Although birth control measures are highlighted as one of the major reproductive health concerns in third world countries, the population suffering from infertility often goes unnoticed. Globally, 12.4% of women who are trying to conceive in the age range of 20–44 years are infertile, accounting for both primary and secondary infertility; south Asia is one of regions with the highest prevalence of infertility ¹ . Despite the fact that about 16,700 infertility cases are enrolled annually in the handful number of facilities available in Bangladesh and many more of the cases remain unreported, the medical, social, financial and psychological burden associated with the condition is almost always under-appreciated in health statistics ² . There is no account of how many infertile couples seek help from fraudulent and pseudoscientific agencies in the country, which may add significantly to the burden ³ . The high cost of infertility treatment happens to be a major barrier for aspiring couples to seek medical help in Bangladesh ² . Failure rates as high as 60%, even in the most developed countries, make the application of in vitro fertilization (IVF) and other assisted reproductive technologies an expensive gamble ⁴ . Moreover, the cause of infertility remains unexplained in 10.95% of cases ² .

To find a solution to this problem, a research group in Spain analyzed mRNA from endometrial biopsy samples of expecting women and identified 238 genes that play a crucial role in implantation ⁵ . Using the results, they developed a customized mRNA array and bioinformatic tool for testing expression of those 238 genes to predict not only the possibility of implantation failure but also the personalized window of implantation (WOI) during which the chance of a successful implantation is maximum. That is called endometrial receptivity array (ERA). As good as it may sound, ERA is not without its own shortcomings. Firstly, it is quite expensive ⁶ . Secondly, availability is limited, only at a handful of institutes in Spain, Japan, the USA and a few other countries. Thirdly and most importantly, even if we can improve the odds for success by analyzing all 238 genes’ expression, the results are not translatable in terms of targeted therapy of infertility because there are simply too many targets. Despite these limitations, there has been no other tool as reliable and precise as ERA to date to predict and improve IVF outcome (i.e., success/failure) ⁷ . Therefore, efforts are being made to discover clever work-arounds to cut both the cost and number of targets with little compromise in terms of predictive values. ERA-directed proteomic studies of uterine biopsies from both groups of women (with implantation success and failure) indicate that secretory endometrium expresses certain cellular adhesion molecules that firmly attach the blastocyst at the site of implantation ⁸ . This cell-cell interaction is pivotal not only for successful implantation but also for maintenance of pregnancy. Proteomics provide a list of few pathways that have a handful of protein molecules ⁹ . These are the key players of implantation. So far, five to ten proteins, including mucin 1 (MUC1) and E-cadherin, have shown potential to constitute a panel for predictive screening of implantation failure ¹⁰ .

The peculiarity of MUC1 is that it acts as a ‘guiding molecule’ for the blastocyst during implantation in order to direct it to the implantation site of the decidua ⁸ . In receptive endometrium, MUC1 expression is generally high, especially in the endometrium during secretory phase, compared to non-receptive endometrium ^{11, 12} . Interestingly, there is always a small patch of decidua where the receptive endometrium almost completely lacks MUC1 expression during the WOI and that marks the site of implantation. Hence the connotation of a ‘guiding molecule’, as if the blastocyst wanders about the decidua until it finds the designated spot to implant marked by the absence of MUC1 ^{8, 13} . In case of non-receptive endometrium as in IVF failures or recurrent spontaneous abortions, global down-regulation of MUC1 throughout the whole of the endometrium fails to elicit the appropriate signal for implantation. However, it should be noted that if implantation is successful then MUC1 expression is quickly depleted and therefore it is one of the negative stains in normal placenta ¹⁴ .

Cadherins constitute a group of glycoproteins responsible for the calcium-dependent cell-cell adhesion mechanism. It has been postulated that E-cadherin exhibits dual attributes consisting of an increased expression during secretory phase and early implantation: epithelial cell adhesiveness by E-cadherin is controlled by intracellular calcium and rising progesterone levels induce calcitonin expression and thus increase the concentration of intracellular calcium, which then suppresses E-cadherin expression at cellular contact sites. It corresponds to the adhesiveness of decidua-blastocyst interface while a subsequent spell of its down-regulation corresponds to epithelial dissociation and trophoblastic invasion ⁸ . Although cyclical variation in mRNA levels of E-cadherin is not reflected at protein level as detected by IHC, there are indeed immunohistochemically detectable differences of E-cadherin expression between receptive and non-receptive endometrium ¹⁵ . Interestingly, progesterone level has somewhat inverse relationship with E-cadherin expression ^{16, 17} . Moreover, it has been demonstrated that targeted mutations in E-cadherin gene are associated with implantation failure ¹⁸ .

We sought to evaluate in this study if there is any correlation between IVF outcome and semi-quantitative immunohistochemistry scores (H-score) of MUC1 and E-cadherin in endometrial epithelium and stroma.

