Criteria for considering studies for this review
Types of studies
All randomised controlled trials (RCTs) of the use of selective estrogen-receptor modulators (SERMs) in the treatment of endometriosis will be included. Crossover trials will be eligible, but only data from the first phase will be included in meta-analyses, as the crossover is not a valid design in this context. Quasi-randomised and non-randomised studies (case control studies, cohort studies) will be excluded.
Types of participants
Women of reproductive age (usually up to 50 years of age) with visually diagnosed or biopsy-proven endometriosis by laparoscopy, or on the basis of international guidelines (Dunselman 2014; RCOG 2006) used to diagnose endometriosis will be included. Women who have endometriosis clinically but with no visual or laparoscopic confirmation will be excluded.
Types of interventions
Trials comparing SERMs via any route versus any other active intervention (for example danazol, GnRHa), placebo or no treatment will be eligible for inclusion.
Any dosage or duration of treatment will be included. Available SERMs include: raloxifene, arzoxifene, levormeloxifene, ospemifene, lasofoxifene, bazedoxifene, centchroman, arzoxifene, tamoxifen, toremifene, droloxifene, idoxifene.
Types of outcome measures
We will consider the outcome measures at the end of the treatment and, when possible, at 3, 6, 9, 12, 18 months following completion of treatment.
1. Relief of pelvic pain, on a visual analogue scale (VAS) (0 (no pain) to 10 (worst pain imaginable) or subjective improvement.
The painful symptoms may take the form of dysmenorrhoea, dyspareunia (pain during or after sexual intercourse) or pelvic or lower abdominal pain. Some women also present with cyclical pain at other sites, relating to endometriosis at extra-pelvic sites. In this review we discuss pelvic pain. Measures of subjective pain relief will be assessed. These include VAS or any other recognised scoring system or qualitative measures such as cured, better, same, or worse.
2. Adverse events resulting from treatment with SERMs (hot flushes, endometrial cancer, vaginal bleeding, thrombosis) either during or following treatment.
3. Quality of life (by quality of life scores)
4. Recurrence rate (symptoms or biopsy-proven)
5. Fertility outcomes (only in women wishing to conceive)
6. Economic outcomes (costs, time taken for the treatment)
Search methods for identification of studies
All published and unpublished RCTs of SERMs versus other interventions will be sought using the following search strategy, without language restriction and in consultation with the Menstrual Disorders and Subfertility Group Trials Search Co-ordinator. Relevant trials will be identified from both electronic databases and other resources.
The following electronic databases, trial registers and web sites will be searched from inception to the present.
Menstrual Disorders and Subfertility Group (MDSG) Specialised Register of controlled trials.
Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library.
English language electronic databases: MEDLINE, EMBASE, PsycINFO and CINAHL.
Chinese language electronic databases: Chinese Biomedical Literature Database (CBM) and Chinese Medical Current Contents (CMCC); using the corresponding Chinese terms for: “endometriosis” AND “SERMs”, “ERs”,“Raloxifene”, etc.
Current Controlled Trials (www.controlledtrials.com/);
World Health Organization International Trials Registry Platform search portal (www.who.int/trialsearch/Default.aspx);
Citation indexes (http://scientific.thomson.com/products/sci/);
Conference abstracts in the Web of Knowledge (http://wokinfo.com/);
LILACS (Latin American and Caribbean Health Science Literature) (http://bases.bireme.br/cgibin/wxislind.exe/iah/online/?IsisScript=iah/iah.xis&base=LILACS&lang=i&form=F);
OpenGrey database (http://opengrey.eu/); and
Google for grey literature.
The MEDLINE search will be combined with the Cochrane highly sensitive search strategy for identifying RCTs, which appears in the searching chapter of the Cochrane Handbook for Systematic Reviews of Interventions. The EMBASE search will be combined with trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN) (http://www.sign.ac.uk/methodology/filters.html#random).
Searching other resources
We will also do the following:
(1) search the references lists of all included studies and relevant reviews to identify further relevant articles;
(2) contact authors and experts in the field for potential studies;
(3) handsearch the following Chinese journals:
Chinese Journal of Obstetrics and Gynecology (from 1953 to present),
Chinese Journal of Practical Gynecology and Obstetrics (from 1985 to present),
Journal of Practical Obstetrics and Gynecology (from 1985 to present),
Maternal and Child Health Care of China (from 1986 to present).
See Appendix 1, Appendix 2, Appendix 3, Appendix 4, Appendix 5 for the search strategies.
Data collection and analysis
We will perform statistical analysis in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Review Manager 5.2 will be used to analyse the data.
Selection of studies
Two review authors will independently examine the full text articles for compliance with the inclusion criteria and will select studies eligible for inclusion in the review. The full text will be retrieved when the information given in the titles, abstracts and keywords suggests that the study uses SERMs as an intervention; the participants have endometriosis; the study has a prospective design and a control group. The authors will correspond with study investigators, if required, to clarify study eligibility (for example with respect to participant eligibility and allocation method). Disagreements as to study eligibility will be resolved by consensus or by discussion with a third author.
The study selection process will be documented with a PRISMA flow-chart.
Data extraction and management
Data will be extracted from the eligible studies using a special data extraction form. Two review authors (Chen YL and Wan Q) will independently extract the data, and any disagreement between review authors will be resolved by discussion with a third review author (Zheng A). The main trial report will be used as the reference and additional details will be obtained from secondary reports, where studies have multiple publications. We will correspond with study investigators for further data on methods and/or results, as required.
