The Science Is Good. But Who Was It Tested On?

By Roswitha Verwer, Founder and CEO of YON E

Femtech has made measurable progress over the past decade. Peer-reviewed research in women's health is expanding, regulatory bodies are increasingly engaging with digital health and biosensing technologies, and investment in the sector is growing. Within this context, one methodological challenge continues to limit the clinical impact of otherwise promising tools: the evidence base underpinning many femtech products lacks the demographic and geographic diversity required for broad generalizability.

This is a methodological gap with documented implications for clinical accuracy, regulatory credibility, and equitable health outcomes.

At YON E, we recently contributed to a peer-reviewed framework published in “Frontiers in Global Women's Health” that addresses this directly. Pillar one of that framework calls for inclusive, multi-site validation and bias-aware regulation. This article explains the scientific rationale behind that recommendation and why it matters for the companies building in this space.

External validity and why it matters

Validation in health technology establishes that a tool performs as intended under defined conditions. External validity asks whether those conditions reflect the real-world populations the tool will serve. A device validated at a single academic medical center in a high-income country may produce evidence that does not generalize to other clinical environments, demographic groups, or infrastructure contexts.

The consequences of limited external validity are documented in adjacent fields. A 2022 study published in “Science Advances” found significant performance disparities in dermatology AI across skin tones, with models trained predominantly on lighter skin showing reduced performance on images from patients with higher melanin concentrations. These findings illustrate a well-characterized methodological risk: when training data does not reflect population diversity, diagnostic accuracy decreases for underrepresented groups.

Reproductive health technology faces the same risk. Published literature cited in our framework, including a 2024 review co-authored by members of the YON E team, suggests that biomarkers relevant to reproductive health monitoring, such as vaginal pH and basal body temperature, may vary across populations due to differences in microbiome composition, environmental exposures, and demographic characteristics. Where validation datasets do not reflect this variation, tools carry a measurable risk of reduced performance for underrepresented groups.

Algorithmic bias as a methodological problem

AI-enabled femtech tools present a specific challenge. Machine learning models derive their predictive capacity from training data. When that data overrepresents certain populations, the model's performance generalizes poorly to groups outside the training distribution. This is a well-characterized limitation in machine learning methodology, documented across medical imaging, risk stratification, and symptom-based diagnostic tools.

A 2025 narrative review examining AI-driven tools for endometriosis diagnosis, cited in our publication, identified demographic homogeneity in training datasets as a primary driver of reduced equity in diagnostic performance. EHR-based predictive models for endometriosis symptom clustering, while demonstrating strong technical accuracy, have similarly been noted to lack prospective validation across diverse populations, raising concerns about generalizability and fairness.

Our framework recommends that femtech developers incorporate mandatory bias audits, pre-market multi-site validation requirements, and post-deployment performance monitoring into their development and regulatory pathways. These are recommendations from our strategic framework, not current regulatory requirements, though they reflect the direction regulatory science is taking as AI-enabled medical devices become more prevalent.

What the evidence base needs

Strengthening the scientific foundation of femtech requires structural changes to how evidence is generated. Sex-disaggregated reporting standards would improve comparability across studies. Pragmatic trial designs and validation partnerships across diverse clinical settings would improve generalizability. Community-based recruitment strategies would address the demographic concentration that currently limits many datasets.

External validity, algorithmic fairness, and demographic representativeness are core methodological standards. Meeting them is what allows a technology to move from a promising prototype to something clinicians, regulators, and health systems can trust at scale.

The scientific case for investment

The commercial implications follow directly from the science. A femtech tool with narrow validation carries a higher regulatory risk profile as scrutiny of AI-enabled medical devices increases globally and a more limited addressable market. A tool validated across diverse populations, clinical settings, and geographies carries stronger evidence of generalizability, a more defensible regulatory position, and access to a significantly larger patient population.

Less than 0.5% of healthcare venture capital is currently allocated to women's health, according to data cited in our publication. Closing that gap requires not only more capital but capital directed toward companies building clinical evidence to the standard that regulators, hospital systems, and academic partners will demand over the coming decade.

At YON E, we are designing our validation strategy with global demographic diversity as a foundational requirement, not a later-stage consideration. We believe that scientific rigor and commercial viability are the same argument.

If that philosophy resonates with you as an investor, we would welcome the conversation.

Roswitha Verwer 

Founder & CEO YON E 

This post is part of a series based on our publication "Aligning Femtech Innovation With Equity: A Strategic Framework for Real World Impact," published in Frontiers in Global Women's Health (2026).

Read the full publication in Frontiers in Global Women's Health here.