Systematic review of patient-specific pre-operative predictors of pain improvement to endometriosis surgery

All authors made a substantial contribution to conception and design, or acquisition of data, or analysis and interpretation of data; and in drafting the article or revising it critically for important intellectual content; and in the final approval of the version to be published. All authors agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. In addition, the authors carried out the following tasks: E B: conception, planning, carrying out systematic review, analysing and writing; B K: carrying out systematic review, analysing and writing; B K Y: carrying out systematic review, writing; C R: conception, planning, carrying out systematic review, analysing and writing; J A: conception, planning, carrying out systematic review, analysing and writing; C M: conception, planning, carrying out systematic review, analysing and writing; A H: conception, planning and writing; K M L: conception, planning, carrying out systematic review, analysing and writing; S B: conception, planning and writing; J D: conception and writing.

Given this study was a systematic review on data in the public domain it is exempt from ethics approval.

A systematic review was performed using a prospectively registered protocol as part of a more extensive investigation (PROSPERO CRD42018108604, 04 September 2018) within the CRESCENDO project (peer and Patient and Public Involvement (PPI)-reviewed, NIHR PB-PG-0317-20018). Findings are reported in line with PRISMA guidelines. The search was performed on PubMed, Cochrane and Embase databases from inception to 19 May 2020 without language restrictions. At the time of protocol writing, no relevant core outcomes were published, though one has since been developed ( Duffy et al. 2020 ). In the absence of predictor variables associated with a favourable surgical outcome published in reviews or guidelines, we chose an inclusive search strategy. The search is detailed in Supplementary material (see section on supplementary materials given at the end of this article). A manual search of reference lists of included articles, as well as backward and forward citation tracking, supplemented the database search. When clarification on data was required, authors were also contacted. Two reviewers (E B and B K) screened titles and abstracts separately for eligible articles and reviewed the full-texts of these articles for final study selection. Disagreements were resolved by discussion between reviewers and with a third reviewer (J A). Our interest was in prognostic factors that can be used to identify women most likely to experience pain relief from laparoscopic surgery for the treatment of endometriosis-related pain. Only patient-specific pre-operative factors were explored, surgery-specific factors were beyond the remit of this review, as the former would be the most relevant for patient counselling before surgery. Thus, the inclusion criteria, using the PECO format ( Morgan et al. 2018 ), were: Patients: women with endometriosis Exposure: Women, for whom the presence of any type of prognostic patient-specific factor was reported (this could be any sociodemographic, lifestyle and disease-related factors). We did not specify the prognostic factor before a priori, but approached the search with an open mind and recorded the prognostic factors that were available in the literature and where the pain outcomes were stratified by those predictors. Comparison: Women without the prognostic factor of interest (e.g. parous women (exposure) nulliparous women (non-exposure)) Outcomes: Improved dysmenorrhoea, dyspareunia, non-cyclical pelvic pain and dyschezia or global pain reported after at least 6 months on the visual analogue score (VAS) or as ‘better’ or ‘improved’ vs ‘not better’ or ‘not improved’ Pain relief after surgery had to be reported stratified by the prognostic factor, to allow, if data were available, for the construction of a 4 × 4 table. This means that studies without a comparative element were not included. Excluded from teh study were recurrence and re-operation rates as measures for surgical outcomes, fertility outcomes, a post-operative follow-up time of less than 6 months (the minimum the research group agreed necessary to judge genuine surgical outcomes), studies comparing different surgical techniques, or laboratory tests as predictor variables, reports without predictor variables, abstracts, case reports, conference proceedings, and review articles. E B and C M independently extracted the data on pre- and post-operative pain scores stratified by risk factors. E B and C R assessed the quality of studies using the Newcastle–Ottawa scale ( Wells et al. 2019 ). Findings were reported as a qualitative synthesis due to a paucity of data and variation in reporting, which precluded meta-analysis.

