Types of studies

Because of the relatively low incidence of the ovarian cancer and the necessity of a follow‐up spanning several decades, our first listed objective, the effect of salpingectomy on the incidence of epithelial ovarian cancer, is not a particularly suitable outcome for a randomised controlled trial (RCT). For this objective, we will therefore consider both RCTs and non‐randomised trials (NRTs) included in this review. Since the risk of bias is greater in NRTs than in RCTs, we will consider only cohort studies (both retrospective and prospective), case‐control studies and cross‐sectional studies. Although the existence of RCTs and quasi‐randomised trials is highly unlikely, we will assess them for inclusion if available.

Because our second and third objectives, the effect of salpingectomy on the incidence of surgery‐related adverse events, and on postoperative ovarian function and ovarian reserve, are suitable outcomes for RCTs, for these objectives we will consider only RCTs included in this review.

Types of participants

Individuals with a population‐based risk of ovarian cancer undergoing surgery for benign gynaecological conditions. We will exclude trials that include the following types of women:

• women with a history of ovarian cancer;

• women with an elevated risk of ovarian cancer based on a proven gene mutation;

• women who have undergone previous bilateral oophorectomy;

• women who have undergone previous bilateral salpingectomy.

The exclusion of women with a proven germline BRCA 1/2 gene mutation is important since there are limited data available to suggest that mutation carriers may undergo an earlier menopause than the general population (Finch 2013).

Types of interventions

We will include both RCTs and NRTs that compare hysterectomy with bilateral salpingectomy to hysterectomy with preservation of the fallopian tubes.

Types of outcome measures

Electronic searches

We will search for papers published in all languages and, if necessary, will obtain translations. We will search the following databases, from their inception until present, using a combined strategy that has been developed:

• Cochrane Gynaecology and Fertility Group (CGF) Specialised Register of Controlled Trials, PROCITE platform (Appendix 1)

• Cochrane Central Register of Controlled Trials (CENTRAL) Cochrane Library platform (Appendix 2);

• MEDLINE (Epub Ahead of Print, In‐Process & other Non‐indexed Citations) Ovid platform (Appendix 3);

• Embase Ovid platform (Appendix 4);

• PsycINFO Ovid platform (Appendix 5);

• CINAHL Ebsco platform (Appendix 6).

Additionally, we will search for ongoing and registered trials (Appendix 7) in the following ongoing trial registries:

• clinicaltrials.gov/ (a service of the US National Institutes of Health);

• who.int/trialsearch/default.aspx (the World Health Organization International Trials Registry Platform search portal).

Searching other resources

We will handsearch the reports of conferences from the following sources: ESGO (European Society of Gynaecological oncology), SGO (Society of Gynecological Oncology), ESHRE (European Society of Human Reproduction and Embryology), EMAS (European Menopause and Andropause Society) and IMS (International Menopause Society). To identify additional trials, we will handsearch the reference lists of all relevant trials obtained by the initial search. We will limit this search to articles and reports published since 1997, as the fallopian tube has been considered as the origin of epithelial ovarian cancer only since 2001 Piek 2001a.

Selection of studies

We will import the titles and abstracts retrieved by the search into the reference manager database Covidence (Covidence). Two review authors (LL, MS) will independently screen the references and check them for duplicates. After obtaining the full text versions of potentially relevant studies, two review authors (LL, MS) will independently assess their eligibility for inclusion. We will resolve any disagreements by discussion and, if necessary, by consulting a third review author (JP). Where the judgement of a review author could be biased due to a conflict of interest, one of the other review authors will assess that particular study. We will document reasons for exclusion and the selection process in a PRISMA flow chart.

Data extraction and management

We will predesign a data extraction form for the extraction of relevant data from included trials. Prior to data extraction, two review authors (LL, MS) will perform an independent trial run of the data extraction form on a sample of studies. Two review authors (LL, MS) will independently extract data on the number of participants, characteristics of participants, characteristics of the intervention with and without salpingectomy, study quality, duration of follow‐up and outcomes. We will resolve any disagreements by discussion and, if necessary, by consulting a third review author (JP). We will attempt to retrieve missing data by contacting the study authors. For studies with multiple publications, we will collate multiple reports of the same study so that each study, rather than each report, is the unit of interest in the review. We will assign these studies a single study identifier.

Assessment of risk of bias in included studies

Two review authors (JW, MS) will independently assess the methodological quality of included studies; we will resolve disagreements by consulting a third review author (JP).

For non‐randomised studies, we will assess the likelihood of bias according to the ROBINS‐I (a tool for assessing the risk of bias in non‐randomised studies of interventions) (Sterne 2016). The hypothetical 'target' trial necessary for the use of the ROBINS‐I would be a large RCT in which women would be allocated to the intervention group (i.e. hysterectomy with bilateral salpingectomy) or the control group (i.e. hysterectomy without bilateral salpingectomy). Baseline information on both groups should include age, parity, the use of oral contraceptives and surgical history. During a follow‐up period of at least 40 years, family history (breast and ovarian cancer), age at menopause, use of oral contraceptives, abdominal surgery and the occurrence of epithelial ovarian cancer should have been documented. If a participant is diagnosed with epithelial ovarian cancer, data on age at diagnosis, tumour stage and histology of the primary tumour should have been collected.

