This systematic review and meta-analysis will follow the PRISMA-P (Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols) [13] for evaluating the effect of testosterone treatment on primary cardiovascular outcomes (clinical events of stroke or transient ischemic attack, nonfatal myocardial infarction, or acute coronary syndrome, emergency coronary artery revascularization, carotid surgery, cardiac mortality, and all-cause mortality) among adult men. Three a priori hypotheses will be tested: types of testosterone treatment (oral administration, transdermal application, intramuscular injection, or implant vs no testosterone), dosage of testosterone received (total dosage [mg] received during the duration of the study vs no testosterone), and follow-up period (at least 6 months vs more than 6 months).

Studies need to meet the following characteristics to be eligible for this systematic review and meta-analysis.

Randomized controlled trials or cluster randomized controlled trials with at least one intervention arm of testosterone and a control group of usual care or no testosterone treatment will be included. Experimental studies with no randomization (such as quasi-experimental studies) will not be included in this study. Observational cohort studies, case-control studies, nested case-control studies, cross-sectional studies, case series, and case reports will also not be included.

Peer-reviewed information published in electronic databases will be searched. Search terms will include testosterone AND (cardiac OR heart OR stroke OR myocardial infarction OR atherosclerosis OR revascularization OR Agatston score OR death OR angina OR carotid OR artery OR coronary OR cerebral OR aortic OR peripheral vascular) anywhere in the text by using truncation and wildcards when available, to accommodate for different spellings. The search modes will be set to Boolean/Phrase when available. In addition, the limiters will be human studies, studies done on males, publications written in English, and publication year set to range from 1980 to 2019 and when available, and peer-reviewed randomized controlled trials. If full-text is not available, the original study authors will be contacted for the full-text of the article published in English language.

CINAHL (Cumulated Index to Nursing and Allied Health Literature), Embase, Medline, Scopus, and Cochrane Controlled Register of Trials will be searched using the advanced search function, where available. The following search strategy will also be applied when searching the Clinical Trial Registry (ClinicalTrials.gov) for trials with completed results. References of the selected articles will also be scanned to ensure literature saturation.

The number of records found from each database and the Clinical Trial Registry will be noted in the format of the PRISMA flow diagram. The full details of the articles will be imported to an EndNote library. Duplicates of identical records will be removed and the number of duplicates removed will be recorded in the PRISMA flow diagram. Two independent reviewers will screen the titles and abstracts of the remaining articles. Articles that are conference abstracts, review articles, observational studies, response or letters to the Editors, not on men, on isolated cells or artery specimens, and irrelevant outcomes will be excluded and the number recorded on the PRISMA flow diagram. If consensus is not reached at the screening phase, there will be discussion between the 2 reviewers or a third independent reviewer may be invited to screen the articles until consensus is reached. The remaining full-text articles will be screened against the selection criteria. Full-text articles that do not meet the eligibility criteria will be excluded and the number recorded on the PRISMA flow diagram. If multiple articles were produced from the same active intervention by the same group of authors at the same institution on the same outcome, only the article with the most complete follow-up data will be included in the review. Discussion between 2 independent reviewers or involvement of a third independent reviewer will resolve disagreement on the selection of the articles. References of the selected articles will also be screened as above to ensure literature saturation. The final number of the selected articles will be noted on the PRISMA flow diagram.

The quality of the studies will be assessed by 2 independent reviewers following the Cochrane Collaboration’s tool for assessing risk of bias. The risk of bias will be assessed in terms of sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting, and bias that may be threats to the validity [14]. Each of these will be assessed as “low risk,” “high risk,” or “uncertain risk” of bias [14] at the study and outcome level. Discussion between 2 independent reviewers and involvement of a third independent reviewer will be required until consensus is reached.

