{"paper_id":"6cef7f17-4d9e-416f-bdd1-f94c26011bdd","body_text":"Endometriosis is currently seen as a systemic inflammatory disease associated with\npelvic pain and infertility, among other symptoms ( Chapron  et al. , 2022 ;  Pirtea  et al ., 2022 ). Though endometriosis is\ntraditionally associated with pelvic manifestations, this disease displays\nmultifactorial and systemic effects with a prevalence estimated between 2% and 10%\nin women in the general population ( Vassilopoulou\n et al ., 2018 ).\nEndometriosis is also present in up to 50% of women with infertility ( Vassilopoulou  et al ., 2018 ;\n Lee  et al ., 2020 ). Some\nof the mechanisms involved in endometrial pathogenesis may cause an exacerbated\ninflammatory state in the uterus and ovaries, thus affecting endometrial\nreceptivity, ovarian reserve and oocyte quality ( Chen\n et al ., 2023 ). In the endometrium, the\ndecidualization program is altered due to estradiol causing an increase in\nprostaglandin E2 production and resistance to progesterone, which affect the\nimplantation rate ( Zhang & Wang, 2023 ).\nSeveral other associations have been reported, such as aberrant gene expression in\nthe endometrium associated with an increased production of inflammatory cytokines\nand chemokines, resulting in differential recruitment and differentiation of immune\ncells, reshaping immune response in the uterus and ovarian microenvironment ( Vallvé-Juanico  et al .,\n2019 ). All such factors contribute to subfertility via pelvic adhesions,\ndistorted pelvic anatomy, and bilateral tubal blockage.\nTherefore, early screening to select patients at higher risk of endometriosis is\nneeded. The question is how to find these patients. Having patients answer a\nquestionnaire is the first step in the diagnostic process ( Chapron  et al ., 2022 ). Validated questionnaires\nfor the early detection of patients at higher risk of endometriosis are currently\navailable ( Bailleul  et al .,\n2021 ;  Chapron  et al .,\n2022 ). Apart from its effect on fertility, endometriosis is associated\nwith dysmenorrhea, dyspareunia and lower abdominal pain; it may also cause dysuria\nand dyschezia, depending on the degree of involvement and location ( Ekine  et al ., 2020 ). Taking\nall this into account, it is important to consider the patient’s clinical symptoms,\nperform adequate physical examination, and order complementary tests including\nimaging-based approaches, such as ultrasound or magnetic resonance imaging (MRI), to\ndiagnose ovarian and deep infiltrating endometriosis ( Chapron  et al ., 2022 ). Unfortunately, imaging-based\napproaches are poor at diagnosing superficial endometriosis, which may require\ndiagnostic laparoscopy ( Goncalves  et\nal ., 2021 ).\nUntil the last decade, diagnostic laparoscopy was routinely performed for the\ndiagnosis and treatment of endometriosis in patients with suspected endometriosis\nwho consulted for pain and infertility. More recently, however, diagnostic\nlaparoscopy has been less prescribed and performed ( Pirtea  et al ., 2022 ). This is due to the accumulated\nevidence suggesting that surgery for endometriosis does not necessarily improve\nassisted reproductive technology (ART) treatment outcomes ( Pirtea  et al. , 2022 ). In fact, reports have\nindicated that surgery may cause further harm by impairing the ovarian reserve\n( Benaglia  et al. , 2017 ).\nContrary to observations made in ovarian stimulation, ART does not worsen\nendometriosis symptoms and has no impact on ovarian endometriomas or deep\ninfiltrating endometriosis ( Somigliana  et\nal ., 2019 ).\nGiven this controversy, it is possible that only a subgroup of patients with\nendometriosis-associated infertility might benefit from laparoscopic treatment and\nimprove their chances of conceiving naturally. In this regard,  Vercellini  et al . (2009)  reported evidence\nindicating that surgery for pelvic endometriosis increased the chances of conceiving\nnaturally by approximately 50% in the 12-18 months after surgery. This was also\nconfirmed by others authors ( Dückelmann\n et al ., 2021 ;  Muzii\n et al. , 2021 ). Thus, while seeing patients with\nclinical suspicion of endometriosis and infertility, we must consider their age,\novarian reserve, tubal patency and male factor among other clinical parameters, to\nthus evaluate their chances of conceiving naturally as wells as the potential\nbenefit of surgical treatment ( Rizk  et\nal ., 2015 ;  Lee  et\nal ., 2020 ;  Dückelmann\n et al ., 2021 ;  Khan\n& Lee, 2021 ;  Muzii  et\nal ., 2021 ).\nEndometriosis-associated infertility is still being debated and more studies are\nrequired, especially considering that the high efficacy of modern-day assisted\nreproductive technology (ART) has led to progressively adopting ART-first\napproaches, particularly for women with endometriosis ( Pirtea  et al. , 2022 ). However, surgery is\nstill recommended for some patients with endometriosis depending on the symptoms\nthey present with and whether they wish to become pregnant. The following questions\nmust be answered: Does laparoscopy play a role in these patients? What other factors\nare involved in the achievement of spontaneous pregnancy by patients with\nendometriosis? This study evaluated the spontaneous pregnancy rate of patients with\nendometriosis after surgical treatment and its possible associations with different\nclinical factors.