The effectiveness of reproductive surgery in the treatment of female infertility: facts, views and vision

Hysteroscopic removal of endometrial polyps detected by ultrasound significantly doubles the clinical pregnancy rate when compared to diagnostic hysteroscopy and polyp biopsy according to one randomised trial including 215 subfertile women with uterine polyps undergoing IUI (RR 2.2, 95%CI 1.6-3.1) (Pérez-Medina et al., 2005). For every three subfertile women with uterine polyps treated by hysteroscopic polypectomy, it is expected that one additional person will have a clinical pregnancy (NNT = 3, 95%CI 2 to 5). In patients with submucosal fibroids with or without intramural fibroids and otherwise unexplained subfertility, hysteroscopic myomectomy doubles the pregnancy rate compared to expectant management (RR 2.2, 95% CI 1.6- 2.9) ( Fig. 7 ) as demonstrated by a meta-analysis of two randomised trials in 298 patients (Casini et al., 2006; Shokeir et al., 2009). For every three women with submucosal fibroids and otherwise unexplained infertility treated by hysteroscopic myomectomy, it is expected that one additional person will have a pregnancy compared to expectant management (NNT 3, 95%CI 2 to5). There is no evidence of significant statistical heterogeneity (Chi² = 0.22; I² = 0%). We did not find RCTs on the effectiveness of hysteroscopic septum resection compared to expectant management or alternative treatments in patients with otherwise unexplained primary subfertility. One randomised trial compared the effectiveness of two methods of hysteroscopic treatment of uterine septa (resectoscopy versus Versapoint electrode) in a mixed population of 160 patients with subfertility and recurrent pregnancy loss (Colacurci et al., 2007) and found no differences in outcome between both techniques. A randomised trial (http://www.studies-obsgyn.nl/trust NTR 1676) studying the effectiveness of hysteroscopic metroplasty in patients with recurrent pregnancy loss is ongoing. There are no randomised trials on the effectiveness of hysteroscopic synechiolysis with pregnancy or live birth rates as primary outcome. We excluded one pseudo-randomised trial on the effectiveness of hysteroscopy in treating intrauterine adhesions (Pabuccu et al., 2008). Furthermore we excluded two randomised trials on the effectiveness of auto-cross linked hyaluronic acid gel in the prevention of intra-uterine adhesions after hysteroscopic adhesiolysis (Acunzo et al., 2003) and after hysteroscopic surgery (Guida et al., 2004) since data on reproductive outcome are lacking. A systematic review (El-Toukhy et al., 2008) with a meta-analysis of two randomised trials (n = 941) (Demirol and Gurgan, 2004; Rama Raju et al., 2006) demonstrated that office hysteroscopy in the cycle preceding a next IVF attempt nearly doubles the clinical pregnancy rate in infertile patients with at least two failed IVF attempts compared to starting IVF immediately (RR 1.6, 95% CI 1.3-1.9). For every seven infertile women with at least two failed IVF attempts treated by office hysteroscopy prior to a subsequent IVF cycle, it is expected that one additional person will have a clinical pregnancy compared to starting IVF immediately (NNT = 7, 95%CI 5 to12). There is no evidence of significant statistical heterogeneity (Chi² = 0.16; I² = 0%). A subgroup analysis in the patients undergoing office hysteroscopy demonstrated no difference in clinical pregnancy rates irrespective whether pathology was detected and treated or not (RR 0.91, 95% CI 0.71-1.2).

