Let’s talk about sex

Many of the biomedical breakthroughs of the twentieth century fueled an ambitious idea: that medicine could one day be tailored to the individual. While the intuition behind personalised care echoes principles already present in classical medicine, it became more and more real in the 20th century. Thanks to the advances in genetics, researchers began to imagine treatments customised to a patient’s biological makeup, giving rise, for instance, to fields like pharmacogenomics. 

This vision was further strengthened by advancements in computer modelling and simulation, offering new ways to represent complex physiological processes mathematically. In parallel, the rapid expansion of the -omic sciences (genomics, proteomics, and metabolomics), facilitated  the comprehension of the biological fingerprint of an individual.

More recently, the explosion of big data in healthcare, coupled with advances in artificial intelligence, has transformed precision medicine from an aspiration to a realistic possibility. Digital twins of patients can integrate biological and physiological data to generate personalised analyses that allow to simulate disease progression, anticipate treatment responses, and support nuanced clinical decision-making.

In principle, these tools offer a powerful pathway toward truly personalised medicine. Yet, despite these great promises, a persistent and long-dating issue remains. For much of its history, Western medicine has been built around a narrow archetype: the white, adult, male. This bias has endured even into the era of big data. Despite the inclusion of hundreds of critical parameters in modern diagnostic and predictive models, women and minority populations continue to be underrepresented in medical research and, as a result, underserved in clinical practice. The “average patient”, implicitly or explicitly, remains a white adult man.

This imbalance has deep historical roots. Many of today’s medical technologies are trained, tested and validated on datasets derived from traditional clinical trials, in which women, non-binary individuals, ethnic minorities, and other underrepresented groups have traditionally been excluded. The consequences are far from abstract. For those left behind, this legacy translates into delayed diagnoses, suboptimal treatments, and poorer outcomes across a wide range of conditions.

As reported by Amalie Holmgaard Mersh in The European Correspondent, heart diseases have been historically seen from a man’s perspective leading to remarkable differences in diagnosis, treatments, and outcomes for women.

In this sense, cardiovascular disorders provide a stark example. They are the leading cause of death worldwide, but affect men and women differently. The so-called “typical” signal of chest pain in myocardial infarction reflects how men experience it, while in women, symptoms are somewhat “atypical” with fatigue, nausea, or back pain. In reality, both of them are typical, but only one of them is currently accepted. This contributes to delayed diagnosis, misdiagnosis, under diagnosis, as well as less impactful treatments, more frequently in women than in men.

Such sex- and gender-related disparities have been documented across a wide spectrum of other disease areas, including diabetes, neurological and mental health disorders, autoimmune conditions, and cancer. Likewise, understanding the gender-specific differences in physiological processes such as brain aging and pain sensitivity, is prevalent. Women seem to also report adverse drug reactions more frequently than men, a sign that female biology has often been inadequately included during drug development. Such challenges further complicate in the representation of intersex, transgender, and non-binary individuals who remain largely understudied in biomedical research.

As highlighted in a 2020 study published in npj Digital Medicine by Davide Cirillo and colleagues, growing awareness of these biases has coincided with the rapid adoption of artificial intelligence in healthcare. As both a problem and solution, scientists caution that  “AI could become a double edged sword”. When designed thoughtfully, it can help correct longstanding inequities by explicitly accounting for sex and gender differences. When imprudently deployed, it risks amplifying the very biases embedded in historical data. Much depends on how algorithms are constructed, trained and validated.

Interestingly, the authors make a distinction between desirable and undesirable biases. Although the term bias has gained a negative connotation, the authors argue that certain biases are essential to the realisation of personalised medicine. For instance, a desirable bias is the deliberate incorporation of biologically meaningful differences, such as sex-specific anatomy or physiology, to improve diagnostic accuracy and personalise treatments. By contrast, an undesirable bias arises when models systematically disadvantage certain groups through omission, misrepresentation, or unjustified assumptions.

