Types of studies

We included randomised controlled trials (RCTs). We excluded cohort and cross‐sectional studies.

Types of participants

We included all adults aged 18 years and over, regardless of gender, with stroke as defined by the World Health Organization (WHO) (Hatano 1976). We also included the partners of adults with stroke.

We included studies in which the cohort included a mix of stroke and non‐stroke survivors if at least 75% of the study population consisted of stroke survivors. When less than 75% of the study population consisted of stroke survivors, we included these studies if data related only to the stroke population were separately reported.

Types of interventions

We included trials comparing pharmacological treatments, mechanical devices, or complementary medicine interventions with placebo and other non‐pharmacological interventions with usual care. We also included trials with different forms of non‐pharmacological interventions (such as a comprehensive individualised psycho‐educational programme versus a basic/standard education pamphlet) for treating sexual dysfunction in stroke survivors.

Interventions could include but were not limited to:

• pharmacological interventions such as phosphodiesterase‐5 inhibitors, intracavernosal injections, intraurethral suppositories, and hormonal therapy;

• non‐pharmacological interventions, including:

  • mechanical devices (such as vacuum pumps, penile implants, penile prostheses, and lubricating gels);

  • psycho‐educational interventions (such as counselling and psychotherapy). These could range from highly structured interventions provided by an appropriately trained health professional (medical practitioner, social worker, occupational therapist, sex therapist) to education alone provided through educational pamphlets from credible sources (such as by stroke associations);

  • established stroke consumer or professional associations as opposed to education from friends who have not undergone appropriate training to provide such education; and

  • physical therapy (such as physiotherapy for bed mobility) by appropriately trained personnel; and

• complementary medicine interventions such as gingko biloba and ginseng.

We excluded uncontrolled and open drug intervention trials. Therefore, we would exclude a trial comparing pharmacological intervention to usual care. It is expected that the control comparator would be in the same category as the intervention. Pharmacological and complementary medicine interventions would be compared with placebo delivered via the same route as the intervention. However, other non‐pharmacological interventions, such as counselling, could be provided at different intensities or in different formats (face‐to‐face versus group, written versus verbal). The key characteristic of an intervention is that the intention was to improve sexual dysfunction. We excluded studies in which the intervention might indirectly improve sexual dysfunction but was given primarily for another reason (such as an antidepressant for depression).

Types of outcome measures

Electronic searches

We conducted searches on 27 November 2019. We searched the Cochrane Stroke Group trials register and the following electronic databases.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 11), in the Cochrane Library (Appendix 1).

• MEDLINE Ovid (from 1950) (Appendix 2).

• Embase Ovid (from 1980) (Appendix 3).

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) EBSCO (from 1982) (Appendix 4).

• Allied and Complementary Medicine (AMED) Ovid (from 1985) (Appendix 5).

• PsycINFO Ovid (from 1806) (Appendix 6).

• Physiotherapy Evidence Database (PEDro) (from 1999) (http://www.pedro.org.au/) (Appendix 7).

• Center for International Rehabilitation Research Information and Exchange (CIRRIE) Database of International Rehabilitation Research (http://cirrie.buffalo.edu/search/index.php) (Appendix 8).

• Database of Abstracts of Reviews of Effects (DARE), in the Cochrane Library (latest issue) (Appendix 9). Please note that DARE is no longer updated as of March 2018.

• ProQuest Dissertations & Theses Database (Appendix 10).

• OT Search by the American Occupational Foundation and the American Occupational Therapy Association (www1.aota.org/otsearch/) (Appendix 11).

• Occupational Therapy Systematic Evaluation of Evidence (OTseeker) (www.otseeker.com/) (Appendix 12).

• National Rehabilitation Information Center REHABDATA Database (www.naric.com/research/rehab/) (Appendix 13).

• SPORTDiscus EBSCO (Appendix 14).

We developed the MEDLINE search strategy with the help of the Cochrane Stroke Group Information Specialist and adapted it for the other databases (Appendix 2). All search strategies deployed were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane for identifying RCTs and controlled clinical trials (as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019)).

Searching other resources

To identify further published, unpublished, and ongoing trials, we:

• handsearched the reference lists of included trials and review articles about sexual rehabilitation after stroke;

• tracked citations using Web of Science Cited Reference Search for all included studies; and

• contacted experts active in this field (including authors of included trials and excluded studies identified as possible preliminary or pilot work).

Selection of studies

Two review authors (LN, HS) independently screened titles and abstracts of references obtained as a result of our searching activities and excluded obviously irrelevant reports. We retrieved full‐text articles for the remaining references and screened full‐text articles, identified studies for inclusion, and recorded reasons for exclusion of ineligible studies. We resolved any disagreements through discussion. If required, we would have consulted a third review author (AB). We collated multiple reports of the same study so that each study ‐ not each reference ‐ would be the unit of interest in the review. We recorded the selection process and completed a Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) flow diagram when appropriate.

Data extraction and management

Two review authors (LN, HS) independently extracted data from included studies. We grouped studies by intervention type (pharmacological, non‐pharmacological, and complementary medicine) when possible.

We used a pre‐designed data extraction form to extract the following data from the included studies.

• Participants: number of participants, age, gender, baseline functional status, or level of impairment.

• Methods: inclusion criteria, time since stroke, and type, nature, and location of lesion.

• Interventions: description of interventions given to each treatment group including duration, type, dose, route of delivery, and frequency. For interventions provided by therapists,

  • we documented the discipline of the person providing the intervention (e.g. physician, occupational therapist, physiotherapist, psychologist, social worker).

