The dual diversity crisis in physiology

Physiological research has achieved extraordinary technical sophistication. High-density electrode arrays decode motor neuron activity with unprecedented precision (Farina et al. 2010). Continuous glucose monitors track metabolic dynamics second-by-second (Rodbard 2016). Cardiac magnetic resonance imaging reveals myocardial strain patterns invisible a decade ago (Barison et al. 2025). Mass spectrometry quantifies individual protein turnover rates in skeletal muscle (Davies et al. 2024). These methodological advances represent genuine progress. Yet sampling practices have remained static over the past two decades. Recent high-impact papers exemplify this pattern: sophisticated methods applied to demographically homogeneous samples that reveal substantial individual heterogeneity but cannot explain it. A 2025 motor control study uses cutting-edge decomposition algorithms on 24 males aged 22 ± 1.7 years, finding that half show one recruitment pattern and half show the opposite – but cannot investigate why (Weinman et al. 2025). A 2025 study published in Nature Biomedical Engineering employs continuous glucose monitoring to identify metabolic subphenotypes but in only 32 participants, precluding analysis of how sex, age or hormonal status predict phenotype distribution (Metwally et al. 2025). A 2024 systematic review documents 0- to 2.7-fold variation in muscle protein synthesis responses but finds ‘substantial heterogeneity’ that demographic factors could explain if samples included diversity (Davies et al. 2024). This creates what I term the dual diversity crisis (Figure 1): demographic homogeneity (Crisis 1) prevents investigating the biological heterogeneity our methods reveal (Crisis 2). The sophisticated measurements we make reveal substantial individual variation, but homogeneous sampling prevents determining whether this variation relates to sex, age, hormonal status, training history, muscle architecture, or other demographic and phenotypic factors versus remaining truly idiosyncratic. We cannot distinguish population-specific patterns from individual variation because our samples lack the demographic range needed to test these hypotheses. Demographic homogeneity (Crisis 1) prevents investigating the biological heterogeneity that sophisticated methods reveal (Crisis 2), creating a validity threat. Recent examples (2024–2025) from motor control, metabolic, exercise and cardiovascular physiology demonstrate this pattern is current and field-wide. This is not about ‘health disparities’ or ‘inclusion’ as secondary considerations. This is about basic physiology: if physiological mechanisms differ systematically by sex, age or metabolic state, then excluding or ignoring these factors does not reduce confounds – it introduces them. We confound population-specific patterns with universal mechanisms. The consequences of this dual crisis extend beyond methodological concerns to clinical outcomes. In cardiovascular medicine, guidelines developed primarily from male samples have contributed to systematic misdiagnosis of women: women with myocardial infarction are twice as likely as men to receive mental health diagnoses, with symptoms attributed to anxiety rather than cardiac pathology (Maserejian et al. 2009). Women under 55 experiencing heart attacks present with more variable symptom combinations than male-derived protocols recognize (Brush et al. 2020), contributing to delayed diagnosis and worse outcomes. Similarly, pharmacogenomic algorithms developed in European-ancestry populations have shown reduced efficacy or increased adverse events when applied to African American patients: the COAG warfarin dosing trial demonstrated that genotype-guided dosing – optimized for European alleles – reduced time in therapeutic range for Black participants compared to clinical dosing alone (Kimmel et al. 2013). These are not hypothetical concerns but documented harms resulting from the assumption that findings from demographically narrow samples apply universally. Weinman et al. (2025) used dual high-density electrode arrays and sophisticated decomposition algorithms to identify motor unit modes in tibialis anterior. The technical sophistication is exemplary. The sample – 24 males aged 22 ± 1.7 years from a single university – is typical. The most striking finding: 50% of participants showed Mode 1 recruitment increasing with force while 50% showed the opposite pattern. This main effect reversal represents a fundamentally different motor control strategy for identical tasks. Yet the homogeneous sample prevented investigating whether sex, age, muscle architecture, training history or other measurable factors predict this striking pattern. The homogeneous sample prevented investigating what predicts this pattern. Sex cannot be tested – all participants were male, despite documented sex differences in motor unit properties (Hunter 2014). Age effects cannot be examined – the range spans under 2 years. Training history, muscle architecture and functional outcomes went unmeasured. The study achieved technical sophistication in one dimension – motor unit recording – but eliminated both the demographic variance (sex, age and training background) and the comprehensive phenotypic assessment (muscle architecture, hormonal status, training history, functional outcomes) needed to understand why individuals show fundamentally opposite motor control strategies. The result: a striking finding with no path to mechanistic understanding. A recent study published in Nature Biomedical Engineering exemplifies the metabolic manifestation of this crisis (Metwally et al. 2025). The authors used continuous