Why sex inclusion is not enough

Psychiatry is currently undergoing a paradigm shift towards mechanism-informed, precision care. One mechanism we can no longer treat as optional context is reproductive neuroendocrine dynamics: cyclical and transitional changes across the hypothalamic-pituitary-ovarian axis and its links to stress, sleep and immune systems. The field has made real progress in enrolling women and in staging reproductive state in animals, but we still too often design clinical studies that average across endocrine states or describe them too vaguely to compare across cohorts. The result is an evidence base that can be sex-aware yet still hormone-blind, leaving symptom variability unexplained and heterogeneity bigger than it needs to be. If we want a precision psychiatry that aims to work reliably for women, neuroendocrine dynamics must be treated as core context, not an inconvenient covariate.

Over the past two decades, sustained advocacy and policy shifts, including messaging from the National Institutes of Health Office of Research on Women’s Health, have pushed the field from routine exclusion towards routine expectation: include women, report sex and justify design choices. ^{Reference Heidari, Babor, De Castro, Tort and Curno1,Reference Arnegard, Whitten, Hunter and Clayton2} In preclinical neuroscience, that has meant directly confronting male-only samples and, increasingly, documenting oestrous stage rather than treating it as a reason to avoid female animals. In human research, requirements to consider sex as a biological variable and reporting guidance have raised the floor on transparency. ^{Reference Arnegard, Whitten, Hunter and Clayton2} Crucially, this progress has already yielded a sizeable body of work in sleep, behaviour and neuroimaging that detects menstrual cycle and reproductive stage effects, even when single time-point oestradiol or progesterone measures do not fully account for those differences. ^{Reference Arnegard, Whitten, Hunter and Clayton2}

However, even in this more sex-aware landscape, endocrine context remains inconsistently reported. ‘Cycle phase’ may be named without specifying how it was determined, hormone use may be collapsed into a single variable despite major differences between regimens, and timing labels (‘follicular’, ‘luteal’) are sometimes too imprecise to map onto the actual assessment window. ^{Reference Larsen, Frokjaer and Ozenne3} These are not small clerical details: they determine whether a null finding is informative or simply uninterpretable, and whether results can be pooled across studies rather than left as isolated observations. More broadly, a survey of six neuroscience and psychiatry journals comparing 2009 with 2019 found that, despite a 30% increase in papers including both sexes, only 19% used an optimal design for discovering sex differences and only 5% analysed sex as a discovery variable. ^{Reference Rechlin, Splinter, Hodges, Albert and Galea4} If the field still rarely designs studies to detect sex differences, it is unsurprising that within-sex endocrine state is even less consistently measured and reported.

A clear next step is to make studies ‘hormone-integrated’, not only sex-inclusive. Hormone-integrated studies should capture reproductive and hormonal context as structured exposure variables that can be reproduced, compared and synthesised across cohorts. This is not a call to rediscover that gonadal steroids affect the brain; that is already established. It is a call to measure the biologically relevant dimension of the exposure. Accounting for sex as a demographic covariate is not equivalent to accounting for hormone state as a temporally dynamic biological exposure. If phase-linked differences in mood, cognition, sleep or circuit function do not consistently correlate with a single oestradiol or progesterone assay, that is not failure; it is evidence that dynamics (rise, withdrawal, receptor adaptation), downstream neurosteroids or interactions with stress and immune pathways may be the relevant variables.

Operationally, this means defining and, where possible, stratifying by sensitivity phenotypes (for example, predictable symptom worsening during periods of oestradiol rise or progesterone withdrawal) rather than treating cycle stage as a static subgroup. Oestradiol, for example, modulates serotonin, dopamine and glutamate systems central to psychiatric therapeutics, and hormonal signalling intersects with inflammatory pathways increasingly implicated in mental illness. When endocrine context is incompletely specified, mechanisms become difficult to test, and ‘inconclusive’ findings are often embedded in the design.

A minimal endocrine context set

A minimal endocrine context set can fit in one sentence or a compact table. Many investigators already record elements of it; the goal is to make reporting consistent enough that results can travel across studies. For any study that includes participants in these reproductive endocrine contexts, authors should report three items: (a) reproductive stage at assessment (menstrual cycle phase, pregnancy/postpartum interval, perimenopausal/menopausal status or oestrous stage), (b) exogenous hormone exposure (none, or type, dose and regimen, including hormonal contraception and hormone therapy) and (c) staging method and timing (how stage was assigned and when, relative to the assessment window). This set is deliberately minimal. It does not mandate assays; it mandates transparent exposure definition – the prerequisite for interpretable null findings, comparable cohorts and cumulative evidence. Where feasible, pairing endocrine staging with targeted measures (for example, neuroactive steroids, immune markers or neuroimaging readouts) can treat reproductive transitions as ethically permissible perturbations for biomarker discovery. ^{Reference Etyemez, Mehta, Iyer, Özdemir and Osborne5} Optional additions can further sharpen inference, such as ovulation confirmation and clearly timed sampling. ^{Reference Schmalenberger, Tauseef, Barone, Owens, Lieberman and Jarczok6} This framework will also require adaptation for gender-diverse populations, including people using gender-affirming hormones.