\n\nThis retrospective observational study used the anonymized records of patients with\nendometriosis-related infertility who underwent laparoscopic surgery at “Fertilis -\nSanatorio Las Lomas”, from 2014 to 2020 with up to two years of follow up. All\npatients included had indication for endometrial surgery due to their symptoms.\nOf the 303 patients that met the inclusion criteria (age 25-43 years; symptoms and/or\nimage findings consistent with endometriosis; infertility; and laparoscopic\ndiagnosis of endometriosis), 200 also met the exclusion criteria (other laparoscopic\ndiagnosis without findings compatible with endometriosis; history of previous\nsurgeries for endometriosis; bilateral negative tubal patency; thrombophilia;\nrecurrent abortion; and moderate/severe male factor) and were used in statistical\nanalysis.\nThe following data were collected from patient medical charts: fertility treatment,\nage (grouped as <30, 30-34, 35-39 and >39 years), endometriosis, ASRM score,\ninitial symptoms, primary/secondary infertility, time of infertility, tubal and\nuterine quality and time until pregnancy. The rASRM classification was designed to\ncategorize cases of endometriosis via direct visualization of the pelvic organs\nduring laparoscopy or laparotomy into four stages: minimal (I), mild (II), moderate\n(III), and severe (IV). Changes involving the peritoneum, the fallopian tubes and\novaries are used to stage the disease. When using the rASRM system, different points\nare assigned depending on whether the endometriotic lesion is deep or superficial,\nthe size of the endometriotic lesion, and the type (filmy or dense) and extent of\nadhesions involving the fallopian tubes, ovaries, and the pouch of Douglas. The\npoints are added to a total score, and the total score is used to stratify the\ndisease into one of the four stages ( Hudelist\n et al ., 2021 )ureters, bowel and sacral roots.\nAdenomyosis (growth of endometrium in the myometrium, sometimes explained by\ndisruption of the uterine junctional zone.  Table\n1  shows some of the collected information. Patients were initially\ncategorized according to fertility treatment as patients that used ART immediately\nafter surgery (16.5%, ART-p) and patients that waited for a spontaneous pregnancy\n(83.5%). This last group was subdivided into patients that achieved spontaneous\npregnancy within 12 months of surgery (71.8%, SP-p) and individuals unable to\nachieve spontaneous pregnancy who required ART (28.2%, NSP-p).\nDemographic information of the studied patients.\nData was analyzed with GraphPad Prism 9.4 (GraphPad Software) using chi-square, ANOVA\nand T-test depending on each comparison. The decision tree was made using\nRPart-package on R ( R Core Team, 2022 ;  Therneau & Atkinson, 2023 ). Different\nvalues for the decision tree parameters, such as maximum depth and minimal records\nper node, were tested to optimize accuracy and sensitivity.\n\nFrom the initial 200 patients, 16.5% opted for immediate ART after laparoscopic\ntreatment, while the rest opted to wait for spontaneous pregnancy, of which 71.8%\nwere able to achieve it within 12 months ( Figure\n1 ). As we further evaluated the treatment approaches within each age\ngroup, we found that the patients who chose to undergo ART immediately after surgery\nwere overrepresented in the older group (>39 years). This was expected as other\nfactors associated with older age and unrelated to endometriosis might have been\ninvolved in the medical decision to go for ART immediately after laparoscopic\ntreatment ( Figure 2 ). Furthermore, when we\ncalculated the EFI score, we found that patients with ART as the initial conduct had\na lower score than those who opted to wait for spontaneous pregnancy, which supports\nthe idea that other factors might be involved. Interestingly, we did not observe\ndifferences between the patients that achieved spontaneous pregnancy (SP-p) and the\nones that did not (NSP-p) ( Figure 3 ).\nFigure 1 Distribution of patients according to initial fertility treatment and\noverall outcome.\nDistribution of patients according to initial fertility treatment and\noverall outcome.\nFigure 2 Patient age according fertility treatment. ART-p are overrepresented in\nthe 39+ years group, which could be caused by other factors associated\nto age such as low ovarian reserve.\nPatient age according fertility treatment. ART-p are overrepresented in\nthe 39+ years group, which could be caused by other factors associated\nto age such as low ovarian reserve.\nFigure 3 EFI score. Patients prescribed ART immediately after surgery (ART-p)\npresented a significantly lower EFI score in comparison with both groups\nthat opted for spontaneous pregnancy (SP-p, NSP-p). Interestingly, no\nsignificant difference was found between these last two groups.