A review of the 257 meta-analyses of the differences between females and males in prevalence and risk, protective factors, and treatment outcomes for mental disorder

Abstract

Background

Despite decades of research, there are still unanswered questions about whether sex-specific treatment or prevention programs are required.

Method

An overview of reviews that is, an exploratory domain analysis comprising three umbrella reviews, using PRISMA guidelines, of meta-analyses reporting differences between males and females in the (1) prevalence and risk factors for mental disorders, (2) protective factors, and (3) treatment outcomes were conducted. To assist integration across umbrella reviews, sub-themes relevant to the domain were generated based on the findings of the searches. An extensive and transparent synthesis of the types of evidence, the comparative magnitude of effects, and representation is provided.

Results

We searched PubMed, Cochrane Library, and APA PsycINFO from inception to July 2021 and located 257 meta-analyses of 11,038 studies with 619,013,307 participants, reporting results for females and males, of which 205 reported prevalence and risk factors, 29 reported preventative and protective factors, and 23 reported on treatment outcomes. The first umbrella review has three main domains of (1) physical and behavioural conditions, (2) diagnosis and antecedents, and (3) suicide and antecedents, and these domains were further categorised into six (e.g., diabetes.), 14 (e.g., Anxiety), and six (e.g., Suicide) sub-themes respectively. The second umbrella review was categorised into five domains (e.g., coping), and the third umbrella review examined treatment outcomes. The risk of bias was assessed using publication bias reported by authors in selected articles, controlling for study design (e.g., clinical trial versus correlations) and effects (e.g., direct versus indirect). Across the three umbrella reviews, 43.9% of the estimates were greater in females, 28.5% in males, and 27.6% found no differences. The differences were small in 90% of the studies (Effect Size between 0 and 0.5). This review confirmed that the prevalence of internalising disorders was greater in females, and externalising disorders were more common in males. Suicide deaths were higher in males, but suicide attempts and ideation were higher in females; females with mental disorders had more heart disease but similar rates of diabetes and obesity. Sexual and physical abuse are strongly associated with a mental disorder in both females and males. Economic adversity and substance use were associated with the poor mental health of males. Rigorous physical activity was of greater benefit to males. In contrast, coping techniques and social supports were more beneficial for females, and females and males obtained similar benefits from treatment for most conditions.

Conclusion

A key finding was that most forms of treatment were similarly beneficial for males and females, although the study revealed significant gaps in research on treatment effects. However, the available evidence did not support the need for sex-specific treatment programs or services.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-025-06848-7.

Introduction

There are several well-established differences in the mental health of males and females. For example, epidemiological studies have consistently shown higher rates of anxiety and depression in females and higher rates of substance use and gambling disorders in males [1]. In most countries, deaths by suicide are higher for males [2–4], whereas suicide attempts are more common in females [5–7]. Regarding protective factors, physical activity is protective against depression for both males and females [8], but ceasing exercise is more strongly associated with increased depression in females [9]. We know that females do better than males in trauma-focused interventions for post-traumatic stress disorder (PTSD) [10], whereas males do better in interventions to stop smoking [11], but there is no difference between females and males in the outcome of cognitive behaviour therapy (CBT) for anxiety and depression [12]. Together, these findings suggest important differences in the mental health of males and females. However, studies to date have focused on specific areas of mental health, and there have been no attempts to synthesise or integrate our current knowledge of differences and similarities between males and females across all mental health domains.

The lack of a comprehensive review across different areas of mental health outcomes limits our ability to identify key gaps in knowledge of differences between males and females, in particular, the evidence of differences in treatment outcomes for mental disorders. One meta-analysis for the treatment of PTSD found females do better than males [10], whereas another meta-analysis for the treatment of anxiety and depression found no differences [12]. In addition, it is unclear whether programs to prevent mental disorders have an equal effect on males and females. These lead to the more significant scientific questions of whether there are meaningful differences in the aetiology and manifestation of mental disorders experienced by females and males and whether separate prevention programs and treatment services should be developed for females and males.

Definitions of sex and gender

The terms “sex” and “gender” have been used interchangeably in the social sciences [13] despite attempts to establish standard terminology [14, 15]. The terms have different meanings, including whether (a) a person is biologically male or female, (b) a person identifies as a man or woman, and (c) the degree to which a person identifies as masculine or feminine [1]. The Institute of Medicine defines sex as “the classification of living things, generally as male and female according to their reproductive organs and functions assigned by chromosomal complement” and gender as “a person’s self-representation as male or female, or how that person is responded to by social institutions based on the individual’s gender presentation” [16]. The definition of gender in the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM- 5) adds “the psychological, behavioural, and social consequences of one’s perceived gender” [17]. People who are gender diverse and do not identify with the gender binary system of sex assignment of male vs. female, that is, people with genderqueer, binary transgender, and non-binary identities [18, 19], are outside the scope of this current paper. Most research on differences between males and females uses the biological definition [20], and hence, this biological definition and the Institute of Medicine definition will be used in this paper.

Domains of differences between males and females in mental health research

There are well-described differences in how biological and cultural factors influence male and female behaviour that are beyond the scope of this review, which focuses on mental disorders. Instead, we identified three main domains of mental health: (1) the prevalence of mental disorders, e.g., [1, 21–23], and the risk factors associated with the presence of disorder (e.g., [24]), suicide and self-harm [3, 4] and risk factors associated with suicide, e.g., [25–27]; (2) factors that prevent and protect against mental disorders for females and males [8, 28]: and (3) outcome of treatment of mental disorder [10, 12, 29, 30].

A more complete understanding of both the differences and the similarities in mental health between males and females is needed to identify both the modifiable risk factors for mental disorders (e.g., [31–33]) and suicide (e.g., [26, 27, 34, 35]) and also protective factors that could be used to promote better mental health (e.g., [8, 28, 36]). A complete understanding of any differences in mechanisms of change and reasons for different treatment responses is also needed to evaluate treatment effectiveness and properly guide service development [10, 12, 29, 30]. Other important questions include the prevalence of psychiatric disorders in people who are not known to mental health services [23] and whether measures that improve mental health could also improve the physical health of both males and females [37].

Despite the many unanswered questions on this core topic in health, to date, there has been no umbrella review examining all the available evidence. An umbrella review of all the available evidence could confirm differences in the prevalence of mental health disorders and the associated risk factors and whether both prevention programs and treatment services should be sex-specific (i.e., an effect found for males or females only), sex-salient (an effect that is greater for males when compared to females or vice-versa) or sex-neutral (i.e., where there is no significant difference). It is also important to establish whether there is a significant unmet need for mental health care in males and females, reflected in the higher rate of suicide in males and the higher rate of suicide attempts in females and whether there should be separate treatment services for females and males.

Aims of this study

Building an evidence map of the diverse factors, outcomes, and typologies of evidence that constitute sex and sex-related differences in mental health remains a critical research imperative. Such mapping requires the integration of varied mental health-related determinants, outcome measures, and evidence types, alongside their multivariate and multidimensional relationships, into a cohesive framework. In this paper, we take on this challenge and propose a novel multidimensional exploration of differences of females and males across a range of health and behavioural factors associated with sex-related mental health disparities. We conceptualise this methodology as an exploratory domain analysis [38].

This study aims to explore, integrate, and synthesise existing knowledge on the similarities and differences in the mental health of males and females using an umbrella review of meta-analyses. While previous umbrella reviews [38–40] have been facilitated by the increasing number of meta-analyses on primary research within similar diseases or conditions [41], such reviews are often constrained by narrow inclusion criteria, limited scope, and a tendency to silo findings by outcome. As a result, they fail to construct a comparative and testable evidence map.

In contrast, domain analysis is an overview of reviews synthesising data across multiple domains, including diseases, disorders, and interventions, to generate an integrated, testable, and comparative framework of effects. This approach enables a more holistic assessment of sex and sex-related differences across the typology of mental health outcomes [42]. In this paper, we apply a domain analysis by conducting three umbrella reviews to systematically examine sex differences and similarities in mental health across three key domains: (1) the prevalence and risk factors associated with mental health disorders and suicide, (2) protective factors and prevention of mental health disorders, and (3) treatment of mental health disorders. We aim to synthesise evidence on the relative magnitudes of these similarities and differences using a heterogeneous sample of meta-analytical studies, focusing on implications for mental health treatment in males and females. There are no stated hypotheses because the aim is to explore, integrate, and synthesise the data and report on meta-themes using the domain analysis framework. We adopted a gender-neutral approach that focuses on sex similarities and is significantly predictive for females and males [39] and a gender-responsive approach that focuses on differences between males and females, examining sex-salient or sex-specific factors [40].

Method

Study registration

The reviews were conducted and reported using the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA; [41, 42]) and Cochrane Collaboration [43, 44] statement guidelines. The protocol was not registered on PROSPERO [45] as the register was not accepting new registrations when the study commenced due to a large backlog of studies.

