Sperm Selection Using Microfluidic Techniques Significantly Decreases Sperm DNA Fragmentation (SDF), Enhancing Reproductive Outcomes: A Systematic Review and Meta-Analysis.

This meta-analysis was conducted following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines [ 17 ] and registered in the PROSPERO database (ID: CRD42024582267). This registration not only ensures the transparency and accuracy of the research process but also allows other researchers to access information about the methods and objectives of our study. An automated search was conducted in the PubMed and Scopus databases, along with a manual reference search of the identified studies, according to our protocol and following the PRISMA guidelines. The same search strategy was used in each case: sperm AND microfluidic AND human AND (“reproductive results” OR “DNA fragmentation”). All (human) study types were considered, except for systematic reviews or meta-analyses. These searches were conducted between October 2023 and October 2024. Using the described search strategy, a total of 196 studies were retrieved (158 from PubMed and 38 from Scopus), excluding literature reviews. Of these, 115 publications were excluded based on title or abstract. Of the remaining 81, 21 were removed as duplicates. After full-text review, 39 studies (articles, abstracts and posters from conferences) were included, while the final exclusions were due to ambiguous or non-useful results for statistical analysis, failure to compare with conventional sperm selection techniques (SU/DG) or full text unavailability. From the 39 publications, 9 were prospective randomized studies and 30 non-randomized (with 16 prospective and 14 retrospective studies). The studies that were included were those in which (i) the authors performed a comparation between seminal parameters after semen samples (from male patients, volunteers or under fertility treatments) were processed either using microfluidics or a conventional technique (SU/DG); (ii) the authors performed a comparation between seminal parameters and/or reproductive outcomes from patients undergoing AI/IVF/ICSI cycles after sperm selection using either microfluidics or a conventional technique. Studies that did not present a comparison, had no control group (SU/DG), there was important missing data or presented ambiguous or non-useful data for the statistical analysis were excluded from the selection. Relevant data were extracted and entered into an Excel database, in which the information was structured for statistical analysis. Data included both semen parameters (SDF (%), concentration (M/mL), motility (%) and morphology (%)) and reproductive outcomes (%) from ART (fertilization rates/metaphase II (MII) oocyte microinjected, euploidy/number of biopsied blastocysts, implantation/embryo transfer (ET), pregnancy/ET, miscarriage per cycle and per pregnancy, live birth rate (LBR) per 1st ET, 1st cycle, all ET and concluded cycle). Article inclusion was reviewed by two researchers (A.G.I. and M.C.), with a third (N.G) as a tiebreaker in case of disagreement. All authors have read and agreed to the published version of the manuscript. The statistical analysis was conducted using Review Manager 5, developed by the Cochrane Collaboration [ 15 ]. Data from the selected studies were recorded and classified by either seminal parameters or reproductive outcomes. Control (SU and/or DG) and experimental procedures (microfluidics) were compared to obtain odds ratios (ORs) or mean differences (MDs), as appropriate, along with the corresponding 95% confidence intervals (CIs). Continuous variables (such as seminal parameter measurements) were grouped using the inverse of the variance model, with the results represented as mean differences. For non-continuous variables, the results were presented as ORs. The Mantel–Haenszel model was applied considering differences significant at p -values < 0.05. To assess heterogeneity and variations between studies not attributable to sampling errors, the I 2 statistic was used. Values above 50% were considered indicative of substantial heterogeneity. Due to the inherent variability found across the studies, only random effects were considered, regardless of the I 2 values found. Homogeneity was generally low, as determined by the different research designs and methodologies across the studies evaluated, and also as indicated by I 2 , which were almost always above 50%. For this reason, random effects models were preferred when conducting the statistical analysis. Funnel plots were generated to explore potential publication bias in the studies through a qualitative assessment of their symmetry. Consequently, forest plots to assess subgroup analyses were performed (microfluidics chip type: ZyMot, ZyMot Fertility, MD USA; Fertile Plus, Koek Biotechnology, Turkey; Other, custom devices by authors and other commercial brands) for each seminal or reproductive outcome to identify potential sources of heterogeneity. The quality in terms of risk of bias of the publications included in the meta-analysis was assessed using the RoB 2 [ 18 ] and ROBINS-I [ 19 ] tools. Each evaluation included multiple domains assessing specific aspects of bias, allowing the determination of both individual domain risk and the overall risk of bias for each publication. The RoB 2 tool was applied when publications were randomized trials, while ROBINS-I was used for non-randomized studies. The results of the quality assessment are presented in Figure A2 and Figure A3 , respectively.

