Improving clinical outcomes through attention to sex and hormones in research

Abstract

Biological sex, fluctuations in sex steroid hormones throughout life, and gender as a social construct influence all aspects of health and disease. Yet, for decades, most basic and clinical studies included only male individuals. As modern health care moves toward personalized medicine, it is clear that considering sex and hormonal status in basic and clinical studies will bring precision to the development of novel therapeutics and treatment paradigms. Toward this end, funding, regulatory and policy agencies now require inclusion of female animals and women in basic and clinical studies. However, inclusion of female animals and women often does not mean that information regarding potential hormonal interactions with pharmacological treatments or clinical outcomes is available. All sex steroid hormones can interact with receptors for drug targets, metabolism and transport. Genetic variation in receptors or in enzymatic function might contribute to sex differences in therapeutic efficacy and adverse drug reactions. Outcomes from clinical trials are often not reported by sex, and, if the data are available, they are not translated into clinical practice guidelines. This Review will provide a historical perspective for the current state of research related to hormone trials and provide concrete strategies that, if implemented, will improve the health of both men and women.

TOC

Sex differences can have important implications in research and medical care. This Review will outline our understanding on sex differences, and will offer strategies that, if implemented, will provide new information to personalize and improve health care for men and women.

Introduction

Sex as a biological variable, variation in sex steroidal hormones throughout life, and gender as a social construct influence all aspects of health and disease. Yet, for decades, most basic and clinical studies included only male animals or participants. This practice, historically, was influenced by unconscious bias reflecting that science and medicine were dominated by men. For example, women were classified as ‘protected subjects’ due to concern for the potential teratogenic effects of pharmacological interventions. This classification led to the exclusion of women in late-phase clinical trials. Women and female animals were also excluded from numerous studies because of concern that hormonal cyclicity would increase variability in the data and complicate the resulting outcomes and their interpretation. These converging concepts resulted in the establishment of physiological and pharmacological norms based solely on data from male individuals, which was then applied to female individuals. Deviations from this ‘norm’ were not considered as representing underlying physiological differences between male and female individuals, but rather as aberrations from the norm.

Fortunately, these biases are being addressed: there are more women in scientific and medical professions (Women in Science and Engineering Statistics; https://www.nationalacademies.org/cwsem/women-in-science-and-engineering-statistics); it is now mandated that women be included in clinical studies, with pregnancy testing as part of the inclusion criteria; there are safety monitoring boards to evaluate risk:benefit criteria for both women and men¹; factors that contribute to variability in data derived from both male and female experimental animals are being identified^2,3; and the study of sex differences has emerged as a scientific discipline (see the Organization for the Study of Sex Differences (https://www.ossdweb.org/); International Society of Gender Medicine (http://www.isogem.eu/). However, although funding and regulatory agencies now have policies that require inclusion of female animals and women in basic and clinical studies, information on hormonal status is often not collected and potential sex steroid hormonal interactions with drug metabolism are not considered. Data are often still not reported by sex, and in many papers, the percentage of male and female participants is not provided. Moreover, if data on sex differences are available, they are not commonly translated into clinical practice guidelines. This Review will provide a historical perspective of the current state of research into sex differences, will discuss basic physiological and pharmacological concepts into sex and hormonal interactions, will outline findings from major hormone treatment trials and will offer strategies that, if implemented, will provide new information to personalize and improve health care for men and women. Note that, unless otherwise specified, the terms women and men refer to ciswomen and cismen.

Historical policies and global initiatives

The landmark publication by the Institute of Medicine, now the National Academy of Medicine, “Does Sex matter?” in 2001 triggered a series of initiatives to incorporate sex as a biological variable into basic and clinical study design, data interpretation and data publication⁴. The overarching conclusion of that publication was that biological sex influences health and disease from “womb to tomb”. Since that publication, there has been steady, albeit slow, progress toward accepting and implementing this concept into research practices. In 2002, the World Health Organization issued a WHO Gender Policy that was followed in 2003 by the European Commission Directorate-General for Research and Innovation, stating the need for “questioning systematically whether, and in what sense, sex and gender are relevant in the objectives and in methodology of projects”⁵. Scientific advocacy groups and professional societies continued to promote and advocate for funding for sex differences research and to encourage professional journals to require the reporting of the sex of cells, of animals used in basic science studies and of participants in clinical trials^6-8. In 2014, the European Commission on Gender Equality and International Development Research and Innovation⁹, the Canadian Institute of Health Research and the United States National Institutes of Health reaffirmed the critical importance of considering sex and gender in basic research design, implementation, data interpretation and reporting. By considering sex and gender, scientific quality would be improved by increasing transparency and reproducibility, and, thus, increasing technological innovation that can be translated to improved societal relevance of research^10-13. Specifically, studying sex differences and reporting data disaggregated by sex helps to identify differences in clinical presentation, drug toxicity and outcomes. Furthermore, analyzing and reporting data disaggregated by sex facilitates meta-analyses, informs sample size calculations used in clinical trial design and informs the design of other studies and cohorts that can validate results^12,13.

Fundamentals: chromosomes and hormones

Consideration of sex chromosomes.