Search strategy and terms

Comprehensive PubMed, the Cochrane Library, and APA PsycINFO searches were conducted from inception to July 2021. The dates these databases were searched were up to 31 July 2021. A broad search strategy was developed [46] by examining terms used in previous reviews [27, 47, 48] and by consensus between members of a research team consisting of three clinical psychologists (NT, BD, RK), a psychiatrist (ON), a statistician (EK) and a post-doctoral fellow in statistics and psychology (LS), with no restrictions of study populations or age. Studies in languages other than English were considered if an English translation of the abstract was available. The terms searched in the electronic databases were:

Risk and prevalence

(Meta-Analysis or Systematic Review).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh] AND (man or men or human males or masculinity or males or gender differences or sex differences or gender comparisons or gender comparator or sex comparisons or sex comparator).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh] AND (depression or psychological distress or mental health or stress or psychological stress or distress or anxiety or suicide or barriers or access to care).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh];

Preventive and protective factors

Exp Meta-Analysis/or exp Systematic Reviews/) AND (man or men or human males or masculinity or males or sex or gender differences or sex differences or gender comparisons or gender comparator or sex comparisons or sex comparator).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh] AND (Coping or Social Support or Resilience or Prevention or Wellbeing or Protective Factors or Mentoring or Rites of Passage).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh];

Treatment outcomes

(Meta-Analysis or Systematic Review).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh] AND (man or men or human males or masculinity or males or gender differences or sex differences or gender comparisons or gender comparator or sex comparisons or sex comparator).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures, mesh] AND (Treatment or Psychotherapy or Therapy or Intervention, or Evidence-based practice or mental health services or treatment outcomes, or treatment effectiveness evaluation or randomised control trials or clinical trials).mp. [mp = title, abstract, heading word, table of contents, key concepts, original title, tests & measures].

Inclusion criteria

The inclusion criteria, defined a priori [28, 49, 50], were any meta-analyses and systematic reviews that reported differences between males and females and similarities in mental health outcomes according to sex (Tables 1, 2 and 3). This criterion included studies where sex as a variable was of primary interest (i.e., direct) or as a covariate in moderator analysis (i.e., indirect).

Table 1.

Inclusion Table For Umbrella Review 1 – Prevalence and Risk Factors

Inclusion
Study Design	Meta-Analyses and Systematic Reviews
Method Participants	Epidemiological Surveys/Questionnaires Children, Adolescents and Adults (18 years and over)
Intervention	NA
Comparator or Control	NA
Primary outcomes	Non-Significant and Significant Sex/Gender differences of prevalence of mental health disorders and risk factors reported (i.e., Depression or Psychological distress or mental health or stress or psychological stress or distress or anxiety or suicide or barriers or access to care)

Table 2.

Inclusion Table For Umbrella Review 2 – Prevention and Protective Factors

Inclusion
Study Design	Meta-Analyses and Systematic Reviews
Method Participants	Standardised and Non-Standardised Questionnaires Children, Adolescents and Adults (18 years and over)
Intervention	NA
Comparator or Control	NA
Primary outcomes	Non-Significant and Significant Sex/Gender differences of prevention and protective factors associated with better mental health (i.e., Coping or Social Support or Resilience or Prevention or Wellbeing or Protective Factors or Mentoring or Rites of Passage)

Table 3.

Inclusion Table For Umbrella Review 3 – Treatment

Study Design	Meta-Analyses and Systematic Reviews
Method Participants	Standardised Questionnaires Children, Adolescents and Adults (18 years and over)
Intervention	Intervention Group (Psychological Treatment)
Comparator or Control	Control Group
Primary outcomes	Non-Significant and Significant Sex/Gender differences of treatment of mental health conditions (i.e., Treatment or Psychotherapy or Therapy or Intervention, or Evidence-based practice or mental health services or treatment outcomes, or treatment effectiveness evaluation or randomised control trials or clinical trials)

For Prevalence and Risk Factors, they were (1) Sex difference meta-estimate reported (between groups standardised mean difference, or SMD), Odds Ratio (OR), Relative Risk (RR), Correlational coefficient related to sex (R). Studies that reported no significant differences between males and females upon testing were also included even if a numerical estimate was not provided or missing; (2) All age groups (child, adolescents, and adults) (3) Risk factors for mental health, including suicidal thoughts, suicide attempts, suicide deaths, socio-economic status (SES), unemployment, marital status, and education etc., and (4) Studies of mental health disorders and associated risk factors, and (5) Prevalence rates of mental health disorders for females and males (e.g., anxiety, depression, and substance dependence, schizophrenia, etc.) which included prevalence meta-analyses that were not drawing on population-based data only.

For Prevention and Protective Factors, the inclusion criteria were (1) studies that reported sex difference meta-estimates upon testing and non-significant differences (2) All age groups (children, adolescents, and adults); and (3) Studies that included mental health prevention and protective factors (e.g., coping, social support, exercise, etc.).

For Treatment outcomes, the inclusion criteria were: (1) studies that reported sex difference meta-estimates upon testing and non-significant differences, (2) all age groups, including both children and adolescents and adults, (3) evidence-based psychological treatment (e.g., Cognitive Behavioural Therapy, Mindfulness-Based Therapy, Motivational Interviewing etc.) of a mental health disorder or comorbid physical and mental disorder, and design of clinical trials included open trials, RCTs and repeated measures.

Exclusion criteria

Studies examining trans-gender studies, male-only and female-only studies, and studies where differences between males and females were not reported or not tested, or the results were unclear or ambiguous, were excluded. Measures taken to identify or clarify missing or unclear information were to review primary study reports. We prioritised simple and redacted data over imputed or simulated data, and for transparency, the original and converted effect sizes are noted in the online Supplemental Excel File and Supplemental Table 2.

Identification of relevant studies

The lead author and post-doc researcher independently screened the titles and abstracts of all the identified records of the searches of the electronic databases. Publications identified as meeting inclusion criteria were then examined in full text. Records were coded by the same two reviewers and rated as ‘statistically significant or non-significant sex difference effect sizes,’ ‘maybe,’ and ‘no difference reported.’ Differences in coding were addressed by consensus after a joint examination.

Data extraction

The identified studies were extracted by the lead author (RK), and half of the articles were independently checked by two post-doctorate researchers using a data extraction Excel spreadsheet that listed the authors, year of publication, number of included studies (k), sample size (N), reported estimate of differences between males and females, 95% Confidence Intervals, risk of publication bias, direct (sex as a variable of primary interest) vs. indirect (sex as a covariate in moderator analysis) effect, design (i.e., clinical trial vs. correlation), age group, and a brief description of findings. Differences in data extraction were coded in red on the spreadsheet by two post-doctorate researchers. This was followed by a series of meetings with the lead author and post-doctorate researchers, where discrepancies were checked by referring to the original publication, and agreement on correct data was made.

As the analysis was completed on a meta-analytical level and not at the primary study level, the approach used with overlapping studies was to extract all meta-estimate data to address our research aim of integrating data across several domains.

Data inclusion and extraction

The broad viewpoint of the varied and many factors

We designed our data extraction process to meet our aim of being broad and inclusive. To reflect this aim, we included studies even when complete statistical information was unavailable, achieving representation from studies that did not specifically focus on differences. We extracted and included data from published meta-analyses that (i) reported a non-significant sex difference test conducted even when test statistics, or where effect sizes were not reported, (ii) included effects from studies that did not test for a sex effect but provided minimally sufficient statistical information to convert a sex difference into effect sizes using standard methodology [52], and combined such incomplete studies alongside (iii) studies that reported both the test and detailed statistics on a sex effect.

This broad inclusion of varied evidence types required us to prioritise inclusion and scale over the granularity and statistical rigour of the included studies. When authors of a study reported a non-significant sex difference, we adopted the author’s conclusion even when our independently converted effects implied the presence of a sex effect (i.e. sex differences that are greater than 1.96 standard errors, corresponding to a 95% confidence interval and a p < 0.05 test). Similarly, we conducted a statistical test of sex effects even when the study authors did not attempt this. We also estimated standard errors for effects when these were not reported outright. We retained the authors' original conclusion in the few studies in which our independent analysis did not match the author’s conclusion.

Our effects were extracted and converted using the standard statistical methodology outlined by the Campbell Initiative for data conversion [52]. Our data is available in the online Supplemental file and Supplemental Table 2, which includes the demarcation of tests and the type of original reported statistics next to the converted effects.

Together, these steps aim to operationalise our objective of achieving a broad, wide-ranging perspective on the evidence. We prioritise inclusion over traditional exclusion criteria based on rigorous standards. In this way, our choice of methods resembles a satellite image, offering broad coverage with lower resolution but delivering valuable high-level insights.

Risk of bias assessment

Most of the included meta-analyses in the current overview did not use a tool to assess the risk of bias. Thus, to determine the methodological limitations of eligible meta-analyses, we relied on the following: (a) authors’ ratings of original meta-analyses on the risk of publication bias [51]. Ratings of included studies are given in Tables 4, 5, 6, 7 and 8; (b) sex as a direct (primary variable of interest) versus indirect (moderating variable of interest) effect; (c) methodological design on outcomes (i.e., clinical trial vs correlational design); and (d) the impact of the type of effect size metric on mental health outcome was examined.

Table 4.

