Why better diagnostics, longitudinal data, and female-specific biology are changing medicine's understanding of risk
Medicine can sequence a tumor, visualize individual cells, predict the three-dimensional structure of a protein and monitor a heartbeat from a wristwatch.
Yet many of its foundational assumptions about women's health remain surprisingly crude.
For decades, we assumed this is a problem of exclusion: too few women in clinical trials, too little investment in women's diseases and too little attention to sex-specific biology.
The real issue could be that women's health suffers from a measurement problem!
We have frequently measured women using tools, thresholds, disease definitions and clinical endpoints that were not designed around female biology. We also have treated hormonal transitions as background noise, reproductive history as a speciality concern and symptoms that do not fit conventional models as "atypical."
Rather than focusing solely on representation, may be it is time to rethink and redesign the scientific frameworks through which women's health is studied.
The "default patient" is being retired
The old model of medical research most of the time began with an implicit reference patient: male, hormonally stable and biologically consistent over time. Women were then added as a subgroup.
That approach assumes that female biology is a variation on the standard model. However, across cardiovascular disease, autoimmunity, neurological disease, cancer and inflammatory disorders, we are seeing a different reality: sex can influence disease mechanisms, symptom presentation, immune activity, drug exposure, toxicity and treatment response.
Women's stronger type 1 interferon responses, for example, may offer protection against some infections while increasing susceptibility to autoimmune disease. X-chromosome biology may help explain why conditions such as lupus disproportionately affect women. Hormonal transitions can modify immune activation, vascular tone, pain, metabolism and tissue repair. Menopause may change the trajectory of diseases far beyond the reproductive system. (Ref: Symposium: Advances in the Development of Therapeutics and Diagnostics for Women's Health: https://pbss.org/eventDetails/1074)
The point is not that every condition requires a completely separate branch of medicine for women and men. We just need to acknowledge that sex and hormonal state can be biologically consequential variables! And this distinction becomes especially important when medicine relies only on thresholds.
A standard angiogram may show no obstructive disease while dysfunction stays in the smaller coronary vessels. Two patients can have the same clinical score while carrying very different levels of underlying pathology. The same drug concentration can produce different benefits or risks depending on disease stage, immune state and hormonal context.
This leads to the problem of Correct number, but wrong interpretation!
From snapshots to trajectories
One of the clearest shifts looks like the movement away from single measurements and population averages toward biological trajectories.
Laura Esserman's work in breast cancer illustrates what changes when risk is treated as individual rather than average. In the WISDOM study, screening strategies incorporated clinical risk, breast density, inherited variants and polygenic risk rather than recommending the same schedule for every woman. According to their report, approximately 30% of women with pathogenic variants had no family history that would have identified them through conventional triage.
This is an example of the fact that how easily high-risk individuals can disappear inside apparently reasonable screening rules.
Risk-based care is not simply "more screening." For some people it may mean earlier or more frequent assessment. For others it may mean avoiding unnecessary procedures. Precision should reduce both neglect and overtreatment.
What menstruation can teach medicine about inflammation
One of the most striking reframes in this topic came from the use of menstrual biology as a model of inflammation and repair.
Ridhi Tariyal talked about this topic. Menstruation has traditionally been treated as a reproductive event, an inconvenience or a variable to be controlled. But biologically, it is a recurring, accessible example of tissue breakdown, immune activation, clearance and regeneration.
In a healthy cycle, inflammation transitions into repair!
We know that many inflammatory therapies are designed to suppress immune activity. They are less effective at restoring the programs that rebuild tissue. A disease can therefore become quieter without becoming resolved.
The work described by their team proposes that longitudinal menstrual samples may reveal where an inflammatory process stays along the path from breakdown to repair. Menstrual fluid contains immune, stromal and epithelial information and can be collected repeatedly without surgery. The reported Inflammatory Resolution Score was designed to identify biological state rather than simply disease severity or elapsed time. In a cohort of patients with ulcerative colitis, the score reportedly provided information that was largely distinct from the standard clinical score and helped separate patients with different probabilities of steroid-free remission.