Statistical analysis was done using the latest updated software provided by the Cochrane Collaboration (Rev Man 5 version 5.0.24, April 16th 2010). Dichotomous data were extracted in 2x2 tables. After consulting with a biostatistician we decided to express the results of individual trials and of the meta-analyses as risk ratios (RR) with 95% confidence limits (95% CI) using a fixed effects model (Mantel- Haenszel method). A sensitivity analysis comparing the use of odds ratios (OR) to risk ratios did not yield differences in the direction of the observed treatment effect, but in general the estimations of the treatment effect were more conservative with the use of RR. To facilitate clinical interpretation, results were re-expressed as numbers needed-to-treat (NNT) with 95% confidence intervals (CI). The NNTs were calculated from the RR in the meta-analyses using appropriate mathemical formulae available from the Cochrane Handbook for Systematic Reviews of Interventions. Statistical heterogeneity was assessed with the Chi-square test and the I² test. We estimated the risk of bias at the study level and across studies using the risk of bias tool provided by the Cochrane Collaboration based on the randomisation sequence generation, concealment of allocation, blinding of patients/ physicians and outcome assessors, selective reporting of outcomes, whether or not incomplete data were addressed and the probability of other forms of bias. We assessed the quality of the included trials by providing levels of evidence using the software provided by the Cochrane Collaboration (GRADE profiler version 3.2.2.20090501). An alternative grading system was used by allocating a ‘level E’ for ‘evidence of an effect’ when there was evidence of a significant difference between the interventions studied for the pregnancy or live birth rate or a’level G’ for ‘gap in evidence’ when there was insufficient evidence of effectiveness or harm. A label ‘E&G’ was allocated when there was some evidence of a significant effect along with some gaps for the primary outcomes.

For the current systematic review no written protocol was registered. We aimed to identify randomised clinical trials (RCTs) or systematic reviews of such trials on reproductive surgery in infertile women with pregnancy or live birth rate as the primary outcome. To do so, we searched the Cochrane Library (1970 to June 1st 2010) for relevant trials in the CDSR, DARE, CENTRAL or HTA databases. We also searched the National Library of Medicine’s MEDLINE using a combination of textwords and MeSH terms for “laparoscopy”, “hysteroscopy” ,“infertility”, “pregnancy rate” and “live birth rate” (1966-June 1st 2010) and Excerpta Medica EMBASE (1974 to June 1st 2010). The search strategies in the appropriate thesaurus for each database were developed by a librarian at the University Biomedical Library Campus Gasthuisberg, Katholieke Universiteit Leuven and are described in the addendum. The search strategy included the filters from the Cochrane Collaboration website (Haynes et al., 1994; Dickersin et al., 1994), developed for the detection of randomised controlled trials with reported sensitivities of 95 and 99% (Shojania and Bero, 2001; Robinson and Dickersin, 2002; Glanville et al., 2006). We also searched the Current Controlled Trials (metaRegister) at http://www.controlled-trials.com/ for registered relevant trials. The last up-dated search was done on June 1st 2010. Language restrictions were not applied, and the searches were done simultaneously and independently by two authors (JB and TD).The reference lists of all known primary articles were examined independently by the same two authors (JB and TD). We also used the ‘related articles’ function of PubMed, as well as the reference lists of detected articles to look for relevant studies. We included RCTs reporting on pregnancy or live birth rates, but excluded non-randomised trials, studies not reporting on reproductive outcome or done in a population with gynaecological problems other than infertility as well as trials on diagnostic accuracy, technical feasibility, patient compliance and cost-effectiveness. Study selection was done by two authors independently based on reviewing the full text article (JB and TD). The κ-value of inter-reviewer agreement on the final inclusion of the relevant RCTs was 0.82. In case of disagreement, the opinion of a third author was asked (SW) until consensus was reached. For data extraction and critical appraisal of the methodology, the PRISMA Statement was followed (Moher et al., 2009). The characteristics of the different study populations, the control and study intervention as well as other relevant study characteristics were extracted from the full text articles. The corresponding authors were contacted in case of unclear study methodology or to obtain missing data. Two authors (JB and TD) independently extracted relevant study data. The critical appraisal based on the internal validity, the magnitude of the treatment effect and the general applicability was done by using standardised work sheets for quality assessment of randomised controlled trials from the website of the Dutch Cochrane Centre ( http://dcc.cochrane.org/sites/dcc.cochrane.org/files/uploads/RCT.pdf ). The internal validity was based on the description of the randomisation sequence generation, allocation concealment, blinding of physicians/patients and outcome assessors, the assessment of incomplete data and whether selective reporting or other forms of bias were likely or not.

