{"paper_id":"faedc964-0a9a-4b7f-a81c-8ef58ec2ccfb","body_text":"O R I G I N A L A R T I C L E Open Access\nPrediction of unsuccessful endometrial\nablation: a retrospective study\nK. Y. R. Stevens 1,2* , D. Meulenbroeks 1, S. Houterman 3, T. Gijsen 4, S. Weyers 2 and B. C. Schoot 1,2\nAbstract\nBackground: Endometrial ablation (EA) is a frequently used treatment for abnormal uterine bleeding, mainly due to\nthe low risks, low costs and short recovery time associated with the procedure. On the short term, it seems successful,\nlong-term follow-up however, shows decreasing patient satisfaction as well as treament efficacy. There even is a post-\nablation hysterectomy rate up to 21%. Multiple factors seem to` influence the outcome of EA. Due to dissimilarities in and\nvariety of these factors, it has not been possible so far to predict the success rate of EA based on pre-operative factors.\nTherefore, the aim of this study is to develop two prediction models to help counsel patients for failure of EA or necessity\nof surgical re-intervention within 2 years after EA.\nMethods: We designed a retrospective two-centred cohort stu dy in Catharina Hospital, Eindhoven and Elkerliek\nHospital, Helmond, both non-university teaching hospitals in the Netherlands. The study population consisted of\n446 pre-menopausal women who underwent EA for abnorm al uterine bleeding, with a minimum follow-up time\nof 2 years. Multivariate logistic regression analysis was used to create the prediction models.\nResults: The mean age of the patients was 43.8 years (range 20 –55), 97.3% had complaints of menorrhagia,\n57.4% of dysmenorrhoea and 61.0% had complaints of intermittent or irregular bleeding. 18.8% of patients\nstill needed a hysterectomy after EA. The risk of re-intervention was significantly greater in women with\nmenstrual duration > 7 days or a previous caesarean section, while pre-operative menorrhagia was significantly associated\nwith success of EA. Younger age, parity≥ 5 and dysmenorrhea were significant multivariate predictors in both models.\nThese predictors were used to develop prediction models, which had a C-index of 0.71 and 0.68 respectively.\nConclusion:We propose two multivariate models to predict the chance of failure and surgical re-intervention\nwithin 2 years after EA. Due to the permanent character of EA, the increasing number of post-operative failure\nand re-interventions, these prediction models coul d be useful for both the doctor and patient and may\ncontribute to the shared decision-making.\nKeywords: Prediction model, Endometrial ablation, Abnormal uterine bleeding, Patient counselling.\nArticle\nThe use of EA as treatment for abnormal uterine bleeding is\nrapidly increasing. This surgical outpatient procedure offers\na minimally invasive alternative for hysterectomy in case\nnon-surgical treatment is not effective. The success of EA is\nmainly based on the short recovery time, low risks and low\ncosts associated with the procedure. [1–5]I nc o n t r a s tt ot h e\nshort-term success, long-term follow-up shows decreasing\npatient satisfaction as well as treatment efficacy [ 6–13]. A\ncommon complaint of patients after EA is pain (20 –23%),\nwhich often leads to re-interventions [7, 8]. Besides the oc-\ncurrence, persistence or aggravation of pain, another reason\nfor re-intervention can be persisting bleeding disturbances\n[7, 10]. Retrospective cohort data reveal a post-ablation hys-\nterectomy rate up to 21% [6, 8–10, 12, 14–16].\nSeveral factors influencing failure of EA have been\nreported. It has been shown that the probability of\nsuccess increases with older age at the time of\n© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0\nInternational License ( http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and\nreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to\nthe Creative Commons license, and indicate if changes were made.\n* Correspondence: Kyr.stevens@gmail.com\nPartly presented as abstract at:- The 24th Annual International Congress of the\nEuropean Society of Gynaecological Endoscopy, Budapest Hungary 2015- The 26th\nAnnual International Congress of the European Society of Gynaecological\nEndoscopy, Antalya Turkey 2017- The 46th global congress of the American\nAssociation of Gynaecologic Laparoscopists, Washington D.C. United States 2017\n1Department of Obstetrics and Gynaecology, Catharina Hospital,\nMichelangelolaan 2, 5623 Eindhoven, EJ, The Netherlands\n2Women’s Clinic, Ghent University Hospital, Corneel Heymanslaan 10, 9000 Ghent, Belgium\nFull list of author information is available at the end of the article\nGynecological SurgeryStevens et al. Gynecological Surgery            (2019) 16:7 \nhttps://doi.org/10.1186/s10397-019-1060-1\n\nintervention. [ 6, 8–12, 14, 17, 18]P r i o rs t u d i e sd e m -\nonstrated different negative influencing factors, such\nas the duration of pre-operative menstruation, dys-\nmenorrhea, the position of the uterus and the thick-\nness of the endometrium [ 6, 7, 9–13, 15, 19, 20].