The search returned 14,366 citations; additional backward and forward citation tracking returned one additional paper. After the removal of duplicates, 34 full-text papers were obtained. Of these, 29 were excluded after inclusion and exclusion criteria were applied ( Fig. 1 and Supplementary Table 1). We included five studies ( n  = 606) ( Abbott et al. 2003 , Chopin et al. 2005 , Banerjee et al. 2006 , Milingos et al. 2006 , Ghai et al. 2020 ), two retrospective ( Chopin et al. 2005 , Ghai et al. 2020 ), three prospective. All were from specialist clinics from the global north and included all endometriosis stages (study details Table 1 ). The results reported are presented in Table 2 and the findings of the included studies are detailed in Table 3. Figure 1 PRISMA flow diagram. Table 1 Characterisation of studies. Reference Design Cohort studied Average age (years) Parity Setting Study size and attrition EM location EM stages Prev. treatment Operative approach Hist. confirm. Follow-up time from baseline, y Pre- and post-operative outcome measures LAP ‡ EM Follow-up Abbott et al. (2003) POC ES 1–4 31 (20–48) 0: 80/175 (60%) 2 UK UH 261 176 135 OE: 38%; USL 88% Stage 1: 28%; Stage 2: 28%; Stage 3: 17%; Stage 4: 41% Analgesia: 70%; HT: 70% Prev. LAP: 70% CLET: 100% Yes 5 (3.2 (2–5) ⁑ ) VAS (10 cm) presented as median and IQR for dysmenorrhoea, dyspareunia, non-menstrual PP, dyschezia) Chopin et al. (2005) ROC DE † 31.7 ± 5.4 0.3 ± 0.61 (0–3) 1 French UAH 241 241 132 USL: 59.9%; vagina: 18.9%; bladder: 9.8%; intenstine: 12.1%; multiple locations: 39.4% Stage 1: 20.5%; Stage 2: 34.1%; Stage 3: 24.2%; Stage 4: 21.2% HT: 56%; Prev. LAP: 0.9 ± 1 CLET: 87.1%; LT: 12.9% Yes 3.7 ± 2.0 (median: 3.3) VAS (10 cm) for each pain type (dysmenorrhoea, dyspareunia, CPP, dyschezia, lower urinary tract symptoms) Banerjee et al. (2006) POC ES 1–4 and no EM NS 1 UK tertiary EM centre in DGH 108 88 46 OE: 14%; USL: 59%; recto-vaginal: 43%; pouch of Douglas: 43%; intestine: 47% Stage 1: 39%; Stage 2: 16%; Stage 3: 9%; Stage 4: 36% NS CLET: 100% Yes 18 months post-surgery Pre- and post-operative global pain scores; these are a sum of VAS (0-5) for dysmenorrhoea, dyspareunia, non-cyclical pelvic pain, menstrual dyschezia, non-menstrual dyschezia, menstrual backache, non-menstrual backache Milingos et al. (2006) POC ES 1–4 NS 1Greek UH 258 101 95 OE:61.8%; USL: 59%; recto-vaginal septum: 10.3% Minimal and mild: 21.4%; moderate and severe: 71.6% Prev. LAP: 14.7% CLET; LT: 6.3% no 6 months post-surgery VAS (10) for each pain type dysmenorrhoea, dyspareunia, non- menstrual pelvic pain. Pain scores grouped as 0, 1-5, 6-7, 8-10 Ghai et al. (2020) RS* ES 1-4 NS 1 UK tertiary EM centre in DGH 102 96 100 ND Stages 1–3 ‘superficial’: 48%; stage 4 with bowel involvement ‘severe’: 52% GRA: 6/12 Stages 1–3: laser destruction or excision; Stage 4: excision (97.1 % CLEP); LT: 2.9% Severe: yes; superficial: NS 12 months post-surgery Pre- and post-operative pain component measured with EPH30 questionnaire *Analysis of data from previous databases (1, 2); † with infiltration of USL or Bladder/intestine/vagina; ‡ women with CPP who had laparoscopy; ⁑ mean (range). CLET, complete laparoscopic excisional treatment; DGH, district general hospital; EM, endometriosis; ES, endometrial stage; GRA; gonadotropin receptor antagonist; Hist. confirm., histological confirmation; HT, hormonal treatment; LAP, laparoscopy; LT, laparotomy; ND, not detailed; NS. Not stated; OE, ovarian endometrioma; POC, prospective observational cohort; Prev. previous; ROC, retrospective observational cohort; RS, retrospective secondary; UAH, university affiliated hospital; UH, university hospital. PRISMA flow diagram. Characterisation of studies. *Analysis of data from previous databases (1, 2); † with infiltration of USL or Bladder/intestine/vagina; ‡ women with CPP who had laparoscopy; ⁑ mean (range). CLET, complete laparoscopic excisional treatment; DGH, district general hospital; EM, endometriosis; ES, endometrial stage; GRA; gonadotropin receptor antagonist; Hist. confirm., histological