The process of literature search and selection is described in Figure 1 . We retrieved 106 possibly relevant articles in MEDLINE, 334 possibly relevant articles in EMBASE and 170 relevant reviews, clinical trials and abstracts of reviews from the Cochrane Library. We identified 37 possibly relevant trials in the Current Controlled Trials Register. After screening of the abstracts or titles of non duplicate 592 possibly relevant publications, 97 full text articles were assessed for eligibility. Finally, we included 63 randomised trials on reproductive surgical techniques and pregnancy outcome. An overview of these trials and a summary of their findings grouped according to the specific pathology or clinical setting are presented in table 1 . DR: delivery rate FR: fecundity rate PR: pregnancy rate c: cycles m: months CPR: cumulative pregnancy rate LBR: live-birth rate OPR: ongoing pregnancy rate THBR: take home baby rate TPR: total pregnancy rate TDR: term delivery rate PR/P: pregnancy rate per patient PR/C: pregnancy rate per cycle CPR/P: clinical pregnancy rate per patient LBR/C: live-birth rate per cycle LBR/P: live-birth rate per patient

Compared to IVF, reproductive surgery has the potential to restore the natural procreation of the subfertile couple leading to several conceptions after one successful intervention. Repeated success can however only be achieved by effective interventions. A randomised controlled trial is the current gold standard of examining the effectiveness of interventions By consequence, clinical research in the field of reproductive surgery should ideally be guided by high quality randomised trials whenever there is uncertainty about effectiveness of a specific surgical intervention. The evidence provided by the meta-analysis on treating minimal or mild endometriosis in women with unexplained infertility shows a beneficial effect in favour of the excision/ ablation and adhesiolysis, despite the fact that the two major trials show different results. The long term risks versus benefits ratio in treating infertile patients with endometriotic cysts by the excision technique should be addressed, whether or not in the IVF setting. The use of IVF for tubal pathology at the expense of reproductive surgery should not be continued without adequate randomised trials studying the harms and benefits of both treatments head-to-head. We need randomised trials studying the effectiveness of the laparoscopic and/ or hysteroscopic removal of intramural fibroids close to the junctional layer of the myometrium or with impression on the uterine cavity in patients with unexplained subfertility and prior to IUI or IVF treatment. A trial on the effectiveness of hysteroscopic removal of uterine septa in patients with recurrent pregnancy loss is currently ongoing. The effectiveness of the anti-adhesion barriers in restoring the normal fertility potential in patients with severe intra-uterine adhesions should be studied by randomised trials. Before promoting hysteroscopy as a screening tool in the infertile population undergoing ART, we should wait for the results of the randomised trial on the effectiveness of hysteroscopy before a first IVF or IUI attempt. All future RCTs should focus not only on the beneficial short term effects of the intervention, but should address the possible detrimental long term effects on female fertility. This is the only sound way to measure the ‘true’ effectiveness of a reproductive surgical intervention.