Importantly, bias can enter medical technologies at multiple levels. For instance, bias can be present in the underlying training data that exclude or misrepresent certain groups, or can be embedded in the model design, with irrelevant feature selection or inappropriate stratification of gender. These issues reflect a broader historical legacy, including the continued reliance on predominantly male animal models in preclinical research. Such inherited biases should not be allowed to shape the emerging field of in silico medicine.

Encouragingly, examples of a more inclusive approach are also emerging. The study published in eLIFE, co-authored by the VPH member Blanca Rodríguez, from the University of Oxford, illustrates how computational medicine can be used to address, rather than reinforce, sex-based disparities.

The study focuses on the electrocardiogram (ECG), a cornerstone of routine clinical assessment of cardiac electrical activity. Previous experimental and computational work has shown that ECG signals are strongly influenced by heart-torso anatomy, sometimes more than by electrophysiology alone. The authors demonstrated that sex and body mass are key determinants of anatomical differences in heart size, orientation and torso volume.

Given that ischemic heart disease remains the world’s leading cause of death, and that acute myocardial infarction is more frequently missed in women, the impact is substantial. Women typically exhibit ECG features that are different from men.

Sex differences in electrocardiogram (ECG). Taken from Prajapati, Chandra, et al. “Sex differences in heart: from basics to clinics.” European Journal of Medical Research 27.1 (2022): 241.

As illustrated in the picture above, women typically exhibit several distinct ECG features compared to men, including a lower ST-segment, different T-wave amplitudes, and a longer QT interval. These physiological differences can complicate ECG interpretation, contributing to diagnostic uncertainty and unequal clinical management.

Using cardiac magnetic resonance images from the UK Biobank, Rodríguez and colleagues analysed data from 1476 individuals, including both healthy subjects (of which 54.3% were females) and patients with prior myocardial infarction (of which 19.8% were females). The authors developed a neural-network pipeline to automatically reconstruct three-dimensional heart–torso models from MRI data. Computational analysis of these personalised models revealed systematic sex-related anatomical differences and demonstrated how anatomy, in both health and disease, significantly shapes ECG biomarkers.

Notably, the study showed that roughly half of the shorter QRS duration observed in women could be explained by smaller cardiac cavity volumes. Moreover, anatomical factors exert a stronger influence on post-infarction T-wave changes in women than in men.

This work lays the foundation for correcting ECG interpretation based on individual anatomy, an essential step towards truly personalised diagnostics. More broadly, it exemplifies the desirable bias in action: the integration of sex-specific information to improve clinical precision.

As in silico medicine continues to advance, inclusive and carefully designed approaches such as this will be essential. Precision medicine cannot fulfil its promise if it remains precise only for some.

PS: This important topic strongly resonated within our Young Scientists community, as highlighted by Laura Lafuente-Gracia (University of Liège), runner-up for the VPH 2024 Best Student Award, who presented the contribution “Minding the gap: sex differences influence bone fracture healing”. Her work showed that, although bone repair involves different cellular activities in men and women, these processes ultimately lead to comparable healing outcomes.

FURTHER READINGS:

• Sex-specific human electromechanical multiscale in-silico models for virtual therapy evaluation– Journal of Molecular and Cellular Cardiology Plus, 2025;
• Female anatomies disguise ECG abnormalities following myocardial infarction: an AI-enabled modelling and simulation study– arXiv, 2025;
• Anatomical basis of sex differences in the electrocardiogram identified by three-dimensional torso-heart imaging reconstruction pipeline– eLIFE, 2025;
• Sex-specific cardiometabolic multimorbidity, metabolic syndrome and left ventricular function in heart failure with preserved ejection fraction in the UK Biobank– Cardiovascular Diabetology, 2025;
• Uncovering the persistent gap: The ongoing challenge of integrating sex and gender in biomedical research– Journal of Biomedical Research, 2024;
• Beyond the gender data gap: co-creating equitable digital patient twins– Frontiers in Digital Health, 2025;
• Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare– npj Digital Medicine, 2020;
• The Gender of Biomedical Data: Challenges for Personalised and Precision Medicine– Somatechnics, 2019;
• Heart health is a man’s world – so far– The European Correspondent, 2026.