• Outcomes: we documented primary and secondary outcomes relevant to this review. If a study used different methods for measuring the same outcome, we noted the outcome to be used for any subsequent analysis. We noted any important confounding variables. If more than two intervention groups were included in the study, we noted the method of including these groups in any subsequent analysis. The two review authors resolved any data extraction discrepancies through discussion. If disagreement persisted, we would have consulted a third review author (AB). We extracted data on whether adverse events were explicitly reported.

Assessment of risk of bias in included studies

Two review authors (LN, HS) independently assessed risk of bias for each study using the 'Risk of bias' tool outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2019). We resolved all disagreements by discussion. If required, we would have involved another review author (AB). We assessed the risk of bias according to the following seven domains.

• Random sequence generation.

• Allocation concealment.

• Blinding of participants and personnel.

• Blinding of outcome assessment.

• Incomplete outcome data.

• Selective outcome reporting.

• Other bias.

We graded the risk of bias for each domain as high, low, or unclear and provided information from the study report together with a justification for our judgement in the ’Risk of bias’ tables.

Measures of treatment effect

We grouped studies by intervention type (pharmacological, non‐pharmacological, and complementary medicine). Given the heterogeneity of interventions, we were not able to perform a meta‐analysis. If it had been possible, we would have used the Cochrane Review Manager software to carry out statistical analyses to determine treatment effect (Review Manager 2014). For dichotomous variables, we would have calculated the treatment effect using a fixed‐effect or random‐effects model and reported it as odds ratios (ORs) with 95% confidence intervals (CIs). For continuous data, we would have calculated the treatment effect using standardised mean differences (SMDs) and 95% CIs when different studies used different scales to assess the same outcome, and we would have calculated mean differences (MDs) and 95% CIs when studies used the same method of measuring outcomes. We would have done this for all outcomes.

Unit of analysis issues

Unit of analysis issues did not arise with the included studies. If such issues had arisen, we would have addressed these as follows: if studies reported change values and the baseline value was not available, we would have used these data in meta‐analyses but planned sensitivity analyses to investigate the effect of including these data. We would have analysed adverse events as dichotomous variables. We would have done this for all outcomes.

Dealing with missing data

When an included study did not report a particular outcome but it had been included in the battery of measures administered, we contacted the study authors to ask for the original data. If we were unsuccessful in obtaining the data, we did not include that study in the analyses of that outcome.

When an included study had missing data (e.g. reports means but not standard deviations for follow‐up data), we contacted study authors to ask for the missing data. If we were unsuccessful, we took logical steps to enter an assumed value. Such steps included estimating a standard deviation based on a reported standard error, estimating a follow‐up standard deviation based on a baseline value, using the median as a proxy for the mean, and using a multiple of 0.75 times the interquartile range or 0.25 times the range as a proxy for the standard deviation values (Hozo 2005). We conducted sensitivity analyses to investigate the effect of entering assumed values.

Assessment of heterogeneity

If relevant, we would have used the I² statistic to measure heterogeneity among the trials in each analysis. When there was substantial heterogeneity (as defined by I² > 50%), we would have performed a meta‐analysis using a random‐effects model. If I² was less than or equal to 50%, we would have performed fixed‐effect meta‐analysis.

Assessment of reporting biases

We attempted to avoid reporting biases by using a comprehensive search strategy that included searching for unpublished studies and searching trials registers. We also assessed the completeness of outcome data.

Data synthesis

When we considered studies to be sufficiently similar, we would have conducted a meta‐analysis by pooling the appropriate data using RevMan 5.3 (Review Manager 2014).

GRADE and 'Summary of findings'

It was not possible to perform quantitative meta‐analysis due to the heterogeneity of studies; we used the GRADE approach instead to assess the quality of evidence (Higgins 2019). The GRADE approach defines quality of studies as the extent to which one can be certain that an estimate of effect is close to the quantity of interest. It specifies four levels of quality for a body of evidence for a given outcome: high, moderate, low, and very low. GRADE assessments of certainty are determined through consideration of the following five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Factors that may decrease the quality level of the included studies therefore include limitations in methods suggesting high likelihood of bias, indirectness of evidence such as with indirect populations, unexplained inconsistency of results, imprecision of results (e.g. with wide confidence intervals), and high probability of publication bias.

The main findings of the review, together with key information concerning the quality of evidence, are presented in the 'Summary of findings' tables. These findings are listed by intervention comparison in three categories (pharmacological interventions, non‐pharmacological interventions, and complementary medicine interventions), with outcomes divided into primary outcomes (sexual function or sexual satisfaction, or both) and secondary outcomes (quality of life, psychological well‐being, satisfaction with intervention, sexual knowledge, marital/relationship/partner satisfaction, and serious adverse events/death). When outcomes have not been addressed, we have reported this in the 'Summary of findings' table as "not addressed".

Subgroup analysis and investigation of heterogeneity

We would have explored heterogeneity by conducting additional subgroup analyses to investigate the effects of:

• time since stroke;

• type of intervention;

• level of impairment at baseline; and

• adherence with additional intervention.

Sensitivity analysis

We would have carried out a sensitivity analysis (when necessary) to explore the effects of the following methodological features.

• Allocation concealment: we would have re‐analysed data, excluding trials with inadequate or unclear allocation concealment.

• Masking of outcome assessor: we would have re‐analysed data, excluding trials without or with unclear masking of outcome assessor.

• Missing outcome data: we would have re‐analysed data, excluding trials with inadequate or unclear methods of dealing with missing outcome data.