\nMean±SEM; Anova, sidak post test *** p <0.001,\n**** p <0.0001.\nEFI score. Patients prescribed ART immediately after surgery (ART-p)\npresented a significantly lower EFI score in comparison with both groups\nthat opted for spontaneous pregnancy (SP-p, NSP-p). Interestingly, no\nsignificant difference was found between these last two groups.\nMean±SEM; Anova, sidak post test *** p <0.001,\n**** p <0.0001.\nThen, we focused on the patients that opted to wait for a spontaneous pregnancy. When\nwe evaluated how long it took to achieve pregnancy, we found that the individuals in\nthe SP-p group achieved pregnancy in 5.7 months, while the subjects on the NSP-p\ngroup took almost 1.8 times longer (10.2±3.7  vs .\n5.7±3.6 months,  p <0.0001) ( Figure 4A ). Interestingly, when only the time since the treatment change\nfrom waiting for spontaneous pregnancy to ART was considered, we found that patients\nachieved pregnancy within a similar time than the ones in the SP group\n(4.9±3.7 months), suggesting that patients in the NSP group might have\nbenefited if they had been identified earlier ( Figure\n4B ).\nFigure 4 Time to achieve pregnancy. (A) Patients in the NSP group took\nsignificantly longer to achieve pregnancy since surgery than subjects in\nthe SP group. However, (B) this difference disappears if only the time\nsince the change in treatment is considered. Mean±SEM; Anova,\nsidak post test **** p <0.0001.\nTime to achieve pregnancy. (A) Patients in the NSP group took\nsignificantly longer to achieve pregnancy since surgery than subjects in\nthe SP group. However, (B) this difference disappears if only the time\nsince the change in treatment is considered. Mean±SEM; Anova,\nsidak post test **** p <0.0001.\nWith this in mind, we decided to look for differences in the other recorded\nparameters between the SP-p and NSP-p groups that might be useful to predict the\noutcome of patients who chose to wait. We did not find significant differences in\nthe ASRM score, though a higher proportion of patients with lower ASRM scores (I)\nwere in the SP (21.2%)  vs . the NSP (11.4%) group. Interestingly, we\nfound that individuals in the ART-p group tended to have higher ASRM scores (III and\nIV) than the patients who chose to wait for a spontaneous pregnancy ( Figure 5A ). Of all other studied variables, only\ntubal quality showed a significant difference, with a higher percentage of regular\nquality on the NSP-p group (22.8%  vs . 8.5%,\n p <0.05) ( Figures 5B ,  5C ).\nFigure 5 A- ASRM score by patient group. Patients in the ART-p group showed a\ntendency toward higher ASRM scores (III and IV), while individuals in\nthe SP-p group showed a tendency toward lower ASRM scores (I and II). No\nsignificant difference was found. B- Comparison of uterine cavity\nquality between patients in the NSP and SP groups. No differences were\nfound. C- Comparison of tubal quality between patients in the NSP and SP\ngroups. NSP had lower tubal quality. Chi-square test.\nA- ASRM score by patient group. Patients in the ART-p group showed a\ntendency toward higher ASRM scores (III and IV), while individuals in\nthe SP-p group showed a tendency toward lower ASRM scores (I and II). No\nsignificant difference was found. B- Comparison of uterine cavity\nquality between patients in the NSP and SP groups. No differences were\nfound. C- Comparison of tubal quality between patients in the NSP and SP\ngroups. NSP had lower tubal quality. Chi-square test.\nSince none of the studied variables alone was able to identify patients in need of\nART, we performed a multivariate analysis. As age could be associated with other\nfactors and considering that in our study all patients in the NSP group were aged\nbetween 30 and 39 years, we chose to focus on patients younger than 40 years old,\nsince they might be the ones that benefit the most from ART. Using R, we obtained a\ndecision tree with 81.3% accuracy and 53.3% sensitivity on the original data set\n( Figure 6 ).\nFigure 6 Decision tree. A multivariable approach was used to predict the patients\nthat required ART after waiting for spontaneous pregnancy. Figure shows\nthe optimized decision tree, accuracy 81.3%, sensitivity 53.3%.\nDecision tree. A multivariable approach was used to predict the patients\nthat required ART after waiting for spontaneous pregnancy. Figure shows\nthe optimized decision tree, accuracy 81.3%, sensitivity 53.3%.\n\nAs indicated in previous clinical studies about the management options for\nendometriosis-related infertility, our data also pointed to an increase in\nspontaneous pregnancy after surgical treatment ( Dückelmann  et al ., 2021 ;  Muzii  et al ., 2021 ). This suggests that there\nis a group of patients (normal ovarian reserve, normal patency and mild male factor)\nwho might benefit from laparoscopic infertility treatment associated with\nendometriosis to improve their chances of conceiving naturally ( Rizk  et al. , 2015 ;  Muzii  et al. , 2021 ).