Predicting success of intrauterine insemination using a clinically based scoring system

Homologous intrauterine insemination (IUI) with or without ovarian stimulation has historically been used as first-line treatment for infertility. Today, due to increasing per cycle success rates of over 30% in achieving pregnancy through in vitro fertilization (IVF) [1], many couples choose this option as an initial treatment. In contrast, based on a multi-center retrospective analysis from 22 European centers evaluating more than 175,000 IUIs, IUI achieved a pregnancy rate of only 8.6% per cycle, albeit with large variability between centers ranging from 4 to 13%. Nonetheless, in women with unexplained infertility < 38 years, comparable live birth rates after ovarian stimulation with IUI and IVF have been reported [2]. As IUI is a relatively simple procedure associated with far lower risks and potential complications, and costs than IVF, it would be clinically useful to predict which patients have a high chance of pregnancy with IUI.

Although models predicting pregnancy after IUI have been previously published [3‐7], they are not widely used in clinical routine practice. In fact, they are all based on semen analysis according to 1999 WHO criteria [8], and ovarian hyperstimulation aiming for two or three dominant follicles at ovulation induction (OI) prior to IUI. Considering the publication of new WHO criteria for semen analysis in 2010 [9] and the evidence showing that strict cancelation policies in case of multifollicular growth are effective in reducing the risk of multiple gestations following IUI [10‐12], these models might be outdated. Recently, another predictive model has been published by Souter et al. [13] but its focus lies on ovarian stimulation and ovulation induction in infertility due to PCOS and unexplained infertility, not on IUI.

Based on these premises, we aimed to determine the per cycle pregnancy rate following IUI as well as the cumulative pregnancy rate after repeated cycles in a retrospective cohort of 758 couples undergoing 1437 IUIs between the years of 2009 and 2017 at our department. Our objective was to identify predictive factors for achieving pregnancy, including demographic and clinical data, infertility diagnosis and sperm parameters. We aimed to develop a simple clinical score based on these routinely obtained parameters, that would facilitate the counseling of infertile patients by providing evidenced-based individualized recommendations.

Between 2009 and 2017, 758 women aged 18–45 years (mean 33.4 ± 5.5; of these, 451 women < 35 years, 307 women ≥ 35 years) collectively underwent 1,437 IUI cycles (Table 1). Of the included couples, 255 (33.6%) suffered from female-related infertility, 152 (20.1%) from reduced male fertility, and 154 (20.3%) from both female and male infertility. The remaining 197 (26.0%) couples had infertility of unknown origin. Observed female aetiologies were: anovulatory infertility (26.9%), endometriosis (10.4%), unilateral tubal factor (8.2%) and anatomical alteration (5.5%). Most women were normal weight, with a median BMI of 22.3 kg/m² (IQR 20.5, 25.8). Mean AMH was 2.40 ng/ml, with 160 (24.7%) women having an AMH < 1 ng/ml. In 51.1% of the analyzed IUI cycles, women underwent a low-dose stimulation with gonadotropins, while in 27.0% clomiphene was used (0.8% received both). In 21.1% of the cycles, IUI was performed in the natural cycle, without prior follicular stimulation. 306 (40.4%) men were diagnosed with pathospermia according to WHO criteria [9] at baseline sperm analysis. On the day of IUI, 97 (14.8%) had a TPMSC < 5 mil, 133 (20.3%) between 5 and 15 mil, and 425 (64.9%) > 15 mil.

Among included IUI cycles, 10.9% resulted in a pregnancy and 19.4% of couples achieved pregnancy after one or more treatment attempts. Most couples had one to three cycles, with only 56 (7.4%) undergoing four or more cycles before achieving pregnancy or discontinuing IUI treatment (Table 2). As a result, we observed only six pregnancies in women with four or more IUI cycles. Rather than a decrease in the per cycle success rate, this seems to reflect the low number of patients continuing beyond three cycles. In fact, the number of observed pregnancies for each cycle lay within the 5% and 95% percentiles of expected outcomes (Table 2) and Fisher’s exact test indicated no variation in cycle success rate (p = 0.673). Successful treatment with IUI was statistically significantly associated with lower female age, infertility due to anovulation or unknown female infertility reason, higher levels of AMH and TPMSC > 15 mil (Table 1). The use of different medications (clomiphene, gonadotropins) for ovarian stimulation showed no significant influence on pregnancy rates.

