Outcomes of Ultimpro Testing Implementation for Patients With Unexplained Recurrent Implantation Failure: A Single-Center Retrospective Cohort Study.

This study was approved by the McGill University Health Center (MUHC) ethics board, study number MUHC REB #2025‐10617.

This is a single center retrospective cohort study of patients followed at the McGill University Health Center Reproductive Immunology clinic (MRIC). The MRIC is the reference center in the province of Quebec of the evaluation of patients with suspected immune‐mediated reproductive failure. We aimed to determine if Ultimpro improved ongoing pregnancy rates (OPR) in patients with unexplained RIF. This study was approved by the MUHC research ethics board (MUHC REB #2025‐10617). All methods were performed in accordance with relevant guidelines and regulations. This manuscript was written according to Strengthening the Reporting of Observational studies in Epidemiology (STROBE) guidelines. All patients undergoing Ultimpro testing and ET between January 2021 and May 2023 were included. Patients were divided into two groups. Group 1 included patients with RIF fitting our inclusion criteria (Table  1 , Figure  2 ). Criteria for Ultimpro testing are: ≥ 3 consecutive high‐quality (≥ 3BB Gardner's criteria) previous blastocyst transfer failures, age ≤ 42 (<45 if using oocyte donation), body mass index (BMI) < 35, non‐smokers, endometrium ≥ 7 mm at time of transfer, absence of hydrosalpinges, normal uterine cavity, absence or correction of endocrinopathy (medically controlled hypo/hyperthyroidism, hyper‐prolactinemia, diabetes) and absence of HIV, hepatitis B/C and syphilis. All patients in Group 1 were required to have at least one high quality blastocyst (Gardner ≥ 3 BB) remaining for transfer. A significant number of patients without RIF or not fitting our Ultimpro criteria requested testing against medical advice. These patients were included in the study, but as a separate group (Group 2) (Figure  2 ). The first ET treated with Ultimpro recommendations was used as the index ET. Patients for whom the index embryo lost significant quality upon thaw (Gardner grade C or lower for the trophectoderm or the inner cell mass) were excluded. Only patients for whom the index ET is known were included in the analysis. In patients who conceived, only those pregnant ≥ 32 weeks gestation at the time of analysis were included. Inclusion and exclusion criteria, definitions. Non‐smoker Having ≥ 1 high quality embryo (Gardner ≥ 3BB) left to transfer) Non‐compliance with recommended treatment Intravenous Immunoglobulin Granulocyte colony stimulating factor (G‐CSF) Low molecular weight heparin (LMWH) Glucocorticoids Calcineurin inhibitors Study cohorts. Patients referred to the MRIC for uRIF were presented with the option of performing Ultimpro testing. Patients accepting Ultimpro testing were provided with a kit containing the collection tube. The referring fertility specialist then performed endometrial biopsy on a hormone replacement therapy cycle (on Days 6–8 of progesterone) or on a natural cycle (LH + 8–11 days) without glucocorticoids or acetylsalicylic acid. Samples were shipped to the Matrice Lab Innove laboratory according to manufacturer instructions. Ultimpro treatment recommendations were followed for each patient for their index ET, only patients who were compliant to treatment were included in our study (Figure  1 ). Although Ultimpro suggests repeat biopsy with treatment application, repeat biopsies were not performed because of time and financial constraints to the patient. We used an unmatched RIF control group (Group 3) from a previous study [ 32 ] as a comparator for Group 1 outcomes. Briefly, 1061 patient files were screened from the MRIC (170 patients) and the MUHC Reproductive Center (891 patients) and included patients undergoing ET between January 2020 and December 2021. We included 94 controls fitting or inclusion criteria (≥ 3 previous high quality ET failures and respecting our inclusion criteria (Table  1 , Figure  2 )). The last blastocyst transfer on record during the study period was included as the index ET. BMI was not available for all control patients, but weight was. We excluded patients with recoded weight over 95 kg (BMI 34.5 for an average 165 cm woman). Patients were also excluded if they had previously been treated with immunomodulatory treatment for any of their ETs (including glucocorticoids, intravenous immunoglobulin, intralipids, or calcineurin inhibitors). Importantly, there was no overlap between Group 3 and the Ultimpro cohort (Groups 1 and 2). All data were collected via retrospective chart review using institutional databases. Only patients for whom the index ET is known were included in the analysis. In patients who conceived, only those whose outcomes (live birth or miscarriage) are known were included in the analysis. Recurrent implantation failure was defined as ≥ 3 previous consecutive high quality blastocyst transfer failures (Gardner grade ≥ 3BB). Patients with primary RIF, secondary RIF, and RIF with donor oocytes were included in Groups 1 and 2 (Table  1 ). Primary outcomes included successful pregnancy, ET failure or miscarriage. A biochemical pregnancy was defined as a positive quantitative hCG blood test without presence of an intra‐uterine sac. Clinical pregnancy was defined as the presence of an intra‐uterine gestational sac. Successful pregnancy was defined as either a ≥ 32 weeks ongoing pregnancy or live birth occurring ≥ 24 gestational weeks. Unsuccessful