{"paper_id":"24184396-7454-4606-8885-78ca270fae45","body_text":"The endometrial receptivity array (ERA) test is a new player amongst the plethora of\nadd-ons ‘solutions’ for repeated implantation failure (RIF), but it seems to be following\nthe same path as other add-ons which were deemed unproven due to failure to demonstrate an\nincrease in success rate, mainly the live birth rate (LBR). Using an invasive biopsy of the\nendometrium (a sort of scratching with the help of Pipelle) at the time of the presumed\nwindow of implantation (WOI), the ERA test attempts to evaluate the gene expression profile,\nwith the hope of identifying the specific transcriptomic signature, and offering a more\nprecise timing for personalised embryo transfer (pET).\nThe ERA test divides the results of the endometrial biopsy into receptive and\nnon-receptive, and further to pre-receptive or post-receptive, by using very complex\nstatistics ( Ruiz-Alonso, 2013 ,  2014 ;  Garcia-Velasco, 2015 ; Mahajan, 2015 ) with the intention of freezing all embryos for future transfer\nwhile enhancing or delaying the WOI artificially, according to the results.\nOffering a diagnostic test to better synchronise the embryo and the uterus is compelling,\nhowever, there are many questions that need to be answered before considering the clinical\nresults of the first RCT that was recently published ( Simón  et al. , 2020 ). The ERA test was offered\ncommercially as a successful tool to increase precision of the WOI in presumed RIF, beyond\nthe ineffective histological dating ( Díaz-Gimeno\n et al. , 2013 ), when it was only in the early stages of\nexploration and before real proofs were presented to indicate an increase in IVF success. In\naddition, in the original case-study, only 17 patients were included, of whom 10 had one to\ntwo failures, 4 had three failures and the remaining 3 had more than three failures ( Ruiz-Alonso  et al. , 2013 ). This\nwas hardly a population of RIF by any definition ( Ben\nRafael, 2020a ). It is noteworthy that we have recently condemned RIF as an invalid\ndiagnosis to denote ‘failure to implant’, since no two failures are similar, and failure\nmight be due to completely different reasons each time. The idea that several failures can\nbe allocated under one name (RIF) and receive special attention and treatment, when not is\nreally an iatrogenic entity ( Somigliana  et\nal. , 2017 ;  Ben Rafael,\n2020a ), was also denounced in an editorial in Human Reproduction ( Evers, 2016 ).\nThe ERA test raises multiple questions which were not answered by the RCT. First, the array\nmethodology has now largely been replaced by next generation sequencing (NGS). Second, how\nmany genes should be included in the test? Many groups using microarray have found a\ndifferent number of genes that were up or down regulated during the WOI. This, by itself,\nhints at the volatility and dependency of the results on the methods used to test the genes,\nthe complexity of the mathematical model, the methods and timing of biopsy, corrective\nmethods to the biopsy material, relationships to preovulatory progesterone levels (which are\noverlooked in most studies) and the validity of displacing the WOI in successive cycles by\nonly 12–24 h. In addition, the implantation process includes crosstalk between the\nendometrium and the embryo before and during invasion ( Diedrich  et al. , 2007 ). This includes many\ndistinct embryonal stages like apposition, adhesion and invasion which are regulated by many\ngenes over a restricted period of days. The biopsies for gene expression evaluation are done\non an endometrium that has not been affected by the embryo-endometrial crosstalk, which\nrepresents an obvious limitation of the whole concept. Furthermore, any hormonal\nmanipulation of the WOI in subsequent cycles can affect gene expression differently, which\nmay result in high rates of intra-patient variability in repeated testing on the same\npatients.