The fundamental origin of all sex differences in health and disease results from the presence of the sex chromosomes (XX for female and XY for male individuals, for most mammals). Genes on these chromosomes direct the development of the reproductive organs, and ultimately gonadal production of the sex steroids (FIG. 1). The X and Y chromosomes have evolved from a pair of autosomes over the past 300 million years¹⁴. Although 98% of the ancestral genes survived on the human X chromosome, only 3% survived on the Y chromosome^15-17. It has been speculated that the reason for the survival of the 3% was for sex determination and spermatogenesis. However, studies have now demonstrated that some of the ancestral genes on the Y chromosome might influence the expression of genes on the autosomes.^18,19 In addition, the ancestral genes on the Y chromosome seem to be critical for immunity and inflammation, and might reflect the necessary double-dosage of genes needed for survival (i.e. same gene on both the Y and X chromosome for XY males)²⁰. The expression of genes on the single X chromosome in males (hemizygous exposure of the X allele) is a source of male-female differences in the incidence of some phenotypic traits in populations, such as colour-blindness, muscular dystrophies and embryonic lethality. The presence of only one X chromosome means that X-linked mutations will result in a fully dominant phenotype in male individuals but range from mild to no phenotypic changes in female individuals.

Figure 1. Sex Steroids influence cellular processes through both genomic and non-genomic regulation.

Sex differences result from differences of the sex chromosomes, which direct development of the reproductive organs independently of the presence of the sex steroids that are produced from the gonads. Sex steroids might directly alter gene expression through ligand-bound receptor binding to hormone response elements on specific genes that influence the production of membrane receptors, enzymes and structural proteins (that is, genomic regulation). Alternatively, sex steroids might bind to surface membrane receptors, which regulate ion channels or enzyme activity directly, such as that of nitric oxide synthase within caveolae (that is, non-genomic regulations). Changes in intracellular enzyme activity might affect the expression of genes that do not contain hormone response elements, such as those genes with protein products that regulate the cell cycle and apoptosis). These pathways represent an indirect mechanism by which sex steroids influence gene regulation. Hormonal effects that are non-reversible on removal of the steroidal hormones include the development of the reproductive organs. Examples of effects that are reversible on removal and re-administration of the hormones include physiological changes such as those occurring during pregnancy or loss of bone mineral density at menopause and subsequent restoration with use of menopausal hormone treatments. AKT, RACα serine/threonine-protein kinase; AR, androgen receptor; DHEA, dehydroepiandrosterone; ER, oestrogen receptor; eNOS, endothelial nitric oxide synthase; GPR30, G protein-coupled oestrogen receptor; IGF1 insulin-like growth factor 1; mRNA, messenger ribonucleic acid; NO, nitric oxide; PI3K, phosphoinositide 3-kinase. Adapted from ref.50, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).

In order to achieve gene dosage equivalency between female and male individuals, and to ensure that the expression of X genes is not twice that of autosomal genes, compensatory mechanisms, including X-inactivation and upregulation of X-expression, have been adapted. X chromosome inactivation in female individuals, via methylation²¹, is random. It results in an uneven or mosaic pattern of expression of genes from the maternal or paternal X chromosome that might cluster and show variation from tissue to tissue²². It is unclear how variability of cellular phenotypes resulting from the mosaicism affect health or risk of disease in XX female individuals. Further, because the maternal and paternal X alleles vary in genetic sequence, X inactivation can be a notable source of variability between XX and XY cells. XX cells express maternal alleles about 50% of the time, whereas XY cells express maternal alleles 100% of the time²³.

Among XX female individuals, it is estimated that up to 23% of the X-chromosomal genes routinely escape X-inactivation and another 10% may or may not escape inactivation for a given reasons not yet understood^24-26. As a result, some genes on the X chromosome are more highly expressed in XX cells than in XY cells and cause either increased or decreased risk of a phenotype²⁴. Factors that disrupt X inactivation in XX cells lead to an overexpression of X genes that are then out of balance with autosomal genes in gene networks²⁴. Alternatively, genes in the pseudoautosomal region (PAR) of X and Y chromosomes are often overexpressed in XY cells compared with XX cells because some genes in XX cells are not expressed due to X inactivation²⁷. Research into the factors that contribute to X inactivation and parental imprinting across the lifespan on health and disease is ongoing^28-30. As an example, findings suggest that X inactivation is important in female cancers. Ovarian and breast cancer cell lines show loss of X inactivation and an interaction between X inactivation and BRCA1 has been reported^31-33. X inactivation also has a role in ovarian tumour biology and affects clinical prognosis³⁴.

The expression of genes on the sex chromosomes might be influenced by sex hormones, thus representing an interaction between the two sex-biasing mechanisms³⁵. These interactions need additional study. An example of a sex chromosome effect that influences a sex steroids response is the gene for the androgen receptor, which is located on the X chromosome and escapes X inactivation. Because male individuals carry one X chromosome, variation in this gene will have a greater effect (that is, androgen insensitivity) than in female individuals. Because female individuals might have heterozygous expression of this variant gene, expression of the phenotypic characteristic, that is, alopecia, will show variability across the population. The genes for the oestrogen receptors (the binding affinity for the ligands of these receptors varies, with 17β oestradiol having greater affinity than oestrone) are located on autosomes (ESR1 and ESR2 on chromosomes six and 14, respectively). The differences in sex chromosomes, X inactivation, genetic variants in sex steroids receptors and changing patterns of sex steroids across the life-span provide the physiological background for all sex differences in regulatory mechanisms in health and disease.