\nChronic inflammation can impair ovarian or endometrial function, leading to disorders\nof folliculogenesis or implantation ( Benaglia\n et al ., 2017 ;  Pirtea\n et al. , 2022 ). Endometriosis usually develops with\ndiminished ovarian reserve due to the presence of an inflammatory microenvironment.\nThe identification of progesterone resistance in an eutopic endometrium leads to an\nestrogenic state that affects endometrial receptivity ( Lessey  et al ., 1996 ;  Zeitoun & Bulun, 1999 ;  Kao\n et al ., 2003 ;  Burney\n et al ., 2007 ;  Lessey\n& Kim, 2017 ). Although we did not find a significant correlation\nbetween ASRM score and spontaneous pregnancy, a tendency toward lower scores in\nassociation with better outcomes was identified. A higher proportion of high ASRM\nscores was observed among the patients who chose to wait for spontaneous pregnancy\ninstead of undergoing ART immediately, possibly indicating the presence of other\nassociated factors not considered in this study.\nAlthough the majority of the patients who opted to wait for spontaneous pregnancy\nafter endometrial surgery achieved it within 12 months, we found a group of\nindividuals that was not able to get pregnant spontaneously and eventually required\nART. Consequently, these patients had a longer time from surgery to pregnancy.\nInterestingly, when we looked at how long these patients took to achieve pregnancy\nsince the start of ART, we found that they took a similar amount of time than those\nwho achieved it spontaneously. If identified earlier, the patients who required ART\nmight have achieved pregnancy by five to ten months earlier. This is not only\nrelevant from the psychological point of view ( Assaysh-Öberg  et al ., 2023 ;  Tetecher  et al ., 2024 ), but\nalso from an endometrial perspective, since laparoscopic surgery for endometriosis\nis not curative, with 40-45% of women having recurring disease, which may, again,\ninterfere with fertility ( Vercellini  et\nal ., 2009 ).\nIn order to identify the patients that will require ART at the clinic, we studied\nseveral clinical parameters, including age, ASRM score, tubal and uterine cavity\nquality, among others. Although we found an association between some of these\nparameters and patient outcomes, none was able to identify patients in need of ART.\nAs a result, we performed a multivariate analysis. Considering that the goal was to\nidentify patients at the clinic, we chose to develop a decision tree algorithm. This\ntype of algorithm presents several advantages. It is not only very easy to use at\nthe clinic, but it also provides for machine learning opportunities, it is\nstatistically driven, flexible and can find patterns hidden in the data ( Therneau & Atkinson, 2023 ). One of the main\npoints about the flexibility of such type of algorithm is that it considers that the\nsame clinical parameter might lead to different outcomes depending on other\nparameters. For example, patient age may lead to different predictions depending on\nhow long the patient has been infertile for.\nWe generated a decision tree with 81.3% accuracy and 53.3% sensitivity from the\noriginal set of data. Our decision tree requires only four parameters (time of\ninfertility, tubal quality, age and ASRM score) and can be worked through in less\nthan a minute without other tools or calculations, which makes it ideal for\nimplementation at the clinic and a tool that might result in shorter waiting times\nuntil pregnancy for a significant part of the patients with endometriosis-related\ninfertility. Further studies with more patients and variables might further improve\nthe proposed decision tree.\nAnother tool to evaluate which is the best approach for patients with endometriosis\nis the Endometrial Fertility Index, or EFI ( Adamson\n& Pasta, 2010 ). The EFI score system has been developed using a wide\nvariety of endometrial patients and validated several times, proving to be\nespecially useful for patients with poor prognosis ( Adamson & Pasta, 2010 ;  Adamson,\n2013 ). In contrast, the decision tree algorithm developed herein focuses\non patients that have a good EFI score and aims to complement the EFI by helping to\nidentify those patients that, even with a good EFI score, will probably require ART.\nThe average EFI score of the patients used in the model was 7.27, ranging from 4 to\n10.\nThe management of endometriosis-associated infertility is still a topic of\ndiscussion, especially in what concerns the role of surgery ( Rizk  et al. , 2015 ;  Lee  et al ., 2020 ;  Bailleul  et al ., 2021 ;  Muzii  et al ., 2021 ). The results presented\nherein support other studies that suggested that surgical treatment for\nendometriosis might improve spontaneous pregnancy rates. Furthermore, we propose\nthat the early identification of patients in need of ART to achieve pregnancy after\nthe surgery will decrease the time between surgery and pregnancy and thus improve\noverall outcomes.\n\nThe decision tree obtained in the present study might be a useful tool to identify\npatients with good EFI scores who might need ART after endometrial surgery.","source_license":"public-domain-us","license_restricted":false}