Results are prepared in four steps. Firstly, sorting and matching of case and control groups are done for probable confounding variables. Secondly, routine histomorphological evaluation of endometrial biopsy samples was done which includes endometrial dating. Thirdly, immunohistochemical markers in endometrial biopsy samples were evaluated with H-scores and optimum cutoffs were estimated. Lastly, correlations of immunohistochemical markers with IVF outcomes were calculated and statistical inferences were made. In summary, correlations between IVF outcome and H-scores of immunohistochemical markers MUC1 and E-cadherin are weakly positive and statistically insignificant.

Out of 21 patients, 17 patients (81%) had IVF failure (not pregnant) and four patients (19%) had IVF success (pregnant). After case-control categorization, the former group was considered case and the latter group was control. All the participants were from middle class to upper middle class according to socioeconomic strata. The women of the case group had a mean age of 34.00 ± 2.31 years while the mean age of the control group was 33.75 ± 4.79 years; the difference between the two groups were not statistically significant. There were no significant difference between them in terms of BMI, FSH, LH, FSH:LH ratio, menstrual and obstetric history which were considered to be the potential confounding factors. Their husbands were all within normal range of relevant reproductive parameters. None of the patients used any contraceptive method ever, except barrier method, and actively trying to conceive for at least two years. Degree of expression of MUC1 and E-cadherin as detected by immunohistochemistry was considered to be the exposure (H-scores ranging from 0-300) and post-IVF pregnancy status was considered as the outcome (pregnant or non-pregnant) in the present study.

It is interesting to note that a peculiar staining pattern was observed with E-cadherin where the surface epithelia stained with markedly greater intensity than the glandular epithelia ( Figure 2). Most of the samples exhibit similar staining pattern with E-cadherin.

Comparison between case and control groups by unpaired t-test show statistically non-significant difference for mean H-scores of MUC1 as well as E-cadherin although MUC1 has a larger effect size (0.91) compared to that of E-cadherin (0.69). Calculation of inter-observer variation between two independent observers (Pathology Consultants) was measured on 10 randomly chosen tissue sections out of 21 samples by Chronbach’s alpha correlation technique ²¹ . Inter-observer variation was found to be within acceptable limits for each of the stains ²² .

Due to the variability of data obtained from two subsets of samples, namely, collected before 20 ^th day and collected on or after 20 ^th day, the statistical analysis for derivation of the cutoff scores were calculated separately for the two subsets on each of the markers. At every attempt to find the optimum cutoff, a receiver-operator characteristics (ROC) curve was drawn first and then its corresponding Youden’s indices (specificity + sensitivity – 1) were plotted if the ROC curve had an area under curve (AUC) greater than zero. The point with the maximum value of the Youden’s index was chosen as the optimum cutoff because it maximized the specificity and sensitivity of the marker in question ²³ . Instances where ROC curve had an AUC of zero unit, Youden’s index could not be meaningful, thus yielding the corresponding cutoff value undefined. AUCs of ROC curves were calculated using trapezoid rule and maximum Youden’s indices were obtained by direct measurement.

ROC curve for MUC1 suggested the existence of optimum cutoff values for samples collected before, on or after the 20 ^th chronological day since AUC>0 held for all the instances ( Figure 3 and Figure 4). The cutoff for samples collected on or after 20 ^th day was 120 ( Figure 5) and the rest of the samples yielded a cutoff of 190 ( Figure 6). MUC1 status is to be considered negative if the H-score is less than or equal to the cutoff; if it is above the cutoff then the status is positive.

Further calculations concluded that MUC1 status thus had a weak correlation with the IVF outcome, which was statistically insignificant ( Figure 7; Table 4).

ROC curve for E-cadherin suggested the existence of optimum cutoff values for samples collected on or after the 20 ^th chronological day but not before, since AUC>0 and AUC=0 held for those two instances, respectively ( Figure 8 and Figure 9). The cutoff for samples collected on or after 20 ^th day was 75 ( Figure 10). E-Cadherin status is to be considered negative if the H-score is less than or equal to the cutoff; if it is above the cutoff then the status is positive. Subsequent calculations determined that E-cadherin status had a weak correlation with the IVF outcome, which was statistically insignificant ( Figure 11; Table 4).

Various confounding variables were matched between case and control groups in the present study ^{24, 25} . Age and BMI are two such anthropometric variables, which were matched in similar studies where comparison was being done between receptive and non-receptive endometrium ⁹ . However, ratio of sample size between case and control groups in the present study is about 5:1 which is not shown with most of the studies with similar design where the ratio is almost 1:1. This deviation is in part due to relatively lower IVF success rates that ultimately resulted in fewer number of successes accomplished within the study period. Nevertheless, the imbalance between the two groups in the present study was taken into account statistically by applying adjustments for unequal variance wherever appropriate.