Historically, genome-wide association studies (GWAS) focused on autosomes and did not examine sex chromosomes³⁶. These studies also did not examine sex differences. One reason was due to inadequate sample sizes for well-powered analysis of each sex. Another reason was the lack of statistical approaches needed to analyze the haploid Y chromosome and to account for dosage compensation and inactivation of the X chromosome³⁷. In the past few years, statistical methods have been developed to analyze the X chromosome, but there is a dearth of methods for the analysis in the Y chromosome in GWAS^38,39. This deficiency is primarily because the Y chromosome does not recombine outside the PAR, which leads to considerable co-variation of Y chromosome single nucleotide polymorphisms across the entire chromosome²⁷. As a result, other scientific methods might be needed to better understand the genes on the Y chromosome. As an example, mechanistic animal studies that can prove what the specific genes do on the Y chromosome, with subsequent extrapolation to humans to correlate variation in Y genes with similar phenotypes to those studied in humans, will provide much needed information. Indeed, the identification of the different genetic processes and accounting for sex chromosomes that might affect the risk of health and disease in men and women is imperative for individualized preventive and treatment plans.

Sex hormones and drug metabolism.

Sex steroid hormones regulate cellular processes through the modulation of gene transcription (that is, genomic effects (FIG. 1)). These genomic effects require co-regulators for specific binding of the ligand-bound hormone receptor to hormone-receptor response elements on genes. These effects become apparent over several hours or days as changes in expression of proteins, alterations in receptors and cell structural elements, and the development of the reproductive organs. Organizational effects of hormones are those that are not reversed once the hormones are removed. The phenotypes that are reversible with removal of the hormones are termed activational effects.

In addition to directly modulating gene expression through hormone response elements on genes, sex steroids also regulate cellular processes through binding to surface steroid membrane receptors (non-genomic effects) with the activation or inhibition of intracellular pathways or surface ion channels (for example, K⁺ activated calcium channels) and actions within the mitochondria^40-43. These effects are rapid and the duration of the effect will depend on the kinetics of the ligand-receptor binding and the specific pathways that are activated or inhibited. Examples of non-genomic effects include the abrupt production and release of nitric oxide in response to acute application of 17β-oestradiol to vascular endothelial cells, regulation of intracellular levels of calcium, and modulation of neuronal activity and insulin secretion^44-47. The non-genomic effects of sex steroids on intracellular signalling cascades might affect the regulations of genes that do not contain steroid response elements, thus representing an indirect mechanism by which steroid hormones might exert a genomic effect⁴⁷. The dramatic hormone-mediated physiological changes associated with pregnancy and the subsequent delivery of the foetus, as well as the phenotypes resulting from removal of the gonads, or phenotypic changes associated with hormone treatments with gender-confirming procedures, exemplifies these activational effects of the sex steroids⁴⁷.

Sex steroids can alter the efficacy of therapeutic interventions via multiple mechanisms, and can contribute to some adverse effects of drugs^48-53. As a result of their genomic effects, sex steroids can influence the expression of receptors or components of intracellular pathways targeted by drugs. For example, 17β-oestradiol modulates the expression of the production of receptors for adrenergic and serotonergic neurotransmitters, which could influence the efficacy of certain classes of beta-blockers and selective serotonin receptor uptake inhibitors in women as they age^54,55. Unfortunately, few studies have examined how 17β-oestradiol modulation across the lifespan could affect the prescribing of an optimal dose, overall efficacy, or adverse effects. Ideally, clinical trials of serotonin receptor uptake inhibitors and other medications should include children and adolescents and premenopausal and postmenopausal women. For men, androgen deprivation therapy or use of testosterone could also affect efficacy of serotonin receptor uptake inhibitors⁵⁶.

In addition, sex steroids could compete with the pharmaceutical agent for receptor or enzyme binding sites or transporters for the agent into the liver or target tissue. Statins represent a class of drugs that are prescribed extensively to reduce serum levels of cholesterol and low density lipoprotein (LDL) cholesterol; high levels of these factors are considered risk factors for the development of atherosclerosis and adverse cardiovascular events. However, in a large meta-analysis of primary prevention studies of statin therapy to reduce the incidence of coronary heart disease, the therapy reduced the risk of adverse cardiovascular events in men but not women^57,58. Current prescribing guidelines for statin therapy are based on serum levels of LDL cholesterol and the presence of atherosclerosis and diabetes mellitus. However, the guidelines are not sex-specific^59,60. All statins are transported into the liver, in part, by organic anion transporting polypeptides (OATP) transporters. There are several genetic variations in the OATP transporters that affect bioavailability and distribution of statins. These variants have not been assessed separately in men and women. As a result, it remains unclear whether genetic variants are associated with variability in statin response, including adverse effects^50,61. Oestrogens compete for the OATP transporter and, thus, could modulate the effectiveness of the drug. In addition, 17β-oestradiol and oestrone are metabolized by the same cytochrome P450 enzymes and undergo sulfate and glucuronide conjugation, as do statins⁵⁰.