confirmation; HT, hormonal treatment; LAP, laparoscopy; LT, laparotomy; ND, not detailed; NS. Not stated; OE, ovarian endometrioma; POC, prospective observational cohort; Prev. previous; ROC, retrospective observational cohort; RS, retrospective secondary; UAH, university affiliated hospital; UH, university hospital. Presentation of results. Considering the risk of bias for the five studies, one scored low ( Chopin et al. 2005 ), one medium ( Ghai et al. 2020 ), and four ( Abbott et al. 2003 , Banerjee et al. 2006 , Milingos et al. 2006 , Ghai et al. 2020 ) high using the Newcastle–Ottawa tool. While scoring highly in other domains, three studies scored low on ‘comparability’ which may be the result of poor reporting rather than poor study design. We found reports which stratified post-surgical pain relief by disease severity and anatomical site. There were no reported data on the predictive role of sociodemographic factors (for instance age and parity). Chopin et al. (2005 ) retrospectively reported data from a continuous series of women (age not stated) from a French university-affiliated hospital who reported pain (dysmenorrhoea, deep dyspareunia, chronic pelvic pain (CPP) or a pain combination) and deep endometriosis (DE) affecting at least a uterosacral ligament (USL). Of the 241 recruited women with laparoscopy-proven DE, 132 were included with complete follow-up. Only women with a histological lesion of ≥5 mm depth were included. Banerjee et al. (2006 ) recruited women (age not stated), with symptoms suggestive of endometriosis from CPP clinics in a district hospital with tertiary level endometriosis care. One hundred and eight women were recruited; 88 women had histologically confirmed endometriosis, two women had no endometriosis. Of the 88 with endometriosis, 44 women had complete datasets and were analysed. Milingos et al. (2006 ) recruited 274 women with CPP of ≥6 months from university fertility and laparoscopy clinics, of whom 258 underwent laparoscopy, excluding women with a pouch of Douglas obliteration or requiring hysterectomy. One hundred and one women were visually diagnosed with endometriosis during laparoscopy. Abbott et al. (2003) reported 254 women referred with pain symptoms suggestive of endometriosis to two specialist units of whom 132 were included in the analysis. The mean age was 31 years (20–48), 6% were nulliparous, 70% had required analgesia for pain and 73% had hormonal treatment. Seventy per cent had at least one prior diagnostic or operative laparoscopy. Ghai et al. (2020 ) reported a secondary analysis of existing databases ( Kent et al. 2014 , 2016 ) of198 women who had endometriosis surgery. In the group with severe endometriosis, 2.9% were converted to laparotomy. The authors do not report demographics but state no difference between responders and non-responders in age and stage of endometriosis within superficial and deep endometriosis ( Ghai et al. 2020 ). Three studies recruited in England ( Abbott et al. 2003 , Banerjee et al. 2006 , Ghai et al. 2020 ) one in Greece ( Milingos et al. 2006 ) and one in France ( Chopin et al. 2005 ). Authors state CPP as an indicator for surgery; one study also includes fertility ( Ghai et al. 2020 ). In studies that stated the data ( Abbott et al. 2003 , Chopin et al. 2005 ), women averaged 31 years and a large proportion were childless ( Table 1 ). Ethnicity and other sociodemographic factors were not reported. All studies reported dropouts ( Table 1 ). In one study ( Banerjee et al. 2006 ), this involved half of the women who had laparoscopically confirmed endometriosis. Apart from Banerjee et al. (2006 ), all other authors listed previous surgical and medical treatments ( Table 1 ). In all studies, the aim was for complete laparoscopic endometriosis removal. The surgical approach depended on the depth and location. Milingos et al. (2006 ) described ablation of implants (not further specified), lysis of adhesions and excision of fibrosis/endometrioma. Six cases of ‘frozen pelvis’ were converted to open hysterectomy