\nHowever, due to dissimilarities in and variety of the\nfactors previously described, it has not been possible\nso far to predict the success rate of EA based on\npre-operative factors. Patien t counselling is therefore\ndifficult.\nThe aim of this study was to develop two predic-\ntion models to counsel patients for failure of EA and\nfor surgical re-intervention within 2 years. In addition,\nwe established the hysterectomy rate, the additional\ntreatment rate and the patient satisfaction after EA.\nMethods\nThis retrospective two-centred cohort study included\npatients with EA for complaints of abnormal uterine\nbleeding in two non-university teaching hospitals in the\nNetherlands (the Catharina Hospital in Eindhoven and\nthe Elkerliek Hospital in Helmond). In both hospitals,\nsimilar ablation techniques were used between 2004 and\n2013, namely Cavatherm® (Veldana Medical SA, Morges,\nSwitzerland), Thermablate® EAS (Idoman, Ireland) and\nGynecare Thermachoice® (Ethicon, Sommerville, US).\nPrevious research showed that these techniques were\nequal in effectivity [ 13, 21]. The study was approved by\nthe local medical ethical review board. All patients gave\ninformed consent.\nPatients\nPatients were identified in the Electronic Patient Care\nSystem using the following search terms: endometrial\nablation, balloon-coagulation endometrium, coagula-\ntion uterus and endoresection: hysteroscopic exten-\nsive. The retrieved cases were verified by means of\nchart review.\nPatients were excluded if they were post-menopausal\nat the time of treatment, if they had or were suspected\nof having an endometrial malignancy or if they had uter-\nine cavity deformations (anomalies, fibroids, adenomyo-\nsis or a polyp).\nFollow-up period after EA was at least 2 years, since\nearlier research showed that most re-interventions took\nplace in this post-operative 2-year period [ 8, 15, 20–24].\nFollow-up ended on the day of hysterectomy, in case of\ndeath or on April 15, 2015.\nData extraction\nTwo researchers extracted all the data from individual\npatient files. Patients were requested to complete a ques-\ntionnaire concerning follow-up information. In case of\nnon-response, patients were contacted by letter again\nand ultimately by telephone. The questionnaire com-\nprised questions based on significant factors previously\npublished [ 6, 9–15, 17–20]. Charts and patient responses\nwere used to obtain post-procedural information on\nmenstrual pattern, patient satisfaction, additional treat-\nment and pathology results in case a hysterectomy was\nperformed.\nAbnormal blood loss as procedure outcome was de-\nfined by a combination of intermittent or irregular\nbleeding and heavy menstrual bleeding (HMB) following\nEA. Treatment prior to EA was defined as any treatment\nfor abnormal uterine bleeding performed prior to sur-\ngery. Satisfaction was evaluated on a four-point scale (1:\nvery satisfied, 2: satisfied, 3: dissatisfied, 4: very dissatis-\nfied). During analysis, we combined the answers scoring\n1 and 2 points as ‘satisfied’ and those scoring 3 and 4\npoints as ‘dissatisfied’.\nOutcomes\nThe primary aim of this study was to identify signifi-\ncant predictors of failure of EA and surgical\nre-intervention within 2 years after EA by construct-\ning a prediction model for each outcome. Failure was\ndefined as pelvic pain, abnormal blood loss or dissat-\nisfaction after the procedure. Secondary outcomes of\nthis study were hysterectomy rate, patient satisfaction\nand percentage of additional treatment after EA, for\nexample hormonal treatment, re-ablation or endomet-\nrial resection.\nStatistical analysis\nStatistical analysis was performed using the statistical\npackage IBM SPSS statistics, software version 21.0\n(IBM Corp., Armonk, NY, USA). Continuous variables\nwere presented as mean and standard deviation or\nmedian and minimum-maximum, depending on nor-\nmality. Categorical variables were reported as\nfrequencies.\nUnivariable logistic regression analysis was used to de-\ntermine which predictive factors were significant. Corre-\nsponding odds ratios (OR) and 95% confidence intervals\n(CI) were given.\nPredictive factors with a p value <.10 were used in the\nmultivariable analysis. A manual selection process was\ndone by progressively excluding the variable with the\nhighest p value.\nThe p value of 0.10 was chosen because, as Steyerberg\net al. stated, an incorrect exclusion of a variable would\nbe far more detrimental than considering to put in a fac-\ntor too many [ 25, 26].\nPossible interaction between the significant predictors\nin the model was tested using interaction terms. Further-\nmore, multicollinearity was tested.\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 2 of 9\n\nThe overall fit of the model was tested using the\nC-index (area under the curve). A value of 1.0 for the\nC-index implies a perfectly produced model, where\nevery prediction made with the variables in the model\nis true. However, a value of 0.5 implies that the\nmodel gives information that is equal to that given by\nthe probability on its own. Values over 0.7 indicate a\ngood model, whereas values over 0.8 indicate a strong\nmodel [ 27, 28].