The consequence of these interactions could affect the outcome of therapy, as was exemplified 20 years ago in the Estrogen for Prevention of Atherosclerosis study (EPAT)⁶². In that study, increases in carotid intima–medial thickness (CIMT) were measured in postmenopausal women randomized to receive oral micronized 17β-oestradiol (1mg per day) or placebo. Women received lipid-lowering medication if their LDL cholesterol levels exceeded 4.15 mmol/l (160 mg/dl). The rate of increase in CIMT was slower in women treated with 17β-oestradiol than in those treated with placebo. However, the positive response of 17β-oestradiol was attenuated among women simultaneously receiving a lipid-lowering medication⁶². Similar interactions could be anticipated for other systems regulated by 17β-oestradiol, such as beta adrenergic drugs and angiotensin converting enzyme inhibitors⁵⁰. Currently, the interactions between drugs and sex hormones are not systematically considered in reporting outcomes of pharmaceutical trials. The absence of consistent reporting of sex differences in outcomes often limits understanding of efficacy and potential adverse reactions to the medication.

Chromosomal versus hormonal effects.

Gonadectomy with subsequent hormonal treatment is used in experimental settings to study the activational effects of the sex hormones, and to begin to differentiate the effects of sex steroids from those of sex chromosomes. To further distinguish the contribution of the sex chromosomes from that of sex steroids in physiological functions in laboratory animals, a ‘four core genotype’ has been developed in which the testis determining gene (Sry) on the Y chromosome is deleted, resulting in XY animals developing ovaries; conversely, placing Sry on an autosome in an XX animal results in the development of testes²⁷. Using these intact and gonadectomized animals allows the contribution of sex chromosomes and sex hormones to be investigated separately when looking at sex differences in specific functions.

Hormonal variations: abrupt endocrine dysfunction.

17β-oestradiol is the major oestrogen secreted by the premenopausal ovary. It is a metabolic product of testosterone produced by Leydig cells in the testes, granulosa cells in the ovary and in small amounts from the adrenal glands. Thus, both testosterone and 17β-oestradiol are present in male and female individuals, albeit in different proportions, and the production level varies across the life span. Levels of androgens increase at puberty in male individuals and typically decline somewhat slowly with age. For female individuals, levels of oestrogens increase at puberty and with pregnancy. Levels of ovarian hormones fluctuate for several years prior to the cessation of menses, which is followed by a slow decline in hormone levels with age. Initially, studies of menopause in female animals incorporated bilateral ovariectomy as an experimental model of menopause. However, the sudden abrupt endocrine disruption resulting from such a procedure did not replicate the slower decline in ovarian function and ovarian hormones (including testosterone) that occurs with age in women. Alternatively, chemically induced reductions in ovarian function might be a better model of these changes in experimental animals^63,64. However, these experimental studies of bilateral ovariectomy for female animals, and orchiectomy for male animals, have contributed to an improved understanding of the short-term and long-term health effects of these procedures on both sexes.

Hysterectomy is the second most common major surgical procedure performed among women living in the USA, after caesarean section⁶⁵. About 64% of all hysterectomies are performed on women aged 35–54 years⁶⁶. Among women who undergo hysterectomy, 23% aged 40–44 years and 45% aged 45–49 years undergo bilateral oophorectomy at the same time as a prophylaxis against ovarian cancer⁶⁷. As a result of this practice, it is estimated that approximately one in eight women have their ovaries removed before reaching natural menopause⁶⁸. The incidence of hysterectomy and bilateral oophorectomy vary by race and ethnicity, and the long-term effects of either surgery are not well understood^69,70.

Although the incidence of premenopausal bilateral oophorectomy has started to decline in some regions⁷¹, there is increasing concern that bilateral oophorectomy might have harmful long-term effects that can negate the benefit conferred by protection from ovarian cancer^72,73, particularly among the majority of women who were at low risk for ovarian cancer. Indeed, prophylactic premenopausal bilateral oophorectomy has been associated with increased all-cause mortality^74,75, dementia^76,77, cardiovascular disease^78,79, skin ageing⁸⁰, sexual dysfunction⁸¹ and overall accelerated ageing^82,83. Thus, it is critical to record whether a woman, as a patient or clinical trial participant, has had a bilateral oophorectomy prior to natural menopause because it can identify women at increased risk of many health outcomes and trial outcomes. Some woman might have received, or currently receive, oestrogen therapy, which might negate some of the negative effects of premenopausal bilateral oophorectomy, especially for women who underwent the procedure prior to the age of 45 years. In contrast to premenopausal bilateral oophorectomy, the short-term and long-term effects of bilateral oophorectomy after natural menopause for non-cancer indications are even less well known. Because these women have gone through menopause and have low oestrogen and progesterone levels, the effects of the surgery have historically been not as concerning. However, even after menopause, the ovaries produce testosterone, androstenedione and dehydroepiandrosterone, the depletion of which might also affect health outcomes⁸⁴.