In this study, we were able to show that while the pregnancy rate per IUI cycle was relatively low (11%), repeated cycles led to an average overall success rate of approximately 19%. Per cycle and cumulative pregnancy rates varied considerably by couples’ characteristics. Among female factors, younger age and higher AMH were positively associated with pregnancy. Additionally, we found that women with infertility due to anovulation and those with unknown female reason for infertility had higher pregnancy rates compared to women with endometriosis, unilateral tubal factor, or anatomical alterations. TPMSC at insemination also correlated with IUI success. Based on these findings, we propose a clinical scoring system to predict a couple’s probability of achieving pregnancy. Couples are assigned 0, 1 or 2 points for each prognostic factor, adding up to a total score ranging from 0 to 5. After 3 cycles, couples in our cohort with a score of 5 had a cumulative probability of pregnancy of approximately 45%. In contrast, couples with a score of 0 only had a cumulative probability of 5% Our per cycle pregnancy rate agreed with previous studies, which reported rates between 7 and 15% [19‐23]. However, our average per patient pregnancy rate was lower than the 29–43% reported by the same authors [21, 23, 24]. This might have been due to a lower number of IUI cycles per couple in our study. On average, only 1.9 cycles were performed in our cohort compared to 2.4–2.9 cycles in previous studies. An insufficient number of treatment cycles has been shown to reduce the per patient pregnancy rate significantly [25].

Only few couples in our cohort underwent 4 or more IUI cycles and we observed only a handful of pregnancies following the 4th cycle. This seemed to reflect the low number of patients continuing past 3 cycles rather than a genuine decrease in per cycle success rate. This is in line with the results of a recent study by Muthigi et al., reporting consistent pregnancy rates up to 6 total IUI cycles [26]. In contrast, other previous studies described a plateau of IUI success rates after 3 or 4 cycles. Khalil et al. reported only marginal increases in the cumulative birth rate after 4 cycles. However, their analysis did not account for a similar decrease in the number of patients that underwent 4 or more cycles [20]. Aboulghar et al. reported a drastically decreasing pregnancy rate in cycles 4–6 (5.6%) compared to the first 3 cycles (16.4%) [27]. While the authors correctly accounted for decreasing patient numbers, their reported success rates were unusually high in early cycles and particularly low in later cycles, compared to the wider literature [20, 21, 23, 24]. Nevertheless, it remains possible that there is a small selection effect as couples with higher chances get pregnant early and drop out. Based on our data, we believe that this effect may be less pronounced than previously reported and may be partially compensated by a detailed selection of couples considered for IUI treatment.

