Q&A: Prof Michelle A Williams on the future of women's health research
To mark International Day of Women and Girls in Science, we hear from Professor Michelle A Williams ahead of her LSE Health and Department of Health Policy Annual Lecture on 9 March 2026.
As Professor of Epidemiology and Population Health at Stanford University, Professor Williams has spent decades advancing our understanding of women's reproductive and perinatal health through groundbreaking research that bridges traditional epidemiology with cutting-edge molecular biomarkers and digital health innovations. In this Q&A, she reflects on the transformative potential of large-scale digital studies, the path from evidence to policy change, and why interdisciplinary approaches are essential for achieving health equity worldwide.
From biology to global health systems
Your academic path spans biology, civil engineering, and epidemiology. How has this shaped your approach to women's health research?
I sometimes describe my path as unconventional, but in retrospect every turn was preparation for the work I do now. My early training in developmental biology and genetics at Princeton gave me a molecular vocabulary: an instinct for mechanism, for asking why something happens at the cellular and physiological level, not just whether it happens in a population. Civil engineering then taught me to think in systems: to design public health structures that are robust, adaptable, and built to be sustainable. That systems-thinking, along with my exposure to statistics and biostatistics at Tufts, turned out to be exactly the mental architecture I needed for epidemiology at Harvard.
When I entered epidemiology, I brought this conviction that the discipline should function as a platform, not a silo. A well-designed epidemiological study should be built to accommodate successive layers of scientific inquiry. You start with rigorous study design and data capture. Then you integrate biochemistry. Then genetics. Then genomics. And now, digital technology and artificial intelligence. Each layer deepens what you can ask and whom you can study.
That layering philosophy has defined my entire career. When we built our placental abruption research network across six hospitals in Lima, Peru, we didn't just design a case-control study; we designed a platform that could absorb molecular biomarker studies, genetic analyses, and novel case-crossover designs as the science evolved. The same principle guided the Apple Women's Health Study, where we engineered a digital infrastructure flexible enough to collect survey data, sensor data, and wearable-derived metrics simultaneously from over 120,000 women.
Women's health challenges are inherently multi-causal. They span biology, behaviour, environment, economics, and policy. You need a research architecture that can hold that complexity, and my interdisciplinary path gave me the blueprints.
The research landscape
Why do conditions like preeclampsia and gestational diabetes remain critically understudied?
These conditions remain understudied for reasons that are equal parts scientific and societal. Scientifically, pregnancy is an extraordinarily complex physiological state: the interplay of two genomes, a temporary organ in the placenta, dramatic metabolic and cardiovascular adaptations, all unfolding over a relatively compressed time window. But the societal dimension is just as important: women's health, and reproductive health in particular, has been chronically underprioritised in funding, in research infrastructure, and in the training of the next generation of investigators. For too long, the medical research enterprise treated pregnancy as a niche concern rather than the profound window into lifelong health that it actually is.
Our parallel case-control studies across three continents (the United States, Zimbabwe, and Peru) were among the first to document the presence of dyslipidaemia, chronic systemic inflammation, oxidative stress, and endothelial dysfunction in preeclamptic women globally. This demonstrated that the pathophysiological signatures of preeclampsia are shared features of the disease regardless of geography, ethnicity, or healthcare context.
Perhaps the finding I'm most proud of concerns physical activity. Through the Omega Study, a 10-year prospective pregnancy cohort funded by NICHD, we documented that women who engaged in regular leisure-time physical activity before and during pregnancy had significantly reduced risk of both preeclampsia and gestational diabetes. This work is widely credited with helping to reverse the longstanding clinical recommendation that pregnant women should adopt a sedentary lifestyle. That was a case where epidemiology didn't just generate knowledge; it changed clinical practice.
We also made important contributions to biomarker discovery, using prospectively collected early-pregnancy specimens to identify molecular signatures that predate clinical disease, pointing towards early detection and intervention.