For men, abrupt endocrine dysfunction via androgen deprivation therapy can occur medically through use of gonadotropin-releasing hormone agonists (GnRHa) for the treatment of prostate or testicular cancer, or via surgical bilateral orchiectomy. Testicular cancer is the most commonly diagnosed cancer among men aged 20 to 40 years, and prostate cancer is the second most common cancer in all men⁸⁵. Both long-term GnRHa use and orchiectomy have been associated with an increased risk of bone fracture^86-88, diabetes mellitus and cardiovascular diseases^89,90. Over the past decade the use of surgical bilateral orchiectomy has declined with the increased use of GnRHa⁹¹; now, 7 out of 10 men with prostate or testicular cancer would opt for GnRHa given the choice⁹². However, evidence suggests that long-term administration of GnRHa, defined as exposure of 35 months or more, is associated with higher risk of fractures, peripheral arterial disease, venous thromboembolism, cardiac-related complications and diabetes mellitus compared with bilateral orchiectomy⁹³. It is not currently known whether there are interactions between other medications and GnRHa or whether the efficacy and adverse effects of specific medications differ for men with versus without bilateral orchiectomy.

Additional considerations.

In designing and interpreting studies to investigate effects of exogenous hormone treatment in experimental animals or humans, it is essential to consider three important points. First, androgens and oestrogens are present in both female and male individuals but the concentrations of the hormones vary with age. Second, receptors for the sex steroids are present in all tissues of the body. Third, the expression of the receptors might vary from tissue to tissue and might differ depending on exposure of the receptors to the steroids. Thus, in addition to the type of treatment, the timing of initiation of the treatment, the dose of the intervention, the responsiveness of the target tissue to the treatment and the duration of the treatment will affect the outcomes.

Hormonal therapy

Hormonal contraceptives.

The composition of oral contraceptives has dramatically changed over the years, particularly in regard to the synthetic progestin and the use of lower doses of both oestrogens and progestins (see TABLE 1)^94,95. In addition, new hormone delivery systems have been developed, including implants, intrauterine systems and injectables. As a result of these changes, it has been difficult to comprehensively evaluate the adverse effects of oral contraceptives and other hormone delivery systems, or their potential interaction with other pharmacological treatments.

Table 1.

List of formations of oral contraceptives and risk of thrombosis by generation^a

Hormonal treatment	Risk of thrombosis
First-generation <50 μg ethinyl oestradiol with the progestins noretynodrel, norethisterone and norethisterone acetate	6–12/10,000 women
Second-generation <50 μg ethinyl oestradiol with the progestins norgestrel or levonorgestrel	5–7/10,000 women
Third generation <50 μg ethinyl oestradiol with the progestins desogestrel or gestodene norgestimate	9–12/10,000 women
Fourth-generation <50 μg ethinyl oestradiol with the progestins drospirenone, dienogest or nomegestrol acetate	9–12/10,000 women
Progestin-only (norethisterone, ethynodiol diacetate, levonorgestrel, desogestrel and lynestrenol)	2–3/10,000 women

^aDerived from Table 1 of Gialeraki, et al.⁹⁴

Oral contraceptives modify the endogenous production of sex steroids and, as a consequence, disrupt the hypothalamic–pituitary–ovary regulatory feedback pathways. As with all oral medications, these hormonal products will enter the enterohepatic circulation for first-pass metabolism in the liver. In addition, oral contraceptives might also interfere with the transport of other medications, modulate the production of triglycerides and lipoproteins, and increase inflammatory cytokines and proteins associated with coagulation^50,96,97. These latter effects might increase the risk of venous and pulmonary thrombosis, and this risk has changed over the generations of formulations used for oral contraceptives (TABLE 1)⁹⁴. The extent of first-pass metabolism depends on the drug, especially for progestins. For example, first-pass metabolism is high for norethisterone, desogestrel and norgestimate, but there is none for levonorgestrel⁹⁸. In addition, women might use oral contraceptives for varying periods of time and often switch formulations. These factors, together with pregnancy history and use of menopausal hormone therapy (MHT), confound the interpretation of the longitudinal effects of oral contraceptive use. At a minimum, a record of oral contraceptive products, dose and duration should be noted for women participating in clinical studies and, when possible, be considered as a confounding variable⁹⁹.

Menopausal hormone therapy.

In contrast to women who experience the abrupt hormonal drop following bilateral oophorectomy, women who undergo natural menopause experience fluctuations in ovarian hormones for years before the cessation of menses¹⁰⁰. These fluctuations are followed by declines in levels of oestrogens and increases in levels of follicle stimulating hormones, which for many women result in vasomotor symptoms, changes in mood, sleep disturbances and sexual dysfunction. Because the incidence of several ageing-related chronic conditions (including cardiovascular disease, osteoporosis and dementia) increases after menopause, it was hypothesized that oestrogens were protective and that the loss of oestrogens placed women at increased risk of these diseases and for accelerated ageing in comparison to premenopausal women. Observational and epidemiological studies of MHT conducted in the late 1980s and 1990s identified physiological benefits of these treatments for reducing the incidence and severity of osteoporosis, altering the ratio of high to low density lipoprotein cholesterol, and reducing the risk of cardiovascular disease^101-105. However, women enrolled in these MHT studies were not a homogenous group, with some women having undergone natural menopause and some with unilateral or bilateral oophorectomy at various ages before or after the natural age of menopause^{62,101-103,106,107}.

