

Conversely, in the antagonist protocol, initial assessment involved ultrasonography and baseline hormonal evaluation on the third day of the menstrual cycle. Subsequently, treatment with recombinant FSH was initiated, analogous to the procedure outlined in the GnRH agonist long protocol. The administration of the GnRH antagonist cetrorelix (0.25 mg, Cetrotide; Lot No. CT202011; Merck-Serono, Geneva, Switzerland) commenced on the sixth day of the stimulation cycle and continued daily until the administration of the hCG trigger. Oocyte retrieval was planned 36 h following the injection of 5000 IU or 10 000 IU of hCG (Lot No. hCG202008; Livzon Pharmaceutical Group, Zhuhai, China) when two or more follicles attained a diameter of 17 mm. Fertilization was assessed 16–18 h post conventional ICSI. Embryos were transferred on the third day post-retrieval.

Semen routine analysis

Semen specimens were obtained in sterile containers following a period of sexual abstinence ranging from 2 days to 7 days. Subsequent to collection, the samples underwent preparation and were analyzed for sperm concentration, forward progressive motility, and total motility rate in accordance with the 5^th edition of the World Health Organization (WHO) manual.19 To be classified as normal controls, the semen quality parameters had to meet or exceed the reference values for human semen characteristics as established by the WHO, which include a total sperm count of ≥39 × 10⁶ per ejaculate, sperm concentration of ≥15 × 10⁶ ml⁻¹, progressive sperm motility of ≥32%, and a proportion of morphologically normal sperm of ≥4%.19

Whole-exome sequencing (WES) and Sanger sequencing

WES was performed on DNA samples from all patients to identify genetic variants associated with MMAF. DNA libraries were prepared using the TruSeq DNA Library Prep Kit (Illumina, San Diego, CA, USA), followed by exome enrichment with the SureSelect Human All Exon V5 (Agilent Technologies, Santa Clara, CA, USA) and TruSeq Exome Enrichment Kit (Illumina). Sequencing was conducted according to standard protocols. Identified variants were cross-referenced against established databases, including ClinVar, dbSNP, the Genome Aggregation Database (gnomAD), and the Human Gene Mutation Database (HGMD), to determine novelty or prior reporting. Patients were initially stratified into two groups: carriers of known pathogenic or likely pathogenic variants, and non-carriers. Further subclassification was based on the clinical significance of variants per American College of Medical Genetics and Genomics (ACMG) guidelines.20

Sanger sequencing was employed to validate selected variants in probands and family members. Primers were designed to amplify target regions, and sequencing was performed using BigDye Terminator version 3.1 chemistry (Applied Biosystems, Foster City, CA, USA) on an ABI 3730xl sequencer (Applied Biosystems).

Sequencing platform and computational analysis

Sequencing was carried out on the Illumina HiSeq 2000 platform (Illumina) for high-throughput parallel processing, with reads aligned to the human reference genome (hg19 assembly). Computational analysis involved genome alignment using the Burrows–Wheeler Aligner (BWA), followed by duplicate read removal with SAMtools. Read realignment and base quality score recalibration were performed according to Genome Analysis Toolkit (GATK) best practices (version 2; Broad Institute, Cambridge, MA, USA). Variant calling and annotation were conducted using SAMtools and ANNOVAR (Wang Genomics Lab, University of Pennsylvania, Philadelphia, PA, USA).

Variant selection criteria

Variants were filtered stringently to prioritize those with potential pathogenicity. Eligible variant types included missense, nonsense, frameshift, and splice-site alterations. Population frequency thresholds were set at less than 1% across multiple databases, including dbSNP, the 1000 Genomes Project, the Exome Aggregation Consortium (ExAC), the NHLBI Exome Sequencing Project (ESP6500), and an in-house genomic database.

Data collection and variables

Baseline characteristics were extracted from medical records, including female age (categorized as <30 years, 30–35 years, and ≥35 years, according to the previous study21), male age (input as continuous for exploratory analysis), and male MMAF genotype. Genotyping was performed via WES, classifying patients as: Gene Type 0 (no identifiable pathogenic variants: “negative”); Gene Type 1 (DNAH1 mutations); or Gene Type 2 (other mutations, e.g., DNAH5, cilia- and flagella-associated protein 43 [CFAP43], cilia- and flagella-associated protein 74 [CFAP74], or miscellaneous including cilia- and flagella-associated protein 65 [CFAP65], dynein axonemal heavy chain 8 [DNAH8], etc.). Procedural details, such as sperm retrieval method and ICSI cycle outcomes, were recorded prospectively in the center’s database.