\n\nA partial list of gene numbers that were found to be dysregulated at the WOI varies from 63\ngenes ( Tapia  et al. , 2008 ),\n238 genes (in commercial ERA test,  Ruiz-Alonso\n et al. , 2013 , recently dropped to 236 genes), 303 genes ( Macklon, 2017 ) and 616 genes that were\nupregulated ( Huang  et al. ,\n2017 ); to 313 genes of which 92% were down and 8% up-regulated ( Koler  et al. , 2009 ). Others have\nfound 91 genes significantly increased more than 2-fold in their expression, and 115 were\ndecreased more than 2-fold in endometriosis patients ( Kao  et al. , 2003 ). Furthermore, earlier studies\nby the same group behind the commercial test were also different. They found that 147 genes\nwere significantly dysregulated in the refractory endometrium ( Horcajadas  et al. , 2006 ) or 218 and 133 genes\nthat changed on day hCG + 7 versus LH + 7 accordingly ( Horcajadas  et al. , 2008 ), compared to the\ncurrent 238 genes in the ERA test. The last number already hints that gene profile at LH + 7\nis not comparable to hCG + 7 as maintained by the ERA test.\nThe different number of genes here and in other publications is not easy to reconcile, and\ntherefore requires a sophisticated statistics analysis to form a clinical conclusion ( Garcia-Velasco  et al. , 2015 ).\nThey are not self-explanatory for the clinician’s (or client’s) judgment, bearing witness to\nthe randomity of the search for a credible signature of the receptive, pre-receptive or\npost-receptive endometrium. This issue cannot be separated from the question of consistency\nof the methods used and the repeatability of the test in the same patients.  Cho  et al.  (2018)  have shown a\nlarge intra-patient variability in the results of gene profiles which might be due to the\nmethods used or the hormonal status.\nThe same group ( Garcia-Velasco  et\nal. , 2015 ) had their own doubts on the issue. In a study of endometrium\nin endometriosis patients, it was claimed that ‘endometrial receptivity’ is a multifactorial\nprocess of which the studied gene expression is but one factor, as ‘other genes that may not\nhave been studied, epigenetic aberrations or even pathologic proteomic profiles might\nprovide further insight’. Unexpectedly, women with endometriosis who were postulated to have\nan endometrial receptivity defect and progesterone resistance ( Fox  et al. , 2016 ), did not differ in their ERA\ntest results ( Miravet-Valenciano  et\nal. , 2017 ).\nTo add to the confusion, it is maintained that it is unlikely that a single endometrial\ncause underlies RIF ( Macklon, 2017 ) since\nmost failures are due to the embryo quality ( Diedrich  et al. , 2007 ), and studies show that after 3–4 IVF\ncycles, only a few women remain not pregnant. This notion supports the idea that a faulty\nendometrium is infrequently the cause of RIF ( Ben-Rafael, 2020a ). Also, it has been shown that RIF is not associated per se with\nabnormal endometrial integrin expression, a marker of implantation, or the expression of\nendometrial integrins a1, a4 and aVb3, which appear to have no prognostic value in\nsubsequent IVF treatments ( Coughlan  et\nal. , 2013 ).\nGiven that various pathologies have been found in endometrial biopsies ( Crum  et al. , 2003 ), and that\n‘histological dating alone lacks the sensitivity to identify a definable defect in\nendometrial development and, therefore, in the implantation process’,  Kliman and Frankfurter (2019)  claimed that dating is not the\nreason for implantation failure.\n\nTechnology is evolving fast. It was recently claimed that all the past and current studies\nusing array technology might not be accepted as accurate anymore. ‘It is now well accepted\nthat sequencing technique NGS is more comprehensive in coverage and precise in\nquantification of global gene expression profiles’ ( McGettigan  et al. , 2017 ;  Huang  et al. , 2017 ). Hence, all the studies on the ERA, including\nthe recent RCT ( Simón  et al. ,\n2020 ), which have used array test become difficult to interpret, and should be\nre-evaluated based on NGS.\n\nCan a ‘blind’ manual procedure of endometrial cells collection be standardised to derive\neach time consistent and representative cell populations? Pipelle, depending on the\noperator, may result in variable amounts and depth of cell collection, hence it is not\nsurprising that the group that developed the test has recently recognised this issue and\nprovided a correction that identifies the relative contribution of the epithelial and\nstromal cells in the biopsy to the gene expression profile. They have offered a partial\nsolution by using ‘computational deconvolution’ which is a statistical, mathematical\ncorrection method, to evaluate the relative contribution of the major cell types to the\ntranscriptome, but not of the less frequent cells ( Suhorutshenko  et al. , 2018 ) which create another limitation of\nthe newly offered corrective method. It should be noted that ‘computational deconvolution’\nwas not used in the RCT which started before its introduction.