What have your international collaborations revealed about how women's health challenges differ across global contexts?
The most striking lesson from working across four continents is that the biology of these conditions is remarkably consistent, but the context in which they unfold is profoundly different. Preeclampsia looks biochemically similar in Seattle and in Harare, but whether a woman receives timely diagnosis, adequate treatment, or survives the condition at all depends entirely on the systems around her.
In Peru, we discovered novel risk factors for placental abruption (mood and anxiety disorders, migraine) that hadn't surfaced in North American studies. That wasn't because the biology was different; it was because the population, the clinical infrastructure, and the study design we deployed were different in ways that made these associations visible for the first time. In Thailand, we were able to evaluate maternal periodontal disease as a risk factor for preeclampsia and preterm delivery, and we were the first group to systematically assess how different operational definitions of periodontal health status alter the associations you observe.
International research also taught me about humility and reciprocity. The most productive collaborations are not extractive but partnerships in which local investigators, clinicians, and communities shape the research questions and share in the scientific credit.
Innovation in women's health research
The Apple Women's Health Study enrolled over 120,000 participants through a smartphone app. What can digital epidemiology reveal that traditional research methods cannot?
Digital epidemiology fundamentally changes three things: scale, continuity, and granularity. Traditional clinical studies typically enrol hundreds to a few thousand women, collect data at discrete time points (perhaps once a trimester, perhaps annually), and rely heavily on retrospective recall. The Apple Women's Health Study, which I co-designed with Apple, the National Institute of Environmental Health Sciences, and colleagues at Harvard, has enrolled over 120,000 women who contribute data continuously through their devices. That is a qualitative, not just quantitative, difference.
Consider menstrual cycle research, which has been astonishingly understudied. Before digital tools, the typical study might ask women to recall their cycle length and regularity retrospectively. With continuous, sensor-assisted, longitudinal tracking, we can observe real-time variation (cycle to cycle, season to season, year to year) and correlate it with lifestyle exposures like sleep, physical activity, nutrition, and stress. We can study the dynamics of reproductive health, not just snapshots of it.
Digital epidemiology also democratises participation. Our participants are not limited to women who live near an academic medical centre or who can take time off work for clinic visits. They participate from wherever they are, whenever is convenient. That dramatically improves the geographic and socioeconomic diversity of the evidence base, which for a field that has historically relied on narrow, clinic-based convenience samples is transformative.
The challenge, of course, is rigour. Digital data are abundant but noisy. Making valid scientific inferences from passively collected sensor data requires careful epidemiological thinking about measurement error, selection bias, and confounding. That is where the discipline of epidemiology remains indispensable, even as the tools on top of it evolve.
How is combining molecular biomarkers with digital health studies transforming our understanding of reproductive health?
For a long time, molecular biology and population-level epidemiology operated in parallel but rarely converged. Bench scientists studied mechanisms in small samples; epidemiologists studied associations in large cohorts. Each approach had blind spots. Molecular studies lacked population context; classical epidemiological studies lacked mechanistic depth.
What we've done is build a bridge. Using our Perinatal Biospecimen Repository (carefully collected, preserved early-pregnancy specimens from large prospective cohorts), we've been able to interrogate pre-diagnostic samples using multiple molecular platforms simultaneously. We can now ask: what molecular profiles in early pregnancy predict preeclampsia or preterm delivery months before clinical signs appear? And how are those profiles shaped by modifiable factors like diet and physical activity?
Combining these molecular data with the high-dimensional behavioural and environmental data streaming from digital health platforms creates something genuinely new: the ability to study how lived experience gets biologically embedded in real time, at population scale. That is not just a technical advance; it is a conceptual one. It reframes reproductive health as a dynamic, multi-level process and opens the door to precision prevention: tailored interventions based not just on clinical risk scores but on an individual's molecular and digital health profile.
From evidence to impact
Can you share an example where your research has directly influenced clinical practice or health policy?