with bilateral oophorectomy and were excluded. Banerjee et al. (2006 ) described excision with monopolar diathermy; rectovaginal and bilateral USL lesions were removed en-bloc, ovarian endometrioma drained and excised, vaginal and bladder endometriosis fully excised and bowel endometriosis treated with shaving or disc resection. Chopin et al. (2005 ) described a ‘see and treat approach’ and excision of all endometriosis lesions ± ureterolysis. Endometrioma were excised; superficial implants were coagulated. Bladder and USL lesions were excised, and vaginal endometriosis was treated with laparoscopically assisted resection. Intestinal lesions were treated by laparoscopy or laparotomy ( n  = 16 not further specified). Abbott et al. (2003 ) reported a previously published excisional technique without hormonal pre-treatment ( Garry et al. 2000 ). Ghai et al. (2020 ) described laser ablationor ultrasonic excision of superficial endometriosis. All DE patients had bowel involvement treated with bowel shaving, disc excision or anterior resection. Women received six months of reoperative gonadotropin-releasing hormone antagonist. In two studies ( Chopin et al. 2005 , Milingos et al. 2006 ), a proportion of complex cases were either converted to or planned as laparotomies. Milingos et al. (2006 ) excluded six cases due to conversion to open surgery. Apart from Milingos et al. (2006 ) and Ghai et al. (2020 ), who included ablation of endometriosis, all others report histological confirmation. The duration of follow up ranged from 6 months ( Milingos et al. 2006 ) to 9 years ( Abbott et al. 2003 ). Three studies scheduled follow up at a single timepoint: Milingos et al. (2006 ) 6, Ghai et al. (2020 ) 12 and Banerjee et al . (2006 ), at 18 months. Chopin et al. (2005 ) reported a mean follow-up of 3.3 years (range 1.0–9.1) and Abbott et al. (2003 ) a meanfollow-up of 3.7 years (range 2–5). Follow-up rates were 94% ( Banerjee et al. 2006 ), 76% ( Abbott et al. 2003 ), 72% ( Ghai et al. 2020 ) for severe endometriosis, 54% ( Chopin et al. 2005 ) and 52% ( Milingos et al. 2006 ). Dysmenorrhoea, dyspareunia, and also non-menstrual pelvic pain or CPP (used synonymously) were measured in all studies. Additional symptoms were menstrual and non-menstrual dyschezia, menstrual and non-menstrual backache and lower urinary tract symptoms. Apart from Ghai et al. (2020 ) to Banerjee et al. (2006 ), researchersused the 10 cm VAS for pain. Banerjee used a 0–5 cm VAS for pain and calculated one global score for each participant. Milingos et al. (2006 ) grouped the VAS results measured as 0 cm, 1–5 cm, 6–7 cm and 8–10 cm, when testing the correlation between pain and endometriosis severity. Furthermore, they created the binary measurement ‘improved’ vs ‘non-improved’ (reduction of ≥2 points) when comparing the post-operative pain reduction of minimal/mild with moderate/severe endometriosis. Ghai et al. (2020 ) measured pain changes using the EPH-30 questionnaire, defining any pain decrease as improvement. Regarding endometriosis severity ( Table 1 ), all studies included all endometriosis stages. The proportion of moderate and severe endometriosis combined is highest in the Milingos study (71.6%) ( Milingos et al. 2006 ), followed by the Abbott study (58%) ( Abbott et al. 2003 ) and in the Chopin ( Chopin et al. 2005 ) and Banerjee Banerjee et al. (2006 ) studies, who both report 45%. Ghai et al. (2020 ) reports DE (excluding moderate severity endometriosis) at 48%. The high proportion of severe endometriosis is likely due to recruitment from specialised centres. The two predictor variables by which outcomes were stratified were endometriosis severity and anatomical site. Within the group of severe endometriosis, Ghai et al. (2020 ) reported higher pre-operative pain and lower feeling of control scores associated with response to surgery. Study findings are shown in Table 3 . Study findings. *Pain scores (median VAS baseline versus follow-up 2–5 years) were all significantly reduced for conditions presented; † values presented as mean ± s.d. (range). EM, endometriosis; LUTS, lower urinary tract symptoms. All five included studies reported endometriosis severity; four considered either AFS stages 1–4 ( Abbott et al. 2003 , Chopin et al. 2005 ) or depth of invasion (superficial/deep) in the relevant analysis ( Banerjee et al. 2006 , Ghai et al. 2020 ). Milingos et al. (2006 ) dichotomised severity into minimal/mild (AFS scores < 16) and moderate/severe endometriosis (≥16), Ghai et al . (2020) reported superficial (stages 1–3) vs severe disease (stage 4). Abbott et al. (2003 ) reported the median and interquartile ranges of the pre-operative and post-operative pain scores for different pain types and compared pain scores for endometriosis stages 1–4 before and after surgery. This study did not compare pain reduction before and after surgery between different stages of endometriosis (such as a test for trend). The reduction in dysmenorrhoea is consistently highly statistically significant ( P  < 0.001) across all endometriosis stages, but women with stage 4 endometriosis showed the highest magnitude in pain reduction across the pain types (dyspareunia <0.0001, non-menstrual pelvic pain <0.0001, dyschezia 0.002). Other stages, while still showing a significant reduction in pain showed lower levels of statistical significance. Only patients with stage 3 endometriosis showed no evidence of pain reduction from dyschezia ( P  = 0.12). Milingos et al. (2006 ) reported higher pre-operative scores for dysmenorrhea and dyspareunia in moderate/severe (group 2) than minimal/mild endometriosis (group 1) ( P  = 0.014 and P < 0.0001, respectively). The authors compared changes in pain scores pre- to post-operatively in two analyses. First, ‘change in pain score’ was depicted graphically for minimal/mild and moderate/severe endometriosis. Without providing numerical data, the authors reported the magnitude for pain score reduction for dyspareunia to be higher in the group with moderate/severe endometriosis ( P  = 0.04). Differences for dysmenorrhoea and for non-menstrual pelvic pain were not statistically significant ( P  = 0.082 and P  = 0.56, respectively). For dysmenorrhoea, the differences may have been clinically significant, as the authors reported a benefit. Secondly, the authors looked at subgroups of women, who had ‘improved’ pain scores for dysmenorrhoea ( n  = 52), dyspareunia ( n  = 38), and non-menstrual pain ( n  = 30) after surgery (≥2 cm VAS reduction) and compared the proportions with minimal/mild and moderate/severe endometriosis. Regarding women with improved dysmenorrhea ( n  = 52), 43% had minimal/mild and 66% moderate/severe endometriosis ( P  = 0.0037). Of the women reporting improved deep dyspareunia ( n  = 38), 33% had minimal/mild and 67% had moderate/severe endometriosis (‘not significant’) and for non-menstrual pain ( n  = 30) 67% had minimal/mild and 56% had moderate/severe endometriosis (‘not significant’). Banerjee et al. (2006 ) reported a global pain score for three groups of women: no endometriosis, isolated superficial endometriosis and DE ± superficial endometriosis. Global scoring was 35 maximum points, the sum of 0–5 points for each of dysmenorrhoea, dyspareunia, non-cyclical pelvic pain, menstrual dyschezia, non-menstrual dyschezia, menstrual backache, non-menstrual backache. The surgeon visually distinguished between superficial peritoneal and deep infiltrating/nodular lesions. Data indicate a correlation between deep/superficial classification and AFS staging (chi-square test of association: X2(3) = 25.8 P  < 0.001). Pre- and post-operative global pain scores were compared using a paired T-test in all three groups, women without endometriosis ( n  = 2; P  = 0.30), with only superficial endometriosis ( n  = 17; P  = 0.43), and with DE ± superficial endometriosis ( n  = 27; P  = 0.004). The authors concluded surgery did not reduce pain scores in superficial endometriosis but was valuable in DE. We agree but note the small group size. Ghai et al. (2020 ) reported a significantly higher proportion of women treated for