The number of articles retrieved and excluded will be recorded in the format of the PRISMA flow diagram. The characteristics of the included study and participants will be noted in a table, grouped by the types of testosterone therapy received. Data extraction will be piloted from a small group of studies by 2 reviewers. One reviewer will then independently extract the data from all the selected articles, which will be verified by the second reviewer. There is no plan of individual patient data meta-analysis; therefore, no further data will be sought from the original researchers. Depending on the number of selected articles, the mean or risk ratios of the outcomes will be presented in a number of tables. Data to be extracted for the review and meta-analysis will include the following: (1) general information of the article (authors, year of publication, country of origin, source of funding), (2) study characteristics (aims of the study, study design, duration of the study, recruitment criteria, sampling technique, unit of randomization [participant or general practitioner]), (3) baseline participant characteristics by intervention group and control group (sample size, age, ethnicity, comorbidities, testosterone level, plaque volume, body mass index or weight classification, smoking status), (4) intervention (generic and trade name of the testosterone therapy, dosage, frequency of administration, and duration of intervention), (5) comparator (usual care or type of placebo for the control group), and (6) outcomes (number lost to follow-up, number of cardiovascular events, odds ratio, risk ratio, hazard ratio, mean [SD] change in volume/score, and effect size). If the data of the key outcomes are not available, the authors of the selected articles will be contacted for the required information.

The quality of the evidence will be assessed against Cochrane’s domains of risk of bias [15]. Meta-biases due to study design or methodological biases, reporting, and publication bias will be assessed. Publication bias will be assessed using Egger’s test and funnel plots [16]. Heterogeneity across studies will be assessed through visual inspection of the forest plot, I² statistic, and τ² [13]. A fixed-effect meta-analysis using the inverse-variance method will be performed if there is minimal heterogeneity [17]. A random-effect model following the DerSimonian and Laird method may be performed to account for heterogeneity (I² statistic >50%) across studies [18]. Possible small sample bias will be assessed by comparing the fixed effect estimate against that from the random effects model. The primary outcomes are the presence of clinical events of any type of stroke or transient ischemic attack, nonfatal myocardial infarction or acute coronary syndrome, emergency coronary artery revascularization, carotid surgery, cardiac mortality, and all-cause mortality. The secondary outcome is any atherosclerotic cardiovascular disease, which includes ischemic heart disease, coronary heart disease, coronary artery disease, cerebrovascular and carotid diseases, aortic aneurysm, and peripheral arterial disease. The number of events regarding primary and secondary outcomes among those who received any testosterone intervention will be compared to that of the control group of usual care or no testosterone intervention and the likelihood of events presented as risk ratios with 95% confidence intervals. Findings regarding coronary atherosclerosis will be reported as changes in the volume of calcified plaque and noncalcified coronary atheromatous plaque or coronary calcium score or Agatston score between baseline and endpoints of the studies (mean [SD]). Carotid atherosclerosis will be reported as the change in the carotid intima-media thickness and volume of carotid plaque between baseline and endpoint of the studies (mean [SD]). The presence of aortic or carotid inflammation on positron emission tomography, in absence of vasculitis, will be reported as the evidence of aortic or carotid atherosclerosis.

Three a priori hypotheses will be tested using stratified meta-analyses. These hypotheses are types of testosterone treatments (oral administration, transdermal application, intramuscular injection or implant vs no testosterone), dosage of testosterone received (total dosage [mg] received during the duration of the study vs no testosterone), and follow-up period (at least 6 months vs more than 6 months). Sensitivity analyses will be performed by excluding studies with high risk of bias and those with pharmaceutical funders. High-risk bias includes any bias rising from issues with allocation concealment (selection bias), blinding (performance bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias) [14], or a combination of these issues. The strength of evidence on the effect estimate will be reported as “strong confidence” (no further research will affect the confidence of the effect estimate reported in the current findings), “moderate confidence” (further research may affect the confidence of the effect estimate reported in the current findings), or “uncertain” (further research will most definitely affect the confidence of the effect estimate reported in the current findings).

All data extracted will be in aggregated form with no access to identifiable individual data. No ethics approval is required. This protocol has been registered on PROSPERO (Prospective Register of Systematic Reviews, registration number: CRD42019134278). Findings will be shared via PROSPERO and a peer-reviewed journal.