\nThe regression model was internally validated with\nbootstrap resampling ( n = 5000) [ 29–33]. Regression co-\nefficients of the model were multiplied by the shrinkage\nfactor to correct for over-optimism of the original\nmodel.\nResults\nIn this study, 762 patients were identified. After examin-\nation of patient records, 33 patients were excluded; 30\npatients did not completely fulfil the inclusion criteria\n(e.g. malignancy, cavity deformations) and 3 patients had\nan incomplete ablation procedure.\nThe remaining 729 participants were contacted, of\nwhom 283 did not respond despite our best efforts. This\nresulted in 446 included patients, which represents a re-\nsponse rate of 61% (Fig. 1).\nThe baseline patient characteristics are listed in\nTable 1. The mean age of the patients at the time of EA\nwas 43.8 years (SD ± 5.5, range 20 –55). The mean BMI\nwas 26.5 kg/m 2 (SD ± 4.7). A mean number of parity of\n2.2 (SD ± 1.0) was observed; 13.7% of the women had\nundergone a previous caesarean section.\nMenorrhagia was present in 97.3% of patients, 61%\nhad complaints of intermittent or irregular bleeding and\n57.4% had complaints of dysmenorrhea. In 39.4% of pa-\ntients, the duration of the menstruation was longer than\n7 days (Table 1).\nFig. 1 Enrolment and allocation of patients\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 3 of 9\n\nRe-intervention model\nIn the study group, 11.9% ( n = 53) of the patients\nneeded a surgical re-intervention within 2 years after\nEA. Univariate analyses showed that the following\npre-operative variables we re significantly associated\nwith a higher probability of getting a surgical\nre-intervention within 2 years after EA ( p < .05): age\n(OR 0.93, 95% CI 0.89 –0.98), dysmenorrhea (OR 2.83,\n95% CI 1.44 –5.55), length of menstruation > 7 days\n(OR 1.87, 95% CI 1.04 –3.37), parity ≥ 5 (OR 5.84,\n95% CI 1.27 –26.83), previous caesarean section (OR\n2.98, 95% CI 1.52 –5.83) and pre-treatment (OR 0.40,\n95% CI 0.16 –0.95) (Table 2). These pre-operative vari-\nables were included in the multivariate analyses.\nIn the final prediction model after multivariate\nanalysis, the following pre-operative variables were\nsignificant: age (OR 0.95, 95% CI 0.90 –1.00), dys-\nmenorrhea (OR 2.48, 95% CI 1.21 –5.07), length of\nmenstruation > 7 days (OR 2.05, 95% CI 1.10 –3.82),\nprevious caesarean section (OR 2.21, 95% CI 1.05 –\n4.64) and parity ≥ 5 (OR 7.63, 95% CI 1.51 –38.46)\n(T able 2). The C-index of the model was 0.71.\nNo two-way interaction and multicollinearity between\nthe variables was detected.\nThe shrinkage factor of 0.823 was used to correct the\nmodel.\nTable 1 Baseline patient characteristics ( N = 446)\nCharacteristic * Value**\nAge (years) 43.8 ± 5.5\nBody mass index (kg/m 2) 26.5 ± 4.7\nDysmenorrhea 57.4%\nFollow-up time (days) 1693.8 ± 871.9\nDuration of menstruation > 7 days ( n = 429) 39.4%\nIntermittent or irregular bleeding 61.0%\nLength of the uterus (cm) ( n = 402) 9.1 ± 1.1\nMenorrhagia 97.3%\nParity (no.) 2.2 ± 1.0\nPrevious caesarean section 13.7%\nSmoking (n = 445) 21.6%\nSterilisation (n = 444) 26.1%\nUterus position ( n = 296)\nAnteverted 72.3%\nRetroverted 23.6%\nMidposition 4.1%\n*n = 446 unless otherwise mentioned\n**Mean ± SD or a percentage\nTable 2 Pre-operative predictors of re-intervention after endometrial ablation\nUnivariate analysis Multivariate analysis\nVariable Odds ratio 95% CI p value Odds ratio 95% CI p value β\nAge (years) 0.93 0.89–0.98 <.01 0.95 0.90–1.00 .06 − 0.052\nBody mass index (kg/m 2) 0.99 0.93–1.05 .68\nDysmenorrhea 2.83 1.44–5.55 <.01 2.48 1.21–5.07 .01 0.097\nDuration of menstruation > 7 days 1.87 1.04–3.37 .04 2.05 1.10–3.82 .02 0.718\nIntermittent or irregular bleeding 1.56 0.84–2.89 .16\nMenorrhagia 0.67 0.14–3.12 .61\nMyomas 0.67 0.30–1.48 .33\nParity (no.) 0.88 0.66–1.17 .38\nParity ≥ 5 5.84 1.27–26.83 .02 7.63 1.51–38.46 .01 2.032\nPre-treatment* 0.40 0.16–0.95 .04 0.49 0.20–1.22 .13 –\nPrevious caesarean section 2.98 1.52–5.83 <.01 2.21 1.05–4.64 .04 0.794\nSmoking 1.36 0.70–2.62 .36\nSterilisation 1.59 0.86–2.94 .14\nTotal endometrial thickness\nThin, 0–3 mm 0.64 0.18–2.29 .49\nNormal, 4–12 mm 1.00 ––\nThick, > 13 mm 0.96 0.36–2.58 .93\nUterine cavity length of the uterus (cm) 1.16 0.88–1.54 .29\nUterus position\nAnteverted 1.00 ––\nRetroverted 1.23 0.54–2.79 .63\nMidposition 2.77 0.70–10.96 .15\n*Any form of treatment (medicamentous or surgical) prior to the EA\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 4 of 9\n\nThe final model after application of the shrinkage fac-\ntor is as follows:\nFailure model\nIn the study group, 35.8% ( n = 160) of the EA failed.\nUnivariate analyses showed that the following\npre-operative variables we re significantly associated\nwith a higher probability of failure of EA ( p <. 0 5 ) :\nage (OR 0.93, 95% CI 0.89 –0.96), dysmenorrhea (OR\n2.14, 95% CI 1.42 –3.23), menorrhagia (OR 0.27, 95%\nCI 0.08 –0.91) and parity ≥ 5 (OR 11.17, 95% CI 1.33 –\n93.60) (Table 3). The pre-operative variables with a p\nvalue p < .10 were also included in the multivariate\nanalyses; these were total endometrial thickness and\npre-treatment.