Summary of sub-theme 1: Studies of Physical and Behavioural Conditions [1, 21, 24, 32, 37, 48, 55, 56, 64–86, 137, 214–220]

 Note. ADHD=Attention Deficit Hyperactivity Disorder; BiPolar=Bipolar Disorder; BMI=Body Mass Index; CEA=Childhood Emotional Abuse; CPA=Childhood Physical Abuse; CPN=Childhood Physical Neglect; CSA=Childhood Sexual Abuse; Dep=Depression; Dept=Department; ES=Effect Size; IBD=Irritable Bow Disorder; IBS=Irritable Bow Syndrome; Methuse=Methamphetamine use; NS=Non-Significant Result; OR=Odds Rato; RR=Relative Risk; Sig=Significant Result; SMD=Standardised Mean Difference; TSH=Thyroid-Stimulating Hormone.

Table 5.

Summary of sub-theme 2: Studies of Diagnoses and Diagnoses Antecedents [1, 9, 21–23, 31, 33, 47–50, 56, 78, 81, 88–134, 136, 138–140, 208, 209, 221–273]

Note. Corr = Correlation; CSA = Childhood Sexual Abuse; ES = Effect Size; GAD = Generalised Anxiety Disorder; GD = Gender Difference; MDD = Major Depressive Disorder; MRR = Mortality Rate Ratio; NS = Non-Significant: OR = Odds Ratio; PD = Personality Disorder; QOR = Quality of Life; RR=Relative Risk; Sig = Significant; SMD = Standardised Mean Difference; SMI = Severe Mental Illness; SXS = Symptoms; YLD = Years of Life Lived with Disability; YLLS = Years of Life Lost

Table 6.

Summary of Sub-theme 3: Studies of Suicide and Suicide Antecedents [2–7, 25–27, 34, 35, 49, 102, 141–161, 163–170, 274–276]

Note. 12m = 12 months; ES = Effect size; NS = Non-Significant result, NSSI = Non-Serious Self-Injury; OR = Odds Ratio; PD = Personality Disorder; RR=Relative Risk; Sig = Significant result; WHO = World Health Organisation

Table 7.

Summary of Meta-theme 2: Studies of Prevention and Protective Factors [8, 28, 36, 171–190, 198, 277–282]

Note. CLBP = Chronic Low Back Pain; Corr = Correlation; ES = Effect size; IPV = Intimate Partner Violence; m = months; NS= Non-Significant result; OR = Odds Ratio; R= Spearman Correlation; RR=Relative Risk; Sig = Significant result; SMD = Standardised Mean Difference; US or USA = United States of America

Table 8.

Summary of meta-theme 3—Studies of Treatment [10–12, 29, 30, 98, 174, 191–204, 283–286]

Note. 12MFU = 12-month follow-up; CBCT = Cognitive Behavioural Couple Therapy; CBT= Cognitive Behavioural Therapy; ES = Effect size; MB= Mindfulness-Based; NS = Non-Significant result; OR = Odds Ratio; pt = patient; Sig = Significant result; SMD = Standardised Mean Differences; TX = Treatment.

Data categorisation and themes

To assist with data synthesis and integration, each umbrella review was categorised by the lead author (RK) and statistician (EK) into layers of themes based on what emerged from the frequency of the studies (i.e., n > 2) within each umbrella review. Infrequent studies (n < 3) were categorised as other. For umbrella review one, layer 1, the domain of prevalence and risk was categorised into layer 2, the three sub-themes of Physical and Behavioural Conditions, Diagnoses and Antecedents, and Suicide and Suicide antecedents. These sub-themes of layer two were further categorised into layer 3. For Physical and Behavioural Conditions, layer 3 included the following six categories: (1) diabetes risk factors, (2) heart disease risk factors, (3) other risk factors, (4) addiction prevalence, (5) weight risk factors, and (6) substance abuse risk factors. For diagnoses and antecedents, layer 3 included the following 14 categories: (1) anxiety prevalence, (2) anxiety risk factors, (3) depression prevalence, (4) depression risk factors, (5) behavioural disorder prevalence, (6) behavioural disorder risk factor, (7) cognitive prevalence, (8) other prevalence, (9) PTSD prevalence, (10) PTSD Risk Factors, (11) Severe mental illness (SMI) prevalence, (12) SMI Risk Factors, (13) Other Risk Factors, and (14) Stress Prevalence. For Suicide and Suicide Antecedents, layer 3 included the following six categories: (1) Death by Suicide Prevalence, (2) Death by Suicide Risk Factors, (3) Suicidal Attempts Prevalence, (4) Suicidal Attempts Risk Factors, (5) Suicidal Ideation Prevalence, (6) Suicidal Ideation Risk Factors.

For umbrella review two, layer 1, the domain of prevention and protective factors was categorised into layer 2, the five sub-themes of (1) Physical Activity, (2) Coping, (3) Other, (4) Social Support, and (5) Wellbeing. For umbrella review three, layer 1, the treatment domain was not categorised and remained as treatment.

Handling of overlapping studies and effect multiplicity

We note that some meta-analyses contributed multiple effects, sometimes across different themes and sometimes within the same theme. Multiple counting of the same primary studies can potentially overestimate some effects.

To conduct a large-scale yet low-resolution synthesis of the evidence, we prioritised an inclusive approach that acknowledges some overlap in listing different sources of evidence. Other common solutions, such as statistical streamlining using techniques like multi-level random intercepts, were also considered. However, the primary focus of this analysis was to capture a broad range of mental health findings in a heterogeneous sample of meta-analyses rather than refining effect estimates from a narrow set of homogeneous studies.

For clarity, each table reports the total number of studies included across all meta-analyses and the number of unique studies contributing to different effects. We also note the total effect sizes and effect sizes originating from the same meta-analysis. This is detailed in the effect size overlap in each section, with further information provided in Supplemental Table 3.

Data synthesis

Investigating sex-related differences across studies of the prevalence of diagnosis and risk factors associated with diagnosis, protective factors, and treatment resulted in the aggregation of studies, topics, and metrics (e.g., percentages, odds ratio, relative risk, correlation coefficients, and effect sizes). The varied estimates of sex-related differences from various studies were scaled into standardised effect size metrics that streamline and compare the varied magnitude of sex-related differences.

In this review, we aimed to provide an inclusive synthesis of how sex differences in mental health outcomes have been reported across meta-analyses rather than to generate pooled meta-analytic estimates. Where possible, we converted between odds ratios (ORs), relative risks (RRs), and correlations to facilitate comparison [52]. However, a minority of studies reporting prevalence rates were retained in their original form, typically as risk ratios, given that most of the reported prevalence meta-estimates lacked sufficient information for effective conversion (e.g., base rates or event counts), used risk metrics not suited to OR conversion (e.g., rates per 100,000), or reported differences in relative physical quantities (e.g., biomarkers rates) where conversion was impractical. We acknowledge that ORs and RRs offer related but distinct perspectives on relative risk. However, while RR and OR metrics differ in their mathematical properties, they reflect efforts to quantify the magnitude and direction of sex-based disparities and present the data in its most direct form. To ensure transparency, we clearly distinguish ORs, RRs, and prevalence estimates in all tables and supplementary materials. We argue that this approach provides a more comprehensive picture of the relative impact of sex across diverse analytic strategies while remaining faithful to the original reporting of each study. Interpretation of direction, magnitude of effects, and significance of both ratio metrics was streamlined in all tables and figures. In most studies, scaling involved the conversion of relative risk estimates from the umbrella reviews into odds ratios and the conversion of odds ratios and correlation statistics into standardised effect size estimates [52]. Estimates from the included studies that examined the prevalence of diagnoses and associated risk factors with diagnosis used a combination of odds ratio metric and relative risk ratios (marked RR) to streamline and benchmark the effects.

Estimates of correlation coefficients were used for preventative and protective factors. Estimates of standardised mean differences were used for outcomes that compare treatment-related differences. If no effect size metrics were reported, we extracted information that would allow the calculation of an effect size estimate, including both effect size and a measure of effect size uncertainty (i.e., confidence interval).

Each of the meta-estimates extracted was further characterised as demonstrating statistically significant sex-related differences if a significant (or non-significant) statistical test was identified in the original manuscript. If this information was not reported, estimates of sex-related differences were determined as significant if the corresponding 95% confidence interval of an effect size exceeded the margin of zero group differences (e.g., zero for correlation coefficients or one for odds ratio).

To provide a summary of the similarities, differences, and subthemes within the mental health literature, we presented the ratio of studies that report on outcomes for females doing better than males in the following: (1) studies that report no significant differences; (2) and studies that report on outcomes for males doing better than females; and (3) where males did better within each of the sub-themes of the domains identified by the authors. Further information about the identified studies is available in the comprehensive online supplemental file and tables (see Supplemental Tables 1-3).

To explore the magnitude of the differences, we used Cohen [53] classified effect size for small differences (effect size (ES): 0.0–0.5, Correlation (R): 0.0—0.24, Odds Ratio Males or Relative Risk ratio for males (OR|RR-M: 1.0–2.5), Odds Ratio Females (OR|RR-F; 0.4–1.0)), moderate differences (ES: 0.5–0.8; R: 0.24 − 0.37; OR|RR-M: 2.5–4.3; OR|RR-F: 0.23–0.4) and large differences (ES: > 0.8; R: > 0.37; OR|RR-M: > 4.3; OR|RR -F: < 0.23). To further enhance and streamline the visualisation of all three types of effects in standard plots, such as forest plots, the ranges of all three effects were converted to a Fisher’s Z correlation matrix in all plots. All metrics are presented in the supplemental materials.