Assessment of risk of bias in included studies
The included studies will be assessed for risk of bias using the Cochrane risk of bias assessment tool (www.cochrane-handbook.org) to assess: selection bias (random sequence generation, allocation concealment); performance bias (blinding of participants and personnel); detection bias (blinding of outcome assessors); attrition bias (incomplete outcome data); reporting bias (selective reporting); and other biases. Two authors will assess risk of bias with any disagreements resolved by consensus or by discussion with a third author (Zheng A). We will correspond with the trialists to identify any within-trial selective reporting. We will seek published protocols and compare the outcomes between the protocol and the final published study. The risk of bias table will be included in the table 'Characteristics of included studies'. All judgements will be fully described. The conclusions will be presented in the 'Risk of bias' table and will be incorporated into the interpretation of review findings by means of sensitivity analyses.
Measures of treatment effect
For dichotomous data (for example recurrence and adverse events) we will record the number of participants experiencing the event in each group of the trial. Odds ratios (ORs) with 95% confidence intervals (CI) will be used to report dichotomous data. For continuous outcomes (for example pain scores) we will extract the final value and standard deviation of the outcome of interest, and the number of women assessed at the endpoint in each treatment arm and at the end of follow-up. Mean differences (MD) with 95% CIs will be used to report continuous data. In the case of outcomes with continuous data in different scales, we will use the standardized mean difference (SMD) with 95% CI. Where data to calculate ORs or MDs are not reported, we will utilise the most detailed numerical data available that may facilitate similar analyses of included studies (for example test statistics, P values) or report the results in narrative form only.
Unit of analysis issues
The primary analysis will be per woman randomised. Fertility outcomes will only be measured among women seeking pregnancy. For certain fertility outcomes (for example miscarriage or multiple birth) the per pregnancy data will be used.
Dealing with missing data
We will try to obtain missing data from the original investigators through email or telephone. The data will be analysed on an intention-to-treat basis as far as possible (the analysis should include all the randomised participants in the groups to which they were originally randomly assigned). Where missing data are unobtainable, imputation of individual values will be undertaken for the primary outcomes only. If studies report sufficient detail to calculate mean differences but provide no information on the associated standard deviations (SDs), the outcome will be assumed to have an SD equal to the highest SD from other studies within the same analysis. For other outcomes, only the available data will be analysed.
Assessment of heterogeneity
We will check the included trials to see if the participants, interventions and outcomes in the included studies are similar enough to consider pooling in a meta-analysis. If so, then the rest of the section will refer to statistical heterogeneity after pooling. Tests for heterogeneity will be carried out using the Chi2 test, with significance set at P < 0.1. The I2 statistic will be used to estimate the total variation across studies due to heterogeneity, where > 50% is high-level heterogeneity (Higgins 2011. An I2 value of over 50% will be taken to indicate substantial heterogeneity and we will investigate this using the sensitivity and subgroup analyses described below. When heterogeneity was not significant (P > 0.1, I2 < 50%) among subgroups, a fixed-effect model will be used. Where heterogeneity is significant (P < 0.1, I2 > 50%) a random-effects model will be used.
Assessment of reporting biases
We will aim to minimise publication bias and other reporting biases by ensuring a comprehensive search for eligible studies and by being alert for duplication of data. Potential publication bias will be assessed using a funnel plot or other corrective analytical methods if there are 10 or more studies in an analysis (Egger 1997). Study reporting bias will be detected by seeking published protocols and comparing the outcomes between the protocol and the final published study.
If sufficient clinically similar studies are available, the data will be pooled in meta-analyses. For continuous outcomes (pain relief measures) the MD between the treatment arms at the end of follow-up will be determined if all trials measured the outcome on the same scale, otherwise the SMD will be calculated. An increase in the odds of a particular outcome will be displayed to the right of the centre line. While the precise scaling of diary data may vary when using a VAS score, the trials can use the mean VAS score or report a reduction in the VAS score (with a 50% reduction being considered clinically significant). Different comparisons will be analysed separately:
all SERMs versus placebo or no treatment;
all SERMs versus alternative active therapy, stratified by all SERMs versus medical treatment and versus surgery.
Subgroup analysis and investigation of heterogeneity
Where data are available, we intend to explore the following potential sources of heterogeneity using subgroup analyses:
individual types of SERMs versus placebo, no treatment, or each alternative active therapy;
duration of treatment (three months, three to six months, six to 12 months, at least 12 months).
When substantial heterogeneity is present, we will explore possible explanations including individual study risk of bias, dose of SERM.
We will consider sensitivity analyses for the primary outcomes to determine whether the conclusions are sensitive to arbitrary decisions made regarding the eligibility and analysis. These analyses will include consideration of whether conclusions would have differed if:
studies that had a high risk of bias were excluded
studies that used different rating scales to assess symptom relief such as unpublished rating scales or scales with no established reliability or validity were excluded.
Summary of findings table
We will prepare a summary of findings table using GRADEPro or Guideline Development Tool software. This table will evaluate the overall quality of the body of evidence for the primary review outcomes (relief of pelvic pain and adverse events following SERM treatment) using the GRADE criteria (study limitations (that is risk of bias), consistency of effect, imprecision, indirectness and publication bias). Judgements about evidence quality (high, moderate or low) will be justified, documented and incorporated into reporting of results for each outcome.