We found four main predictive factors that should be taken into account when counseling infertility patients regarding IUI: female age, AMH, female infertility diagnosis and TPMSC. It has long been known that female fertility declines with advancing age due to a decrease both in oocyte quantity and quality [28]. Accordingly, lower IUI success rates at increasing female age have been reported by several previous studies [19, 21, 23, 24, 29‐33]. Less clear is the association between IUI success rates and AMH. This hormone has been proven to be a reliable marker of ovarian reserve [34, 35], typically declining at increasing age [36, 37]. It is, therefore, difficult to understand to what extend the observed effects of AMH are biased by age. However, AMH seems to reflect the decreasing number of oocytes more than their quality. In fact, Morin et al. [38] were able to show equivalent live birth rates per euploid embryo transfer in women with diminished ovarian reserve undergoing IVF as compared to age-matched controls, indicating that the pathophysiological mechanisms mediating a decline in oocyte quantity may differ from the ones leading to poor oocyte quality. Regarding IUI, the results reported in the literature are still conflicting, some authors describing AMH as good predictor of pregnancy rates [32, 39‐41], while other studies don’t find any difference in IUI success rates based on AMH [42‐45]. Among these, a recent retrospective cohort study showed no statistically significant difference of live birth rates in patients < 35 years with AMH < 1.0 ng/ml undergoing IUI compared to age-matched controls with AMH ≥ 1.0 ng/ml  [45]. However, their data still showed a tendency versus lower conception rates in the group with low AMH. When considering only the first IUI cycle performed per couple—as it was done in the cited study—in our cohort, we still observed lower pregnancy rates in the group of patients < 35 years with AMH < 1.0 ng/ml then in the same age group with higher AMH. In patients ≥ 35 years, this effect was less pronounced, but still present. Based on these findings and the similarity in AIC, BIC, and c-indices of Models 1 and 2, we decided to include AMH in our predictive score. In accordance with previous studies we also found higher per cycle and per patient pregnancy rates for women with unexplained infertility or anovulatory cycles compared to those with mild endometriosis, unilateral tubal factor, or anatomical alterations [20, 21, 24, 29, 46, 47]. This is clinically plausible, as endocrine disorders leading to anovulation or suboptimal follicle maturation can be overcome by ovarian stimulation used in combination with IUI, while anatomical and immunological factors cannot be compensated by IUI. Finally, sperm characteristics and more precisely TPMSC were found to be predictive of IUI success. However, due to the retrospective design, only pre-wash semen parameters were available. It has previously been reported that the minimum TPMSC for the performance of an IUI should be > 5 mil [21, 48]. Recently, Muthigi et al. found maximal pregnancy rates at IUI with a TPMSC ≥ 9 mil [26]. These data referred to TPMSC after semen preparation, whereas the correlation between pre-wash TPMSC and live birth rate after IUI was previously found to be low [49]. In contrast, in this study we observed a significant influence of pre-wash TPMSC on pregnancy rates after IUI. Future studies should further investigate this correlation, as pre-wash semen parameters are much more widely available in routine care.

Previous studies reported the number of dominant follicles as a further predictive factor for IUI success rates. Several authors described higher pregnancy rates in the presence of at least 2 preovulatory follicles [21, 24, 50]. Accordingly, the number pf preovulatory follicles is considered in all the previously published models predicting IUI success [3, 4, 6, 7]. However, in recent years multiple pregnancies have become a rising concern. It has been shown that strict cancelation policies in case of multifollicular growth are effective in reducing the risk of multiple gestations following IUI [10‐12]. These findings limit the applicability of the cited predictive models. Based on our national guidelines [51], our institute cancels all cycles with recruitment of > 2 follicles > 16 mm. We, therefore, a priori excluded the number of leading follicles from our IUI success score. In an additional univariate sensitivity analysis (not shown), pregnancy rate was not significantly influenced by the number of follicles. Subgroup analysis showed a significant difference between the average number of leading follicles in successful versus unsuccessful IUI cycles only for patients with endometriosis (unsuccessful IUI cycle: 1.1 ± 0.4 leading follicles, successful IUI cycle: 1.5 ± 0.5; p = 0.041). It is possible that in case of mild endometriosis the presence of more than 1 leading follicle at OI and IUI increases the chances of pregnancy. This is plausible as endometriosis has been reported to be associated with reduced oocyte quality [52]. However, the number of patients with endometriosis in our cohort was very small (n = 74), limiting the interpretability of this finding.

While previous studies found higher pregnancy rates when IUI was preceded by ovarian stimulation with gonadotropins compared to clomiphene or letrozole [3, 4, 53], we didn’t observe significant differences in IUI success based on stimulation protocol. This might be related to differences in the study population, as different stimulating agents have been shown to be superior depending on female infertility diagnosis [12, 53‐55]. Moreover, previous studies reporting better IUI outcomes with gonadotropins did not control for the number of preovulatory follicles [53]. This is important as -depending on their dosage—gonadotropins can lead to multifollicular growth more easily, possibly introducing bias into this finding.