severe endometriosis responding to surgery ( n  = 86/96) than for superficial disease (77/102; P  = 0.0089). Women with severe endometriosis were more likely to respond if they had higher pre-operative EPH-30 pain scores (median: 66, range: 24–83) vs lower scores (median: 50; range: 20.5–63.6) and lower scores for ‘feeling of control’ (60.25; range: 47.7–72.7 vs 62.5, range: 45.8–70.8). All authors, apart from Ghai et al. (2020 ), detailed endometriosis location; USL endometriosis was listed in four studies, ovarian endometrioma in three ( Abbott et al. 2003 , Banerjee et al. 2006 , Milingos et al. 2006 ) and rectovaginal septum ( Milingos et al. 2006 ) and intestinal endometriosis in two ( Chopin et al. 2005 , Banerjee et al. 2006 ). Chopin et al. (2005 ) reported pre- and post-operative pain scores stratified by location: USL, vagina, bladder and intestine. Pre- and post-operative differences in pain scores were compared for each location, but locations were not compared with each other. Removal of USL endometriosis ( n  = 78) resulted in highly significant reduction across all five pain types (dysmenorrhoea: P  < 0.001, deep dyspareunia: P  < 0.001, dyschezia: P  = 0.001, lower urinary tract symptoms: P  = 0.011, and non-cyclical pelvic pain: P  < 0.001). Vaginal ( n  = 25) and intestinal ( n  = 16) endometriosis excision was associated with a significant reduction of four pain types (dysmenorrhoea: P  = 0.001 and P  = 0.004, respectively, deep dyspareunia: P  = 0.001 and P  = 0.015, respectively, dyschezia: P  = 0.007 and P  = 0.033, respectively, and non-cyclical pelvic pain P  = 0.022 and P  = 0.027, respectively), but not lower urinary tract symptoms ( P  = 0.0679 and P  = 0.0697, respectively). Removal of bladder endometriosis ( n  = 13) resulted in a significant reduction in dysmenorrhoea P  = 0.022, deep dyspareunia P  = 0.0117 and lower urinary tract symptoms P  = 0.022, but not dyschezia P  = 0.0697. Non-cyclical pelvic pain reduction could not be ascertained due to missing data. Two excluded studies for which we obtained full texts merit further discussion. Sutton et al. (1994 ) was excluded due to limited presentation of results. Seventy-four women from gynaecology clinics with symptoms suggesting endometriosis were included. Visual assessment at laparoscopy showed minimal ( n  = 29), mild ( n  = 28) and moderate ( n  = 6) endometriosis, which was destroyed with laser and not histologically confirmed. Follow-up was 3 and 6 months post-operatively. Of the 74 recruited women 63 completed the study. Women recorded the intensity of global pain on a 10 cm VAS and also ‘how pain had changed’. The proportion of women with pain alleviation stratified by endometriosis stage was graphically displayed, without numerical values or significance testing. The proportion of women with stage 3 endometriosis is depicted at 100 ‘percentage better’ whereas the percentage in stage 1 endometriosis is depicted below 50 ‘percentage better’. The authors were unsuccessfully contacted for their raw data. However, they concluded that the severity of pain experienced by endometriosis patients may be used to predict their response to surgery. A retrospective cohort study by Harris et al. (2020 ) recruited 972 women who underwent therapeutic laparoscopy for confirmed endometriosis. In total 398 women had complete follow-up reported 6/52 weeks post-operatively. This study was excluded because of short follow-up. Global pain was recorded as ‘pain improvement/resolution’ vs ‘no improvement’. The proportion of women with improvement/resolution was higher if women: were ‘not Caucasian’ ( n  = 188, 67.7%) vs ‘Caucasian’ ( n  = 90, 32.4%) – OR: 0.60, CI 0.37–0.99, P  = 0.046; were operated on by a specialised endoscopic gynaecologist ( n  = 75, 83.0%) vs not ( n  = 15, 16.7%) – OR: 0.42, CI: 0.18–0.94, P  = 0.036; had a history of CPP ( n  = 29, 55.8%) vs not ( n  = 23, 44.2%) – OR: 2.0, CI: 1.14-3.76, P  =0.02); had stage 3–-4endometriosis ( n  = 128, 83.1%) vs stage 1–2 ( n  = 26, 16.9%) – OR: 0.35, CI: 0.21–0.57, P  < 0.001.