\nIn the final prediction model after multivariate ana-\nlyses, the following pre-operative variables were signifi-\ncant: age (OR 0.93, 95% CI 0.90 –0.97), dysmenorrhea\n(OR 2.11, 95% CI 1.37 –3.26), menorrhagia (OR 0.21,\n95% CI 0.06 –0.77) and parity ≥ 5 (OR 11.19, 95% CI\n1.30–96.51) (Table 3). The C-index of the model was\n0.68.\nNo two-way interaction and multicollinearity between\nthe variables was detected.\nThe shrinkage factor of 0.904 was used to correct the\nmodel.\nThe final model is as follows:\nOther results\nOur results showed that 82.6% ( n = 368) of patients were\nsatisfied with the outcome of EA, and 86.8% ( n = 387) of\npatients would recommend EA to a friend.\nOf the satisfied group, 14.6% ( n = 54) of patients had\na new medical therapy or a surgical re-intervention.\nFurthermore, 32.7% ( n = 146) of the total population\nhad an additional treatment after EA, varying from hor-\nmonal to surgical intervention.\nThe hysterectomy rate was 18.8% ( n = 83), and 61% ( n\n= 51) of this group had surgery within 2 years after EA.\nA total of 22.9% ( n = 102) of the study population had\nadditional surgical treatment, 52% ( n = 53) of whom\nwithin 2 years after EA. Besides the number of smokers,\nthere was no significant difference between the baseline\ndata and the hysterectomy rates between the responders\nand the non-responders.\nDiscussion\nMain findings\nThis study identified predictors for the outcome of EA\nas a treatment of abnormal uterine bleeding; this re-\nsulted in two prediction models, one for the probability\nof a surgical re-intervention within 2 years after EA\n(C-index 0.71) and one for the probability of failure of\nEA (C-index 0.68).\nExplaining the models\nThe significant factors seem to be in line with the previ-\nously published literature. [ 6, 11–15, 17–20]\nAn EA procedure at a younger age increases the risk\nof failure due to the longer interval until menopause.\nThis increased time interval can also increase the risk of\nnew complaints or re-intervention. In our model, age\nwas used as a continuum, so the probability can be cal-\nculated more specifically based on the exact age of the\nindividual patient.\nThe significant factor of high parity ( ≥ 5) is probably\ndue to a larger multiparous uterine cavity, which is less\ncongruent with an optimal fit of the ablation devices.\nHowever, we did not find a univariate significant differ-\nence in uterine cavity length.\nPrevious caesarean section as a significant negative\nrisk factor can possibly be explained due to abnormal\nbleeding caused by uterine scar defects. It is possible\nthat the device cannot make complete contact with\nthe entire surface, especially in the inner part of the\nniche, leading to incomplete EA due to residual active\nendometrium [ 34]. Furthermore, in our models,\npre-operative dysmenorrhea is associated with higher\nrisk of failure and surgical re-intervention.\ny ¼ 1\n1 þ e− 3:485− age/C2 0:063ðÞ þ dysmenorrhea yes ¼1n o ¼0ðÞ /C2 0:677ðÞ þ parity ≥ 5 yes ¼1n o ¼0ðÞ /C2 2:183ðÞ − menorraghia yes ¼1n o ¼0ðÞ /C2 1:400ðÞð Þ\ny ¼ 1\n1 þ e−ð−0:896− age/C2 0:046ðÞ þ dysmenorrhea yes ¼1n o ¼0ðÞ /C2 0:008ðÞ þ parity ≥ 5 yes ¼1n o ¼0ðÞ /C2 1:781ðÞ þ duration of menstruation >7 yes ¼1n o ¼0ðÞ /C2 0:629ðÞ þ previous caesarean section yes ¼1n o ¼0ðÞ /C2 0:700ðÞ Þ\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 5 of 9\n\nAdenomyosis has been suggested to be a factor influ-\nencing the increased occurrence of (post-ablation)\npelvic pain [ 35–38]. Pain is a subjective outcome\nmeasure. On the one hand, the level of pain can be\nexplained by the coping mechanism of the patient; on\nthe other hand, if a patient experiences many\npre-operative complaints, the cause can be multifac-\ntorial (e.g. coping, dysmenorrhea, adenomyosis, endo-\nmetriosis [ 37–41].\nPerforming ablation in patients with a certain extent\nof uterine pathology (fibroids, adenomyosis) can be\nseen as a risk for success of therapy [ 2, 34, 41, 42].\nHowever, sensitivity and specificity of the diagnostic\ntools for determining these myometrial diseases are\nstill low.As expected, thin endometrium is a positive\npredictor for ablation success due to the increased\nchance of complete penetration of heat during the\nEA. In multivariate analysis, however, this no longer\nwas a significant factor.\nMenorrhagia, defined as the subjective estimation of\nheavy bleeding (e.g. increased blood clots, overall\nbleeding quantity), is a patient characteristic that\nseems to fit the success profile for EA. This can be\nexplained by the primary expected effect of EA:\nreduction of endometrial surface and subsequent\nbleeding.\nFurthermore, we observed that pre-treatment leads to\na univariate outcome of significantly higher risk of fail-\nure or re-intervention. Multiple treatments prior to EA\ncan be an indication of the complexity of the underlying\ncause of the uterine disorder.