Potential confounders and effect modifiers were identified through a statistical screening process. Candidate variables (female age, male age, male genotype, and ovarian stimulation protocol) were evaluated for: (1) association with TESA group assignment (via Chi-square or Mann–Whitney U tests, as appropriate, and standardized mean differences); (2) association with the primary outcome of fertilization rate (via Pearson or Spearman correlations); and (3) potential effect modification. Variables meeting exploratory criteria were selected for inclusion in adjusted and stratified analyses. This resulted in the retention of female age and male genotype, while male age and ovarian stimulation protocol were excluded due to lack of significant signals (see screening results in Supplementary Table 1).

Supplementary Table 1.

Screening results for covariate selection

Covariate	Association with TESA Group (P; standardized diff)	Association with fertilization rate (correlation coefficient; P)	Interaction with TESA (regression interaction P)	Selection decision
Female age (continuous)	P=0.28; std diff=0.17 (marginal)	Fertilization: r=−0.18; P=0.012 (sig)  Cleavage: r=−0.14; P=0.051 (marginal)  On-time Day-3 embryo: r=−0.09; P=0.21 (NS)  Pregnancy: rho=−0.06; P=0.42 (NS)	Fertilization: P=0.045 (sig) Cleavage: P=0.088 (marginal)  On-time Day-3 embryo: P=0.32 (NS)  Pregnancy: P=0.56 (NS)	Included
Male genotype (categorical)	P=0.03 (Gene Type 0); std diff=0.32 (Gene Type 0); P=0.04 (Gene Type 1+); std diff=0.28 (Gene Type 1+)	Fertilization: rho=0.15; P=0.038 (sig)  Cleavage: rho=0.12; P=0.094 (marginal)  On-time Day-3 embryo: rho=0.05; P=0.48 (NS)  Pregnancy: rho=0.08; P=0.28 (NS)	Fertilization: P=0.15 (NS)  Cleavage: P=0.022 (sig)  On-time Day-3 embryo: P=0.41 (NS)  Pregnancy: P=0.37 (NS)	Included
Male age (continuous)	P=0.33; std diff=0.14 (non-significant)	Fertilization: r=−0.09; P=0.21 (NS)  Cleavage: r=−0.07; P=0.33 (NS)  On-time Day-3 embryo: r=−0.04; P=0.58 (NS)  Pregnancy: rho=−0.03; P=0.67 (NS)	Fertilization: P=0.48 (NS)  Cleavage: P=0.62 (NS)  On-time Day-3 embryo: P=0.71 (NS)  Pregnancy: P=0.82 (NS)	Excluded
Ovarian stimulation protocol (categorical)	P=0.52; std diff=0.11 (non-significant)	Fertilization: rho=0.07; P=0.35 (NS)  Cleavage: rho=0.05; P=0.49 (NS)  On-time Day-3 embryo: rho=0.10; P=0.16 (NS)  Pregnancy: rho=0.09; P=0.23 (NS)	Fertilization: P=0.62 (NS)  Cleavage: P=0.74 (NS)  On-time Day-3 embryo: P=0.39 (NS)  Pregnancy: P=0.51 (NS)	Excluded

Pearson’s r was used for continuous outcomes and Spearman’s rho for binary outcomes. Covariates were screened based on at least one of the following criteria: 1: imbalance between TESA and non-TESA groups (standardized difference ≥0.10); 2: significant or marginal association with reproductive outcomes; 3: evidence of effect modification (interaction P<0.10). TESA: testicular sperm aspiration; Std diff: standardized difference; sig: statistically significant; NS: not significant

Outcome measures

The primary outcome measures focused on early embryonic development and pregnancy outcomes. The fertilization rate was defined as the proportion of injected oocytes that formed two pronuclei (2PN) 16–18 h after ICSI. The cleavage rate was defined as the proportion of 2PN zygotes that reached the 2- to 4-cell stage by Day 2. The on-time Day-3 embryo development rate was defined as the proportion of cleaved embryos that met laboratory criteria for timely development on Day 3, including the presence of 6–8 cells, less than 20% fragmentation, even blastomeres, and no multinucleation. Pregnancy outcome was assessed as clinical pregnancy, defined by the presence of a gestational sac on transvaginal ultrasound at 6 weeks of gestation per embryo transfer cycle. All outcome rates were calculated on a per-cycle basis and expressed as percentages, with fertilization rates calculated using injected metaphase II oocytes as the denominator, cleavage rates using 2PN zygotes, on-time Day-3 embryo development rates using cleaved embryos, and clinical pregnancy rates using embryo transfer cycles.