outcomes included miscarriage (<24 weeks gestation after serologic confirmation of pregnancy) or ET failure (hCG < 5, 10 days after ET). Secondary outcomes included maternal side effects from medication, as well as maternal pregnancy and neonatal outcomes. Patients were stratified into Group 1 (uRIF patients fitting our RIF inclusion criteria for Ultimpro testing, (Table  1 )) and Group 2 (patients not fitting our criteria) to preserve group homogeny (Figure  2 ). Index ET outcomes for patients in Group 1 were compared those of Group 3. ET outcomes for patients in Group 2 were compared to expected outcomes of non‐RIF women (stratified by maternal age at oocyte retrieval) from the national Canadian ART database (CARTR 2023) and from a representative Quebec Fertility clinic (Ovo Clinic). Categorical data were presented as percentage, continuous variables were presented as mean and minimum or maximum values. Shapiro–Wilk test was used to assess normal distribution of the quantitative parameters. The Student t ‐test or Mann–Whitney U test was used for continuous variables and the Fisher exact test for categorical data. The Wilson score interval with continuity correction was used to calculate 95% confidence intervals. Analyses were performed using R version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria). The authors have used appropriate statistical methods for analysis. Raw data are available on demand for editorial review. Multivariable analyses were conducted to compare the primary outcomes between groups using three models to account for potential confounders. In model 1, we adjusted for potential confounders (maternal age at time of oocyte retrieval (categorical as ≤35 vs. >35), AMH (normal vs. low; normal was defined as AMH level of ≥ 1 ng/mL or use of oocyte donation), number of previous ET (categorical as <5 vs. ≥ 5) and PGTA (yes/no). For models 2 and 3, we used propensity score analysis. A propensity score was estimated using a multivariable logistic regression model with maternal age at time of oocyte retrieval (used as continuous in order to optimize matching and potentially reduce bias by avoiding matching a 35 year old with a 22 year old since both would be grouped as ≤35), AMH (normal vs. low), number of previous ET (categorical <5 vs. ≥ 5), and PGTA (yes/no). For the model 2, we adjusted for the propensity score (used as a covariate). In model 3, we conducted a propensity score matching analysis. Matching was performed using the SAS macro match.sas and was based on a caliper width of 0.1‐fold the SD of the logit‐transformed propensity scores. The association of the outcome with intervention versus control groups in the matched sample was examined using logistic regression analyses using Generalized Estimated Equations (GEE) with an unstructured correlation. This study was approved by the McGill University Health Center (MUHC) ethics board, study number MUHC REB #2025‐10617.

Between January 1, 2021 and May 31, 2023, a total of 128 patients underwent Ultimpro testing. Seven patients were excluded from analysis (five patients were lost to follow‐up, one patient resorted to a gestational carrier and one patient's last embryo degenerated at thaw). All included patients adhered strictly to Ultimpro recommendations and were followed in clinic until cessation of treatment. Group 1 was comprised of 93 patients with unexplained RIF fitting our criteria (Table  1 ). Group 2 was comprised of 28 patients who did not fit our RIF criteria as follows (each patient may have several): age >42 with autologous oocytes or >45 with oocyte donation ( n  = 5), thin endometrium ( n  = 5), poor quality embryos ( n  = 14), 35 ( n  = 1), partner with balanced translocation, and no pre‐implantation genetic screening of embryos ( n  = 1) (Figure  2 ). Between January 2020 and December 2021, 1061 patient files from the MUHC Reproductive Center and MRIC were screened. Ninety‐four patients met our uRIF inclusion criteria and had not received any previous immunomodulatory treatment; they were included in our control group (Group 3) as previously described [ 32 ] (Figure  2 ). Patients undergoing oocyte donation were excluded to preserve group homogeneity. Baseline patient characteristics are presented in Table  2 . Overall, patients in Group 1 had a higher median maternal age at oocyte retrieval (35 vs. 34, p  < 0.001) and a higher mean number of previous high quality blastocyst transfers (4.74 vs. 4.13, p  = 0.012) than Group 3. AMH levels were similar between Group 1 and Group 3. Patients in Group 1 had higher rates of pre‐implantation genetic screening (PGS) compared to Group 3 (34/93 [36.5%] vs. 13/94 [13.8%], p  = 0.0004). Baseline patient characteristics. Group 1 N  = 93 Group 2 N  = 28 Group 3 N  = 94 1) 0.0001 a 2) 0.0001 a 3) 0.17 1) 0.037 a 2) 0.036 a 3) 0.023 a 1) 0.22 2) 0.16 3) 0.91 1) <0.0001 a 2) <0.0001 a 3) 0.03 a 1) 0.0014 a 2) 0.0037 a 3) 0.54 1) 0.013 a 2) 0.02 a 3) 0.43 1) 0.5 2) 0.27 3) 0.55 1) 0.0096 a 2) 1 3) 0.0004 a Notes: All p values were calculated using the Student t ‐test for continuous variables (expressed as median and (minimum–maximum value)) or the Mann U Whitney or Student t ‐test for categorical variables (expressed as median and range (minimum‐maximum value)). p values are recorded as: 1) Comparison between Group 1 and Group 2. 2) Comparison between Group 2 and Group 3. 