The broader significance could be that a biological process medicine once dismissed as inconvenient may contain information relevant far beyond gynaecology.
This reversal captures the changing landscape of women's health. Features that were once excluded because they complicated experimental design are becoming sources of scientific insight.
This "confounding variable" may turn out to be the biology we needed to understand.
I found this fascinating!
A word of caution: AI can integrate the data, but it cannot invent the missing labels
Women's health is entering a transformative period just as biomedical science is generating unprecedented amounts of data. Single-cell sequencing, spatial technologies, multi-omics, multimodal analysis, wearables, and electronic health records now allow us to capture biological changes across tissues, life stages, and time.
AI is exceptionally well suited to uncovering patterns within these complex, heterogeneous datasets. But even the most sophisticated algorithm cannot compensate for incomplete, biased, or poorly defined labels. When trained on such data, AI may simply produce a more elegant and convincing version of the same historical blind spots.
The next challenge is to create better datasets: longitudinal rather than episodic, biologically annotated rather than merely large, diverse across ancestry and life stage, and connected to outcomes that matter to patients.
The field needs scientists, clinicians and patients to decide what the data should represent.
The endpoint is not always a cure
Another important change is occurring in how success is defined.
Drug development has traditionally prioritized endpoints that are easy to standardize: biomarker changes, radiographic response, disease progression or survival. These remain essential.
But patients may define meaningful benefit differently.
A treatment that allows someone to work, sleep, care for a child, travel independently or live without debilitating pain may be transformative even when it does not reverse the underlying disease.
The discussion offered an instructive example: some patients with endometriosis may reject therapies that chemically induce menopause because the trade-off feels worse than living with the disease. Meanwhile a trial can demonstrate biological activity and still produce an intervention patients do not want.
This comes from failure to ask the right scientific question.
Patient-reported outcomes, daily function, treatment burden and quality of life should not be decorative additions placed near the end of a development program. They help determine whether the intervention solves the problem it claims to solve.
"Patient-centered" should mean treating lived experience as evidence that shapes the protocol.
The opportunity is larger than reproductive health
Women's health is still too often used as shorthand for fertility, pregnancy and reproductive organs.
Those areas deserve far more research and investment than they have historically received. But they do not define the limits of the field.
Women's health includes conditions that affect women uniquely, disproportionately or differently.
That includes autoimmune disease, cardiovascular disease, neurodegeneration, osteoporosis, depression, inflammatory disorders and treatment toxicities. It includes how hormonal transitions alter disease risk across the lifespan. It includes why the same clinical measurement may have a different meaning in two biological contexts.
Cardiovascular disease is a particularly important example. There is a possibility that women's cardiovascular disease may be more likely to involve microvascular dysfunction that is not captured by conventional tests designed to detect obstruction in larger vessels. The precise prevalence and clinical definitions require careful external verification, but the underlying concern is well established: a reassuring standard test may not always rule out meaningful disease in women.
The future of women's health requires willingness to question the test if the results do not agree with the symptoms.
What progress could look like
Ten years from now, success should also be measured by the structure of medicine in addition to the number of women's health companies launched or the amount of capital invested.
Clinical trials would be designed from the beginning to detect sex-specific differences in efficacy, safety and pharmacology.
Hormonal state and reproductive milestones would be recorded when biologically relevant rather than omitted because they complicate analysis.
Risk models would incorporate trajectories, life stage and individual biology instead of relying only on thresholds derived from population averages.
Diagnostics would be validated in the populations they are meant to serve.
Patient-reported outcomes would influence development decisions alongside molecular and clinical endpoints.
AI systems would be trained on datasets labeled well enough to recognize female biology rather than reproduce its historical absence.
The hopeful part is that many of the necessary tools already exist.
The field now has increasingly affordable omics technologies, adaptive clinical trial models, longitudinal real-world data, patient communities ready to participate and computational methods capable of integrating complex biological information.
This will be a possibility if we are willing to reconsider the framework.
Women's health is becoming one of the defining scientific and healthcare opportunities of our generation because solving it will require us to improve what we know about women and how medicine produces this knowledge in the first place.