Currently, around 15% of couples of reproductive age face infertility, defined as the inability to conceive after one year of unprotected sexual intercourse. This condition may arise due to female, male or combined known factors with a remaining relevant proportion of idiopathic infertility, where no apparent cause can be identified. In some cases, these factors directly cause infertility, while, in others, may diminish the chances of reproductive success [ 1 , 2 ]. In response to the aforementioned situation, assisted reproductive techniques (ARTs), following a proper diagnosis of infertility, aim to increase the chances of achieving a live birth (LB). They do so by replicating the physiological processes that naturally occur in vivo within the female reproductive tract, overcoming infertility factors and increasing the likelihood of reproductive success. However, these techniques’ success rates do not reach their full potential, frequently requiring repeated treatments due to diverse reasons (ovarian stimulation failures, poor embryonic development, implantation failure, recurrent miscarriages, etc.). As such, each step is in continuous development and improvement [ 3 ]. ART procedures require semen sample preparation to select the most competent sperm, focusing on optimal motility and morphology, which typically involve conventional techniques such as swim-up (SU) and density gradients (DGs). These techniques have shown to be safe and effective, primarily selecting sperm based on motility [ 3 ]. However, they do not consider the sperm’s underlying physiology or molecular profile. Both techniques involve sample manipulation, including pipetting and multiple centrifugation steps, which may negatively affect sperm viability by promoting sperm DNA fragmentation (SDF). This is crucial because, just as poor motility hinders sperm’s ability to reach the oocyte, sperm’s DNA damage has been associated with poorer embryonic quality and development, as well as an increased risk of miscarriage [ 1 , 4 , 5 ]. As a result, there has been a growing effort in recent years to develop new sperm selection methods that, in addition to being safe, replicate the physiological process of the female tract more efficiently. These techniques aim to minimize sample manipulation and consequent SDF while selecting sperm based on morphological and molecular features. Such is the use of testicular sperm [ 6 ], Magnetic Activated Cell Sorting (MACS) [ 7 ] or microfluidic chips. Testicular sperm has proven to be effective due to its retrieval prior to exposure to the oxidative environment of the seminiferous tubules and epididymis, where excessive production of reactive oxygen species (ROS) can induce oxidative stress and lead to SDF [ 6 ]. In contrast, techniques such as MACS and microfluidics are based on the selection of spermatozoa after ejaculation. Regarding microfluidic devices, they aim to mimic natural sperm selection by using microchannels, eliminating centrifugation steps and minimizing sample manipulation. Various devices are available, including the widely used ZyMot Multi, ZyMot ICSI (DxNow Inc., Gaithersburg, MD, USA) and Fertile Plus chip (Koek Biotechnology, İzmir, Turkey), while some authors have designed their own laboratory devices [ 4 , 8 ]. The difference between these two commonly used microfluidic chips (ZyMot and Fertile Plus) lies in their internal architecture. ZyMot utilizes a system of parallel microchannels without physical barriers or membranes, requiring spermatozoa to actively swim through narrow pathways from the inlet to the outlet chamber. In contrast, Fertile Plus consists of two chambers separated by a microporous membrane. Sperm selection occurs through passive filtration, where only spermatozoa with adequate morphology and appropriate size can traverse the membrane into the collection chamber [ 9 , 10 , 11 ]. An illustrative representation of each type of chip is provided in Appendix A ( Figure A1 ). In addition to the possible benefits from the use of microfluidic chips in relation to the elimination of the centrifugation steps, their use could also facilitate the workflow and allow freeing up technicians for other tasks. However, their implementation represents an