Sample size reporting, per effect and overall

Several included meta-analyses reported multiple effect sizes per outcome (i.e., multiplicity in 75 of the 257 meta-analysis studies included), resulting in overlapping study samples and potential duplication (135 of the total 435 effects extracted). However, due to incomplete reporting, it was not always possible to determine whether individual studies contributed to more than one effect. To avoid potential double-counting and inflation of the evidence base, we adopted a conservative approach: for each meta-analysis, we retained a single overall estimate of the total number of studies and sample size, which we used as the common denominator in summary statistics (e.g. for the sum of the 257 meta-analysis studies). While we report the number of studies and sample size associated with each individual effect, the total sample size across the review reflects one aggregated estimate per meta-analysis.

Sensitivity analysis for variance estimation methods

In our analysis, standardising variance across the range of studies proved challenging due to incomplete reporting of effect variances, sample sizes, test statistics, various statistical metrics, and combinations of these gaps in different studies (presented in the Supplemental Tables). To be inclusive and provide a comprehensive evidence map, we incorporated and standardised as much of the reported data as possible. However, we acknowledge that more streamlined, assumption-based methods exist, such as estimating effect variance from fixed and random sample variances instead of relying solely on variance derived directly from moderation test parameters. In our study, variance estimates reported directly from the studies were prioritised, as these are generally considered the most reliable reflection of the original data. However, alternative statistical approaches were also incorporated, given the wide range of effect types, the differing degrees of statistical conversion required from the available data, and the inconsistent availability of uniform testing across studies. These approaches included an estimation of variance under a fixed effects model, wherein variance was derived from the sample size and the number of available studies, as well as a second contrast in which variance was estimated under a random effects model assuming a between-study variance of 50%, thereby incorporating a heterogeneity penalty parameter. Although these alternative methods provide a more homogeneous means of assessing variance, they are inherently less direct than the reported values. In the sensitivity analysis, these streamlined estimates were compared with the effects derived directly from observed study variances, allowing the exploration of how different variance estimation approaches influence the interpretation of the effect estimates. This step offers an opportunity to gauge alternative variance estimates and to assess the impact of methodological choices on the overall meta-analytic conclusions.

Bias and tests of moderation

Possible sources of measurement bias were explored across the three umbrella reviews by including five sensitivity analyses. These analyses employed Chi Square tests of proportions to compare the frequency of significant male, female, and non-significant sex-related meta-estimates across several moderating variables and within each theme. First, these tests examined the association of meta-estimate results with the level of publication bias (i.e., low (L; non-significant fail-safe n test and/or symmetrical funnel plots), medium (M; mixed results of significant fail-safe n test and symmetrical funnel plots or vice-versa), High (H; significant fail-safe n test and/or asymmetrical funnel plots), and Insufficient (I)) [51, 54] (i.e., see Supplemental Table 1). Here, the level of publication bias could be explored as a potential confound on the effect of sex on mental health outcomes.

Second, there are outcome differences between studies where sex is examined as the primary variable of interest (i.e., direct effect) compared to studies where sex is examined as a secondary variable of interest, that is, as a moderator of outcomes (i.e., indirect effect). In this way, any methodological differences between direct and indirect effects could be explored as potential confounders of the effect of sex on mental health outcomes (i.e., See Supplemental Table 1).

Third, the effect of differences in age group (i.e., adult versus child/adolescence) on mental health outcomes. In this way, the adjusted models explored the impact of age group as a possible moderator of mental health outcomes (i.e., Supplemental Table 1).

Fourth, the effect of differences in methodological design on outcomes (e.g., clinical trial vs. correlational design). Here, the adjusted models explore the impact of methodological design as a potential confound on the effect of sex on mental health outcomes (Supplemental Table 1).

Fifth, the effect of the type of effect size metric on mental health outcomes was examined. This was to determine if effect size type (i.e., standardised mean difference (SMD), effect size (d), correlation (R), relative risk (RR), or Odds Ratio) as a potential confound on the effect of sex on mental health outcomes (i.e., see Supplemental Table 1).

Together, these analyses aim to subject the results to several methodological biases and confounding tests and test the reliability of the results to our methodological decisions of sensitivity testing. In combination, as a set of multiple steps, including (i) incorporating broad selection criteria while (ii) streamlining and integrating diverse statistics and themes, form the building blocks of our domain analysis and our effort to create an expansive viewpoint of the evidence.

Results

Search results

The electronic searches yielded 3073 titles. After removing duplicates and irrelevant articles, 535 papers were examined in full text. Of these, 257 met inclusion criteria, 205 on prevalence and risk factors (see Fig. 1), 29 on protective and preventative factors (see Fig. 2), and 23 on treatment outcomes (see Fig. 3).

Fig. 1.

Umbrella Review 1 – Prevalence of Mental Health Disorders and Associated Risk Factors

Fig. 2.

Umbrella Review 2 – Prevention and Health Promotion

Fig. 3.

Umbrella Review 3 – Treatment

Study characteristics

Sample size and number of studies

Tables 4, 5, 6, 7 and 8 display the number of studies (k) and sample sizes (n), converted effect sizes across the three umbrella reviews. The 257 meta-analyses reported 435 meta-estimates of sex-related differences, based on 11,038 studies with 619,013,307 participants. There was no duplication of meta-estimates within each meta-analysis. Within each domain, there is an overlap of meta-estimates with no duplication.

Overlap of meta-analytical effect sizes

Repeated meta-analytical effect sizes that measured the same outcomes in a specific domain were deemed overlapping effect sizes. For example, the number of repeated meta-analytical effect sizes reporting on the prevalence of mental health disorders (e.g., depression), risk factors (alcohol and suicide), prevention (exercise and depression), and treatment (CBT for depression) were reported. The greater the number of overlapping meta-analytic effect sizes, the greater the evidence supporting the trend in the domain analyses.

Tables 4, 5, 6, 7 and 8 summarise the similarities and differences between males and females in estimates of mental health outcomes across the three domains. Overall, 77% of the studies on prevalence and risk factors, 66% on preventative and protective factors, and 39% on treatment outcomes reported significant differences.

Prevalence and risk factors

Sub-themes identified by the authors from the examination of the studies were similarities and differences in diagnosis and the association between mental disorders and physical (e.g., diabetes) and behavioural conditions (e.g., substance abuse) according to sex.

Physical and behavioural conditions

The estimates from the included studies summarised in Table 4 and Fig. 4 show mixed results for the effect of sex on the influence of physical conditions on mental health disorders. Overall, for physical and behavioural conditions as a risk factor for developing mental health disorders, 22 estimates (41%) reported a greater risk for females, 18 estimates (33%) reported no sex difference, and 14 estimates (26%) reported greater risk for males.

Fig. 4.

Umbrella Review 1 – Sub-theme 1—Physical and Behavioural Conditions. Note. BiPolar = Bipolar Disorder; BMI = Body Mass Index; CEA = Childhood Emotional Abuse, CEN = Chilhood Emotional Neglect; CPA = Childhood Physical Abuse; CPN = Childhood Physical Neglect; CSA = Childhood Sexual Abuse; HIV= Human Immunodeficiency Virus; IBS = Irritable Bowel Syndrome; NS = Non-Significant Result; Sig = Significant Result; TSH = Thyroid-Stimulating Hormone

Body weight risk factors

For weight-related risk factors, nine (69%) meta-estimates from seven meta-analyses found no differences between males and females in the association of weight-related factors on symptoms of depression [55–61], two meta-analyses (15%) found differences, with females with mental health disorders experiencing more difficulties with weight and obesity when compared to males [62, 63]. One study reported a greater risk for males experiencing stress [64].

Heart disease and mental health

Five meta-analyses found females diagnosed with mental health disorders are at greater risk of heart disease when compared to males [56, 65–68]; two found no sex differences [69, 70], and one found a slightly higher risk of heart disease in males with depression [71].

Diabetes

For people diagnosed with a mental disorder, one meta-analysis found that females were at a slightly higher risk of diabetes than males [72], and another found males were at a slightly greater risk [37]. Three meta-analyses found no sex difference in the risk of developing diabetes [73–75].

Substance abuse

Substance abuse is associated with poor mental health in both males and females. Ten (48%) meta-estimates from six meta-analyses reported a greater risk for males [48, 76–80]. Males were more likely to have a comorbid bipolar disorder, job strain, and a higher risk of mortality. Three meta-estimates (12%) from three meta-analyses reported no differences between males and females for age of onset [81], sleep disorders [82], or genetic risk [32]. Twelve meta-estimates (40%) from six meta-analyses reported a greater risk of mental disorders for females with substance abuse [78, 80, 83–86]. Females with substance abuse disorder were more likely to have experienced childhood abuse or were influenced to use substances by a partner. Mixed results emerged for substance abuse and depression; one study found that males reported higher levels of depression [48], and another study reported females who use amphetamines reported higher depression levels [86].