3) Comparison between Group 1 and Group 3. In bold: variables for which p values were calculated. Abbreviation: AMH, anti‐mullerian hormone; BMI, body mass index; ET, embryo transfer; PCOS, polycystic ovarian syndrome. Statistically significant. Only high‐quality blastocyst (≥ 3BB according to Gardner's scale). Total number of previous embryo transfers including cleavage stage embryos, morula and lesser quality blastocysts (<3BB according to Gardner's scale). AMH was not recorded for the 12 patients undergoing oocyte donation, mean AMH is calculated for the 15 patients undergoing autologous embryo transfer (one missing data). Patients in Group 2 had a higher mean maternal age at ET than both Group 1 and Group 3 (39.7 vs. 36.2, p  = 0.002; 39.7 vs. 34, p  = 0.0001, respectively), lower AMH levels for those undergoing autologous ET (1.65 vs. 3.22, p  = 0.01; 1.65 vs. 3.5, p  = 0.002, respectively), a lower number of previously failed high quality ETs (1.92 vs. 4.74, p<0.0001; 1.92 vs. 4.13, p<0.0001, respectively) and lower total number of previously failed ETs (3.6 vs. 4.74, p  = 0.01; 3.6 vs. 4.85, p  = 0.0012). Mean maternal BMI and proportion of patients with unexplained infertility was similar between all groups. In Group 3, there were 31 live births (31/94, 33%, 95% CI [0.24–0.43]). This is lower than expected compared to 2023 Canadian statistics (OPR per blastocyst transfer 46.2%) and compared to outcomes from a large representative Quebec Fertility clinic (Ovo Clinic, OPR 53.8%) (Table S3 ). In Group 1, there were 46 successful pregnancies (46/93, 49.5% [95% CI: 0.39–0.59]), which was higher than in Group 3 (31/94, 33% [95% CI: 0.24–0.43]; p  = 0.026) (Table  3 ); unadjusted OR 1.99 (95% CI: 1.10–3.59) (Table  5 ) and comparable to expected outcomes according to Canadian 2023 CARTR statistics and to outcomes from a large representative Quebec Fertility clinic (Table S3 ). Because the rate of successful pregnancy per ET likely decreases with the number of previously failed ETs [ 2 ], we stratified patients by RIF severity. Only patients with more severe RIF phenotypes (≥ 5 previously failed blastocyst transfers) had significantly higher success rates compared to control (21/38 [55.2%, 95% CI: 0.38–0.71]) vs. 5/25 [20%], p  = 0.0084). Patients with less severe phenotypes trended toward more successful pregnancies than control, but this was not significant. Without stratifying for RIF severity, patients in the underactive endometrial immune profile had significantly more successful pregnancies than Group 3 (17/26 [64.5%] vs. 30/94 [32%], p  = 0.003). This was not the case for the other three profiles (Table  3 ). Notably, PGS nor maternal age did not affect successful Ultimpro outcomes (Tables  2 and  5 ). Cohort outcomes stratified by severity of RIF and Ultimpro profile. Over‐active Underactive Mixed Balanced 19/41 (46.3) 17/26 (65.4) 7/18 (38.9) 3/8 (37.5) 0.17 0.006 b 0.4 1 Successful pregnancy (%) 95% confidence interval 9/28 (32.1) (0.15–0.52) Over‐active Underactive Mixed Balanced 3/8 (37.5) 2/7 (28.6) 3/11 (27.3) 1/2 (50) Notes: p values were calculated with Fisher exact test. Patients in Group 1 were compared to Group 3. Patients in Group 2 were compared to national averages of ongoing pregnancy per embryo transfer (1). Only high‐quality blastocysts. Statistically significant. Multivariable analyses were conducted to compare rates of successful pregnancy between Groups 1 and 3 (Table  5 ). In Model 1, we adjusted for potential cofounders (maternal age at oocyte retrieval, AMH level, number of previous ETs and pre‐implantation genetic testing for aneuploidy (PGT‐A). In Model 2, we adjusted for the propensity score (used as a covariate) which included maternal age as a continuous variable as well as AMH level, number of previous ET failures and PGT‐A. For both models, we were able to demonstrate a beneficial effect for Ultimpro testing with adjusted OR 2.00 (95% CI: 1.10–3.99) and 1.99 (95% CI: 1.10–3.99), respectively (Table  5 ). Propensity score analysis was also conducted. Unmatched patients were excluded from the analysis and 48 patients were matched. As expected, this loss of sample size affected the statistical significance of our results, but nonetheless, we retain the size effect of our previous analyses (OR 2.28 [95% CI 0.87–5.97], Table  5 ). Miscarriage rates were higher in the over‐active endometrial immune profile group (9/41, 22%) compared to the other three groups (underactive 0/36 (0%), mixed 3/18 (16.7%), balanced 1/8 (12.5%) endometrial immune profiles), but this was not significant ( p  = 0.73 and p  = 1, respectively, for mixed and balanced endometrial immune profiles). Of note, in the over‐active endometrial immune profile group, there were three anembryonic pregnancies and one aneuploidy on product of conception testing (47XX, +21); this may reflect miscarriage due to embryo aneuploidy rather than Ultimpro protocol failure. Compared to miscarriage rates in the general population, which range from 15% to 25% depending on maternal age [ 33 ] miscarriage rates in patients undergoing Ultimpro did not differ significantly (Table  4 ). In Group 1, maternal characteristics were similar for patients who had a successful pregnancy, miscarried or failed their ET; both patients with successful pregnancies and miscarriages were more likely to have unexplained infertility compared to patients who failed to conceive post index ET (Table  S1 ). Ultimpro Cohort outcomes by immune profile. Three anembryonic pregnancies and one aneuploid pregnancy on products of conceptions testing (47, XX + 21). Adjusted analysis. Intervention ( N  = 93) n (%) Controls ( N  = 94) n (%) Model 1 aOR (95% CI) Model 