additional cost for both the laboratory and patients undergoing infertility treatments. Additionally, while some studies have shown that microfluidics may improve sperm parameters and reproductive outcomes [ 4 , 5 , 10 , 12 , 13 ], other authors have reported no significant differences when compared to conventional techniques [ 1 , 14 , 15 , 16 ]. This highlights the lack of consensus, raising questions about the true potential benefits of its integration into daily clinical practice, although the reasons behind these discrepancies may include differences in the patient populations studied, protocols, outcomes measured, sample size and a number of other methodological factors leading to varying results in each study. Despite the lack of consensus and robustness, the use of these systems in laboratories (currently categorized as add-ons due to the above-mentioned reasons) has increased worldwide over the years, being frequently offered to patients with different infertility etiologies [ 9 ]. Hence, there is a need to systematically analyze all available data to determine if incorporating this technique into routine clinical practice provides a proper cost–benefit balance. The aim of this systematic review and metanalysis is to evaluate whether sperm parameters, as well as reproductive outcomes in assisted reproduction cycles, are improved, and by how much, based on the sperm selection technique using either conventional techniques (SU/DG) or microfluidic devices.

As previously mentioned, a total of 39 studies were included in the meta-analysis, and the selection process is shown in Figure 1 (Flow diagram). From the 39 studies, 9 were randomized controlled trials (RCTs) and 30 non-randomized. For each analyzed parameter, all studies presenting data were considered, regardless of whether they were RCTs or non-randomized studies. The main characteristics of the included studies, such as the design, study population, type of microfluidic chip used and measured parameters, are summarized in Table A1 . It is recommended to consider the quality analysis when evaluating the following results. Risk of bias was assessed as low (high quality) in two randomized trials ( Figure A2 ), while seven articles were assessed as having a moderate risk of bias (moderate quality). For non-randomized trials, the quality results exhibited greater variability ( Figure A3 ), with 1 study rated as low risk of bias, 23 publications with moderate risk of bias, and 6 articles rated as serious risk of bias (low quality). Additionally, funnel plots were used to analyze the heterogeneity between the studies included in each seminal parameter/reproductive outcome ( Figure A3 and Figure A4 , respectively). These plots also enabled the assessment of publication bias based on the dispersion of points in the graph. In general, most graphs did not show symmetry in the points, with greater dispersion indicating the presence of publication bias, except for three parameters, with the funnels plots acquiring a pyramidal structure (less data dispersion and absence of publication bias): miscarriage rate per pregnancy ( Figure A6 H), LBR per first transfer ( Figure A6 I) and LBR per first cycle ( Figure A6 J). However, for the last two outcomes, the results are not entirely conclusive due to the limited sample size, as only three articles were included. Starting with the analysis of the different seminal parameters, the following subsections detail the results obtained from the comparison between the two sperm processing techniques. These findings are summarized in Figure 2 . For sperm concentration (M/mL) [ 4 , 10 , 11 , 16 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 ] ( Figure 2 A), microfluids significantly lowered the concentration for each subgroup (ZyMot: p = 0.01, MD = −6.50 [−11.44, −1.55]; FertileChip: p < 0.00001, MD = −20.53 [−26.18, −14.88]; Others: p < 0.00001, MD = −20.3 [−29.16, −11.54]). The same conclusions were drawn in the combined analysis of all chip types ( p < 0.00001, MD = −15.95 [−19.28, −12.61]). The heterogeneity was considerable (I 2 = 99%, p < 0.00001), and the funnel plot ( Figure A5 A) suggests the possibility of publication bias. When analyzing progressive motility (%A+B) [ 10 , 11 , 16 , 20 , 21 , 22 , 27 ], microfluidics showed significant improvement of this parameter ( Figure 2 B) in the ZyMot ( p < 0.0001, MD = 35.09 [18.26, 51.91]) and Other ( p < 0.00001, MD = 14.31 [8.49, 20.12]) groups when compared to SU/DG. Within the ZyMot group, the two analyses conducted by Parrella et al. [ 11 ] led to this difference, as the study by Zaha et al. [ 33 ] crossed the line of no effect (though it had lower precision, with a wider 95% CI). No differences were found between FertileChip and the conventional techniques ( p = 0.85, MD = 2.58 [−23.72, 28.88]). The forest plot when considering all chip types showed higher progressive motility when using microfluids ( p = 0.04, MD = 14.50 [7.84, 21.17]). However, it should be noted that the heterogeneity in this case is also considerable, and the funnel plot ( Figure A5 B) suggests a potential publication bias. As for the total motility (%A + B + C) [ 4 , 10 , 11 , 21 , 22 , 23 , 25 , 27 , 28 , 29 , 30 ] analysis ( Figure 2 C), FertileChip studies once more crossed the line of no effect ( p = 0.66, MD = −9.89 [−53.30, 33.53]), while both ZyMot ( p < 0.0001, MD = 20.26 [14.99, 25.53]) and Other ( p < 0.00001, MD = 13.90 [9.35, 18.45]) showed a significant increase in total motility. When considering all chips, benefits for total motility were found with microfluidics ( p < 0.00001, MD = 10.68 [6.04, 15.31]). Once again, the heterogeneity between studies was high, and publication bias seemed to appear in the funnel plot ( Figure A5 C). Morphology (% normal forms) [ 10 , 11 , 21 , 25 , 27 , 30 , 31 ] was also analyzed ( Figure 2 D) following the aggregated approach, showing a beneficial effect when using microfluidics ( p = 0.0002, MD = 1.41 [0.67, 2.16]). It is worth noting that the study by Parrella et al. [ 11 ], favored either conventional techniques or microfluidics depending on the study population (ICSI cycles vs. ICSI cycles after prior failures at other centers + PGT, respectively). This may help identify which cases could benefit most from its use. Again, considerable heterogeneity was detected (I 2 = 95%) and publication bias was observed through the funnel plot ( Figure A5 D). SDF had the substantially largest number of published studies [ 4 , 5 , 10 , 11 , 12 , 20 , 21 , 25 , 27 , 30 , 31 , 32 ], with a subsequent higher precision, as seen in its 95% CI in the forest plot ( Figure 2 E). SDF detection techniques have varied among authors, with TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) [ 4 , 11 , 25 , 30 , 32 ] and SCD (sperm chromatin dispersion) [ 10 , 20 , 21 , 27 ] being the most commonly used methods. However, some studies have also employed alternative techniques such as SCSA (sperm chromatin structure assay) [ 31 ], COMET (comet assay) [ 5 ] or Halosperm [ 12 ]. Since the subgroups of interest for the present analysis were based on the type of microfluidic chip used, the forest plot does not display a stratification by SDF detection method. Nevertheless, the specific methodology used in each study was duly recorded in the database. Moreover, to enhance the trustworthiness of the analysis, we performed subgroup analysis based on both the type of microfluidic chip and the DNA fragmentation detection technique. However, no significant differences were found in either the p -values or the 95% confidence intervals when compared to the analysis presented in Figure 2 E, in which we incorporated all the studies regardless of the SDF detection method used. Additionally, some methods, such as COMET or Halosperm, were reported in only one study, making subgroup analyses by each SDF detection method impractical due to the decrease in statistical power. In the subgroup analysis, ZyMot chip showed significant differences ( p = 0.02, MD = −20.39 [−37.07, −3.71]). All articles suggested the same findings, except for Kocur et al. [ 25 ], whose data determined that microfluidics did not improve the SDF, although it was very close to the line of no effect ( p < 0.00001, MD = 2.90 [1.77, 4.03]). On the other hand, Mirsanei et al. [ 10 ] found significant differences favoring the use of FertileChip in two different analyses ( p < 0.00001, MD = −15.55 [−16.30, −14.79]). In the Others group, there was greater variability