The laparoscopic treatment of all visible implants of minimal-mild endometriosis in women with otherwise unexplained subfertility is likely to be beneficial since it might increase the chance of a live birth or ongoing pregnancy. The two major trials do however report conflicting results. A beneficial effect of treating minimal-mild endometriosis is in accordance with the pooled data (Hughes et al., 1993; Adamson and Pasta, 1994) from one pseudo-randomised trial (Nowroozi et al., 1987) and two cohort studies (Fayez et al., 1988; Paulson et al., 1991) but is not confirmed by other observational studies (Seiler et al., 1986; Levinson, 1989; Chong et al., 1990). It should be noted that atypical endometriotic lesions have not been included in the earlier studies. The wider eligibility criteria in the Italian study, namely the longer duration of subfertility and the higher prevalence of more advanced endometriosis could have led to the unintentional inclusion of more patients with a less favourable prognosis, explaining the absent treatment effect in the Italian trial. The results of the Italian trial are more correctly interpreted if one accepts its division into two different subgroups based on whether co-treatment with GnRH agonists was given or not. A type II error due to low statistical power may therefore be responsible for the absence of statistically significant differences between the intervention and control groups. The wider eligibility criteria and the co-treatment with GnRH agonists have caused the substantial statistical heterogeneity in the meta-analysis of the two major trials. The underlying pathophysiological mechanism linking minimal-mild endometriosis to subfertility is still largely unknown. Therefore, controversy still exists whether there is a causal link between these lesions and subfertility (Olive and Schwartz, 1993; Vercellini and Crosignani, 1993). The excision of endometriotic cysts is superior to simple drainage and ablation for increasing the spontaneous conception rate. Some authors have observed that ovarian tissue may be inadvertently excised together with the endometrioma wall in the majority of patients (Muzii et al., 2005), which could lead to a reduction in ovarian volume (Exacoustos et al., 2004). There are many observational studies reporting conflicting results concerning the impact of ovarian cystectomy on the ovarian responsiveness (Yazbeck et al., 2006; Nargund et al., 1996; Loh et al., 1999, Ho et al., 2002; Marconi et al., 2002; Alborzi et al., 2007; Horikawa et al., 2008; Canis et al., 2001; Donnez et al., 2001; Wyns and Donnez, 2003). As a consequence there is uncertainty whether ovarian cystectomy of endometriotic cysts despite its favourable effect in the short term, could have a deleterious impact on the ovarian reserve in the longer term. The absence of a treatment effect in favour of the excision of endometriotic cysts smaller than 6 cm prior to IVF on the pregnancy rates compared to starting IVF immediately is in accordance with the results of an earlier observational study (Garcia-Velasco et al., 2004). In the randomised trial (Demirol, 2006) the trend of lower pregnancy rates after ICSI in the patients who were treated with cystectomy could be explained by the longer stimulation period, a higher gonadotropin requirement and a lower oocyte number: the absence of statistically significant differences due to the low number of included patients cannot enable at the present time to draw definitive conclusions on the need to perform an ovarian cystectomy prior to IVF. Laparoscopic ovarian diathermy as a second-line treatment in women with clomiphene-resistant PCOS results at least in equal pregnancy rates and decreases the risk for multiple pregnancy compared to gonadotropin treatment, irrespective whether the technique is used uni- or bilaterally . The underlying physiological mechanism of action might be due to both local and systemic effects, resulting in follicular recruitment, maturation and ovulation (Aakvaag 1985; Armar et al., 1990; Balen et al., 1993; Greenblatt and Casper, 1987). It is however unknown how long the treatment effect of LOD lasts, although repeated spontaneous ovulations and subsequent pregnancies after a first pregnancy or miscarriage have been reported (Farquhar et al., 2002). Ovarian adhesions after the LOD procedure have been described, but their clinical relevance is unclear (Greenblatt and Casper, 1993). The theoretical risk of