Overlapping effect sizes

Of the five meta-analyses reporting Diabetes Risk Factors, three overlapping meta-analytical studies reported the association with depression, and two reported work-related factors.

Of the eight meta-analyses reporting on Heart Disease Risk Factors, five overlapping meta-analyses reported on the association with depression.

Of the 11 meta-analyses reporting on the prevalence of substance and behavioural dependence – three overlapping meta-analyses reported on alcohol use.

Of the 12 meta-analyses reporting weight risk factors, six overlapping meta-analyses reported an association with depression.

Diagnoses and diagnoses antecedents.

Overall, for diagnoses and diagnoses antecedents, 102 estimates (50%) reported a greater prevalence or risk of mental disorder for females, 52 estimates (26%) reported no sex difference, and 48 estimates (24%) reported a greater prevalence or risk for males.

Internalising disorders and associated risk factors

Internalising disorders are defined as mental health disorders characterised by high levels of negative affect, including depressive disorders, anxiety disorders, obsessive–compulsive disorders, trauma and stressor-related disorders and dissociative disorders [87]. Overall, for internalising disorders and associated risk factors, 67 estimates (59%) reported a greater prevalence or risk for females, 27 estimates (24%) reported no sex difference and 20 estimates (17%) reported greater prevalence or risk factors for males.

Prevalence of internalising disorders

The studies summarised in Figs. 5, 6, and 10 with 58 estimates of the prevalence of PTSD, stress, anxiety, and depressive disorders between males and females, many derived from very large population-based epidemiological studies (e.g., 272 million in the global burden of anxiety study [88]). Of the 58 estimates, 47 (81%) found a higher prevalence of internalising disorder in females [1, 21, 22, 48, 88–115], with seven (12%) estimates showing no difference in hypochondriacal symptoms [116], depression in Iranian college students and older adults [117, 118], depression during COVID- 19 [100], and change of depression levels over time [119]. Four estimates (7%) found that males had a greater number of years of life lost associated with anxiety [88], higher levels of depression in Chinese college students [120], poor prognosis after 6–8 months [121], and higher cortisol response to psychosocial stress when older [122].

Fig. 5.

Umbrella Review 1 – Sub-theme 2 – Diagnoses and Antecedents Anxiety/Depression. Note. CSA = Childhood Sexual Abuse; GAD = Generalised Anxiety Disorder; GD = Gender Difference; MDD = Major Depressive Disorder; NS = Non-Significant Result; OCD = Obbsessive Complusive Disorder; Sig = Significant Result; SXS = Symptoms; YLD = Years of Life Lived with Disability; YLLS = Years of Life Lost

Fig. 6.

Umbrella Review – Sub-theme 2 – Diagnoses and Diagnoses Antecedents – Other disorders. Note. ADHD = Attention Deficit Hyperactivity Disorder; CSA = Childhood Sexual Abuse; GAD = Generalised Anxiety Disorder; GD = Gender Difference; MBCT= Mindfulness-Based Cognitive Therapy; MBSR = Mindfulness-Based Stress Reduction; MDD = Major Depressive Disorder; NS = Non-Significant Result; OCD = Obssesive Compulsive Disorder; PD = Personality Disorder; PTSD = Post-Traumatic Stress Disorder; QOL = Quality of Life; RR = Relative Risk; Sig = Significant Result; SMI=Severe Mental Illness; SXS = Symptoms; YLD = Years of Life Lived with Disability; YLLS = Years of Life Lost

Fig. 10.

Sex Similarities and Differences across the three umbrella reviews. Note. NS; Non-Significant; PTSD=Post-Traumatic Stress Sisorder; SMI = Severe Mental Illness

Risk factors associated with internalising disorders

Overall, for risk factors associated with internalising disorders, 20 estimates (36%) reported a greater prevalence or risk for females, 20 estimates (36%) reported no sex difference, and 16 estimates (28%) reported greater prevalence or risk for males.

Risk factors for developing internalising disorders in females and males included inherited and psychosocial factors, such as physical abuse and sexual abuse. One meta-analysis found that physical abuse was a greater risk factor for males developing anxiety and depressive disorders [123]. However, another meta-analysis found no significant sex difference in the risk of developing anxiety disorder after physical abuse, with increased risk for both males and females [124]. For childhood sexual abuse, one meta-analysis reported a non-significant trend that sexual abuse is a greater risk factor in developing anxiety disorders [124], and another meta-analysis found it was a significant factor for depression in females [125]. There were also no differences between males and females in the effect of work conditions [126], perceived discrimination [127], and the effect of partner violence [123] on the likelihood of developing depression. For biological factors, one meta-analysis found that changes in the brain-derived neurotrophic factor (BDNF) were associated with a greater inherited risk of depression in males [31], while another meta-analysis found no difference [128].

Factors that increased the likelihood of males developing depressive disorder when compared to females included separation, being single and being widowed [129], claiming a disability pension [130], substance abuse and the frequency of risk-taking and impulsive behaviours [48]. Factors that conveyed a greater risk for females were screen-time sedentary behaviours [131], weight-related factors [48, 132], cessation of exercise [9], sleep problems [48], hospitalisation [133] and negative thinking about the future [134].

Overlapping effect sizes

Of the 22 meta-analytical effect sizes reporting on anxiety – 15 effect sizes reported on the prevalence of anxiety.

Of the nine meta-analytical effect sizes reporting on anxiety risk factors, three effect sizes reported on the association with abuse.

Of the 27 meta-analytical effect sizes reporting on depression – 27 effect sizes reported on the prevalence of depression.

Of the 40 meta-analytical effect sizes reporting on depression risk factors, four effect sizes reported an association with loneliness.

Externalising disorders and associated risk factors

Externalising disorders are defined as mental health disorders characterised by maladaptive behaviours affecting others and a person’s environment, such as antisocial personality disorder, conduct disorder, attention-deficit/hyperactivity disorder (ADHD) and addictions [135]. For the diagnosis of an externalising disorder or the presence of an associated risk factor, 33 estimates (42%) reported a greater prevalence of diagnosis or risk for males, 21 estimates (27%) reported no sex difference, and 24 estimates (31%) reported greater prevalence or risk for females.

Seventy-six per cent (16 of 21) of estimates on the prevalence of the externalising disorders ADHD [1], conduct disorder [47], substance use disorder [1, 21, 24] and problem gambling [1] were higher among males. Twenty-four per cent (5 of 21 estimates) showed no differences. Two meta-analyses yielded four meta-estimates that found no difference in aggression, victimisation, psychosis, and mania amongst adolescents [102, 136]. Another meta-analysis found no differences between males and females in the prevalence of alcohol use disorder across several countries [137].

Females who have been sexually abused are more likely to develop externalising disorders of conduct disorder [138], substance abuse disorder [84], or severe mental illness [123, 139] when compared to males. Males who have been physically abused, such as being physically attacked, threatened with a weapon, or sustained or witnessed serious injury in an accident are more likely to develop conduct disorder [138], and personality disorders [123].

Factors associated with severe mental illness (psychotic and bipolar disorders)

For severe mental illness, four estimates (45%) reported a greater risk for females, three estimates (33%) reported no sex difference, and two (22%) estimates were greater for males. Identified genetic profiles [33], and psychosocial risk factors of sexual abuse [139] and partner violence [123] all increased the likelihood of females developing severe mental illness. In contrast, a family history of schizophrenia increased the likelihood of a diagnosis of severe mental illness in males [140].

Overlapping effect sizes

Of the eleven meta-analyses reporting on PTSD – four and six overlapping meta-analyses reported on the prevalence of PTSD and the association of genetic factors, respectively.

Of the 23 meta-analytical effect sizes reporting on behavioural disorder risk factors, seven effect sizes reported on the association with physical and sexual abuse, and there were three effect sizes with substance use.

Of the nine meta-analytical effect sizes reporting on severe mental health risk factors, four effect sizes reported on the association with violence and abuse.

Suicide and suicide antecedents

Overall, for suicide and suicide antecedents, 31 estimates (44%) reported a greater prevalence of identified suicide risk factors for females, eight estimates (12%) reported no sex difference and 31 estimates (44%) reported a greater prevalence of suicide risk factors for males.

Suicide

Overall, for risks associated with completed suicide, 19 estimates (66%) reported a greater prevalence or associated risk for males, four estimates (13%) reported no sex difference and six estimates (21%) reported greater prevalence or associated risk for females.

The studies summarised in Figs. 7 and 10 show the prevalence of suicide and suicide antecedents. In males, the rates of completed suicide are higher in nearly all regions [3, 4, 141–146] except in China [147], among Multiple Sclerosis patients [148] and physicians [149]. Risk factors for a greater probability of females completing suicide than males include child abuse [150], substance abuse [34], affective disorder diagnosis [151], and prior contact with mental health services [152] and lithium in drinking water [153]. For males, an increased likelihood of suicide was associated with low education and occupational status [34], recent hospital discharge [2], unemployment [154], cancer [155, 156], ready access to means, the suicide of a friend [146], childhood disorders, any substance use (except in China) [34], alcohol use disorder, personality disorder [25, 151], and a diagnosis of schizophrenia or severe mental illness [157, 158]. Risk factors for suicide in which there were no differences between males and females included inherited factors [159] and ambient temperatures in Eastern Asian countries [160]. The role of divorce as a risk factor for death by suicide suggested that males were at more risk, with one meta-analysis indicating a greater risk for males [35] and another showing a non-significant trend towards a greater risk of suicide for males [49].