among the studies, with some crossing the line of no effect, yet showing statistically significant improvement in the experimental group (MD −5.28 [−8.89, −1.66]). In the overall analysis of all chip types, significant differences were obtained, demonstrating favorable results by lowering the SDF ( p < 0.00001, MD = −9.98 [−13.19, −6.76]) when using microfluidics. Considerable heterogeneity was observed (I 2 = 99%, p < 0.00001), and the funnel plot ( Figure A5 E) appears to suggest the presence of publication bias. Only one retrospective study addressing the use of microfluidics in artificial insemination cycles was found [ 22 ]. Given that it was a single article, no statistical analysis was performed, but it is worth noting that better outcomes were reported when using FertileChip (compared to DG) in terms of biochemical pregnancy rate, clinical pregnancy and ongoing pregnancy, although none of the differences reached statistical significance ( p > 0.05). All publications referred to ICSI cycles, with the exception of Palmerola et al.’s study [ 26 ], which was conducted in in vitro fertilization (IVF) cycles. Consistent with the previous section, the outcome-specific results are detailed in the following subsections. A comprehensive summary of these findings is also provided in Figure 3 . For the fertilization rate (per MII oocyte microinjected, %), as shown in Figure 3 A, 12 studies were included [ 4 , 11 , 14 , 20 , 23 , 34 , 35 , 36 , 37 , 38 , 39 ]. No significant differences were found in any subgroup (ZyMot p = 0.39 OR = 1.08 [0.90, 1.31], FertileChip p = 0.43 OR = 1.23 [0.74, 2.06]; Others p = 0.15 OR = 1.43 [0.87, 2.35]). However, when considering the overall results, significant differences were found ( p = 0.04), demonstrating an improvement in the fertilization rates when using microfluidics, although the OR (1.22 [1.01, 1.46]) presents a confidence interval close to the line of no effect. On the other hand, it is important to note the considerable heterogeneity among the studies (I 2 = 90%, p < 0.00001). A dispersion of the articles is also observed in the funnel plot ( Figure A6 A), which does not form a pyramidal structure, indicating a potential publication bias. For the rate of euploid embryos/number of biopsied blastocysts (%) ( Figure 3 B), seven studies were included in the analysis [ 4 , 11 , 24 , 26 , 35 , 36 , 38 ]. No significant differences were found between the two techniques ( p = 0.77, OR = 1.34 [0.88, 2.04]), despite some authors reporting favorable results when using microfluidics [ 4 , 36 ]. The heterogeneity among the studies remained high (I 2 = 90%, p < 0.00001). Once again, the funnel plot ( Figure A6 B) shows a potential publication bias. For biochemical pregnancy rate/all ET (%), eight publications were found ( Figure 3 C) [ 11 , 16 , 18 , 26 , 29 , 34 , 40 ], and some of the results differed based on the study population [ 13 , 40 ]. No significant differences were found in the biochemical pregnancy rate based on the sperm selection protocol used by FertileChip ( p = 0.99, OR = 1.00 [0.68, 1.47] and Others ( p = 0.21, OR = 1.41 [0.82, 2.44]). Nevertheless, the ZyMot group showed improved biochemical pregnancy rates ( p = 0.04, OR = 3.28 [1.03, 10.45]), with the 95% CI nearly crossing the line of no effect. In the overall microfluidic chip analysis, no significant differences were observed when compared to the conventional techniques ( p = 0.29, OR = 1.23 [0.84, 1.80]), with the exception of Parella et al.’s study [ 11 ]; however, this study shows low precision with an OR of 57 [1.92, 1693.41]). Furthermore, we again encounter a non-pyramidal funnel plot, suggesting publication bias ( Figure A6 C). Regarding implantation rate/number of ET (%) ( Figure 3 D), six studies were included, with significant differences favoring the use of ZyMot [ 11 , 36 ] ( p < 0.00001 (OR = 20.94 [6.13, 71.47])) and Other [ 4 , 20 ] ( p = 0.06 (OR = 7.06 [0.91, 54.78])). However, it is worth mentioning that these studies had the greatest dispersion and therefore lower precision. Neither of the two studies [ 16 , 34 ] using FertileChip showed significant differences between techniques ( p = 0.54, OR = 0.86 [0.52, 1.41]). When addressing the overall