Statistical analyses

Data were analyzed using Python 3.12 (Python Software Foundation, Wilmington, DE, USA) with libraries including SciPy, Statsmodels, Scikit-learn, Pandas, NumPy, Matplotlib, and Seaborn for visualization. Continuous variables were summarized as mean ± standard deviation (s.d.) or range and compared using Mann–Whitney U test (non-parametric, primary method due to potential non-normality), independent t-test, or Welch’s t-test. Categorical variables were compared via Chi-square or Fisher’s exact tests, with Cramer’s V for effect size.

To address confounding, multivariable linear regression models adjusted for female age, male age, and genotype when appropriate, reporting coefficients, R², F-statistics, and P-values. Inverse probability of treatment weighting (IPTW) estimated average treatment effects (ATE), with propensity scores derived from logistic regression. Covariate balance was assessed via standardized mean differences (threshold <0.1) and propensity score overlap. Mantel–Haenszel test provided stratified odds ratios for binary outcomes.

Effect modification was explored through stratified analyses by female age and genotype, with interaction terms in regression models. Power calculations used Cohen’s d for effect size, assuming α = 0.05 and β = 0.20.

Sensitivity and robustness checks included bootstrap, outlier removal, variance inflation factor (VIF) for multicollinearity, and multiple comparison corrections. Assumptions were verified: normality (Shapiro–Wilk), homoscedasticity (Levene’s test), and residual diagnostics. A two-sided P < 0.05 was considered significant, with emphasis on effect sizes and clinical relevance over P-values alone.

RESULTS

Baseline characteristics

A total of 196 ICSI cycles from MMAF patients were included, with 65 (33.2%) in the TESA group and 131 (66.8%) in the non-TESA group. Baseline characteristics are summarized in Table 1. Female and male ages were comparable between groups (P = 0.28 and P = 0.33, respectively), with small standardized differences (0.17 and 0.14, respectively). However, genotype distribution differed significantly: the TESA group had fewer Gene Type 0 cases (41.5% vs 57.3%, P = 0.03, standardized difference: 0.32) and more Gene Type 2 cases (20.0% vs 9.9%, P = 0.04, standardized difference: 0.28). Detailed genotype distributions are provided in Supplementary Table 2, highlighting the predominance of DNAH1 mutations in Gene Type 1 and a diverse array of rarer mutations in Gene Type 2.

Table 1.

Baseline characteristics of study participants

Variable	Non-TESA group (n=131)	TESA group (n=65)	P	Standardized difference
Female age (year), mean±s.d. (range)	31.7±5.2 (21–52)	30.8±5.3 (23–52)	0.28	0.17
Male age (year), mean±s.d. (range)	33.1±5.5 (22–55)	32.4±5.0 (24–51)	0.33	0.14
Genotype distribution
Gene Type 0 (negative), n (%)	75 (57.3)	27 (41.5)	0.03	0.32
Gene Type 1 (DNAH1), n (%)	43 (32.8)	25 (38.5)	0.46	0.12
Gene Type 2 (other), n (%)	13 (9.9)	13 (20.0)	0.04	0.28

Gene Type 0 (negative): no pathogenic or likely pathogenic variants detected by whole-exome sequencing. Gene Type 1 (DNAH1): DNAH1 detected by whole-exome sequencing. Gene Type 2 (other): includes pathogenic or likely pathogenic variants in genes other than DNAH1. TESA: testicular sperm aspiration; s.d.: standard deviation; DNAH1: dynein axonemal heavy chain 1

Supplementary Table 2.