\nExamples of using the models\nIn clinical practice, the models can be used to esti-\nmate the risk of failure for individual patients. For in-\nstance, a 38-year-old patient, para 5, with a previous\ncaesarean section, a menstrual duration of more than\n7 days and complaints of dysmenorrhea and menor-\nrhagia, has 93% chance of failure of EA and 62%\nchance of surgical re-intervention within 2 years after\nEA.\nOn the other hand, a 48-year-old woman, para 2, with\nno previous caesarean section, a menstrual duration\nshorter than 7 days and complaints of menorrhagia but\nno dysmenorrhea, has a chance of 28% failure of EA and\n4% chance of surgical re-intervention within 2 years after\nEA (Table 4).\nTable 3 Pre-operative predictors of failure of endometrial ablation\nUnivariate analysis Multivariate analysis\nVariable Odds ratio 95% CI p value Odds ratio 95% CI p value β\nAge (years) 0.93 0.89–0.96 <.01 0.93 0.90–0.97 <.01 − 0.070\nBody mass index (kg/m 2) 0.99 0.95–1.03 .61\nDysmenorrhea 2.14 1.42–3.23 <.01 2.11 1.37–3.26 <.01 0.749\nDuration of menstruation > 7 days 1.26 0.84–1.89 .27\nIntermittent or irregular bleeding 1.22 0.82–1.83 .33\nMenorrhagia 0.27 0.08–0.91 .03 0.21 0.06–0.77 .02 − 1.544\nMyomas 0.92 0.56–1.49 .72\nParity (no.) 0.88 0.73–1.07 .22\nParity ≥ 5 11.17 1.33–93.60 .03 11.19 1.30–96.51 .03 2.415\nPre-treatment 0.63 0.39–1.03 .07 0.74 0.37–1.47 .39 –\nPrevious caesarean section 1.57 0.90–2.72 .11\nSmoking 0.73 0.45–1.18 .20\nSterilisation 1.30 0.84–2.01 .24\nTED\nThin, 0–3 mm 0.94 0.47–1.85 .85 1.11 0.54–2.30 .78 –\nNormal, 4 –12 mm 1.00 ––\nThick, > 13 mm 0.55 0.29–1.07 .08 0.56 0.27–1.16 .12 –\nUterine cavity length (cm) 1.07 0.89–1.28 .49\nUterus position\nAnteverted 1.00 ––\nRetroverted 1.40 0.79–2.46 .25\nMidposition 1.51 0.46–4.95 .49\n*Any form of treatment (medicamentous or surgical) prior to the EA\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 6 of 9\n\nOther results\nIn accordance with the literature, our results show that\nmost re-interventions take place within 2 years after EA\n[8, 15, 20–24].\nOur study showed that 82.6% ( n = 368) of patients\nwere satisfied with EA, and 86.8% ( n = 387) would rec-\nommend it to a friend. The discrepancy between the\nsatisfaction and the percentage of re-interventions can\nbe explained by the fact that many patients stated that\nthey first wanted to try a minimally invasive therapy,\ninstead of having a major surgery such as a hysterec-\ntomy. If the EA failed for them, they would still recom-\nmend the treatment to others, to possibly avoid a more\ninvasive treatment.\nAs stated in previous literature, satisfaction is a diffi-\ncult and subjective concept and therefore an outcome\nthat is less reliable as an objective parameter for suc-\ncess [ 43–48].\nStrengths and limitations\nThe two-centred aspect of the study ensures its representa-\ntiveness. Furthermore, two researchers reviewed the charts,\nand if unclear answers were given, the patients were con-\ntacted by telephone to filter out wrong or misinterpreted\ndata.\nThe models were developed with the data of 446 pa-\ntients, who responded to our questionnaire. The hyster-\nectomy rate in this group does not differ significantly\nfrom that of the non-responder group. The chance of se-\nlection bias therefore is minimal, although this cannot\nbe completely ruled out.\nThe most important limitation of this retrospective\nstudy is the acquisition of data from patient charts with\na non-validated questionnaire.\nBesides the calculated probability of failure and\nre-intervention within 2 years, there still is a chance of hav-\ning a re-intervention after th is time; this cannot be calcu-\nlated with the models.\nAn external validation of the prediction models is\nneeded; this is currently being performed, using\nretrospective data of similar patient groups in two\nnon-university teaching hospitals in the Netherlands.\nFurthermore, we are currently performing a study to\ninvestigate the impact of the models (and their cor-\nresponding individual percentages) in the decision of\nboth the patients and doctors. The influence of costs\nof the treatment has not been added, although this\nmay influence the choice of the patient or doctor.\nTherefore, this option has been added to a follow-up\nquestionnaire of this study.\nWe are aware of the fact that some of the devices in the\nstudy are no longer used or have been updated; therefore,\nin the external validation, Novasure® and Thermachoice\nIII® will be added.\nPrevious research however showed that these tech-\nniques were similar effective [ 13, 21].\nInterpretation in light of other evidence\nWhen comparing existing literature concerning the suc-\ncess rates of EA, there seems to be some inconsistency\nin the importance of variables, especially when multi-\nvariate analyses were performed. Bongers et al. reported\nthat dysmenorrhoea seems less important in predicting\nthe outcome of EA in relation to other variables when\nperforming multivariate analyses [ 9]. In contrast to this\nstudy, the multivariate prediction model produced by El\nNashar et al. showed young age, high parity, history of\nsterilisation and pre-operative dysmenorrhea as signifi-\ncant prognostic factors for failure of EA. [ 6]\nTo illustrate the discrepancies in literature, a\ncase-control study by Peeters et al. reported that the\noutcome is not predicted by age and sterilisation, but by\npre-operative dysmenorrhea, submucous myomas and\nlarge-sized uteri. [ 19]\nEl Nashar et al. created a model to predict ‘failure’ of EA.