2 aOR (95% CI) Intervention ( N  = 48) n (%) Controls ( N  = 48) n (%) Model 3 OR (95% CI) Notes: Model 1: Adjusted for potential confounders (maternal age at time of oocyte retrieval (categorical as ≤35 vs. >35), AMH (normal vs. low), number of previous embryo transfer (categorical <5 vs. ≥ 5). and PGTA (yes/no). Model 2: Adjusted for the propensity score (used as a covariate). Propensity score included maternal age at time of oocyte transfer (used as continuous), AMH (normal vs. low), number of previous embryo transfer (categorical <5 vs. ≥ 5), and PGTA (yes/no). **Number of patients in intervention = 92/number of controls = 91. Model 3: Propensity score matched analysis using logistic regression analyses using Generalized Estimated Equations (GEE) with an unstructured correlation. **Number of patients in intervention = 48/number of controls = 48. In Group 2, there were nine successful pregnancies (9/28, 32.1%), 14 patients failed ET (50%), and five patients miscarried (17.9%). Since patients in Group 2 did not have RIF, we compared their outcomes to national Canadian averages of ongoing clinical pregnancy per ET (2023 CARTR statistics). In patients not fitting our criteria for RIF, Ultimpro did not improve rates of successful pregnancies (Table  3 , Table S3 ). Like Group 1, the miscarriage rate in Group 2 is comparable to age‐expected rates in the general population. For Group 1, 10 patients reported pregnancy complications (10/46, 21.7%). Median maternal age was 36 (range 29–42), median number of previously failed ET was 5 (range 3–8) and median BMI was 21.7 (range 18.6–30.1); one patient had a normal profile (1/8, 12.5%), four patients had an overactive profile (4/41, 9.7%), four patients had an underactive profile (4/26, 15.4%) and one patient had a mixed profile (1/18, 5.5%). Birthweight was available for 30 neonates (Table S4 ). For Group 2, six uncomplicated pregnancies and live births were recorded. Birthweights were available for five neonates; no neonatal complications were reported. Of the three ongoing pregnancies, one patient (overactive profile) has developed steroid‐induced GDB in her first trimester requiring insulin, she is currently 37 weeks pregnant. For the Group 3, birthweights were available for the six patients that delivered within the MUHC. Pregnancy complications were reported in four patients (4/31 (13%)) (Table S4 ).

Unexplained recurrent implantation failure will affect up to 10% of couples undergoing IVF [ 34 , 35 ]. Although the expected OPR per blastocyst transfer approach 46% in Canada (2023 CARTR report), this percentage is likely lower for patients after a uRIF diagnosis, with historical cohorts (including our control group) estimating 12%–30% per subsequent ET [ 35 , 36 ]. Immune‐mediated RIF (IM‐RIF) is mainly a diagnosis of exclusion based on lack of alternative explanation, failure of previous therapy, and sometimes on variances in peripheral blood NK cell testing. Because of limited ability to select patients with a diagnosis of IM‐RIF, studies evaluating the performance of immunomodulatory therapies to improve RIF outcomes have not shown convincing evidence of therapeutic benefit [ 7 ]. By assessing endometrial NK cells and factors known to influence their behavior during the WOI, Ultimpro may thus provide a more representative view of in‐situ endometrial immune events contributing to IM‐RIF and a better personalized approach to immune therapy. However, Ultimpro is an invasive test carrying low albeit present risks (infection, uterine perforation, pain), is expensive (1600–2000$) and can cause delays for ET (1–3 months). Although studies performed to date seem encouraging, they are observational and sometimes lack appropriate controls [ 27 , 28 , 29 , 37 ]. Furthermore, studies are all performed by the company marketing Ultimpro, providing potential biases. Thus, there is a need to (1) identify if and which patients would most benefit from Ultimpro testing and (2) determine if the risk‐cost/benefit ratio is clinically reasonable. The McGill University Health Center Reproductive Immunology Clinic (MRIC) is the Quebec reference center for patients with suspected IM‐RIF. In collaboration with referring colleagues from Quebec‐wide fertility clinics, we decided to pilot Ultimpro testing as of January 2021. Our criteria for Ultimpro testing was stringent (Table  1 ), as we aimed to identify patients most likely to have an underlying diagnosis of IM‐RIF by excluding cofounding variables which can contribute to implantation failure (Group 1). We accounted for embryo quality (only patients whose index ET was a high blastocyst (Garner grade ≥ 3BB) were included, endometrial preparation, body mass index and maternal age at oocyte retrieval and ET as well as presence of pre‐implantation genetic screening (PGS). We did not assess endometrial receptivity as ERA (Igenomix) testing do not seem to improve pregnancy rates [ 38 , 39 , 40 ] and have fallen out of favor in Quebec fertility clinics. Our control population (Group 3) was selected using the same criteria as we wanted to determine ET outcomes in patients with uRIF without additional immune treatment. Overall LBR per index ET was lower for Group 3 (31/94, 33%) compared to Canadian (2023 CARTR Statistics) OPR per ET (Table S3 ) and in line with previous reports [ 35 , 36 ]. We also noted an inverse correlation between probability of live birth and number of previously failed ETs (Table  3 ). There were some differences between Group 1 and Group 3. Maternal age at