Fig. 7.

Umbrella Review 1 Sub-theme 3 – Suicide and Suicide Antecedents. Note. 12 m = 12 months; NS = Non-Significant Result; NSSI = Non-Suicidal Self-Injurious Behaviours; PD = Personality Disorder; Sig = Significant Result; SMI = Severe Mental Illness; WHO = World Health Organisation

Overlapping effect sizes

Of the 12 meta-analytical effect sizes reporting on death by suicide, 12 effect sizes reported on the prevalence of death by suicide.

Of the 29 meta-analytical effect sizes reporting on risk factors associated with death by suicide, five effect sizes reported on the association with substance use.

Suicide attempts

Overall, for suicide attempts and associated risks, 13 estimates (72%) reported a greater prevalence or associated risk for females, three estimates (17%) reported no sex difference, and two estimates (11%) reported greater prevalence or associated risk for males.

Several meta-analyses confirmed that females attempt suicide at a higher rate than males [5, 6, 102, 145, 146, 161–165], except among Chinese college students, where the rates of non-fatal self-harm were higher in males [166], and another meta-analysis found no differences between males and females [145]. A risk factor for males attempting suicide at higher rates than females was having a history of childhood sexual abuse (CSA) [27], although another meta-analysis reported no differences between males and females [167]. Obesity was a risk factor for females attempting suicide at higher rates than males [26].

Overlapping effect sizes

Of the 14 meta-analytical effect sizes reporting on suicidal attempt rates, 14 effect sizes reported on the prevalence of suicidal attempts.

Of the four meta-analytical effect sizes reporting on risk factors associated with suicidal attempts, two effect sizes were reported on the association with abuse.

Suicide ideation

For suicide ideation, nine estimates (82%) reported a greater prevalence or risk for females, one estimate (9%) for males and one estimate (9%) reported no differences between males and females. Although six estimates from five meta-analyses found that suicidal ideation was more common in females than males [5, 6, 161, 165, 168], one meta-analysis found more suicidal ideation in boys [169]. Adolescent females in the WHO-defined region of the Americas had a higher rate of suicidal ideation than adolescent males, whereas no sex difference was reported in the African region [170].

Overlapping effect sizes

Of the seven meta-analytical effect sizes reporting on suicidal ideation rates, seven effect sizes reported on the prevalence of suicidal ideation.

Prevention and protective factors

Among the studies of physical activity, coping and social support, and the use of professional support for mental health, 30 estimates (43%) reported greater use of protective factors among females, 24 estimates (34%) reported no difference, and 16 estimates (23%) reported more protective factors for males.

For the effect on mental health and the use of coping skills, defined as accepting responsibility, exercising self-control, positive reappraisal, and adaptive problem solving, ten estimates (59%) reported the coping skills in females to be more protective, seven estimates reported no difference (41%), and no estimates reported males made greater use of coping skills as a protective factor for their mental health. Five meta-analyses that reported seven meta-estimates found no sex difference in the effect of using coping skills as a protective factor for mental health [171–175]. Three of the above five meta-analyses reported greater use of communication, positive reappraisal, active coping, and structured problem-solving in females but also a greater propensity for rumination, avoidance and self-blame than males [171, 174, 176].

The studies summarised in Figs. 8 and 10 show that vigorous physical activity was more protective for cognition, mood and suicide attempts in males [177–179]. In contrast, less vigorous activity showed no differences between males and females [8, 178, 180].

Fig. 8.

Umbrella Review 2—Prevention and Protective Factors. Note. IPV = Intimate Partner Violence; m = months; NS = Non-Significant Result; PTSD=Post-Traumatic Stress Disorder; Sig = Significant Result; US or USA = United States of America

In terms of using social support to improve mental health, 15 estimates (65%) reported that females had more protective factors, six estimates (26%) reported no difference, and two estimates (9%) reported that males used social support more as a protective factor.

Females reported less social isolation [176], less internet addiction [181], higher levels of spousal, emotional, and family support [182] and greater use of professional support [183]. Males reported a higher level of same-sex cooperation [184] and non-specific forms of social support [176]. Meta-analyses of the protective effects of social support on sexual health [185], chronic low back pain [186], depression in children and adolescents [28], and psychopathology [187] found no differences between males and females.

In measures of well-being, two estimates (43%) reported greater levels of happiness and satisfaction with life in females [188], two estimates (28.5%) reported no difference [188, 189], and three estimates (28.5%) reported that males had both lower levels of psychological symptoms [182] and a greater level of well-being [188, 190].

Overlapping studies

Of the 13 meta-analytical effect sizes reporting on physical activity, eight effect sizes reported on the effect of physical activity on depression.

Treatment outcomes

For the outcomes of a range of treatments for mental disorders, six estimates (21%) reported a greater response in females, 17 estimates (61%) reported no sex difference, and five estimates (18%) reported greater improvement in males.

The studies summarised in Figs. 9 and 10 indicate that there was no significant difference in the outcome of treatment of mental disorders between males and females, other than in treatments to address anger [191], weight loss and use of exercise [192, 193], PTSD [10, 194], tobacco smoking [11], relationship difficulties [174], and heart disease [195]. For treatment of alcohol abuse, the only meta-analysis to find that females did better than males was a 1992 study that showed better results at 3–12 month follow-up but reversed after 12 months [196]. Six subsequent meta-analyses found no sex difference in the effect of treatment for alcohol misuse on mental health [29, 197–201]. Moreover, no significant differences between males and females were found in meta-analyses that examined the efficacy of psychological treatment of depression [12, 202, 203], anxiety [30], and self-control [204]. Females did better in the treatment of PTSD [10, 194] and anger [191]. Males did better than females in treatment that targeted smoking [11], weight loss and exercise [192, 193], and heart disease [195].

Fig. 9.

Umbrella Review 3 – Treatment. Note. CBCT = Cognitive Behavioural Couple Therapy; CBT = Cognitive Behavioural Therapy; MB-CBT = Mindfulness Based—Cognitive Behavioural Therapy; MFU = months follow-up; NS = Non-Significant Result; PTSD=Post-traumatic Stress Disorder; Sig = Significant Result; TX = Treatment

Overlapping Studies

Of the 28 meta-analytical effect sizes reporting on the treatment of mental health disorders, 10 and three effect sizes reported on the treatment of alcohol abuse and Cognitive Behavioural Therapy (CBT) of depression, respectively.

The magnitude of the differences and similarities within and across the three umbrella reviews

Of the 435 estimates (see Fig. 10) reported in 257 meta-analyses included in the three umbrella reviews, 394 (90%) reported no difference or small differences (effect size (ES): 0.0–0.5, Correlation (R): 0.0 − 0.24, Odds Ratio Males (OR-M: 1.0–2.5), Odds Ratio Females (OR-F; 0.4–1.0)), 30 estimates (7%) reported moderate differences (ES: 0.5–0.8; R: 0.24 − 0.37; OR-M: 2.5–4.3; OR-F: 0.23–0.4) and only 11 estimates (3%) reported large differences (ES: > 0.8; R: > 0.37; OR-M: > 4.3; OR-F: < 0.23).

Of the 337 estimates (see Fig. 10) reported in 205 meta-analyses of the first umbrella review on diagnosis and associated risk factors, 300 (89%) reported no difference or small differences, 27 estimates (8%) reported moderate differences and 10 estimates (3%) reported large differences.

Of the 70 estimates (see Fig. 10) reported in 29 meta-analyses of the second umbrella review on protective factors and prevention, 67 (96%) reported no difference or small differences, two estimates (3%) reported moderate differences and one estimate (1%) reported large differences.

Of the 28 estimates (see Fig. 10) reported in 23 meta-analyses of the third umbrella review on treatment, 27 (96%) reported no difference or small differences, one estimate (4%) reported moderate differences and zero estimates (0%) reported large differences.

Bias and tests of moderation

The results of the sensitivity analyses across the three umbrella reviews found the following: [1] The subset of Diagnosis and Diagnosis Antecedents within the Umbrella review 1 Risk and Prevalence found that sex as a primary variable of interest, that is, direct effect (p = 0.028), age group children (p = 0.01), and studies reporting correlation effect size (p < 0.0001) confounded the effect of sex on the mental health outcome of prevalence and risk of mental health disorders; (2) The Umbrella Review 2 Prevention and Protective Factors found that methodological designs of clinical trials (p =. 002) and when effect size metric type is not reported (p =. 008) confounded the effect of sex on the mental health outcome of prevention and protective factors; (3) The Umbrella Review 3 Treatment found no variables that confounded the effect of sex on treatment outcome.

These sensitivity analyses infer that the pattern of significant and non-significant sex-related meta-estimates was robust across variation in the publication bias of the studies considered, variation in the source of the study (i.e., direct and indirect effects; clinical trials vs. observational studies), consideration of the results by stratification of major age groups, and differing methodologies for quantifying sex-related effects (effect size metrics).