We will assess the eligible studies for bias due to confounding, selection of participants, classification of interventions, missing data, measurement of outcomes and selection of the reported result. We have identified the following domains as potential confounders, which therefore should preferably be similar among study groups: age, parity, family history of ovarian or breast cancer, use of oral contraception and history of tubal ligation. We have identified no co‐interventions that could potentially confound the results. Confounding might result in considerable heterogeneity between studies and requires adequate methods to control for it, such as stratification of regression modelling with propensity scores or covariates. We will assess the appropriateness and quality of these methods critically. We will compare NRTs to their published protocol, where available, so as to assess selective or incomplete reporting.

We will assess the risk of bias in randomised studies using the Cochrane's ‘Risk of bias’ assessment tool (Higgins 2011) and will include the following domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective reporting. If available, we will compare the published protocols of selected studies to the reported outcomes so as to assess selective or incomplete reporting bias.

Measures of treatment effect

For NRTs, we will extract and report both unadjusted and adjusted effect estimates. For cohort studies, we will calculate a hazard ratio (HR). We expect there to be a long duration of follow‐up for the epithelial ovarian cancer outcome, which might result in selection bias over time. We will, therefore, calculate HRs for different time‐points. For case‐control studies, we will calculate odds ratios (OR) with 95% confidence intervals (CIs) by extracting the number of participants in each treatment arm that experienced the outcome of interest, and the number of participants assessed per outcome.

For dichotomous data extracted from RCTs (i.e. adverse surgical events), we will calculate ORs with 95% CIs. For continuous data (i.e. postoperative hormonal status), we will estimate mean differences (MDs) with 95% CIs for variables with a normal distribution where the same measure was used to assess the outcome. If the included studies used different measures to assess the same outcome, we will use the standardised mean difference (SMD). For skewed continuous variables, we will extract mean values and standard deviations. If necessary, we will transform the data prior to meta‐analysis according to one of three methods presented by Higgins 2008a.

If the data necessary to calculate ORs or MDs are not available, we will make use of the most detailed numerical data available that may facilitate similar analyses of included studies. We will attempt to retrieve missing data by contacting the study authors.

Unit of analysis issues

We will perform the primary analysis per woman included in the studies.

Dealing with missing data

For NRTs, we will conduct sensitivity analyses to assess how robust our conclusions are to assumptions about missing data (Higgins 2008b).

For RCTs, we will analyse the retrieved data according to the intention‐to‐treat principle. In case of missing data, we will contact the original researchers to obtain the missing data. If these attempts do not provide us with extra data, we will use only the data that are available. We will perform an additional per‐protocol analysis, if applicable.

Assessment of heterogeneity

For NRTs, we expect heterogeneity, and thus we will base our assessment of heterogeneity on consideration of the different study designs and analysis details.

To examine whether meta‐analysis is possible for RCTs, we will assess the heterogeneity of the included studies using the I² statistic. This heterogeneity is important, since it indicates whether the studies are similar in clinical and methodological characteristics. We will consider an I² value of 50% or higher as indicating substantial heterogeneity.

Assessment of reporting biases

In view of the difficulty of detecting and correcting for publication bias and other reporting biases, we will aim to minimise their potential impact by performing an extensive search for eligible studies and by being alert for the duplication of data. If there are 10 or more studies in an analysis, we will use a funnel plot to explore the possibility of small study effects (the tendency for estimates of the intervention effect to be more beneficial in smaller studies). To prevent language bias, we will exclude no studies based on language. If the studies prove to be exceptionally difficult to translate, we will ask the authors to provide a summary of their methods and results. We will compare the studies, and authors and their affiliations so as to avoid multiple publication bias.

Data synthesis

If the selected studies are considered to be similar enough for meta‐analysis, we will combine the data using a fixed‐effect model. We will make the following comparisons.

1. Incidence of epithelial ovarian cancer after hysterectomy with bilateral salpingectomy versus incidence of epithelial ovarian cancer after hysterectomy without bilateral salpingectomy.

2. Surgical outcomes of hysterectomy with bilateral salpingectomy versus hysterectomy without bilateral salpingectomy.

3. Ovarian reserve after hysterectomy with bilateral salpingectomy versus hysterectomy without bilateral salpingectomy.

Subgroup analysis and investigation of heterogeneity

Where possible, depending on the availability of the data, we will perform the following subgroup analyses.

1. Effect of salpingectomy on the incidence of epithelial ovarian cancer in the following subgroups:

a. premenopausal versus postmenopausal women.

2. Effect of salpingectomy on the incidence of epithelial ovarian cancer in the following subgroups:

a. nulliparous versus parous women.

3. Incidence of epithelial ovarian cancer in women who have a history of tubal ligation versus incidence of epithelial ovarian cancer in women who have no history of tubal ligation.

4. Incidence of surgery‐related adverse events depending on surgical approach:

a. abdominal approach versus laparoscopic approach;

b. vaginal approach versus laparoscopic approach;

c. abdominal approach versus vaginal approach.

Sensitivity analysis

For the primary outcomes, we will conduct sensitivity analyses to determine whether the review conclusion would remain the same if:

• eligibility had been restricted to studies without high risk of bias (which we define as those with no high risk of bias in any domain);

• a random‐effects model had been adopted;

• NRTs had been excluded (only applicable if RCTs have been included).