oocyte retrieval (35 vs. 34, p  = 0.023) and the mean number of previous blastocyst transfer failures before index ET (4.74 vs. 4.14, p  = 0.012) was higher in Group 1 compared to Group 3; Group 1 likely represents a group with a worse prognosis than Group 3. Nevertheless, patients treated according to Ultimpro recommendations had higher rates of successful pregnancies overall (46/93 [49.5%] vs. 31/94 [33%], p  = 0.026). Importantly, patients in Group 1 had higher rates of PGS compared to Group 3 (34/93 [36.5%] vs. 13/94 [13.8%], p  = 0.0004). However, PGS did significantly affect pregnancy outcomes (Table S5 ). When stratifying by severity of RIF phenotypes, only patients with ≥ 5 previously failed blastocyst transfers benefitted significantly from Ultimpro (Table  3 ). Although there was a trend for benefit in groups <5 previous blastocyst transfer failures, this was not significant. It is possible that only patients with a more severe RIF phenotype have true immune‐mediated RIF and correcting the underlying deficit alone will improve outcomes. Therefore, the significantly higher pregnancy rates seen in Group 1 may be solely driven by patients with ≥ 5 previous ET failures. Another explanation is that our study is underpowered to detect differences in outcomes between Group 1 and Group 3 for patients with milder RIF phenotypes; the baseline LBR in the control group with <5 previous ET failures is relatively high (37.6%), and more patients would be needed to detect differences. Another interesting finding in this study is that, without stratifying for RIF severity, patients with underactive endometrial immune profiles had significantly more successful pregnancies than controls (17/26 [65.4%] vs. 30/94 [32%], p  = 0.003). Ultimpro recommendations for such a profile include mid‐luteal phase scratching on cycle prior to ET and luteal phase hCG support. Although previous studies have shown that endometrial scratch therapy may have marginal effects for improving IVF outcomes [ 7 , 41 ], perhaps this intervention would be most beneficial in well‐selected patients. Additionally, only 60% of patients in Group 1 received glucocorticoids (all patients with overactive and mixed endometrial immune profiles) (Figure  1 , Table  4 ). Glucocorticoids would likely have suppressed the normal endometrial immune events required for implantation in patients with balanced and underactive endometrial immune profiles, thus decreasing implantation rates. This is an important concept as previous studies evaluating the utility of glucocorticoids for RIF did not select patients based on immune profiles and may have inappropriately concluded that steroids are not a useful adjunct for uRIF [ 42 , 43 ]. Lastly, we included all patients who had Ultimpro testing but did not meet our testing criteria as a separate group (Group 2). These patients were excluded because they either did not have RIF ( 35, balanced parental translocation). As expected, treatment recommendations according to Ultimpro results did not improve the rate of ongoing pregnancy per ET compared to overall Canadian statistics. Interestingly, the complication rate was higher in Group 1 compared to Group 3 (Table S4 ). Two patients receiving glucocorticoids (GC) developed gestational diabetes (GDB) at 28 weeks requiring insulin (two other patients with an underactive profile developed GDB treated with diet, likely unrelated to Ultimpro treatment as there was no GC exposure). In patients receiving GC, steroid taper is performed at 6 weeks gestation (lasting 3–4 weeks). No patients receiving GC developed GDB in their first trimester. We cannot exclude that GC treatment in the first trimester predisposes to GDB in the third trimester. None of the patients in our cohort delivered macrosomic infants, no deliveries were complicated by shoulder dystocia. Placenta‐related complications were higher in Group 1. Two pregnancies were complicated by intra‐uterine growth restriction (IUGR). However, one patient had a balanced immune profile and did not receive any adjunctive treatment, and the other (underactive profile) delivered a neonate later diagnosed with a monogenic disorder explaining the IUGR. Two patients developed term preeclampsia without severity features (one with an underactive profile with conception post endometrial scratch and the other with an overactive profile receiving GC early in pregnancy). Lastly, one patient (overactive profile) had a stillbirth. She presented to hospital with acute fatty liver of pregnancy, thrombocytopenia, and acute kidney failure at term. She was on salicylic acid and low molecular weight heparin (known positive lupus anticoagulant, no previous history of miscarriages or clotting) and had received GC in early pregnancy. Her case was reviewed in a multidisciplinary morbidity and mortality meeting, and it was concluded that her medications did not contribute to her adverse pregnancy outcome. Studies performed in patients with recurrent pregnancy loss do not show that glucocorticoid use in early pregnancy adversely affects pregnancy outcomes [ 44 , 45 ]; but if used at higher doses and continued throughout pregnancy, there are increased risks of gestational diabetes and pre‐term birth [ 46 ]. Although pre‐conception/early pregnancy glucocorticoids administration is likely safe, patients should be counseled and stringently monitored for gestational diabetes. Although we did not observe an increased risk of pre‐term birth in Group 1, patients should be counseled that this is a possibility with