Publication bias for the first umbrella review, prevalence and risk, yielded 40.3% low risk, 9.7% medium risk, 7.3% high risk and 42.2% insufficient evidence. For the second umbrella review, prevention and protective factors; 43.8% low risk, 12.5% medium risk, 12.5% high risk and 28.1% insufficient evidence. For the third umbrella review, treatment, 22.75% low risk, 22.75% medium risk, 0% high risk and 54.5% insufficient evidence.

Discussion

In this paper, we introduce and critically examine the challenges and opportunities associated with evidence mapping to construct a comprehensive perspective on the diverse types of evidence regarding sex-related differences in mental health outcomes. Among these challenges are measurement complexities, particularly the difficulty of integrating formative constructs that capture the determinants of mental health, including risk factors, protective factors, treatment pathways, and evidence gaps across multiple mental health outcome categories. Additionally, we highlight the statistical challenges inherent in synthesising complex and heterogeneous bodies of literature. Combined, this study provides a high-level examination of the existing empirical meta-analytic literature under an exploratory domain analysis methodology [38], offering a systematic and integrative assessment of the current state of knowledge.

This domain analysis, consisting of three umbrella reviews of 257 meta-analyses of differences between males and females in mental health, which included over 11,000 primary studies with a total sample size of more than 619 million people, found 43.9% of the estimates were greater in females, 28.5% were greater in males, and in 27.6% there were no differences between males and females. The following key differences and similarities were identified, starting with similarities:

Areas of similarity between males and females

An important finding was that differences between males and females in mental health outcomes are generally small, with little difference in the response of both females and males to a range of therapeutic preventions and interventions. Of the 435 estimates reported in 257 meta-analyses, 394 (90%) reported no difference or small differences, 30 estimates (7%) reported a moderate effect size, and only 11 estimates (3%) reported a large difference in effect size. Moreover, large differences have been reported because of the over-representation of one sex or a specific diagnosis in primary studies, and the differences were not as large after statistical adjustment by meta-analysis (e.g.,[29]). In particular, most meta-analytic studies of protective and preventive factors (96% of estimates) and treatment outcomes (96% of estimates) show minimal differences. We found that while there are differences in prevalence and associated risk factors, many of those differences are quite small and were smaller in the measured effect of prevention and protective factors and even smaller in treatment outcomes.

Areas of differences

Males are more likely to externalise, and females internalise

Prevalence of diagnoses

Consistent with the large body of research on the prevalence of mental health diagnosis (e.g., [1, 205]), females have higher rates of internalising disorders such as anxiety, depression and PTSD than males (83% of estimates show a higher prevalence). In contrast, males have higher rates of externalising disorders of substance use disorder, problem gambling and ADHD (77% of estimates indicate a higher prevalence of those disorders in males). Risk factors related to mental health disorder-specific differences between males and females are well-researched. For example, the higher rates of depression among females occur because of biological susceptibility, higher exposure to psychosocial stressors, adverse early life experiences, and greater stress arising from economic and social adversity [206, 207]. However, the influence of those risk factors has not been compared for externalising and internalising disorders as a whole, and it is not clear whether the known risk factors have shared associations with both internalising and externalising disorders and whether intervention can reduce the occurrence of both types of disorders.

One model examining shared associations is the Invariant Dimensional Liability Model of Gender Differences in Mental Disorder Prevalence [205], which assumes that mental health disorders can be classified into two latent dimensions: internalising and externalising disorders. This model hypothesises no sex invariances, that both sexes have equivalent psychopathology structures, and that the differences in the latent structures account for the differences in prevalence rates. The findings of this review provide some support for the model, with the personality traits of disinhibition and neuroticism correlating with externalising and internalising disorders, respectively. Moreover, the model suggests targets for prevention and treatment interventions, with disinhibition targeting males and neuroticism, negative rumination and cognitive distortions targeting females [205].

Loneliness is a significant risk factor for depression and death in males, whereas weight-related factors are significant risk factors for depression in females

Four meta-analyses showed that psychosocial risk factors related to loneliness and isolation, such as separation and being widowed, being single or having a disability pension, were greater risk factors for males in both depression [129, 130, 208] and mortality [208]. Five meta-analyses found that weight-related risk factors such as obesity, diet, screen time, sedentary behaviour and stopping exercise [9, 48, 131, 132, 209] were significant risk factors associated with depression in females. One meta-analysis reported no differences between males and females in the risk of anxiety and depression related to sexual abuse and physical abuse [124]. In contrast, one meta-analysis found physical violence to be associated with a greater risk of anxiety in males [123], and another found childhood sexual abuse associated with greater risk of depression in females [125]. Loneliness as a greater risk for depression and death for males supports the sexual selection evolutionary theory that females have more negative emotionality and effortful control than males, facilitating the development of empathy, interpersonal sensitivity and manipulation of interpersonal contexts, which act as protective factors against loneliness for females [210].

Females with mental health conditions had higher rates of heart disease but similar rates of diabetes and obesity to males

This study is the first to our knowledge to examine the range of similarities and differences between the sexes in the effect of comorbid physical disorders on the prevalence of mental health conditions. Overall, females with a comorbid physical condition are at greater risk of having a mental disorder (47% of the estimates) compared to males (17% of estimates). Females with mental health disorders are at a greater risk of heart disease (63%) compared to males (13%) [56]. For diabetes and weight-related risk factors, 60% and 75% of the meta-analyses’ estimates showed no differences, respectively, which suggests that males and females have similar metabolic comorbidity.

Death due to suicide is higher in males, but rates of suicide ideation and attempts are higher in females

Again, not surprisingly, in light of a large body of existing research (e.g., [3, 4, 151, 211]), males with mental health disorders are more likely to die by suicide (75% of the estimates), except for physicians [149], MS patients [148], and in China, where a single meta-analysis found that before 2000 deaths by suicide were higher in females. No significant difference was found in Chinese studies conducted after 2000, attributed to the improved status of females and the reduced availability of lethal chemicals in rural areas [147]. These results provide some support to the High Traditional Masculinity (HTM) model proposed by Coleman and colleagues, which includes the more masculine traits of competitiveness, emotional restriction and acceptability of violence and anger, [211, 212] predicting increased use of externalising behaviour, such as impulsive lethal self-injury.

Also consistent with previous research [5, 161], higher rates of disclosed suicidal ideation (100% of the estimates) and suicide attempts (86% of the estimates) were reported among females with mental disorders. The exception was among Chinese college students [164], where males made more suicide attempts.

Physical and sexual abuse are the most frequently reported risk factors associated with poor mental health outcomes (i.e., death by suicide, depression, anxiety, opioid use disorder and severe mental illness) for both males and females. Substance abuse and economic factors (i.e., unemployment, divorce, occupation, education, and level of societal deprivation) are the most frequently reported risk factors associated with poor mental health outcomes (i.e., death by suicide and severe mental illness, bipolar disorders, depression, and mortality) for males

The first umbrella review of meta-analyses confirms that childhood trauma and abuse have a significant adverse effect on the mental health of both males and females. It is the most frequently cited risk factor associated with a range of mental disorders. Females who have experienced sexual abuse are more likely to die by suicide, develop conduct disorder, develop opioid use disorder, and have a severe mental illness when compared to males who have been sexually abused, but there were mixed results. One meta-analysis found no difference in suicide attempts by those who had been sexually abused [167], but one meta-analysis found that females who had been sexually abused were at greater risk of completing suicide [150], and another meta-analysis found that males were more likely to make suicide attempts [27], which is the reverse of the overall findings.

Substance use is the most frequently cited risk factor for death by suicide. Males with substance abuse problems are more likely to die by suicide, other than the first half of the one study from China, where substance use was associated with an increased risk among females, along with lower educational attainment and income [34]. Males with substance use disorder are more likely to have depressive disorders and severe mental illness than females. These results provide further support to the High Traditional Masculinity (HTM) [211, 212] that predicts increased use of externalising behaviour, such as substance abuse.

For justice-involved youth, illegal drug use is associated with more recidivism in males, whereas family substance abuse is associated with increased recidivism among females. The results highlight the need to prevent exposure to early life physical and sexual abuse for both males and females.

Physical activity, coping, and social support are protective for both males and females. however, physical activity is more protective for males and social support for females

Physical activity, adaptive coping techniques, and making use of social support are protective against mental disorders for both males and females. This finding supports the “knowing-how” hypothesis that people who have gender-role flexibility have a broader coping repertoire [213]; for example, the protective technique of recruiting support, which is traditionally viewed as a ‘feminine’ protective factor, is more protective for females, as 65% of the estimates found greater protective factor in the outcomes of females who seek support. Resorting to physical and other activity, traditionally viewed as a ‘masculine’ coping style, was more beneficial to males (62% of estimates). Females used coping strategies such as specific active coping and structured problem-solving to a slightly greater degree than males, but males were slightly less likely to ruminate and self-blame. Overall, having a flexible array of masculine and feminine coping and preventative strategies that adapt to different situations benefits both males and females.