Detailed genotype distribution

Genotype	Mutation type	Non-TESA Group (%)	TESA Group (%)	Total
Gene Type 0	Negative	75 (100)	27 (100)	102
Gene Type 1	DNAH1	43 (100)	25 (100)	68
Gene Type 2	DNAH5	1 (7.7)	1 (7.7)	2
	CFAP43	1 (7.7)	1 (7.7)	2
	FSIP2	1 (7.7)	1 (7.7)	2
	Other†	10 (76.9)	10 (76.9)	20
	Subtotal	13 (100)	13 (100)	26

^†Other mutations include CFAP74, CFAP65, DNAH8, CT55, TCC29, CFAP47, AKAP3, DNAH6, DAB2IP, SOHLH, DNHD1, AKAP4, CEP78, USP9, CCDC39. Percentages are calculated within each gene-type subgroup. Gene-type classification was based on predefined MMAF-related genetic categories. MMAF: multiple morphological abnormalities of the sperm flagella; TESA: testicular sperm aspiration

Propensity score analysis confirmed good covariate balance post-weighting, with all standardized differences <0.1 and overlap in propensity scores (0.175–0.512). The model’s area under the receiver operating characteristic curve (AUC) was 0.638, indicating moderate predictive ability.

Primary outcomes

TESA’s effects on key outcomes are detailed in Table 2. Fertilization rates were significantly higher in the TESA group across multiple methods (multi-method: P = 0.0116; adjusted coefficient: +5.63%, P = 0.022; IPTW-ATE: +5.9%, 95% confidence interval [CI]: 0.010 to 0.108, P = 0.022), with a medium effect size (Cohen’s d: 0.371). Cleavage rates showed limited evidence of improvement (multi-method: P = 0.044; IPTW-ATE: +3.8%, 95% CI: −0.006 to 0.097, P = 0.013; Cohen’s d: 0.246), but adjusted models were non-significant (coefficient: +3.34%, P = 0.238). No significant effect was observed for the on-time Day-3 embryo development rate across analytical methods (95% CI: −0.030 to 0.140, multi-method: P = 0.251; Cohen’s d: 0.195). Similarly, pregnancy outcomes did not differ significantly between the non-TESA group and TESA group (χ² = 1.046, P = 0.593; Cramer’s V = 0.073; Figure 1).

Table 2.

Core effects of testicular sperm aspiration on primary outcomes

Outcome	Mean difference (%)	P	Effect size (95% CI) c	Conclusion
Fertilization rate				Strong positive
Adjusted modela	+5.63	0.022	0.371 (0.010 to 0.108)
IPTW analysisb	+5.9	0.022	-
Cleavage rate				Limited evidence
Adjusted modela	+3.34	0.238	0.246 (−0.006 to 0.097)
IPTW analysisb	+3.8	0.013	-
On-time Day-3 embryo development				Insufficient evidence
Adjusted modela	-	0.251	0.195 (−0.030 to 0.140)
Pregnancy outcome				Negative result
Chi-square test	-	0.593	0.073 (Cramer’s V)

^aEstimates derived from multivariable linear regression adjusted for female age and male genotype. ^bATE calculated based on propensity score weighting. ^cStandardized effect sizes are presented as Cohen’s d for continuous outcomes and Cramer’s V for categorical outcomes. CI: confidence interval; IPTW: inverse probability of treatment weighting; TESA: testicular sperm aspiration; ATE: average treatment effect; -: not applicable or no reliable estimate because of data limitations.

Figure 1.

Overview of outcomes and subgroup effects. (a) Distributions of fertilization rate, on-time Day-3 embryo development rate, and cleavage rate for non-TESA and TESA groups (box plots with individual points). (b) Clinical pregnancy outcomes by group (not pregnant, miscarriage, and pregnant). TESA: testicular sperm aspiration.

Multivariable regression supported these findings: for fertilization rate (R² = 0.041, F = 2.705, P = 0.047; TESA coefficient: 0.0563, P = 0.022) and for cleavage rate (R² = 0.047, F = 3.126, P = 0.027; TESA coefficient: 0.0334, P = 0.238). Mantel–Haenszel adjusted effects were 0.0585 (Z = 2.290, P = 0.022) for fertilization and 0.0590 (Z = 2.497, P = 0.013) for cleavage. VIF indicated mild collinearity (female age: 3.95; Gene Type: 1.69; TESA: 3.01; Figure 2).

Figure 2.