GC use early in pregnancy. Similarly, endometrial scratching prior to ET does not seem to significantly affect pregnancy outcomes [ 47 ]. It is therefore unlikely that the adverse pregnancy or obstetrical outcomes in Group 1 are attributable to treatments administered in early pregnancy, but not impossible. However, whether the higher complication rate in Group 1 is due to the underlying etiology of the infertility or to the treatment administered prior to conception/early pregnancy remains to be determined. Indeed, patients with uRIF may represent an at‐risk obstetric population irrespective of treatment administered and should be treated as such once pregnant. Unfortunately, longitudinal studies evaluating pregnancy outcomes in patients with uRIF are lacking. Perhaps an unfavorable endometrial microenvironment at conception may translate into placental pathology later in pregnancy; treatment aiming to correct the endometrial defect at implantation may not be sufficient to prevent adverse pregnancy outcomes. Long term follow‐up (from conception until parturition), particularly for patients with more severe RIF phenotypes, should be incorporated in all studies assessing treatment efficacy in patients with uRIF. Our study has both strengths and limitations. In terms of strengths, patients were stringently selected, ensuring the most homogenous patient population possible. Patients are followed until delivery, providing the opportunity to describe not only OPR, but also LBR and pregnancy/neonatal complications. In terms of limitations, it is a retrospective, observational study which may hold undetected biases. Furthermore, the interpretation of test results is performed by Ultimpro using proprietary norms. Although the Ultimpro test was developed respecting a rigorous scientific method [ 19 , 23 , 48 , 49 , 50 , 51 ] and clinical implementation [ 27 , 28 , 30 ], thresholds for identifying immune dysregulation are not detailed, and there has been no independent validation of these thresholds or proof of inter‐laboratory reproducibility. Treatment recommendations are provided by MatriceLab Innove according to established algorithms and not by the treating physician. Group 3 is small and unmatched (reflecting the low incidence of true RIF (≤5%) [ 6 ]), but patients were selected with the same criteria, thus providing an adequate comparator for Group 1. Although most maternal characteristics were similar between Group 1 and 3, Group 1 had a higher number of mean ET failures prior to index pregnancy. Group 1 may thus represent a patient population with worse prognosis (which would emphasize the positive effect of Ultimpro). Group 3 was selected from an earlier timepoint (2020–2021) than Group 1 (2021–2023), but success rates per transferrable embryo did not significantly change during this time (2020–2023 Ovo/MUHC Reproductive Center statistics, CARTR 2020–2023 statistics). A matched control group within the same timeframe as the Ultimpro cohort would have been ideal. However, since 2021–2022, most Quebec fertility clinics refer patients with ≥ 3 ET failures to the MRIC. It would now be difficult to find a comparable RIF control cohort within Quebec without crossover with the intervention group. Additionally, patients are referred to the MRIC from over 10 different fertility clinics in Quebec and Ontario, we were unable to include information of IVF cycle parameters, stimulation/ET protocols or conditions of embryo culture as this data are often not available in the patient's chart. In this study, we cannot account for variability in laboratory practices and their effect on ET success rates. Of note, the rate of pre‐implantation genetic screening (PGS) was higher in Group 1 than Group 3 but did not impact outcomes (Table S5 ). Next, while overall patient numbers were robust (93 patients in Group 1 and 94 patients in Group 3), the number of patients with moderate‐severe RIF (≥ 4 previously failed high quality blastocyst transfers) was lower. Nevertheless, we were able to show a beneficial effect of Ultimpro for such patients. Another limitation is that the rate of pre‐implantation genetic testing for embryo aneuploidy was low (PGT‐A) (36% in Group 1 and 13.8% in the Control group). Although “true RIF” should incorporate ET failure with euploid embryos, PGT‐A testing is controversial for patients <35 as OPR is similar per ET with and without PGT‐A (2023 CARTR statistics). Additionally, because PGT‐A is not covered in Quebec, it is rarely performed for first IVF cycles or in patients <38 years of age. Further, we did not perform products of conception (POC) testing in all patients who miscarried and cannot determine if miscarriages occurring with Ultimpro recommendations represent a treatment failure or embryo aneuploidy. Most miscarriages were biochemical or <6 weeks; one patient had an 8‐week loss after positive viability scan but lost her pregnancy spontaneously at home. Only one patient who miscarried was able to perform POC testing and had a proven aneuploid loss (47XX, +21). Lastly, we are using OPR (pregnancy with fetal cardiac activity at 6–8‐week ultrasound) from 2023 CARTR/2023 Ovo Clinic statistics and not LBR as a comparator for Ultimpro outcomes. Ideally, we would be using LBR; OPR may overestimate favorable IVF outcomes. However, there is a 1–2‐year delay in the reporting IVF cycle birth outcomes and 10%–20% of IVF cycles for which birth outcomes are not reported. Based on previous publications [ 52 ] and discussion amongst our group, we decided that