Males and Females obtain similar benefits from treatment. however, the uptake of treatment for anxiety, depression, and PTSD is higher for females, and the uptake of treatment for substance abuse is higher for males

An important finding confirmed by most of the meta-analyses is that males who engage in mental health treatment do just as well as females (61% of estimates). Females do better in PTSD treatment, relationship counselling, and anger management. Males do slightly better in treatments that target weight loss and tobacco use. However, when baseline symptom levels are controlled for, as in alcohol treatment, these small differences are further reduced [29]. In the studies of treatment for smoking and PTSD, the baseline levels of PTSD symptoms and smoking were not controlled, which might explain the difference. Also, another confounding factor could be the lack of control for the proportion of males and females in treatment, which is higher for females in PTSD and males in smoking cessation. Future research on increasing the uptake of substance abuse treatment for females and increasing the uptake of treatment for anxiety, depression and PTSD in males will need to control for the differences in proportions of both the sex ratios in study participants and the prevalence of the respective disorders. Another possible future direction could be the transdiagnostic treatment of internalising and externalising disorders, focusing on different cognitive processes [205]. For example, female participants' focus would be on the treatment of internalising disorders focusing on rumination and negative mood. In contrast, males' focus would be on treating externalising disorders, focusing on self-control.

There is extensive research on the prevalence and risk factors for mental disorders, but less is known about the impact of prevention and intervention programs that target risk factors, in particular, interventions to address loneliness in males and the effect of measures to prevent sexual abuse in both males and females are not well researched

The vast majority of both the meta-analytic studies (205 of 257 samples, or 80%) and the primary research on differences between males and females research has been on prevalence and risk, with fewer meta-analyses examining research into prevention and protective factors (11%), and even fewer on differences between males and females in treatment outcomes (9%). In particular, a significant gap was identified in research on differences between males and females in the mechanisms of preventing and treating mental disorders.

As an example, there has been no meta-analysis of the mental health outcomes of non-specific support and same-sex cooperation in a group format to reduce loneliness in males. There are also no meta-analyses that have reported on attempts to address disinhibition in males on the likelihood of resorting to substance use in the presence of mental disorders or on measures to prevent physical or sexual abuse despite the adverse effect on mental health outcomes.

With regards to treatment, there were 12 estimates reporting differences between males and females in outcomes of treatments for substance abuse, four examining outcomes in treatment of depression, two each for treatments of PTSD and weight loss in people with mental disorders, and one meta-analysis each examining treatment of dyadic relationships, sexual dysfunction, anger, anxiety, parenting and mental disorders, fatigue and cancer, cardiac outcomes in the presence of mental disorder, and the effect of differences between males and females in self-control.

There were also apparent gaps in examining the effect of underlying psychological processes, such as externalisation or rumination, on participation and outcome in treatment. Using the example of alcohol treatment, alcohol consumption is measured, but not the effect of the intervention on psychological mechanisms of disinhibition or coping style.

Sensitivity analysis for variance estimation methods

The sensitivity analysis collated in Supplemental Table 2 demonstrated that both the reliance on fixed-effects and random-effects variance estimates were notably narrower—and thus more liberal—than those obtained from studies that directly tested the gender effect. This finding is not surprising, as the direct extraction methods benefit from the richer, more detailed information available within individual studies. In contrast, the streamlined approaches rely solely on sampling variance, which tends to underestimate the true variability in sex-related differences. Together, these results suggest that direct extraction yields a more homogeneous set of estimates and provides a more accurate reflection of sex-related effects than methods based solely on sampling variance.

Research gaps

The review found comparatively less research measuring the effect of preventative measures, including measures to prevent the emergence of disorder and reduce disability and to examine sex-specific differences in substance use, gambling, and suicidal behaviour. The review also identified a need for a meta-analytic study of differences between males and females in the outcome of specific treatments for anxiety disorders, severe mental illness (e.g., schizophrenia and bipolar disorder), ADHD and binge eating disorder. [27, 167].

The study also revealed some contradictory findings in meta-analyses that presumably had access to the same primary research. The quality of meta-analyses would be expected to improve over time, given the greater ease of conducting searches, the greater number of original studies to include, and the establishment of protocols for performing meta-analytic studies, which may address these contradictory findings over time. For example, the last decade of research on psychological treatment of alcohol misuse has clarified contradictory findings and shown no differences between males and females. An important implication of this finding is the need for transparency of methods, disclosure of primary and synthesised data, and the need to replicate even well-conducted reviews.

Moreover, despite efforts to aggregate a wide range of results, a gap in the current study is the limited ability to assess the interconnections between the various topics in the included studies. Despite its size and span, the synthesised findings leave unresolved questions regarding the available data's overlap, interactions, and inevitable methodological limitations. These unresolved issues raise concerns about the validity of the findings. This gap can be addressed in future studies, incorporating independent relational data that analyses some of the issues raised in this paper, thereby providing more detailed relational insights. This is particularly relevant given the differences in the statistical methods used in the sensitivity analyses, which underscore the need to directly test raw data rather than relying solely on differing statistical assumptions.

Future research could use increasingly available published raw data to confirm the findings of meta-analyses and examine the effect of a broader range of variables.

Implications

Clinical implications

The main implication of these reviews is the demonstrated need for more prevention and treatment programs based on these findings, males and females who engage in the traditional masculine (e.g., physical activity) and feminine (e.g., recruiting social support) protective behaviours and treatment do better. Access is an important factor, not only to treatments but to information, resources, and population-level interventions. This review revealed surprising gaps in our knowledge of treatment outcomes. While more research needs to be done in prevention and treatment, this review confirms the need for the greater use of available technology to help both males and females access evidence-based treatment and resources.

Another implication is the need to explore differences between males and females in the mechanisms of change inside and outside therapy. For example, no meta-studies explore the differences between males and females in what people learn, how they improve and the changes they make while engaged in psychological treatments. There may be differences between males and females in mechanisms of change that are not captured by the current methodology.

Health policy implications

The main policy implication of this domain analysis is the need for routine measurement and reporting of the effect of measures employed to prevent and treat mental disorders, as well as the need to report separate outcomes for males and females. A large part of the literature on differences between males and females appears to reflect either opinion or preliminary findings based on gender stereotypes or results of smaller studies that are not replicated in meta-analysis rather than integrated evidence from properly conducted research. The results do confirm that any measures that can prevent physical and sexual abuse will improve mental health at a population level. Similarly, there is abundant evidence that community-wide measures to reduce substance use and promote physical health and social support will reduce suicide and improve the mental health of both males and females.

Strengths and limitations

A strength of this study is the sheer breadth and scale of the search for topics that have been the subjects of meta-analyses. This synthesis is possibly the most extensive synthesis of therapeutic research to date. Also, the paper summarised several key themes in differences between males and females, but the extent of the similarities and differences are more readily understood from the table and graphic representation of the similarities and differences in the mental health of males and females. However, there were some significant limitations. The focus on sex rather than gender as a binary two-dimension variable excludes the diverse ways of expressing gender. Although there is emerging research on gender diversity there was limited data in the meta-analyses located for this study. The protocol could not be registered to make the study protocol publicly available for review before publication. The searches focussed on mental health and may have overlooked differences between males and females in other well-being pathways, such as differences reported in sociological research. Only a minority of published meta-analyses included in the study confined themselves to controlled trials and, despite overlapping samples, reported contradictory results, suggesting some methodological shortcomings in the available literature. Because of the nature of the study, there was limited ability to gauge the magnitude of effects given the heterogeneity of the studies and the potential overlap of primary studies in the more homogenous samples, which is why we focused on counting and mapping the trends rather than assessing the magnitude of these differences in overlapping studies. Domain analysis was also unable to account for the unequal weighted effects due to the overlapping individual studies within meta-analyses, as the domain analysis approach was focused on the meta-trends of differences between males and females in mental health outcomes rather than the effects of individual studies. Whilst the authors provide some patterns and selective emphasis, the readers are invited to delve further into the tables, graphs and supplemental material and gauge the range of effects, the comparative magnitude of influences and the suggested gaps that emerge from this analysis of a whole domain. Finally, umbrella reviews combine the results of studies of quite different samples and cannot examine between factor effects without having access to individual patient data meta-analyses. Making de-identified individual subject data available to other researchers may improve the reliability of published research and allow further analysis beyond the original research questions.

Conclusion

The striking findings of this study have been that in a large number of research studies across three domains, there are few significant differences between females and males, and despite the differences in prevalence and risk factors between males and females, identified protective factors and existing treatments are similarly beneficial for both males and females. This study confirms that psychological treatments are effective for both females and males, and the main challenge is to improve access and uptake of care. The available information does not support the need for separate mental health services for men and women. However, the domain analysis identified significant gaps in the available research, particularly a limited number of randomised control trials specifically examining gender differences and gender-tailored therapies.

Authors’ contributions

RK conceived and designed the study, conducted the literature review, interpreted the data, and drafted the manuscript. EK conducted analyses and drafted the analysis sections of the manuscript. NT, BD, EK, LS, ON reviewed the design, analysis, interpretation of data and the manuscript. All authors have contributed to, read and approved the manuscript.

Funding

No Funding.

Data availability

Sequence data that support the findings of this study have been deposited in the supplementary information files.

Declarations

Ethics approval and consent to participate

Not Applicable.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.