Although endometrial dating by histomorphological features are indeed helpful, it must be noted that the method has its pitfalls. Basal-only sampling, tangential cuts resulting in false crowding of glands, difficulties in detecting spiral arteries and predecidual change, epithelial-stromal discordance, polyps, inflammation etc. often hinder the accuracy of endometrial dating. Also, some of the samples in the present study lack surface epithelium due to the small quantity of the tissue collected, which was somewhat unavoidable (discussed later). All those factors, compounded by subjective nature of the methodology, makes histological dating of endometrium a challenging feat for the pathologist ²⁶ . Despite the difficulties, best efforts were made to ensure the dating in the present study was as accurate as possible, which included consultation with expert pathologists, checking and double-checking the findings and cross-referencing with multiple authentic sources.

H-score, acronym for ‘histological’ score, was initially developed to quantify immunohistochemical staining of certain tumors but later its use diversified to serve many other arenas of immunohistochemistry including the assessment of endometrial receptivity ⁹ . The current study utilized this method to provide a uniform platform for the immunohistochemical markers to be assigned a number that directly reflects its staining characteristics and also makes it more amenable to the robust statistical tools applied subsequently. Due to lack of consensus regarding the roles of immunohistochemical markers and their optimum cutoff values for useful categorization as far as the investigator could search for in English scientific literature, the present study attempted to develop such a scheme tailored for this study from scratch according to the fundamental statistical principles underlying some of the most widely used scoring and categorization systems in diagnostic pathology ²⁷ .

How should the endometrial expression pattern of MUC1/EMA be in order to facilitate implantation remains an open question since non-congruent results regarding this are frequently found in English scientific literature ^{8, 10} . The present study attempted to address the question by incorporating MUC1 into the immunohistochemistry panel but found its correlation with IVF outcome to be weak and statistically insignificant.

Although correlation of E-cadherin H-score with IVF outcome was not statistically significant in the present study, higher H-scores of E-cadherin were found to be more associated with IVF failure which resonates with the findings of other similar studies ^{8, 16} . An interesting finding of the present study about E-cadherin was the marked difference of its relative intensity between surface epithelia and glandular epithelia where the surface epithelial staining was almost consistently higher in intensity than glands. A similar finding is noted in a study where such differential staining is appreciated in non-receptive endometrium, except the marker was PGRMC1 and not E-cadherin ⁹ . Perhaps further inquiry into such peculiarities would unravel more mysteries in the ever-expanding field of reproductive biology.

Although this study was confined to endometrial samples at a single point of the menstrual cycle which may contribute to non-representative sampling, further research may be designed incorporating endometrial biopsy collection at more than one day of the patient’s menstrual cycle. This might be achieved by iatrogenic thickening of endometrium by progesterone and/or luteinizing hormone administration, preferably six months to one year prior to the IVF embryo transfer cycle to avoid possible complications in implantation process as practiced in case of ERA ⁹ . Besides the prediction of IVF outcome, this modified technique may provide necessary information to calculate the WOI for individual patient that might be used to transfer embryo during IVF, thus improving the odds of IVF success. However, hormone-induced changes in immunohistochemical expression of the markers would have to be taken into account if that method is followed ²⁶ .

Due to ethical and methodological constraints, the amount of endometrial tissue obtained by scratch procedure was generally less than that of routine endometrial curettage. This added to the difficulty in endometrial dating and in appreciating the potential heterogeneous nature of endometrial tissue. One sample had to be kept out of the study pool simply due to sample inadequacy. Moreover, endometrial sample had to be collected just once per patient, thus eliminating the possibility to evaluate samples at more than one point in a menstrual cycle for a patient. Hormonal influence on endometrial tissue may alter the morphological and immunohistochemical findings ^{9, 26} . Therefore in the present study, samples were collected prior to commencement of hormone treatment. Since the endometrial ‘scratch’ biopsy procedure was performed on each of the patients of both case and control groups, any possible effect of the procedure itself would have been cancelled out between the two groups. Since this study was partly motivated by an attempt to complement the genomic approach for IVF outcome prediction known as ERA, it would be ideal to be able to cross-validate the findings by ERA of each of the samples. However, financial and time constraints prevented the investigators to pursue this aspect.

It is anticipated that the study will be used as a guide for further research and will provide essential information to design a low-cost alternative diagnostic tool to ERA for predicting IVF failure as well as individualized WOI. Thus, it may help reduce the financial and associated burdens of women undergoing IVF or other infertility treatment.

Ethical approval was obtained from the Institutional Review Board of Bangabandhu Sheikh Mujib Medical University (approval no. BSMMU/2017/1576). Written informed consent for publication of the patients’ details was obtained from the patients.

The underlying and extended data is publicly available on Open Science Framework: MUC1 and E-cadherin immunohistochemistry of endometrium cannot predict the outcome of in vitro fertilization: A case-control study, https://doi.org/10.17605/OSF.IO/UAMV4 ²⁸ .