The clearest example is our work on physical activity and pregnancy. When I began studying this in the 1990s, the prevailing clinical wisdom was that pregnant women should rest, that physical exertion was risky. There was a deep cultural and medical conservatism around pregnancy and exercise. Our research, particularly the series of papers from the Omega Study, provided rigorous prospective evidence that regular leisure-time physical activity before and during pregnancy was associated with significantly reduced risks of preeclampsia and gestational diabetes.
What made translation possible was a combination of factors. First, the evidence was replicated across multiple study populations and continents, which gave it credibility. Second, we actively engaged with clinical and public health communities to communicate the findings in accessible terms. And third, the timing aligned with a broader shift in medicine towards lifestyle-based prevention. Our work became part of the evidentiary foundation that professional organisations used to revise their guidance on exercise during pregnancy.
This experience also taught me that the gap between evidence and practice is not primarily a knowledge gap. Clinicians and policymakers often face structural barriers to acting on evidence, which is why science communication, health equity, and systems reform are inseparable from the research enterprise itself.
What are the most critical gaps between what the evidence tells us and what health systems actually deliver for women?
I see three critical gaps. The first is in how we define the scope of women's health. Too many health systems still treat women's health as synonymous with maternal health or reproductive care within a narrow window of childbearing years. The evidence tells us that reproductive health is a window into lifelong health: that conditions like preeclampsia and gestational diabetes are harbingers of cardiovascular disease, metabolic syndrome, and accelerated ageing decades later. Health systems that fail to connect these dots are missing a critical opportunity for early intervention.
The second gap is in data infrastructure. We know vastly more about women's health than we did thirty years ago, but most health systems lack the capacity to collect, integrate, and act on the kind of longitudinal, multi-dimensional data that modern research generates. The Apple Women's Health Study demonstrates what is technically possible, but translating that capability into routine healthcare delivery requires investment, political will, and a rethinking of how health systems use data.
The third gap is equity. Even where good evidence and good infrastructure exist, access to evidence-based care is distributed unequally. Low-income communities, racial and ethnic minorities, and women in low- and middle-income countries bear a disproportionate burden of adverse reproductive outcomes. Closing the gap between evidence and delivery requires not just better science, but better communication, cross-sectoral partnerships, better policy, and more equitable systems.
The path forward
What do you see as the most promising frontier for advancing women's health globally over the next decade?
I see the convergence of three forces creating an extraordinary opportunity. The first is the maturation of digital health platforms. We now have the technical capacity to conduct truly global, longitudinal women's health research at a scale and with a granularity that was unimaginable a decade ago. The Global Women's Health Study, for example, will leverage mobile applications, wearable sensors, and quarterly health surveys to follow geographically and racially diverse populations across the reproductive and post-menopausal life course.
The second force is artificial intelligence. AI is the natural next layer on the epidemiological platform. When you have high-dimensional, longitudinal, multi-modal data (survey responses, sensor streams, molecular profiles, clinical records), you need computational approaches that can identify patterns and generate predictions at a scale that human analysis alone cannot match. We are already developing AI-enabled analyses for predicting accelerated ageing and informing clinical screening. The challenge is to deploy these tools rigorously, with epidemiological discipline around bias, confounding, and generalisability.
The third force is institutional and political. There is growing recognition (at the World Economic Forum, in government, in the private sector) that women's health is not a niche medical concern but a global economic and security imperative. Healthy women are the foundation of healthy families, communities, and economies. That recognition is creating new funding streams, new partnerships, and new political will.
If I had to name the single most promising frontier, it would be precision prevention: the ability to identify, early in life and early in pregnancy, which women are at highest risk for which outcomes, and to intervene with tailored strategies before clinical disease manifests. Achieving that will require exactly the interdisciplinary integration I have spent my career building: rigorous epidemiology as the platform, molecular science for mechanistic insight, digital technology for scale and continuity, and AI for pattern recognition. The pieces are finally coming together, and I believe the next decade will be transformative for women's health globally.