\nThe timing of biopsy in relationship to the WOI should also be questioned. It was proven\nthat fertile and infertile women had similar out of phase biopsies ( Coutifaris  et al. , 2004 ), hence without a gold\nstandard for the exact implantation window, the timing of the biopsy (i.e. the WOI) remains\ncontroversial. If we cannot consistently identify the ovulation ( Park  et al. , 2007 ;  O’Connor  et al. , 2006 ), then the timing of\nbiopsy and gene profile might be skewed accordingly. For example, while recent reviews have\nfound that the optimal day of embryo transfer was LH + 6 or hCG + 7 ( Mackens  et al. , 2017 ), the ERA test considers\nLH + 7 and hCG +7 ( Ruiz-Alonso  et\nal. , 2013 ) or hCG +7 ( Horcajadas\n et al. , 2006 ) as equal and optimal. Others also rejected the\nidea that LH+ 7 is equal to P + 5 and claimed that defining the date of biopsy is not simple\nand that it is inconsistent in natural and medicated cycles ( Kliman and Frankfurter, 2019 ). Furthermore, it should be noted\nthat any endometrial hormonal manipulation before embryo transfer (ET) or any ovarian\nstimulation or triggering of ovulation may affect the gene profile differently ( Horcajadas  et al. , 2008 ;  Humaidan  et al. , 2012 ;  Mahajan, 2015 ).\n\nProgesterone is the undeclared elephant in the room. Progesterone is the main driver of the\nsecretory changes that determine the WOI and there are complex correlations between\nprogesterone, implantation and the success rate, and yet the ERA test was presented\nindependently of progesterone levels. Progesterone starts to rise slightly before ovulation\nand a day before oocyte retrieval because of exogenous hCG. This can vary with the number of\nfollicles. Progesterone alters gene expression ( Fatemi and Van Vaerenbergh, 2015 ). A preovulatory rise in progesterone above\n1.5 ng/ml has been associated with lower pregnancy rates ( Bosch  et al. , 2010 ) and dysregulation of over 140\nendometrial genes that are required for normal endometrial function (64 up- and 76\ndown-regulated) while 13 marker genes of receptivity were over regulated ( Labarta  et al. , 2011 ). This\ndictates that any new procedure that is based on gene expression, should consider the\nconcomitant progesterone level.\nIn yet another study, it was proposed that progesterone serum levels of less than 10 ng/ml\non the day of embryo transfer are associated with a lower success rate ( Labarta  et al. , 2017 ) that is\nfully correctable by the addition of progesterone in the same cycle (Labarta; presented\nduring COGI Paris 2019; ww.congresmed.com/COGI). Additionally, others have suggested that ‘P\nlevels of >5 ng/ml that act on an adequately primed endometrium result in endometrial\nluteinisation and receptivity, which does not differ from that achieved by much higher\nlevels’ ( Usadi  et al. , 2008 ;\n De Ziegler  et al. , 2013 ).\nHowever, progesterone levels were not reported in conjunction with gene profile studies, but\nrightfully in the RCT, patients with a preovulatory progesterone rise were excluded.\nWhen planning to correct a pre-receptive or post-receptive endometrium, we need to freeze\nall embryos and rely on hormone replacement cycles, but neither the serum progesterone\nlevels that are required to optimise cycle outcome, nor the optimal length of exposure to\nprogesterone before frozen embryo transfer, have been firmly established ( Van de Vijver  et al. , 2016 ). In\nthe ERA test, 5 days of progesterone administration is considered optimal for implantation\nof Day 5 embryos, and progesterone exposure is assumed to be able to enhance or delay the\nendometrial maturity by precisely ±12 h intervals. This idea is not supported by studies\nwhich have shown that plus or minus one day in progesterone exposure does not affect\nimplantation.  Van de Vijver  et al. \n(2016)  randomised two groups who received a Day 4 embryo transfer after 3 or 5 days\nof progesterone exposure. The pregnancy rate was similar in both groups, although a shorter\nprogesterone exposure was associated with a higher miscarriage rate. So ± 12–24 h was not\nshown to be a problem regarding implantation, which raises a disagreement over the question\nof ‘what difference does one day make?’ ( Ruiz-Alonso\n et al. , 2014 ). Furthermore, the feasibility of manipulating the\nendometrium with such precision has not been proven, suffice to say the variability in the\ncompliance of the ‘progesterone start’ between patients might span more than 12 h.  Kliman and Frankfurter (2019)  rejected the idea\nthat LH + 7 is equal to P + 5 and questioned why such precision in embryo transfer is\ncritical given that the implantation window is at least 3 days in duration.\nAs for luteal support, serum P levels vary widely even when the same preparations are\ngiven, since the uptake, absorption and metabolism of each hormone varies amongst women\n( Yovich  et al. , 2015 ).\nAlso, intravaginal progesterone is thought to have a first pass and effect on the\nendometrium that is beyond the serum levels ( Cicinelli  et al. , 2001 ), but neither serum progesterone levels\nnor endometrial tissue concentrations seem dose proportional ( Paulson  et al. , 2014 ). Taken together these data\nindicate that it is not so feasible to manipulate (enhance or delay) the endometrium in the\ncoming cycle by relying only on the number of hours or days of progesterone administration\nto displace the WOI by ±12–24 h.\n\nThe repeated claim that about 25% of failed IVF cycles are due to endometrial problems is\nnot supported by their own study. According to  Mahajan (2015) , the ERA test is probably applicable to a marginal number of\npatients. They have used the test in a group 186 patients, dividing them into two groups:\nGroup 1 who failed only one IVF cycle and were found to have 15.1% nonreceptive endometrium,\nand Group 2 who failed three IVF cycles and were found to have 27.5% non-receptive\nendometrium. The non-receptive cases underwent a 12 h (only) modification/correction of the\nWOI in the subsequent thawed cycle. The only pregnancy registered after one previous failure\nand correction of the non-receptive endometrium miscarried (1/7 i.e. 14.3%). However, after\nthree failures, the correction of the non-receptive endometrium resulted in a 44.5% (8/18)\npregnancy rate of which one was aborted, an apparent good result. Calculated differently,\nthe overall pregnancy rate in the first cycle was 61%, and in the third cycle before any\ncorrections were made, it was 42%. Assuming that in the second cycle (which was not\nreported) the pregnancy rate was also about 40%, it leaves us with only about 6.5% women not\npregnant after three failed cycles, of whom 27.5% were non-receptive (a similar percentage\nof non-receptive endometrium was suggested after three failures also by others;  Hashimoto  et al. , 2017 ). which\nis about 1.8% of all patients treated who can potentially utilise the test following three\nimplantation failures. Even with a lower pregnancy rate of 30–35% in each cycle for three\ncycles, the test is applicable to only 5–10% of all patients starting, hardly a breakthrough\nconsidering that even after three failures the chances of conceiving are still very good. A\nconfirmation that RIF (three IVF failures) is rare, as is the need for changes including the\nERA test, was provided in a recent study ( Pirtea\n et al. , 2020 ) on 4428 patients who received a thawed euploid\nsingle embryo transfer (SET). The cumulative pregnancy rate after three cycles was 95.2% and\nthe LBR was 92.6%, which supports our above calculation ( Mahajan, 2015 ) that the ERA test, if done after three failures,\nwill apply only to a minority of IVF failures.\n\nRCTs, and even meta-analysis on the different add-ons, have initially shown positive or at\nleast encouraging results, and only when they ae used on a larger scale or with the correct\ndesign, questions and planning, have their true value emerged ( Ben Rafael, 2020a ).