analysis, significant differences were demonstrated supporting the use of microfluidic devices ( p = 0.01, OR = 4.51 [1.42, 14.37]). Considerable heterogeneity was detected (I 2 = 84%, p < 0.00001), and the funnel plot indicates a potential publication bias ( Figure A6 D). Regarding clinical pregnancy rate/all ET (%), 16 articles were included [ 1 , 4 , 8 , 11 , 14 , 16 , 18 , 20 , 26 , 29 , 34 , 36 , 37 , 41 , 42 ] ( Figure 3 E), 2 [ 13 , 41 ] of them showing different results depending on the study population. In the FertileChip subgroup, no significant differences were found ( p = 0.16, OR = 1.28 [0.91, 1.81]), unlike ZyMot ( p = 0.04, OR = 3.21 [1.08, 9.58) or Other ( p = 0.02, OR = 2.23 [1.16, 4.30]), where differences showed an improvement in the clinical pregnancy rate when using these chips. It is worth mentioning that studies on both ZyMot and Other showed high dispersion in their 95% CIs, an indicator of lower precision. In the overall analysis, significant differences were found, showing higher clinical pregnancy rates when using microfluidics ( p = 0.002, OR = 1.73 [1.22, 2.45]). Additionally, substantial heterogeneity between studies was observed (I 2 = 67%, p < 0.0001). Although publications using Fertile Plus appear to be grouped considerably in the funnel plot ( Figure A6 E), those presenting results from ZyMot and custom-made chips are more dispersed, indicating a potential publication bias. For ongoing pregnancy rate/all ET (%), with six articles included [ 4 , 11 , 14 , 26 , 28 , 42 ], significant differences ( Figure 3 F) were found when using microfluidics rather than the conventional methods. However, they were very close to the line of no effect ( p = 0.04, OR = 1.99 [1.03, 3.83]). Additionally, some of the studies showed low precision due to data dispersion [ 4 , 13 ]. There was also substantial heterogeneity (I 2 = 71%, p = 0.004), and the funnel plot did not show a pyramidal structure ( Figure A6 F). Regarding miscarriage rate (biochemical and clinical, %), most of the studies included in the metanalysis were analyzed per pregnancy. Three of the publications did not clearly specify the denominator; therefore, it was inferred as per cycle based on the data presentation. As mentioned, only three articles [ 26 , 35 , 44 ] were found per cycle (one using ZyMot, one FertileChip and one Other). The analysis showed no significant differences in the miscarriage rate per cycle ( Figure A4 A) based on the sperm selection method ( p = 0.35, OR = 0.84 [0.54, 1.31]). As for miscarriage rate per pregnancy (%) ( Figure 3 G), 11 studies were included [ 1 , 4 , 8 , 11 , 16 , 28 , 34 , 35 , 36 , 41 ]. Following the overall focused analysis, no differential effect was observed between techniques ( p = 0.07, OR = 0.84 [0.54, 1.31]). The heterogeneity is moderate (I 2 = 58%, p = 0.009), and the funnel plot ( Figure A6 H) shows a pyramidal structure, ruling out the presence of publication bias. Based on the presentation of the data, this reproductive outcome was measured as LBR per first transfer, per first cycle, per all transfers or per completed cycle. As for LBR per first transfer (%) ( Figure A4 B), only two studies [ 18 , 45 ] provided these results, with the first study showing a larger size effect. The overall analysis suggested that there were no significant differences in the clinical pregnancy rate per first transfer when using either microfluidic devices or conventional techniques, as the OR = 1.60 [0.80, 3.22] crossed the line of no effect ( p = 0.18). The funnel plot shows a pyramidal structure, ruling out the presence of bias ( Figure A6 I). For LBR per first cycle (%) ( Figure A4 C), only two publications were available [ 35 , 44 ], where Escudé-Logares et al. [ 44 ] provided data from two different study populations. An improvement in the LBR when using microfluidic devices ( p = 0.009, OR = 1.59 [1.12, 2.24]) was found, although the size effect was small. In this case, the funnel plot shows a pyramidal structure. However, since only two articles are included, no conclusive statements can be made regarding publication bias ( Figure A6 J). Regarding LBR per all embryo transfers ( Figure 3 H) [ 34 , 35 , 43 ], significant beneficial differences were found in microfluids ( p = 0.03, OR = 1.65 [1.06, 2.55]). The heterogeneity in this section is substantial, I 2 = 64 ( p = 0.02), and the funnel plot indicates a potential publication bias ( Figure A6 K). Lastly, four studies considered outcomes per concluded cycle (%) ( Figure A4 D). One study used ZyMot [ 36 ], one used FertileChip [ 41 ] and two used Other [ 8 , 35 ]. Except for Kocur et al. [ 36 ], the authors did not find significant differences between processing methods in terms of LBR rate per cycle. It is worth noting that Kocur et al.’s study had the lowest precision, with an OR = 64.30 [3.66, 1129.40] that contributed 4.6% to the overall size effect. Regarding the overall analysis, significant differences were found, suggesting an improvement in this rate when using microfluidics ( p = 0.92, OR = 1.03 [0.53, 2.00]). The funnel plot does not present a pyramidal structure, which, in this case, may be due to the limited number of studies available for the analysis ( Figure A6 L). A summary table ( Table 1 ) provides an overview of the results obtained from the statistical analysis of each outcome. It includes the values for each outcome, the corresponding estimator with its 95% CI and the number of studies included in each analysis.

This meta-analysis compiled all available studies evaluating the impact of microfluidic techniques on semen parameters and reproductive outcomes in ART, mainly focused on ICSI cycles. After selection, 39 studies were included for analysis. Significant differences were found across all seminal parameters. Regarding sperm concentration, a significantly lower concentration was observed in the microfluidics group after processing. This effect is to be expected, as microfluidic devices allow a limited sample volume for processing, usually between 500 μL and 850 μL [ 9 ], whereas conventional methods can utilize up to the entire sample volume, accounting for the total sperm count available. This rationale was also noted by most authors [ 13 , 27 ]. In terms of sperm motility, analysis showed a significant improvement in both progressive and total motility when using microfluidics, according to the literature [ 4 , 23 ]. This suggests that, as sperm migrate through the microfluidics chip channels, those with lower motility are progressively filtered out. Consequently, only sperm with optimal motility reached the final stage of the process. Furthermore, better sperm morphology was also observed after using the chips compared to SU/DG, which could also be attributed to the selection process, as they passed through the channels (with poor morphology sometimes being associated with impaired motility). Regarding SDF, the results favored the use of microfluidics chips, which aligns with previous studies [ 10 , 13 ]. This suggests that the damage caused by sample manipulation when following SU/DG protocols could be reduced with the use of microfluidics thanks to the diminished manipulation. However, a few other studies did not find significant differences [ 1 , 25 ], as suggested by the moderate risk of bias assessed in those studies. In contrast, reproductive outcomes yielded more inconsistent results, which may be partly attributed to the substantial variation in the number of available publications and the total sample size across the outcomes. The findings showed an improved fertilization rate following the use of microfluidics, according to previous studies [ 4 , 40 ]. These findings suggest that the improvement in semen parameters, particularly the reduction in SDF, helps select higher-quality sperm for ICSI, thereby increasing the chances of successful fertilization. As for embryo euploidy, no significant differences were found, suggesting that the improvement in semen parameters did not impact the chromosomal content of the embryos. Nonetheless, some authors demonstrated a positive effect in their results [ 4 , 36 ]. Beyond this point, fewer significant differences were observed, which may be due to the limited number of studies that analyzed pregnancy, miscarriage and LBR. Improvements in the semen parameters after the use of microfluidics do not appear to significantly influence the biochemical pregnancy rate, which is consistent with the literature [ 24 , 45 ]. In