\nFailure in this model was defined as bleeding or pain follow-\ning EA, with the necessitating of having a hysterectomy or\nre-ablation. In this model,age as a continuum was not used\n[6]. Comparing outcomes of our study on re-intervention\nand complaints, we observed different significant variables\npredicting the two types of failure of EA. Therefore, we\nmade two prediction models, so patients can be counselled\nfor the chance of failure and for the risk of re-intervention.\nIn this way, they can decide what is most important to them.\nThe models still need external validation.\nConclusion\nProper patient selection is the key for failure or\nre-intervention of EA. Therefore, we propose two multi-\nvariate models to predict the chance of failure and surgi-\ncal re-intervention within 2 years after EA. Due to the\npermanent character of EA, the increasing number of\npost-operative failure and re-interventions, these predic-\ntion models could be useful for both the doctor and pa-\ntient and may contribute to the shared decision-making.\nTable 4 Clinical example\nVariable Patient\n1\nPatient\n2\nPatient\n3\nPatient\n4\nAge (years) 38 42 45 48\nDysmenorrhea Yes No Yes No\nDuration of menstruation > 7 days Yes No Yes No\nMenorrhagia Yes Yes Yes Yes\nParity ≥ 5 Yes No No No\nPrevious caesarean section Yes No Yes No\nChance of failure of EA (%) 93% 84% 50% 28%\nChance of getting a surgical re-intervention\n< 2 years after EA (%)\n62% 26% 17% 4%\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 7 of 9\n\nSignificant factors in these models are age, dysmenor-\nrhea, duration of menstruation > 7 days, menorrhagia,\nparity and previous caesarean section.\nExternal validation of the models is being performed;\nfurthermore, we are performing a study to see the im-\npact of the models in the decision of both the patients\nand the doctor.\nAcknowledgements\nThe authors thank the patients for completing the questionnaires and for\nconsenting to participate in our study.\nFunding\nNone\nAvailability of data and materials\nThe datasets generated and analysed during the current study are not\npublicly available due to privacy, but they are available from the\ncorresponding author on a reasonable request.\nAuthors’ contributions\nKS contributed to the project development, data management, data analysis,\nand manuscript writing/editing. DM contributed to the project development,\ndata collection, and manuscript editing. SH contributed to the data analysis\nand manuscript editing. TG contributed to the data collection and\nmanuscript editing. SW contributed to the data collection and manuscript\nediting. BS contributed to the project development and manuscript editing.\nAll authors read and approved the final manuscript.\nEthics approval\nThe ethical board in the Catharina hospital and in the Elkerliek hospital\nconcluded that ethics approval was not necessary for this study.\nConsent for publication\nNot applicable\nCompeting interests\nB.C. Schoot received fees from Medtronic on an hourly basis for lectures on\nhysteroscopic morcellation. The fees were donated to a foundation that\npromotes research in obstetrics and gynaecology. The remaining authors\nhave no competing interests.\nPublisher’sN o t e\nSpringer Nature remains neutral with regard to jurisdictional claims in\npublished maps and institutional affiliations.\nAuthor details\n1Department of Obstetrics and Gynaecology, Catharina Hospital,\nMichelangelolaan 2, 5623 Eindhoven, EJ, The Netherlands. 2Women’s Clinic,\nGhent University Hospital, Corneel Heymanslaan 10, 9000 Ghent, Belgium.3Department of\nEducation and Research, Catharina Hospital, Michelangelolaan 2, 5623 Eindhoven, EJ, The\nNetherlands.4Department of Obstetrics and Gynaecology, Elkerliek Hospital, Wesselmanlaan\n25, 5707 Helmond, HA, The Netherlands.\nReceived: 18 December 2018 Accepted: 12 March 2019\nReferences\n1. Waddell G, Pelletier J, Desindes S, Anku-Bertholet C, Blouin S, Thibodeau D\n(2015) Effect of endometrial ablation on premenstrual symptoms. J Minim\nInvasive Gynecol 22(4):631 –636. https://doi.org/10.1016/j.jmig.2015.01.023\n2. Laberge P, Leyland N, Murji A, Fortin C, Martyn P, Vilos G et al (2015)\nEndometrial ablation in the management of abnormal uterine bleeding. J\nObstet Gynaecol Canada. 2015;37(4):362 –79.\n3. Bouzari Z, Yazdani S, Azimi S, Delavar MA (2014) Thermal balloon\nendometrial ablation in the treatment of heavy menstrual bleeding. Med\nArch (Sarajevo, Bosnia Herzegovina) 68(6):411 –413. https://doi.org/10.5455/\nmedarh.2014.68.411-413\n4. Miller J, Troeger KA, Lenhart GM, Bonafede M, Basinski CM, Lukes AS (2015)\nCost effectiveness of endometrial ablation with the NovaSure&amp;reg;\nsystem versus other global ablation modalities and hysterectomy for\ntreatment of abnormal uterine bleeding: US commercial and Medicaid\npayer perspectives. Int J Womens Health:59. https://doi.org/10.2147/IJWH.