\nUnlike the seminal non-controlled studies on the ERA test, the recent multicentric open\nlabel RCT ( Simón  et al. ,\n2020 ) did not deal with RIF. They aimed to compare pET to fresh or frozen ET in\nyounger age women, under 37, in their first IVF cycle (about 70%) or after a few failures\n(30%); these are clearly not cases that needed experimental add-ons procedures. For a\nmulticentric study, many issues in the protocol, such as stimulation protocol, vitrification\nmethod and progesterone dose in the transfer cycle were not fully controlled and left to the\ndecision of the participating centre, a fact that must have created an heterogeneity in the\nstudy group. Also, with the current trend to substitute array test with more precise NGS\ntechnology, all previous studies including the RCT, with the old version of array gene\nexpression should be re-confirmed by NGS ( McGettigan\n et al. , 2017 ;  Huang\n et al. , 2017 ).\nIt is beyond the scope of this paper to provide a critical review of the study design,\nexclusion and inclusion criteria, before and after oocyte collection, but it is important to\nindicate that by running the ERA test only on the pET group and not on the control groups,\nthey missed the chance to comment on the most important issue, i.e. what can be expected\nwhen women with a non-receptive endometrium or ‘positive ERA test’ keep trying in the\nsuccessive cycles without any ‘correction’?\nNevertheless, the results of the primary outcome measured in this RCT showed no differences\nin pregnancy rate or LBR by intention-to-treat analysis (an unbiased estimate of the\nefficacy of the intervention) or by first embryo transfer. Only the cumulative LBR after\n12 months was higher.\nA point of caution stems from the fact that despite the exclusion of all bad-risk patients\nand patients with non-satisfactory stimulation and high preovulatory progesterone, the\npercentage of women with displaced WOI (before any failure) was 37.5%, which is higher than\nthe previously reported 27.5% after three implantation failures or 15.5% after one failure\n( Mahajan, 2015 ;  Hashimoto  et al. , 2017 ).\n\nFinally, the freezing embryos for later transfer without a good indication, may be wrong\n( Ben-Rafael, 2020b ). The significantly\nhigher rate of preeclampsia (7.5% vs. 4.9%) and eclampsia (4% vs. 2.5%), including chronic\nhypertension after frozen-thawed embryo transfers and the rates of premature labour after\nthe use of frozen oocytes compared to fresh, are arguments accumulating against freezing all\nembryo ( Sites  et al. , 2017 ).\nA 3-fold increase in hypertensive disorders (4.4% vs. 1.4%;  P  < 0.009)\nafter freezing was also confirmed in PCOS patients ( Chen  et al. , 2016 ). Both hypertensive disorders and large for\ngestational age were confirmed in a meta-analysis ( Maheshwari  et al. , 2018 ).\n\nA complex crosstalk between the endometrium and the embryo during implantation includes\ngenes expression and anatomical, physiological and metabolic changes, all of which can vary\nwith the medical treatment administered. We are far from being able to test for or having a\nfull understanding of these processes, hence, every measurement of one such effect might\nrepresent a tubular rather than perspective outlook. Since most IVF failures are due to the\nquality of the embryos, focusing on the endometrium as a reason for IVF failure is probably\napplicable in only a minority of cases.\nThe hurdle of the ERA test, to prove the existence of a non-receptive endometrium, and\npinpoint the displacement within a few hours frame, assumes that progesterone manipulation\ncan bridge a ± 12–24 h gap, and correct the endometrium, without having a gold standard to\ncompare to any of these effects, all with a single tool namely gene profiling, seems\ninsurmountable. For example, it has long been maintained that ovarian stimulation cycles\nresult in an asynchronous endometrium, which obviously calls for different measures\nincluding ‘freeze all’ to circumvent the ‘problem’, but the problem of asynchronous\nendometrium has never been shown to exist and it is doubtful whether it needs any\ncorrection, except in special cases ( Ben Rafael,\n2020b ).\nIt is not surprising that the only RCT that was published so far ( Simón  et al. , 2020 ) could not show, based on\nintention to treat, any improvement. Furthermore, another recent retrospective study by the\nsame group ( Cozzolino  et al. ,\n2020 ) has concluded that the use of the ERA test in 488 patients who underwent\npreimplantation genetic testing for aneuploidy (PGT-A), ERA or both, could not show any\nadvantage to the use of ERA test. In short, ERA did not appear to improve outcomes in a\ngroup of patients who failed three or more ETs with a total of three or five embryos, and\nstrangely PGT-A per se was effective in those failing three ETs, but was ineffective after\nfailures of transfer of five embryos ( Cozzolino\n et al. , 2020 ). Similarly, others ( Neves  et al. , 2019 ) in a multivariate analysis\nhave confirmed that performing an ERA test did not influence the pregnancy rate, in the\ntested euploid ET arm and was even associated with a diminished pregnancy rate in the donor\nET arm. Taken together, these observations coupled with the lack of intra-patient\nconsistency, bare witness to the randomity of the current state of research in the\nfield.