The current systematic review identified severity of endometriosis as a possible predictor for surgical response based on a small number of studies, mostly assessed as having a ‘high risk of bias’. The review has also shown there is a knowledge gap that needs to be filled. A multicentre RCT to clarify if low stage endometriosis removal causes any improvement in pain scores is planned ( Horne et al. 2019 ). We are also currently producing an algorithm to predict surgical success in women with confirmed or suspected endometriosis (CRESCENDO, NIHR PB-PG-0317-20018) using pre-existing databases ( Daniels et al. 2009 , Byrne et al. 2018 , Khan 2018 ). Given the review findings we recommend that future studies should be designed more robustly and less heterogeneously. An important element is the reporting of pre-defined core outcome sets for endometriosis treatment ( Hirsch et al. 2016 , Duffy et al. 2019 ). With standardised reporting, studies can be adequately compared, synthesised and meta-analysed. A core outcome set for endometriosis has recently been published ( Duffy et al. 2020 ) that includes overall pain, improvement in the most troublesome symptom and quality of life, and its adoption may create more substantive evidence in the future.

Four of the five included studies indicate that stronger pain relief after endometriosis surgery was related to more severe disease prior to surgery ( Chopin et al. 2005 , Banerjee et al. 2006 , Milingos et al. 2006 , Ghai et al. 2020 ). Although the current review returned a limited quantity and quality of evidence, the ‘theme severity of endometriosis’ is consistent across studies and warrants further investigation to determine whether it may be used in the future to counsel women about laparoscopic surgery for endometriosis. Endometriosis severity may be only fully understood during laparoscopy. Nonetheless, there are clinical pointers to DE, such as the severity of symptoms ( Fedele et al. 1992 , Ferrero et al. 2005 ), USL nodularity, and the ‘kissing ovary’ sign on scan, which may be used as surrogate markers for disease severity. More research is needed to quantify the value of using these in treatment decision making ( Matorras et al. 1996 , Ghezzi et al. 2005 ). The strengths of this systematic review include a thorough literature review following PRISMA guidelines and assessment of studies using the Newcastle–Ottawa quality tool. However, due to the limitations of the available data and the high risk of bias scores we are unable to make definitive conclusions about predictors of surgical success. It was surprising to find so few studies focussing on patient-specific predictors of favourable surgical outcomes, given the large number of series that report evidence of a reduction of endometriosis-related pain scores after surgery ( Garry et al. 2000 , Ford et al. 2004 , Wykes et al. 2006 , Angioli et al. 2014 , De la Hera-Lazaro et al. 2016 , Byrne et al. 2018 , Rindos et al. 2020 ) and the large numbers of affected patients. Reviewed studies included women with advanced endometriosis, treated in specialist centres and with reported complete excision. Surgical factors that could influence operative outcomes – such as whether excision is complete – are highly relevant to future research. Studies show less pain reduction in incomplete compared to complete surgery ( Hidaka et al. 2012 , Angioni et al. 2015 , Cao et al. 2015 ). Thus a systematic review of three randomised controlled trials (RCTs) with 335 women indicates superior reduction of dysmenorrhea (mean difference (MD) = 0.99; 95% CI: −0.02 to 2.00; P   = 0.05) and dyschezia (MD = 1.31; 95% CI: 0.33–2.29; P   = 0.009) using excision compared to ablation, but not in dyspareunia (MD = 0.96; 95% CI: −0.07 to 1.99; P   = 0.07) ( Pundir et al. 2017 ). Conversely, a later RCT of 73 women with endometriosis ablation and excision showed no difference in dysmenorrhoea but a difference in dyspareunia at 6 months (mean change −22.96; 95% CI: −39.06 to −6.86; P   = 0.01) ( Riley et al. 2019 ). The studies included in the present review used the r-AFS scoring or a score deduced from it. Whilst the AFS score was designed to predict fertility and puts strong weighting on endometriotic cysts, it may correlate less well with pain ( Vercellini et al. 1996 ), whereas the ENZIAN score ( Haas et al. 2013 , Montanari et al. 2019 ) may have stronger correlation in pain in DE. The location of endometriosis in the USL and its