Mantel–Haenszel stratified effects and model diagnostics. (a) Fertilization rate differences effect (difference defined as the percentage in the TESA group minus the percentage in the non-TESA group) by female age strata (young [< 30 years], middle [30–35 years], and advanced [≥ 35 years]). (b) Cleavage rate differences effect (difference defined as the percentage in the TESA group minus the percentage in the non-TESA group) by genotype strata (Gene Type 0, 1, and 2). (c) Variance inflation factor (VIF) for key covariates. (d) Model goodness-of-fit shown by R² and adjusted R² for fertilization and cleavage models. ^*P < 0.05. VIF: variance inflation factor; TESA: testicular sperm aspiration.

Overall standardized effect sizes and regression coefficients are summarized in Figure 3a and 3b. Power analysis revealed good overall power for fertilization (0.825) but moderate for cleavage (0.519), with the 30–35-year age subgroup showing the highest power (0.436).

Figure 3.

Effect sizes and confounding analyses. (a) Overall standardized effect sizes (Cohen’s d) for fertilization, on-time Day-3 embryo development, and cleavage. (b) Multivariable regression coefficients (95% CI) for fertilization and cleavage models (TESA, female age, and genotype). (c) Stratified Cohen’s d for fertilization (age strata) and cleavage (genotype strata). (d) Bootstrap mean differences with 95% CI for fertilization and cleavage. ^*P < 0.05. NS: not significant; CI: confidence interval; TESA: testicular sperm aspiration.

Stratified and interaction analyses

Stratified results by female age (Supplementary Table 3) showed the strongest TESA effect on fertilization in the 30–35-year group (IPTW-ATE: +9.5%, P = 0.017), with non-significant trends in other groups. Cleavage effects were modest across ages. By genotype (Supplementary Table 4), cleavage rates improved most in Gene Type 0 (IPTW-ATE: +10.1%, P = 0.012), with no benefits in Gene Type 1 or 2.

Supplementary Table 3.

Stratified effects by female age

Age group (years)	Sample size	Fertilization effect (%)	P	Cleavage effect (%)	P
<30	98	+2.4	0.401	+4.8	0.141
30–35	63	+9.5	0.017	+3.4	0.664
≥35	35	+6.7	0.262	+3.2	0.246

Effects are presented as absolute percentage-point differences derived from IPTW-weighted models. Female age was identified as an effect modifier for fertilization rate (Supplementary Table 1). IPTW: inverse probability of treatment weighting

Supplementary Table 4.

Stratified effects by genotype

Genotype	Sample size	Fertilization effect (%)	P	Cleavage effect (%)	P
Gene Type 0	102	+4.6	0.234	+10.1	0.012
Gene Type 1	68	+6.4	0.105	−1.3	0.993
Gene Type 2	26	+8.6	0.189	−4.0	0.646

Stratified analyses were conducted using IPTW-adjusted models. A significant improvement in cleavage rate was observed only in Gene Type 0. IPTW: inverse probability of treatment weighting

These effect-modification patterns were consistent with the stratified analyses, with the strongest fertilization benefit observed in women aged 30–35 years and the strongest cleavage benefit in Gene Type 0 patients (Figure 2 and Supplementary Table 3 and 4).

Sensitivity and robustness analyses

Results were consistent across statistical methods for fertilization (Mann–Whitney U test P = 0.0116; t-test P = 0.0152; Welch’s P = 0.0199) but sensitive for cleavage (Mann–Whitney U test P = 0.0435; t-test P = 0.1073; Welch’s P = 0.0879). Outlier removal slightly altered P-values (fertilization: t-test P = 0.0131, n = 189; cleavage: t-test P = 0.1845, n = 171), indicating robustness for fertilization but vulnerability for cleavage. Bootstrap CIs confirmed effect for fertilization (0.058, 95% CI: 0.005 to 0.105) and cleavage (0.045, 95% CI: −0.010 to 0.095; Figure 3d).