Why hormone integration matters clinically

The strongest argument for hormone-integrated research is practical: incorporating hormonal state into study designs makes psychiatry more clinically useful during periods when women seek care. Psychiatry already operates around hormonally defined reproductive transitions and stages – menarche, the menstrual cycle, the perinatal period and the menopausal transition – but research rarely does so. The gap between how illness is experienced and how it is studied produces findings that generalise poorly to care. Characterising hormone state and defining hormone exposure connect research findings to clinical presentation, turning group averages into time-specific risk and treatment opportunities. ^{Reference Etyemez, Mehta, Iyer, Özdemir and Osborne5}

First, the menstrual cycle: premenstrual exacerbation of established disorders is common and clinically consequential, yet it is still easy to misread as treatment resistance or ‘unexplained relapse’ when symptoms are not tracked against cycle timing. Selective serotonin reuptake inhibitors can be effective for severe premenstrual syndromes using intermittent luteal-phase dosing, an inherently phase-sensitive strategy that only makes sense if timing is taken seriously. In attention-deficit hyperactivity disorder, early work suggests some patients benefit from premenstrual dose adjustments when ovarian steroid shifts alter dopaminergic signalling and medication response. The point is not that every patient needs phase-tailored prescribing today; it is that we cannot build cumulative guidance if studies do not report hormonal state in a consistent way.

Second, pregnancy and postpartum: here, large endocrine shifts coincide with sleep disruption and immune adaptation, and the historical exclusion of pregnant and lactating populations has left clinicians in a high-need, high-uncertainty prescribing environment. ^{Reference Etyemez, Mehta, Iyer, Özdemir and Osborne5} In postpartum depression, neuroactive steroid biology has translated into disorder-specific treatments: brexanolone, an intravenous formulation of allopregnanolone, and the newer oral zuranolone. ^{Reference Wilson, Robertson, Ayre, Hendon, Dawson and Bridges7} These advances broaden the mechanistic conversation beyond oestradiol and underscore progesterone-derived neurosteroids as tractable pathways for precision therapeutics. ^{Reference Wilson, Robertson, Ayre, Hendon, Dawson and Bridges7}

Third, peri- and post-menopause: patients often describe a triad of sleep fragmentation, mood change and cognitive ‘brain fog’. Evidence synthesis suggests menopausal hormone therapy can improve sleep disturbances, often via relief of vasomotor symptoms, and may offer modest, timing-dependent benefits in some cognitive domains. ^{Reference Gravelsins and Galea8} Here, ‘hormone-integrated’ evidence depends on specifying the type of menopause (spontaneous, surgical, induced) and hormone therapy formulation, route and schedule, because these differences plausibly shift brain and mood outcomes. ^{Reference Gravelsins and Galea8} Without hormone-integrated phenotyping, it becomes harder to distinguish complaints that track endocrine stage from comorbidity, and to identify who might benefit from targeted interventions.

This hormone blindness is not only a theoretical gap; it is also a safety issue. Sex differences in pharmacokinetics predict female-biased adverse drug reactions across drug classes. ^{Reference Zucker and Prendergast9} Zolpidem is a cautionary example: lower clearance and greater next-morning impairment in women led to revised dosing guidance only after two decades of routine prescribing. Sex-stratified analyses help, but they are often insufficient when cohorts include hormonal contraception, fertility treatments or menopausal hormone therapy, each altering the endocrine milieu that a sex label quietly assumes. ^{Reference Larsen, Frokjaer and Ozenne3} Psychiatric neuroscience cannot become mechanistic when hormone exposure remains unmeasured.

Making hormone-integrated studies practical

On the design side, small choices can make endocrine context analysable without adding burden. Trials can prespecify endocrine variables (cycle phase, exogenous hormone use, reproductive stage) as potential effect modifiers by using existing tools, ^{Reference Schmalenberger, Tauseef, Barone, Owens, Lieberman and Jarczok6} then build in simple features that prevent convenience recruitment from washing out signal. These can include (a) phase-stratified sampling, so recruitment does not quietly overrepresent one endocrine state; (b) within-person repeated measures across two strategically chosen phases, capturing fluctuation while reducing between-person confounding; and (c) where timing is mechanistically relevant, modelling change over time (transition periods, rate of change, withdrawal sensitivity), rather than only static categories.

These designs are more feasible than they were a decade ago. Scalable tools such as ecological momentary assessment and actigraphy can be paired with endocrine staging to test clinically relevant questions; for example, whether sleep disruption mediates phase-related mood change, or whether cyclical symptom peaks track inflammatory markers. Editorial standards can accelerate uptake by expecting a brief Endocrine Context Statement in Methods sections: state whether endocrine variables were recorded, how they were assigned and why any were not used, aligned with sex and gender reporting guidance. ^{Reference Heidari, Babor, De Castro, Tort and Curno1} A pragmatic stepped introduction could begin with prospective interventional studies, especially randomised controlled trials, where protocols already prespecify modifiers and safety monitoring, then extend the same minimal reporting expectation to observational cohorts, registries and mechanistic studies as endocrine context reporting becomes normalised. This is not bureaucracy; it is how we turn scattered observations into cumulative science.