removal may have a special role in pain relief after surgery ( Chopin et al. 2005 , Chapron & Dubuisson 1996 ), and appears to be closely associated with the symptom of dyspareunia ( Porpora et al. 1999 , Fauconnier et al. 2002 , Montanari et al. 2019 ). The presence of endometriosis is specifically associated with tenderness of the cul-de-sac or USL during examination ( Yong et al. 2017 ). This can help indicate the presence of DE. Debate remains whether surgical removal of endometriosis can relieve non-cyclical pelvic pain. Abbott et al. (2003 ), Chopin et al. (2005 ), and Banerjee et al. (2006 ), but not Milingos et al. (2006 ) reported evidence of improvement of non-cyclical pelvic pain. Pain scores for non-cyclical back pain and non-cyclical dyschezia failed to show evidence of improvement after removal of endometriosis in one paper that included these outcomes ( Banerjee et al. 2006 ). These symptoms may have causes other than endometriosis, as also can non-cyclical CPP that is resistant to laparoscopic endometriosis treatment. The use of post-operative adjuvant hormone treatment (such as the oral contraceptive pill or levonorgestrel intrauterine device) could have been a confounding variable for pain improvement, especially dysmenorrhoea. However, this detail is not provided in the studies included.

Endometriosis is a chronic inflammatory condition affecting 6–10% of women of reproductive age, defined by the presence of endometrial-like tissue outside the uterus, commonly affecting the peritoneum, ovaries and other pelvic organs ( Viganò et al. 2004 ). Endometriosis impacts on many aspects of daily life and is associated with considerable costs to health services and society ( Simoens et al. 2012 ). Commonly, women with endometriosis experience infertility and fatigue as well as pain, the latter often worsening during menses (dysmenorrhoea) and sexual intercourse (dyspareunia). In addition, pain may occur during bowel movements (dyschezia) or in a non-cyclical fashion. There is no cure for endometriosis and current established treatments show an inconsistent response. Laparoscopic removal of endometriosis (therapeutic laparoscopy) remains the mainstay of treatment for endometriosis-associated pain (as described above) as a stand-alone intervention ( Zanelotti & Decherney 2017 ), after the failure of or in conjunction with medical treatment ( Duffy et al. 2014 ). There is a distinction between diagnostic and therapeutic laparoscopies, and clinicians are advised to use a combined ‘see and treat’ approach for most cases ( Ball et al. 2008 ). A recent meta-analysis ( Leonardi et al. 2020 ), which included two studies also reviewed in this paper ( Sutton et al. 1994 , Abbott, 2004 #12), demonstrated that operative laparoscopy was more effective for pain relief at 6 months than diagnostic laparoscopy ( n  = 102; RR 2.65; 95% CI 1.61–4.34, P  < 0.001). Unfortunately, between 20% ( Abbott et al. 2004 ) and 28% ( Sutton et al. 1994 ) of women with endometriosis pain do not respond to therapeutic laparoscopy (pre- and post-operative pain scores are not different), but it is not known which subgroup of women will respond and which will not. A recent meta-analysis ( Leonardi et al. 2020 ) entitled ‘When to Do Surgery and When Not to Do Surgery for Endometriosis’ failed to identify sufficient evidence to answer this question. The location and the severity of endometriosis commonly staged 1–4 using the revised American Fertility Society grading system (r-AFS) (The American Fertility Society 1985 ) may correlate with patients’ symptoms ( Fauconnier et al. 2002 , Sinaii et al. 2008 ) and it could be hypothesized that these factors may also have prognostic value for treatment response. If clinicians knew which subgroup of endometriosis patients benefitted from laparoscopic surgery, they would be better able to counsel their patients and manage their expectations. Access to therapeutic laparoscopy, which is a costly, limited resource associated with anaesthetic and surgical risks, could be better managed. This review aims to determine which women will benefit from therapeutic laparoscopy for endometriosis.

None of the authors declare financial, personal, political, intellectual or religious interests relating to the current paper. Andrew Horne is a Co-Editor-in-Chief of Reproduction and Fertility. Andrew Horne was not involved in the review or editorial process for this paper, on which he is listed as an author.