The types of MHT used, the length of use and the dose used by women participating in those observational studies were also dissimilar. Types of MHT included oral 17β-oestradiol and oral conjugated equine oestrogens (CEE), with and without cyclic or continuous natural progesterone or medroxyprogesterone acetate, a synthetic progestogen^{103,104,108,109}. Some women also used oral contraceptives as an MHT. As with oral contraceptives, oral MHT products undergo enterohepatic first-pass metabolism in the liver and might compete with receptor and enzymatic binding sites for other drugs. This first-pass maximizes effects on circulating lipid levels but might place women at increased risk of venous thrombosis due to increases in inflammatory cytokines and changes in proteins associated with coagulation¹¹⁰. Furthermore, genetic variants in enzymes associated with steroid metabolism affect the bioavailability of 17β-oestradiol or various biologically active metabolic products in the systemic circulation^111-113. As shown in TABLE 2, differences in the bioavailability of these metabolites are particularly evident when comparing oral and transdermal 17β-oestradiol and oral CEE because CEE contains various active metabolites of oestrogen, including oestrone, oestrone sulfate and oestradiol–sulfate. Thus, given the heterogeneity in observational studies and the many biases inherent in observational study designs (that is, selection, healthy-user, compliance, survivor and prevalence-incidence bias), it was not possible to provide guidelines for determining the right dose and formulation for treating menopausal symptoms or for the prevention of cardiovascular diseases and other conditions.

Table 2.

Hormone levels in participants of the Kronos Early Estrogen Prevention Study (KEEPS)

Treatment	Androstenedione (ng/dl)		Testosterone (ng/dl)		Oestrone (pg/ml)		Oestradiol (pg/ml)
	Mean (SD)	Med (IQR)	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)
Placebo (n = 146)	55.5 (23)	52.0 (26)	18.1 (18)	15.0 (10)	28.5 (11)	28.0 (15.0)	7.2 (4)	6.6 (4.0)
Transdermal oestradiol (n = 107)	64.0 (33)	56.0 (36)	17.0 (9)	15.0 (10.0)	52.4 (26)	50.0 (35.0)	47.1 (40)	41 (53.7)
Oral conjugated equine oestrogen (n = 109)	57.5 (23)	53.0 (27)	20.4 (10)	18.0 (12.0)	106.2 (86)	88.0 (90.0)	15.5 (10)	13.0 (12.3)

Hormone levels reported 48 months after being randomly assigned to placebo, oral conjugated equine oestrogens or transdermal 17β-oestradiol. IQR, interquartile range; SD, standard deviation.

To overcome some of these limitations, a prospective, double-blind, placebo-controlled trial, the Women’s Health Initiative (WHI), was initiated in 1991 to provide evidence that use of MHT would provide primary prevention for cardiovascular disease among women aged 50–79 years (mean age of 63.3 years). At that time, the WHI was the largest randomized clinical trial funded by the National Institutes of Health in the USA. The treatment chosen was based on MHT formulations that were most prescribed at the time: oral CEE (0.625 mg per day) with medroxyprogesterone acetate (MPA; 2.5 mg per day) for women with a uterus or oral CEE alone for women who had undergone a hysterectomy. The primary endpoints for the WHI were new heart attack, cardiac deaths with secondary clinical events of fractures, cancers, venous thromboembolism and stroke¹¹⁴.

The WHI was stopped prematurely after a mean of 5.2 years of follow-up because of an increased incidence of breast cancer, venous thromboembolism and stroke in the CEE+MPA group. Since the initial publication of these clinical trial results, the study design has been criticized because the average age of the women was at least 10 years older than women in the observational studies, and beyond the age when women would seek to use MHT for relief of menopausal symptoms. In addition, the cardiovascular health of the women at the time of enrolment was by self-report and some women had preclinical cardiovascular disease^114,115. Unfortunately, when the WHI results were reported, the specific formulation of treatment was not reported as CEE. Instead, the general term oestrogen was used. As a result, it was publicized that the use of oestrogen as an MHT, and not just CEE (with continuous MPA) increased the risk of adverse events. This imprecise designation created confusion and fear among physicians and patients. This confusion has been difficult to combat even with follow-up studies from the WHI indicating that women who initiated treatment soon after menopause and who were at low risk of cardiovascular disease had reduced severity of hot flashes, reduced risk of a cardiovascular event and, for women not using the MPA, reduced risk of breast and colon cancer¹¹⁴.

The WHI marked a tipping point in clinical practice, in that many women stopped using MHT. For women who chose to use MHT, clinical recommendations were for “the lowest dose for the shortest period of time”, even though the WHI did not provide data on dose-response or duration. Furthermore, many women switched to transdermal products, and MPA was no longer considered a progestin of choice to reduce the risk of endometrial cancer. While results from the WHI continue to be published and participants continue to be followed up, it is questionable as to how, or if, those results might apply to women who are currently using other formulations and other doses of oestrogenic compounds and progestogens. The type, mode of delivery and duration of use of oestrogens and progestogens across the menopause transition has not been investigated systematically.