inducing premature ovarian failure needs to be addressed since some observational studies have described a significant reduction of the ovarian reserve after LOD (Weerakiet et al., 2007). Prior to IVF treatment, the laparoscopic removal of an ultrasonographically visible hydrosalpinx doubles the live birth rate compared to starting IVF immediately. This supports the observed negative impact of tubal infertility due to hydrosalpinx on the implantation rates in IVF treatment (Camus et al., 1999). At the present, there is no evidence to support performing bilateral salpingectomy whether or not bilateral hydrosalpinges are present. The pooled data from two RCTs (Dechaud et al., 1998; Strandell et al., 1999) confirm that salpingectomy for hydrosalpinx prior to IVF is effective before a first IVF treatment cycle. Some observational studies have studied the effect of salpingectomy on the ovarian reserve. One clinical controlled trial demonstrated significantly higher baseline FSH levels after salpingectomy as well as a lower ovarian response to stimulation but the pregnancy rates were similar in both groups (Gelbaya et al., 2006). Another clinical controlled trial equally found higher baseline FSH levels after laparoscopic salpingectomy compared to proximal tubal division but pregnancy rates per patient did not differ between both groups (Nakagawa et al., 2008). The possible long term negative impact of salpingectomy on female fertility should be addressed by future RCTs. Alternatively, the occlusion of a hydrosalpinx is as effective as salpingectomy (Kontoravdis et al., 2006; Moshin and Hotineanu, 2006) whereas the ultrasound-guided transvaginal needle aspiration shows a trend in doubling the clinical pregnancy rate (Hammadieh et al., 2008) but its effect was statistically not significant. The negative impact of a hydrosalpinx on the outcome of IVF is hypothetically explained by the intermittent bathing of the uterine cavity with toxic fluid within the hydrosalpinx, which may lower the endometrial receptivity (Akman et al., 1996; Fleming and Hull, 1996; Freeman et al., 1996; Katz et al., 1996; Strandell et al., 1994) possibly by reducing the endometrial expression of β- integrin (Meyer et al., 1997). Alternative hypothetical mechanisms of action include direct embryo toxicity as demonstrated in a murine model (Mukherjee et al., 1996) or a negative impact on oocyte growth and development during early follicular recruitment (Freeman et al., 1996). Randomised trials studying the effectiveness of tubal surgery compared to expectant management and IVF in terms of livebirth rates are lacking, as has been reported by other authors (Pandian et al., 2008). The impact of fibroids on fertility remains controversial (Pritts, 2001; Lefebvre et al., 2003; Vilos, 2003; Griffiths et al., 2006; Somigliana et al., 2007; Vimercati et al., 2007; Somigliana et al., 2008; Klatsky et al., 2008; Pritts et al., 2009) despite an abundancy of observational studies (Seoud et al., 1992; Narayan and Goswamy, 1994; Farhi et al., 1995; Lumbiganon et al., 1996; Eldar-Geva et al., 1998; Marshall et al., 1998; Ramzy et al., 1998; Stovall et al., 1998; Bulletti et al., 1999; Bajekal and Li, 2000; Dietterich et al., 2000; Healy, 2000; Hart et al, 2001; Jun et al., 2001; Surrey et al., 2001; Wang et al., 2001; Check et al., 2002; Donnez and Jadoul, 2002; Ng and Ho, 2002; Yarali and Bukulmez, 2002; Bulletti et al., 2004; Manyonda et al., 2004; Oliveira et al., 2004; Parazzini et al., 2004; Wang and Check, 2004; Wise et al., 2004; Benecke et al., 2005; Gianaroli et al., 2005; Ng et al., 2005; Rackow and Arici, 2005; Surrey et al., 2005; Khalaf et al., 2006; Klatsky et al., 2007). Some observational data suggest that submucosal, intramural and subserosal fibroids interfere with female fertility in decreasing order of importance (Somigliana et al., 2007) whereas other non-controlled studies have suggested that the number, size and distorsion effect of fibroids on the uterine cavity may be more important (Bulletti et al., 1999; Varasteh et al., 1999; Bernard et al., 2000; Fernandez et al., 2001; Oliveira et al., 2005, Khalaf et al., 2006; Mukhopadhaya et al., 2007). Fibroids are believed to interfere with sperm migration, ovum transport and embryo implantation (Richards et al., 1998), which may explain why the hysteroscopic removal of submucosal fibroids doubles the clinical pregnancy rates compared to expectant management in women with otherwise unexplained