\nS75030\n5. Angioni S, Pontis A, Nappi L, et al. Endometrial ablation: first- vs. second-\ngeneration techniques. Minerva Ginecol 2016;68(2):143 –153. http://www.\nncbi.nlm.nih.gov/pubmed/26928420. Accessed 3 Dec 2018\n6. El-Nashar SA, Hopkins MR, Creedon DJ et al (2009) Prediction of treatment\noutcomes after global endometrial ablation. Obstet Gynecol 113(1):97 –106.\nhttps://doi.org/10.1097/AOG.0b013e31818f5a8d\n7. Wishall KM, Price J, Pereira N, Butts SM, Della Badia CR (2014) Postablation\nrisk factors for pain and subsequent hysterectomy. Obstet Gynecol 124(5):\n904–910. https://doi.org/10.1097/AOG.0000000000000459\n8. Thomassee MS, Curlin H, Yunker A, Anderson TL (2013) Predicting pelvic\npain after endometrial ablation: which preoperative patient characteristics\nare associated? J Minim Invasive Gynecol 20(5):642 –647. https://doi.org/10.\n1016/j.jmig.2013.04.006\n9. Bongers MY, Mol BWJ, Brölmann HAM. Prognostic factors for the success of\nthermal balloon ablation in the treatment of menorrhagia. Obstet Gynecol\n2002;99(6):1060–1066. http://www.ncbi.nlm.nih.gov/pubmed/12052600.\nAccessed 3 Dec 2018\n10. Longinotti MK, Jacobson GF, Hung Y-Y, Learman LA (2008) Probability of\nhysterectomy after endometrial ablation. Obstet Gynecol 112(6):1214 –1220.\nhttps://doi.org/10.1097/AOG.0b013e31818c1766\n11. Shaamash AH, Sayed EH (2004) Prediction of successful menorrhagia\ntreatment after thermal balloon endometrial ablation. J Obstet Gynaecol\nRes 30(3):210 –216. https://doi.org/10.1111/j.1447-0756.2004.00189.x\n12. Klebanoff J, Makai GE, Patel NR, Hoffman MK (2017) Incidence and\npredictors of failed second-generation endometrial ablation. Gynecol Surg\n14(1):26. https://doi.org/10.1186/s10397-017-1030-4\n13. Louie M, Wright K, Siedhoff MT (2018) The case against endometrial\nablation for treatment of heavy menstrual bleeding. Curr Opin Obstet\nGynecol. 30(4):287 –292. https://doi.org/10.1097/GCO.0000000000000463\n14. Bansi-Matharu L, Gurol-Urganci I, Mahmood T, Templeton A, van der\nMeulen J, Cromwell D (2013) Rates of subsequent surgery following\nendometrial ablation among English women with menorrhagia: population-\nbased cohort study. BJOG An Int J Obstet Gynaecol 120(12):1500 –1507.\nhttps://doi.org/10.1111/1471-0528.12319\n15. Shavell VI, Diamond MP, Senter JP, Kruger ML, Johns DA (2012)\nHysterectomy subsequent to endometrial ablation. J Minim Invasive\nGynecol 19(4):459 –464. https://doi.org/10.1016/j.jmig.2012.03.013\n16. Kalampokas E, McRobbie S, Payne F, Parkin DE (2017) Long-term incidence\nof hysterectomy following endometrial resection or endometrial ablation\nfor heavy menstrual bleeding. Int J Gynecol Obstet 139(1):61 –64. https://doi.\norg/10.1002/ijgo.12259\n17. Cramer MS, Klebanoff JS, Hoffman MK (2018) Pain is an independent risk\nfactor for failed global endometrial ablation. J Minim Invasive Gynecol 25(6):\n1018–1023. https://doi.org/10.1016/j.jmig.2018.01.020\n18. Bouzari Z, Yazdani S, Naeimi Rad M, Bijani A (2018) Is thermal balloon\nablation in women with previous cesarean delivery successful? TURKISH J\nMed Sci 48(2):266 –270. https://doi.org/10.3906/sag-1707-50\n19. Peeters JAH, Penninx JPM, Mol BW, Bongers MY (2013) Prognostic\nfactors for the success of endometrial ablation in the treatment of\nmenorrhagia with special reference to previous cesarean section. Eur J\nObstet Gynecol Reprod Biol 167(1):100 –103. https://doi.org/10.1016/j.\nejogrb.2012.11.009\n20. Kreider SE, Starcher R, Hoppe J, Nelson K, Salas N (2013) Endometrial\nablation: is tubal ligation a risk factor for hysterectomy. J Minim Invasive\nGynecol 20(5):616 –619. https://doi.org/10.1016/j.jmig.2013.03.011\n21. Lethaby A, Penninx J, Hickey M, Garry R, Marjoribanks J (2013) Endometrial\nresection and ablation techniques for heavy menstrual bleeding. Cochrane\nDatabase Syst Rev (8):CD001501. https://doi.org/10.1002/14651858.\nCD001501.pub4\n22. Sambrook AM, Bain C, Parkin DE, Cooper KG (2009) A randomised\ncomparison of microwave endometrial ablation with transcervical resection\nof the endometrium: follow up at a minimum of 10 years. BJOG. 116(8):\n1033–1037. https://doi.org/10.1111/j.1471-0528.2009.02201.x\n23. Penninx JPM, Herman MC, Mol BW, Bongers MY (2011) Five-year follow-up\nafter comparing bipolar endometrial ablation with hydrothermablation for\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 8 of 9\n\nmenorrhagia. Obstet Gynecol 118(6):1287 –1292. https://doi.org/10.1097/\nAOG.0b013e318236f7ed\n24. Herman MC, Penninx JPM, Mol BW, Bongers MY (2013) Ten-year follow-up\nof a randomised controlled trial comparing bipolar endometrial ablation\nwith balloon ablation for heavy menstrual bleeding. BJOG. 120(8):966 –970.