The Kronos Early Estrogen Prevention Study (KEEPS), which began enrolment in 2004, compared the effectiveness of 4 years of treatment with oral CEE (0.45 mg per day) to transdermal 17β-oestradiol (50 μg per day), both with pulsed oral micronized progesterone (200 mg per day), to placebo for the rate of increase of CIMT. The KEEPS differed from the WHI in that the women were within 3 years of natural menopause and had low risk of cardiovascular disease at baseline, including a coronary artery calcium score of <50 Agatston units^116,117. Neither oral CEE nor transdermal 17β-oestradiol statistically significantly affected the age-related increase in CIMT, but both reduced the incidence and severity of hot flashes and sleep disturbances, and preserved bone mineral density and sexual function compared with placebo¹¹⁸. These results suggested that MHT can be safely used around the time of menopause to manage menopausal symptoms.

A direct comparison of the timing of initiation of treatment relative to menopause was provided by the Early versus Late Intervention Trial with Estradiol (ELITE). In ELITE, women were randomized to receive oral micronized 17β-oestradiol (1 mg per day) with cyclic vaginal progesterone gel (1 mg per day for 12 days) or placebo either within 5 years of menopause (early) or after 10 years of menopause (late)¹¹⁹. Similar to KEEPS, the primary outcome was the rate of increase in CIMT. Unlike KEEPS, ELITE included women who had unilateral oophorectomies. Women who began MHT soon after menopause had statistically significantly lower rates of increase in CIMT than women who began therapy many years after menopause¹²⁰. This finding aligned with several previous lines of research, including basic science studies, epidemiology studies, KEEPS, and a subanalysis of the WHI data^121-126. These previous lines of research suggested that the use of MHT soon after menopause might safely treat menopausal symptoms in women without cardiovascular disease and could also reduce the risk of some chronic conditions of ageing^121-126. Unfortunately, the United States Public Health Service Report of 2017 continues to rely on the initial results of the WHI for MHT recommendations and ignores the findings related to reductions in osteoporotic fractures and some cancers¹²⁷. Many questions remain regarding the change in hormone levels over the menopausal transition, the use of menopausal hormone therapies on multiple health outcomes, and the impact on the efficacy and adverse effects of other medications.

Although women have used testosterone products for a variety of conditions, there are few systematic studies examining the effects of long-term use of these products¹²⁸. The Global Consensus Position Statement on the Use of Testosterone Therapy for Women¹²⁹ identified the key benefits and risks associated with use of testosterone in postmenopausal women. However, it is clear from that consensus statement that additional evidence is needed to provide better recommendations for risk:benefit ratios, dose, and duration of treatment¹²⁹.

Testosterone in male individuals.

Testosterone prescribing has dramatically increased over the past two decades among men 40 years and older¹³⁰. Beginning in 2013, the percentage of men taking testosterone decreased slightly after the publication of findings of an increased risk of cardiovascular outcomes^131,132; however, current use is still much higher than two decades ago, especially after subsequent clinical trials did not replicate the increased risk of cardiovascular events^133,134.

The results of testosterone clinical trials among men for many conditions are mixed. The discrepancies are, in part, due to differences in study design and limitations such as: the enrolment of men who did not have low testosterone levels, use of inconsistent doses of testosterone that included sub-threshold or supra-threshold doses, the lack of clinically significant outcomes, and incorporation of different age groups. In 2003, the Institute of Medicine concluded there was insufficient evidence that testosterone treatment in older men was beneficial¹³⁵. As a result, the United States National Institute on Aging funded a coordinated collaborative to examine the short-term efficacy of testosterone in older men with low testosterone levels. The Testosterone Trials included seven placebo-controlled, double-blind clinical trials of men, with a mean age of 72 years¹³⁶. Collectively, these studies reported that men randomized to receive testosterone had some benefits with regards to sexual function, mood and depressive symptoms, bone density and strength, as well as haemoglobin concentrations and anaemia. However, there was little effect on vitality or walking distance and no beneficial effect on cognition^137-142.

The Testosterone Trials treatment duration was limited to 1 year and only enrolled men aged 65 years and older¹³⁶. Thus, it is not clear how the results would generalize to younger men, men taking testosterone for more than a year or men who have decreased libido but normal testosterone levels. Although men with comorbidities were included, there was no assessment of interactions between the testosterone treatment and other medications for any outcome. To date, testosterone products are only approved by the US Food and Drug Administration for men with hypogonadism and not for age-related declines in testosterone levels. Thus, a good portion of testosterone prescribing is off-label. In addition, studies of commercial insurance data suggest that <50% of men did not meet the guidelines of two laboratory tests of low testosterone before being prescribed supplementation^130,143,144.

Conclusions

Establishing an effective dose of an exogenous hormonal product for a specific physiological outcome might vary depending on an individual’s steroid metabolic profile¹¹⁴. For example, SULT1A1, which encodes sulfatase, is a polymorphic gene with multiple genetic variants and copy numbers. In healthy, women within three years of menopause, the age of natural menopause was associated with the number of copies of the SULT1A1 gene with a r75055 variant¹⁴⁵. As the activity of the enzyme increases with copy number, this observation suggests that women with increased enzyme activity might experience reduced circulating levels of 17β-oestradiol, with an associated dysregulation of hypothalamic–ovarian regulatory processes over the life course. Such women might benefit from increased MHT doses for symptom relief. Currently, a trial-and-error approach is used to find an effective dose of 17β-oestradiol to relieve menopausal symptoms in some women. The development of a genetic panel for assessing genetic variants for enzymes of steroid metabolism will be important to guide clinical decisions and individualize MHT prescribing.