OPR was a robust enough metric to use as a comparator for Ultimpro outcomes. Although our results are encouraging for Ultimpro use in patients with uRIF, it is too early to recommend routine incorporation of Ultimpro in IVF clinics as standard practice. A prospective validation cohort in our Center is currently underway, but a randomized controlled trial (RCT) is needed to define the role of Ultimpro in the management of uRIF patients. However, RCTs are expensive and time consuming; they must be well designed and adequately powered to maximize the information gained. Therefore, observational studies such as this one are important in identifying key variables and populations to target for an eventual RCT. Ultimpro is likely going to prove a useful clinical adjunct for select RIF patients, but it may be detrimental to others (specifically patients without RIF or patients with explained RIF). Potential risks of biopsy, adverse pregnancy outcomes related to treatment and cost of Ultimpro testing must be carefully weighed against perceived benefits; all of which can be addressed by a RCT. Although Ultimpro currently assesses NK cell markers, the uterine immune microenvironment is also enriched in other key immune cell types such as T, B, and dendritic cells. Broadening the scope of the endometrial immune phenotyping may further define relevant immune profiles and their use in improved diagnostic accuracy. Indeed, studies using genomic approaches such as single cell transcriptomics to profile the underlying immune pathways in mid‐luteal phase endometrial samples are currently underway. These may eventually validate Ultimpro biomarkers and broad diagnostic categories but would also shed more light into normal immune events permitting implantation as well as pathogenic events underlying implantation failure. Although our independent pilot study shows promising preliminary results, it may be too soon to recommend incorporation of Ultimpro testing into mainstream clinical practice. Awaiting the results of an independent multi‐centered randomized controlled trial, Ultimpro should be used within a research setting (registry, longitudinal cohort study). This permits periodic practice audits for efficacy but more importantly, for drug side effects as well as maternal, pregnancy and fetal outcomes. Lastly, as physicians who deal with vulnerable and often desperate couples, new technology offering clinical guidance and improved pregnancy rates is attractive. However, careful consideration of whom this testing is offered to is of the utmost importance.

Infertility is a significant health problem, affecting 1/6 couples globally (2023 WHO statistics); up to 10% of reproductive age women in North America seek medical assistance to conceive [ 1 ]. In vitro fertilization has revolutionized the treatment of infertility; however, success rates remain low with 20%–30% live birth rate (LBR) per embryo transfer (ET) [ 2 , 3 ]. Up to 10% of couples resorting to IVF will have recurrent implantation failure (RIF), generally defined as the absence of clinical pregnancy after 3 high quality blastocyst transfers [ 4 ]. Over 50% of such patients will be diagnosed with unexplained RIF (uRIF) after exclusion of anatomical, hormonal and infectious causes, further compounding the psychological and financial burdens these couples face [ 5 ]. Although embryo aneuploidy likely explains a high proportion of uRIF cases [ 6 ], additional factors must be considered especially in young couples with high quality embryos and normal parental karyotypes. Indeed, immune mediated RIF (IM‐RIF) is increasingly recognized as a host‐derived factor that can contribute to reproductive failure. An appropriate endometrial immune response is necessary for organized decidualization, endometrial receptivity, and efficient embryo implantation [ 7 ]. Multiple immune cell types are involved in coordinating this response, with the natural killer cell (NK cell) being the predominant leukocyte in the endometrium during the window of implantation [ 8 ] (Figure S1 ). Uterine NK cells (uNK cells) exhibit significant differences compared to their peripheral blood counterparts (pNK cells). They are CD56 bright CD16 dim , express a low cytotoxic profile and secrete different cytokines and growth factors [ 9 ]. During the luteal phase, uNK cells secrete pro‐angiogenic factors and matrix metalloproteases, participating in spiral artery remodeling and loosening the endometrial extracellular matrix to facilitate embryo implantation [ 8 , 10 ]. The uNK cells are also thought to be involved in embryo recognition, sensing embryo quality [ 11 ] and guiding trophoblast invasion into the endometrium [ 9 , 12 ]. Therefore, it is possible that uNK cell dysfunction (both overactive and insufficient NK cell responses) during implantation contributes to reproductive failure [ 13 ]. Although peripheral blood NK cells do not reflect uNK cell populations [ 14 , 15 ], nor correlate with reproductive outcomes [ 16 , 17 ]; endometrial uNK cell evaluation may offer more insight into implantation failure [ 18 ]. The Ultimpro test (Matrice Lab Innove, Paris, France) offers targeted uNK cell testing to identify immune dysregulation contributing to reproductive failure. Briefly, by studying differential transcriptional [ 19 ] and protein expression profiles [ 20 , 21 , 22 , 23 ] between patients with unexplained RIF and fertile controls, the company has developed biomarkers to assess uNK function and behavior. Endometrial biopsies are performed in the