contrast, microfluidics seems to have a positive impact on implantation [ 4 ], clinical pregnancy [ 1 , 36 ] and ongoing pregnancy [ 4 ]. However, some authors report contradictory findings [ 14 , 24 , 41 , 42 ]. Studies showing statistically significant effects for these outcomes tend to have wider 95% confidence intervals, whereas those with narrower intervals (indicating higher precision) do not report a beneficial effect of microfluidics. Therefore, these findings should be interpreted with caution. Once implantation occurred, no significant differences were observed in the miscarriage rate, whether analyzed per cycle or per pregnancy. Therefore, the improvement achieved in the semen parameters does not appear to be crucial in influencing the results of this outcome, although one author did find a beneficial effect in the microfluidics group [ 45 ]. Finally, for LBR, similar rates were found per first ET and per concluded cycle in both groups (microfluids and SU/DG). In contrast to LBR per cycle and per all ET, where microfluids showed a beneficial effect, which has been supported by previous studies [ 16 , 35 , 36 ]. Partly, the current findings support the use of microfluidics, suggesting that the reduction in SDF and improvement in other semen parameters positively impact the LBR. However, discrepancies in the results demonstrate that no conclusive results can be drawn regarding this outcome, highlighting the need for further studies to analyze it, as the number of publications pertaining to the LBR was the lowest. So far, the improvement in seminal parameters evidenced by microfluidics chip processing has not been fully reflected in reproductive outcomes. Overall, this meta-analysis offers a broader perspective on the findings from previous studies on the topic. It is essential to highlight that, when reviewing the results, factors such as the effect size, the number of publications available, sample size (N), dispersion in the 95%CI and data quality based on risk of bias should be considered, in addition to variations in the study populations or the use of different microfluidics chips (ZyMot, FertileChip, laboratory-made devices or other commercial brands). These differences among publications are limitations of the meta-analysis, as they may influence the conclusions drawn. Based on the analysis conducted in this meta-analysis, the conclusions suggest that microfluidics may offer advantages in the seminal parameters, improving sperm capacitation. This, along with the expectation that reduced sample manipulation leads to less sperm DNA damage, could potentially enhance reproductive outcomes, but further studies are needed. It should also be noted that some chips seem to offer better results overall. It remains uncertain whether these improvements are significant enough to justify its routine clinical use, despite its clear advantage in freeing up technicians to focus on other tasks. Instead, it may be more appropriate to consider microfluidics chips in specific cases such as high SDF, poor semen quality and male factor infertility with a previous history of failed cycles.

In conclusion, microfluidics techniques for sperm capacitation appear to enhance the semen parameters, thereby improving the quality of samples that will subsequently be used for ICSI. While the analysis of specific reproductive outcomes suggests a beneficial effect, certain outcomes appear to remain unaffected, possibly due to the limited number of publications and the results of their quality assessment. Consequently, further research is necessary to clarify the effect of microfluidic devices on reproductive outcomes. From a practical standpoint, microfluidics can streamline semen preparation and reduce laboratory workloads, though this aspect has not been formally evaluated. Nonetheless, their use entails additional costs for clinics and patients. For this reason, rather than being integrated into routine clinical practice, their use should be considered in selected cases (such as patients with high SDF, poor sperm morphology or motility or previous cycle failure). Thus, while their use is promising and should be viewed favorably, it must be implemented with caution and subject to ongoing assessment.