subfertility. Many hypothetical mechanisms have been put forward such as altered contours of the uterine cavity resulting in altered mechanical pressure or abnormal uterine contractility (Bettocchi et al., 2002; Farrugia et al., 2002; Oliveira et al., 2004), local inflammation, focal endometrial vascular disturbances, chronic endometritis, secretion of vasoactive substances or an enhanced intrauterine androgen environment(Richards et al., 1998). Hysteroscopic polypectomy prior to IUI doubles the pregnancy rates but at the present we cannot recommend the systematic removal of all polyps in subfertile women based on one RCT. Observational studies have suggested a possibly higher impact of tubocornual polyps on female fertility (Venturini et al., 1987; Brooks et al., 1990; Lee et al., 1997; Shokeir et al., 2004; Yanaihara et al., 2008). The effect of the size, number and the localisation of endometrial polyps on fertility should be examined as well as possible association between endometrial polyps and endometriosis (Mc Bean et al., 1996; Kim et al., 2003). Two randomised trials have confirmed the effectiveness of auto-cross linked hyaluronic acid gel in the prevention of intra-uterine adhesions after hysteroscopic adhesiolysis (Acunzo et al, 2003) and after hysteroscopic surgery (Guida et al., 2004) but have unfortunately failed to present data on the fertility outcome. The higher pregnancy rates after hysteroscopy even in the absence of intrauterine pathology in women with recurrent IVF failure is an unexpected observation which nevertheless could be explained by the cervical dilatation and/or direct hysteroscopic visualisation of the uterine cavity facilitating embryo transfer (Mc Manus et al., 2000; Mansour and Aboulghar, 2002) or alternatively by an immunological mechanism triggered by the hysteroscopic manipulation or by the effect of the distension medium on the endometrium. The hypothecical immunological mechanism which may similarly explain the increased odds of spontaneous pregnancy after hysterosalpingography (Luttjeboer et al., 2007) is currently under study in an ongoing randomised trial ( NCT 00367367 ) (Geslevich et al., 2006). The results of a registered randomised trial on the effectiveness of hysteroscopy before a first ICSI treatment cycle have not been published yet to the best of our knowledge ( NCT 00830401 ). We included only randomised trials in this systematic review because this provides the least biased measure of the effectiveness of interventions (Benson and Hartz, 2000; Britton et al., 1998; Concato et al., 2000; McKee et al., 1999; Kunz et al., 2001; Johnson et al., 2008). A recent systematic review of Cochrane reviews on gynaecological surgery demonstrated that the treatment effects tended to be overestimated systematically in trials without allocation concealment, although the difference was not statistically significant (Selman et al., 2008). This finding is consistent with the current understanding of the mechanisms of allocation concealment bias (Kunz and Oxman, 1998). A graphical presentation of the risk of bias of all the included studies is presented in Figure 8 . A summary of the risk of bias of all the individual trials included in this systematic review is given in Figure 9 . Nearly 75% of all studies have an adequate randomisation sequence generation, while nearly 50% have adequate allocation concealment and less than 20% have adequate blinding. The overall quality of the included trials in the present systematic review is by consequence mediocre. Therefore, we should be cautious in making definitive conclusions. At the present we should refrain from providing guidelines for clinical practice in reproductive surgery. It seems more appropriate to present levels of evidence for the different clinical interventions as illustrated in the summary of the effectiveness of the interventions in Table 2 . Most surgical trials will inevitably be at high risk for performance bias resulting from the difficulties with blinding surgeons and patients. Moreover, variation in expertise of surgeons with different surgical procedures is an almost unavoidable confounding variable (Johnson et al., 2008), as well as variation in techniques such as the routine use of anti-adhesive barriers. Nevertheless the future holds promise since gynaecological surgery, in contrast to other surgical specialities is being exposed to the scrutiny of RCTs, following the pioneering pathway of reproductive