mid‐luteal phase (8–11 days after LH surge in monitored natural cycles and on Days 6–8 of progesterone in HRT cycles). Immunohistochemistry is performed on formalin preserved tissue sections using an anti‐CD56 antibody. The rest of the sample is used for RNA extraction and an RT‐PCR panel is performed for the markers IL‐15, IL‐18, Fn‐14, and TWEAK. CD56 is a uNK cell surface marker and is used to quantitate the amount of uNK cells present in each biopsy sample. The cytokine IL‐15 is involved in NK cell mobilization, proliferation, and maturation, whereas IL‐18 is a pro‐angiogenic cytokine involved in regulating uNK cell phenotype, endometrial vascular remodeling and trophoblast invasion [ 24 ]. When in excess, IL‐15 and IL‐18 can stimulate NK cell cytotoxicity and contribute to disorganized spiral artery remodeling. TNF Weak inducer of Apoptosis (TWEAK) and its receptor, fibroblast growth factor‐inducible molecule 14 (Fn‐14), have been proposed to suppress NK cell cytotoxicity and regulate IL‐15 and ‐18 when in excess [ 25 ]. The TWEAK/Fn14 axis is also proposed to be involved in tissue regeneration and repair; excessive TWEAK/Fn14 can contribute to tissue damage and aberrant endometrial remodeling [ 26 ]. High IL‐18/TWEAK and/or IL‐15/Fn14 mRNA levels and/or elevated CD56 staining characterize the “over‐active” endometrial immune profile where the endometrial microenvironment is pro‐inflammatory, uNK cell recruitment is excessive and vascular remodeling is disorganized. On the contrary, low IL‐18/TWEAK and/or IL‐15/Fn14 mRNA and/or CD56 staining characterize the “underactive” endometrial immune profile, where uNK recruitment is deficient and the endometrial micro‐environment exhibits insufficient inflammatory stimulus for normal vascular remodeling and implantation to occur. A “mixed” endometrial immune profile denotes a combination of over‐immune or underactive phenotypes whereby NK cells may be immature or insufficiently recruited, but the endometrium has inflammatory characteristics which may induce a deleterious NK cell phenotype [ 27 ]. Depending on the patient's profile, treatment recommendations to normalize immune profile for ET are made by the company and applied for subsequent ETs (Figure  1 ). Of note, CD56 levels, IL‐18/TWEAK and IL‐15/Fn14 ratios are compared to proprietary norms established by the MatriceLAB Innove company. The readout addressed to physicians provides a numerical value for these markers and whether they are higher, lower or within norm. The interpretation of individual patient results is performed by the company. Overview of Ultimpro results and treatment recommendations. CD56 levels as well as IL‐18/TWEAK and IL‐15/Fn14 ratios are compared to proprietary norms established by MatriceLab Innove. Not pictured: Ultimpro recommendations for estrogen and progesterone dosing were followed for each patient. For patients receiving prednisone, gastroprotection (omeprazole 20–40 mg PO die) and osteoprotection (vitamin D and Calcium) were administered; glycemic monitoring was performed for the duration of glucocorticoids (targets: fasting glycemia ≤5.3 mmol/L, 2 h post prandial ≤7.8 mmol/L) and glucocorticoids were gradually tapered over 3–4 weeks after the recommended administration. A retrospective cohort study [ 28 ], a prospective cohort study [ 29 ] and a randomized controlled trial [ 30 ] led by the company suggest a possible benefit of Ultimpro for patients with RIF. However, there has been no independent validation of their CD56 or cytokine ratios, limited evidence of inter‐laboratory reproducibility [ 31 ] and thus far no independent studies assessing clinical performance of the test. Although the ability to diagnose and correct immune dysregulation contributing to implantation failure would likely improve the management of uRIF patients, Ultimpro testing is expensive, invasive, and data amassed so far is insufficient to recommend widespread clinical implementation. An independent randomized controlled trial (RCT) is needed to assess the clinical utility of the test in uRIF patients but first, patient populations most likely to benefit from Ultimpro testing must be identified. In this study, we present a retrospective review of our clinic's 2.5‐year outcomes, comparing success rates of Ultimpro to “expected outcomes” from a historical cohort of patients with uRIF not having undergone Ultimpro testing and not having received immunomodulatory treatments.

No author has any conflicts of interest to declare.

Table 1: Group 1 maternal characteristics by ET outcome Table 2 : Group 1 outcomes by maternal age (at oocyte retrieval if using own oocytes and at maternal age at time of ET if using oocyte donation) Table 3 : 1) Ongoing pregnancy rates (OPR) per transferable frozen blastocyst transfer according to maternal age at oocyte retrieval (2023 CARTR statistics), 2) OPR per transferable frozen blastocyst transfer according to maternal age at oocyte retrieval (Ovo Clinic 2023), 3) Live birth rate per high quality index blastocyst transfer according to maternal age at oocyte retrieval (control group), 4) Successful pregnancy rates per index high quality blastocyst transfer (Group 1); 95% Confidence Interval calculated with the Wilson score interval with continuity correction. FET: frozen embryo transfer *For patients undergoing oocyte donation, the age of the donor at oocyte retrieval was used Table 4: Maternal, pregnancy and fetal outcomes Table 5: Outcomes by PGT‐A testing Figure 1: Immune cell contribution to endometrial function