\nhttps://doi.org/10.1111/1471-0528.12213\n25. Steyerberg EW, Eijkemans MJ, Habbema JD. Stepwise selection in small data\nsets: a simulation study of bias in logistic regression analysis. J Clin\nEpidemiol 1999;52(10):935 –942. http://www.ncbi.nlm.nih.gov/pubmed/\n10513756. Accessed 3 Dec 2018\n26. Steyerberg EW, Eijkemans MJ, Harrell FE, Habbema JD. Prognostic modelling\nwith logistic regression analysis: a comparison of selection and estimation\nmethods in small data sets. Stat Med 2000;19(8):1059 –1079. http://www.\nncbi.nlm.nih.gov/pubmed/10790680. Accessed 3 Dec 2018\n27. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd editio. (ed. New\nYork: John Wiley & Sons;, ed.).; 2000\n28. C-statistics. Definition, Examples, Weighting and Significance: Statistics How To;\nwww.statisticshowto.com/c-statistic/. Published 2018. Accessed 27 Mar, 2016\n29. Collins GS, Reitsma JB, Altman DG, Moons KGM (2015) Transparent\nreporting of a multivariable prediction model for individual prognosis or\ndiagnosis (TRIPOD): the TRIPOD statement. Br J Surg 102(3):148 –158. https://\ndoi.org/10.1002/bjs.9736\n30. Steyerberg EW, Harrell FE, Borsboom GJ, Eijkemans MJ, Vergouwe Y,\nHabbema JD. Internal validation of predictive models: efficiency of some\nprocedures for logistic regression analysis. J Clin Epidemiol 2001;54(8):774 –\n781. http://www.ncbi.nlm.nih.gov/pubmed/11470385. Accessed 3 Dec 2018\n31. Steyerberg EW (2018) Validation in prediction research: the waste by data\nsplitting. J Clin Epidemiol 103:131 –133. https://doi.org/10.1016/j.jclinepi.\n2018.07.010\n32. Steyerberg EW (2009) Clinical prediction models. A practical approach to\ndevelopment, validation and updating. Springer, New York\n33. Bouwmeester W, Zuithoff NPA, Mallett S et al (2012) Reporting and\nmethods in clinical prediction research: a systematic review. PLoS Med 9(5).\nhttps://doi.org/10.1371/journal.pmed.1001221\n34. Moulder JK, Yunker A (2016) Endometrial ablation: considerations and\ncomplications. Curr Opin Obstet Gynecol 28(4):261 –266. https://doi.org/10.\n1097/GCO.0000000000000282\n35. Kalish GM, Patel MD, Gunn MLD, Dubinsky TJ (2007) Computed\ntomographic and magnetic resonance features of gynecologic\nabnormalities in women presenting with acute or chronic abdominal pain.\nUltrasound Q 23(3):167 –175. https://doi.org/10.1097/RUQ.0b013e31815202df\n36. Orazov MR, Nosenko EN, Radzinsky VE, Khamoshina MB, Lebedeva MG,\nSounov MA (2016) Proangiogenic features in chronic pelvic pain caused by\nadenomyosis. Gynecol Endocrinol 32(sup2):7 –10. https://doi.org/10.1080/\n09513590.2016.1232902\n37. Radzinsky VE, Khamoshina MB, Nosenko EN et al (2016) Treatment strategies\nfor pelvic pain associated with adenomyosis. Gynecol Endocrinol 32(sup2):\n19–22. https://doi.org/10.1080/09513590.2016.1232673\n38. Pontis A, D ’Alterio MN, Pirarba S, de Angelis C, Tinelli R, Angioni S (2016)\nAdenomyosis: a systematic review of medical treatment. Gynecol\nEndocrinol 32(9):696–700. https://doi.org/10.1080/09513590.2016.1197200\n39. Rogachov A, Cheng JC, Erpelding N, Hemington KS, Crawley AP, Davis KD\n(2016) Regional brain signal variability: a novel indicator of pain sensitivity\nand coping. Pain. 157(11):2483 –2492. https://doi.org/10.1097/j.pain.\n0000000000000665\n40. Higgins NC, Bailey SJ, LaChapelle DL, Harman K, Hadjistavropoulos T (2015)\nCoping styles, pain expressiveness, and implicit theories of chronic pain. J\nPsychol 149(7):737–750. https://doi.org/10.1080/00223980.2014.977759\n41. Busca A, Parra-Herran C (2016) The role of pathologic evaluation of\nendometrial ablation resections in predicting ablation failure and\nadenomyosis in hysterectomy. Pathol Res Pract 212(9):778 –782. https://doi.\norg/10.1016/j.prp.2016.06.007\n42. Loffer FD (2006) Endometrial ablation in patients with myomas. Curr Opin\nObstet Gynecol. 18(4):391 –393. https://doi.org/10.1097/01.gco.0000233932.\n06060.aa\n43. Al-Abri R, Al-Balushi A (2014) Patient satisfaction survey as a tool towards\nquality improvement. Oman Med J 29(1):3 –7. https://doi.org/10.5001/omj.\n2014.02\n44. Urden LD. Patient satisfaction measurement: current issues and implications.\nLippincotts Case Manag 7(5):194 –200. http://www.ncbi.nlm.nih.gov/\npubmed/12394558. Accessed December 3, 2018\n45. Yellen E, Davis GC, Ricard R. The measurement of patient satisfaction. J Nurs\nCare Qual 2002;16(4):23 –29. http://www.ncbi.nlm.nih.gov/pubmed/\n12125901. Accessed December 3, 2018\n46. Bjertnaes OA, Sjetne IS, Iversen HH (2012) Overall patient satisfaction with\nhospitals: effects of patient-reported experiences and fulfilment of\nexpectations. BMJ Qual Saf 21(1):39 –46. https://doi.org/10.1136/bmjqs-2011-\n000137\n47. Rama Mohan DD, Kanagaluru Sai Kumar D. A study on the satisfaction of\npatients with reference to hospital services. Vol 1.; 2011. http://www.\nzenithresearch.org.in/images/stories/pdf/2011/Dec/ZIBEMR/2_ZIBEMR_\nVOL1_ISSUE3.pdf. Accessed 3 Dec 2018.\n48. Kilbourne WE, Duffy JA, Duffy M, Giarchi G (2004) The applicability of\nSERVQUAL in cross-national measurements of health-care quality. J Serv\nMark 18(7):524 –533. https://doi.org/10.1108/08876040410561857\nStevens et al. Gynecological Surgery            (2019) 16:7 Page 9 of 9","source_license":"CC0","license_restricted":false}