medicine (Johnson et al., 2003, 2008; Vandekerckhove et al., 1993). Despite the difficulty with the methodology and conduct of RCTs in reproductive surgery, we need to set up adequately powered and pragmatic multicentre randomised trials studying the effectiveness of reproductive surgery versus no treatment or alternative treatment. Two possible sources of bias in this systematic review should be addressed. Firstly, the filters for the systematic literature search did not include textwords or MeSH terms for endometriosis, fibroids, polyps and tubal pathology. This might have decreased the sensitivity of our search due to the unintentional omission of smaller trials. A second possible source of detection bias is the fact that our group has already published a narrative review on laparoscopy and a systematic review on hysteroscopy in the treatment of infertility (Bosteels et al., 2007, 2010). An important limitation in the majority of the included trials concerns the choice by the authors to use pregnancy and live birth rates as outcome measures. This is a crude way to assess fertility compared to other parameters such as monthly fecundity rate, cumulative pregnancy rate and time to pregnancy by life table analysis. Future randomised trials on reproductive surgery should not neglect the importance of the time factor in the choice of the most appropriate outcome measure. Moreover, in some clinical settings an ‘expectant management’ group should be used as a control to study the effectiveness of a surgical procedure as proposed by some authors (Hull et al., 1987; Olive et al., 1985).

Worldwide ART has replaced reproductive surgery for tubal factor infertility, limiting its role as first-line treatment (Bosteels et al., 2007). It is not clear whether this change in clinical practice is due to the higher cost-effectiveness of ART compared to reproductive surgery or caused by other factors such as a lack of surgical expertise, patient’s desires to achieve results rapidly or the concern to protect patients from procedure-related complications. In moderate and severe endometriosis radical laparoscopic surgery is often delayed until several ART cycles have failed (Littman et al., 2005). Its surgical treatment necessitates a high level of expertise (Kennedy et al., 2005). As a consequence, laparoscopy is increasingly bypassed in the diagnostic work-up of infertility (Fatum et al., 2002). Some authors report that the exploration by diagnostic laparoscopy in an infertile population either did not reveal any pathology or only minimal or mild endometriosis in 40-70% of cases (Forman et al., 1993). From a prognostic point of view, the test ‘diagnostic laparoscopy’ fails to be an ideal predictor for infertility (Collins et al., 1995; Mol et al., 1999). However, the shift away from reproductive laparoscopic surgery favoring ART is not supported by solid evidence. The position of hysteroscopy in current fertility practice is similarly unclear. Used on an outpatient basis with small calibre hysteroscopes, its technical feasibility and high patient compliance in general gynaecological practice are demonstrated by numerous RCTs (Kremer et al., 2000; Soriano et al., 2000; Unfried et al., 2001; De Angelis et al., 2003; Guida et al., 2003; Litta et al., 2003; Pellicano et al., 2003; Marsh et al., 2004; Shankar et al., 2004; Campo et al., 2005; Sharma et al., 2005; Garbin et al., 2006; Guida et al., 2006; Sagiv et al., 2006; De Placido et al., 2007; Kabli and Tulandi, 2008). However, neither the ease of use nor the low costs should by themselves justify the widespread application of a surgical procedure. Indeed, the number of RCTs demonstrating the effectiveness of hysteroscopy in treating female infertility is limited (Bosteels et al., 2010). Our aim is to study the effectiveness of reproductive surgery in treating female subfertility, by giving an overview of all published randomised trials measuring pregnancy or live birth rates.

ART: assisted reproductive technology CDSR: Cochrane Database of Systematic Reviews CENTRAL: Cochrane Central Register of Controlled Trials CI: confidence interval DARE: Database of Abstracts of Reviews of Effects HTA: Health Technology Assessment Database IVF: in vitro fertilisation IUI: intrauterine insemination MH-F: Mantel-Haenszel, fixed effects model mRCT: metaRegister of Controlled Trials PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses RR: risk ratio RCT: randomised controlled trial