\nDifferent studies have documented a 10-fold spread (from 63 to 616) in the number of genes\nthat are dysregulated at the WOI, which casts doubt on the meaning of any number of genes in\nindicating a correlation with implantation. This might be partly explained by the array\nmethod, which is no longer accepted as accurate and should be replaced by NGS, and the blind\nmethod of biopsy which might be hard to standardise to provide similar cell populations\nevery time. These issues need more research and awareness. Furthermore, without bringing\ninto the equation the different factors that can potentially affect implantation, such as\noestrogen priming, progesterone type and route of administration including local and\nperipheral levels, and the changes they induce on the endometrium, the gene expression\nsignature by itself, even with NGS, cannot reliably reflect the full complexity of the\nimplantation process.\nThe idea of improving implantation through high technology is most welcome, but like any\nnew procedure, it must be reconciled with several known facts and be proven beyond any doubt\nbefore it is widely offered commercially. The suggestion that the endometrium may be delayed\nor enhanced at the WOI, by as little as ±12 or 24 h ( Mahajan, 2015 ;  Valdes  et\nal. , 2017 ;  Simón  et\nal. , 2020 ) almost implies a ‘point of implantation’ rather than a\n‘window of implantation’ and requires a good argument for the advantage it provides to human\nprocreation, rectified with the known facts that embryos can be transferred once or twice\n(double ET) in any of the first 6 days and can linger for days before the actual embedding\nstarts, thus showing a high degree of tolerance to a non-receptive endometrium before it\nturns to be receptive. We also know that two embryos can implant in the same uterus days, or\neven weeks, apart (superfetation or superfecundation).\nClinicians tend to grasp any new idea that can potentially improve results ( Ben Rafael, 2020a ), also to demonstrate that they\nare competitive and updated, but unfortunately even after many years of practice, most\nadd-ons have been deemed not proven ( Macklon\n et al. , 2019 ). New tests, like the ERA, until shaped and proven,\nshould be offered only under research protocols that separate compounding factors,\nconsidering all the above reservations, and keeping in mind that freezing embryos is not\nrisk free. As we have witnessed time and again, a single RCT or even a meta-analysis should\nnot be accepted as a final proof towards its overall utility as a new solution or its\nwide-spread use ( Ben Rafael, 2020a ). It is the\nduty of the clinical societies and peer review journals to follow up on the evidence, and\nfilter new ideas and technological procedures, to avert physicians from repeating false\ndirections or mistakes for longer than necessary.\nAs physicians, it is our continuous duty to appraise any new treatment and diagnostic tool\nand offer an add-on procedure only by indication and if it is proven to provide a better LBR\nunder research protocols. The message to the patients should be that failures occur more\noften than not, and if no special obstacle to pregnancy exists, when stimulation, embryo\nculture and endometrial width appear to be normal, there is no need to resort to unproven\ncostly add-ons, and, if patients agree, they need to persevere with their similar trials for\nfive or more cycles, which will leave only a small fraction of patients in need other,\nalbeit unproven, solutions.\n\nNo new data were generated or analysed in support of this research.\n\nThe author declares no funding was given to this work.\n\nThe author declares no conflict of interest. Author has written and reviewed the\nmanuscript.","source_license":"CC-BY-4.0","license_restricted":false}