There are many hormonal contraceptives and MHT, in a variety of doses, that can currently be prescribed. A better understanding of the metabolism, interaction with genes and short-term and long-term effects of these therapeutics is critically needed to develop guidelines for prescribing and for better individualized medicine approaches. Information on use of hormonal contraceptives, MHT, testosterone, androgen deprivation therapy, surgical history (oophorectomy or orchiectomy), pregnancy history and menopausal status should be routinely collected during medical visits and potential interactions with other medications and the impact on health should be considered. For basic science studies, sex-specific approaches that include data on hormonal status will increase the foundation of knowledge upon which novel therapies and interventions to improve human health can be developed (Box 1). In clinical trials, the numbers of female and male individuals enrolled should reflect the incidence or prevalence of the condition by sex. Information regarding hormonal status, especially for women, should be accounted for, as well as pregnancy history. In addition, results of all clinical studies should be reported by sex, even if there is insufficient power to determine a statistically significant sex difference. Providing data by sex will inform trends in demographics and will enable propensity analyses and meta-analyses. Data that should be stratified by sex and presented in results include number of participants, age, self-reported gender, clinical parameters (for example, body mass index, blood pressure, serum lipids and medications), and outcome variables for the specific study. Specific studies that assess the effect of sex and hormonal status in a range of ethnic groups, including Indigenous peoples, are also needed, not as comparison studies, but as restricted studies to inform specific biological and psychosocial factors that impact the health of these groups to better individualize care.

Box 1. Sex and hormonal factors to consider in the design of basic science studies^a.

Isolated cells and tissues

1. Is the expression of the target receptor, pathway, enzyme, etc. affected by the sex and hormonal status of the donor animal?

2. Is the expression of the target receptor, pathway, enzyme, etc. affected by hormones in the culture media?

3. Is the choice of cell or tissue appropriate for the mechanisms of interest related to human disease in regard to sex, age and hormonal status?

4. Is the number or potency of the pluripotent stem cells affected by gonadectomy of the donor animal either before or after sexual maturity?

5. For cell-based therapies, are outcomes influenced by sex match or mismatch of the donor or recipient?

Experimental animals

1. Does the animal of choice exhibit the phenotype and/or pathology as expressed in humans?

2. If gonadectomy is used, consider the impact of the procedure prior to or after sexual maturity.

3. Consider environmental factors, such as housing conditions, the light cycle, amount of handling, caging density and use soy-based foods, that might affect hormonal regulation.

^a Derived from reference Raz L, Miller VM (2012) Considerations of sex and gender differences in preclinical and clinical trials. Handb. Exp. Pharmacol. 214, 127-147 (2012)¹⁴⁸

The concept of gender that was brought into the biological and medical sciences from sociology is receiving increased attention as a contributor to health. While sex is biological, defined by sex chromosomes and functional sexual organs⁴, gender includes how society influences an individual, how the individual relates to society and how individuals relate to each other (Gender Innovations in Science, Health & Medicine, Engineering, and Environment: https://genderedinnovations.stanford.edu/)¹⁴⁶. There are many attempts to measure gender and the development of better measures is ongoing. Some components of gender, such as family status, lifestyle (for example, risk behaviours, engagement in sports and diet), income and geographical or occupational exposures (for example, acclimatization to altitude, humid or cold environments or exposure to environmental toxins) are collected in clinical trial data, but many are not. The incorporation of these components into the study design is needed because they can be quantified and addressed as confounders or covariates. Including these factors in statistical analysis is not new and provides information upon which new hypotheses or interventions can be developed.

The scientific and publishing communities are in a position to direct how experiments are conducted, what data are collected, and how results are published. Common practices considered to be best practices might not be so, such as accounting for sex as a covariate in trial analysis. Accounting for sex as a confounder does not identify or inform our understanding of how sex influences the outcomes. Although many journals have editorial policies that require authors to report the sex of the experimental material or participants, the implementation of such policies often falls short¹⁴⁷. Each of us must take responsibility to assure maximal information can be provided by our data in order to move science and medicine toward sex-specific guidelines for prevention and care.

Key points.

• Sex as a biological variable, variation in sex steroidal hormones throughout life, and gender as a social construct influence all aspects of health and disease.

• Inactivation of the X chromosome in XX individuals is random and might cluster within a tissue resulting in a patchy pattern or mosaicism of X expression in female individuals; it is unclear how this mosaicism affects disease risk and progression across the lifespan.

• Sex steroids can alter the efficacy of therapeutic interventions via multiple mechanisms, and contribute to adverse drug reactions.

• Many contraceptives and menopausal hormone therapies are available; better understanding of their metabolism and interaction with genes is needed to develop prescribing guidelines and individualized medicine approaches.

• Use of hormones (testosterone, 17β-oestradiol, or other androgen or oestrogenic compounds), surgical history (oophorectomy or orchiectomy), pregnancy history and menopausal status should be routinely collected and considered for potential medication interactions and effect on health.

• Clinical trial data should be reported by sex so as to improve transparency and reproducibility, and to inform future studies and treatment guidelines.