DISCUSSION

The findings from this retrospective cohort study demonstrate that TESA significantly enhances fertilization rates in patients with MMAF undergoing ICSI, with an estimated improvement of 5.6%–5.9% across adjusted models. Evidence for cleavage rate improvement was relatively limited (3.4%–3.8% increase, Cohen’s d: 0.246), sensitive to analytical choices, while no benefits were observed for on-time Day-3 embryo development rate or pregnancy outcomes. These results are consistent with prior reports indicating advantages of TESA in severe male infertility and extend these findings to the MMAF population by incorporating rigorous control for confounding factors and comprehensive analyses of effect modification.8,9,22,23

A key contribution of this study is the identification of clinically relevant effect modifiers: TESA’s benefits on fertilization were most pronounced in women aged 30–35 years (+9.5%), with attenuated and non-significant effects in those under 30 years (+2.4%) and over 35 years (+6.7%). This non-linear pattern may reflect a “sweet spot” in the 30–35-year age group, where oocytes retain good quality but begin to exhibit subtle age-related vulnerabilities that TESA can effectively compensate for.

For cleavage rate, the strongest improvement was observed in patients without known MMAF-related variants (Gene Type 0: +10.1%). This finding suggests that patients with milder or non-genetic MMAF phenotypes may derive greater benefit from testicular sperm. Such an effect may be related to the purported superior DNA integrity of testicular sperm. Interaction analyses confirmed these patterns, highlighting opportunities for personalized assisted reproductive technology (ART) strategies, e.g., prioritizing TESA in mid-aged couples or genotype-negative men.

Methodological strengths include the use of multi-method validation and comprehensive sensitivity assessments, which enhance confidence in the primary finding despite the retrospective design. However, results for cleavage rates were fragile to outliers and method variations, underscoring the need for caution. Statistical power was sufficient for detecting effects on fertilization but suboptimal for subgroups, and the single-center setting limits generalizability. Unmeasured confounders (e.g., sperm DNA fragmentation) and multiple comparisons represent limitations, as does the observational nature, which cannot fully rule out selection bias in TESA assignment. Although no statistically significant differences were observed for on-time Day-3 embryo development or pregnancy outcomes, these analyses were underpowered relative to fertilization. Given the observed baseline rates and the available sample size, the minimum detectable effect for these endpoints was modest; therefore, type II error remains a possibility.

For patients with MMAF, ICSI with ejaculated sperm remains the standard approach in many fertility centers; however, severe tail defects, immobility, or suspected oxidative damage may limit performance. Testicular sperm (TESA/TESE) can reduce epididymal exposure and potentially improve DNA integrity, supporting use when ejaculated sperm are scarce, immotile, or repeatedly underperform. In selected cases with prior ICSI failure, adjunct sperm selection approaches and the cautious use of assisted oocyte activation (AOA) may be considered according to the center’s standard operating procedures (SOPs) and patient-specific risks. While genetic findings can inform preconception counseling, they do not uniformly predict ART outcomes, warranting individualized decisions.

Our data indicate that TESA provides the most clinically meaningful insights with respect to early outcomes (fertilization) and reveal discernible subgroup effects based on female age and genotype negativity. These patterns may guide selective TESA rather than routine application for all MMAF cases.

Clinically, these data support the selective use of TESA to optimize fertilization outcomes in MMAF, provide no evidence in favor of routine application for downstream endpoints, and motivate prospective validation. Future prospective, multi-center trials should validate these stratified effects, explore underlying mechanisms, and assess cost-effectiveness. Ultimately, integrating genotype and female age into clinical decision-making algorithms may advance the implementation of precision medicine in reproductive endocrinology.

CONCLUSIONS

TESA is associated with higher fertilization rates in MMAF patients undergoing ICSI. Evidence for cleavage benefits is modest and method-sensitive, and we observed no statistically significant differences in on-time Day-3 embryo development or pregnancy outcomes under limited power. The presence of age- and genotype-related effect modification suggests that a selective approach to TESA may be the most appropriate, pending validation in larger, prospective studies.

AUTHOR CONTRIBUTIONS

LXP conceived of the study, participated in its design, carried out data acquisition and analysis, and drafted the manuscript. LLF participated in the statistical analysis, including multivariable regression and propensity score methods, and revised the manuscript critically for intellectual content. JHL contributed to data interpretation and effect modification analyses. LQ assisted with data collection and sensitivity analyses. XYQ performed the genotyping and covariate screening. YB contributed to the clinical outcome assessments and stratified subgroup analyses. WHL participated in study coordination, ethical oversight, and final revisions. YHY conceived of the study, oversaw its design and coordination, and helped draft and revise the manuscript. All authors read and approved the final manuscript.

COMPETING INTERESTS

All authors declare no competing interests.