glucose monitors during oral glucose tolerance tests to identify metabolic subphenotypes in prediabetes. The technical innovation is impressive: machine learning models predicted subphenotypes with 89–95% accuracy based on glucose curve shapes. Gold-standard metabolic testing revealed that 34% of participants had dominant muscle or hepatic insulin-resistance phenotypes, while 40% exhibited β-cell dysfunction or impaired incretin action. This represents substantial metabolic heterogeneity demanding investigation. Yet the study included only 32 individuals in total. Sex distribution is unreported. Age range is unreported. Menstrual cycle phase – which alters insulin sensitivity by up to 40% (Escalante Pulido homogeneous sampling just ensures we cannot investigate its sources. This creates a validity threat, not merely a generalizability limitation (Henrich et al. 2010). When findings emerge from narrow samples, we cannot distinguish universal biological principles from population-specific patterns. The homogeneous sample that appears to reduce confounds introduces a critical one: we confound demographic-specific patterns with universal biology. Addressing the dual diversity crisis requires no technical breakthroughs – the tools exist, the methods are mature – but it does require more comprehensive measurement within diverse samples. The examples presented (50–50 split in motor unit modes, metabolic subphenotypes, 2.7-fold variation in MPS) demonstrate substantial biological heterogeneity even in homogeneous samples, underscoring the need for both more measurements to characterize the heterogeneity and diverse samples to investigate its sources. The technical capacity for comprehensive phenotyping exists – high-density EMG arrays, continuous metabolic monitoring, protein turnover measurements – but applying these methods to demographically diverse samples reveals whether observed variance reflects universal biology or population-specific patterns. The barrier is not technical capability but study design choices. The examples presented in this article – 50–50 motor unit recruitment split, metabolic subphenotypes, 2.7-fold variation in muscle protein synthesis – demonstrate that substantial biological heterogeneity exists even within demographically homogeneous samples. This heterogeneity demands explanation, yet our current measurements often fail to provide it. The solution requires both more comprehensive measurement within individuals (to characterize heterogeneity in detail through hormonal tracking, muscle architecture assessment, training histories, genetic polymorphisms, functional outcomes) and demographically diverse samples (to test whether demographic factors predict the patterns). A study of 50 participants (25 male/25 female, ages 20–40) that applies comprehensive phenotyping can investigate whether sex, age, training status, or their interactions predict the heterogeneity – or whether it remains unexplained by these factors and reflects truly individual variation. We need both: comprehensive measurement depth and demographic sampling breadth, applied together. What is needed is reconceptualizing rigor: from demographic homogeneity as methodological virtue to demographic diversity as validity requirement. Practically, this means: (1) report demographic composition completely, including sex, age range, race/ethnicity, training status, and for women, menstrual cycle phase or hormonal status; (2) justify sample composition explicitly – if demographic restriction is necessary, acknowledge it as a validity limitation affecting interpretation; (3) when samples include diversity, analyse it – test whether demographic factors predict the heterogeneity your methods reveal; (4) include functional outcomes to validate those patterns that matter for physiology; and (5) design studies to investigate heterogeneity rather than average over it. This does not require massive sample sizes. A study of 50 participants (25 male/25 female) spanning ages 20–40 can test sex and age effects without increasing cost over a homogeneous sample. Metwally et al.’s innovative CGM approach would have been strengthened by doubling their sample to 64 participants with balanced sex distribution and menstrual cycle documentation – similar cost, vastly improved biological insight. Davies et al.’s systematic review reveals heterogeneity exists; future studies need only design to investigate rather than note it. A critical point warrants explicit emphasis: we are not measuring comprehensively enough within individuals, even in homogeneous populations. The examples presented – massive heterogeneity in motor unit recruitment, metabolic phenotypes and protein synthesis responses – demonstrate that our current measurements, while technically sophisticated, fail to explain the individual variation they reveal. The solution is not to choose between comprehensive individual phenotyping and demographic diversity, but to combine both. A 50-participant study with balanced sex distribution that applies comprehensive phenotyping – detailed training histories, hormonal status tracking, muscle architecture assessment, genetic polymorphism analysis, functional outcome measures – can investigate whether demographic factors (sex, age), training variables, genetic factors, or their interactions predict the heterogeneity. The technical capacity exists: continuous glucose monitoring, high-density EMG arrays, muscle architecture ultrasound, hormonal assays, genetic sequencing. These methods applied comprehensively to demographically diverse samples reveal whether variation reflects sex differences, age effects, training adaptations, genetic polymorphisms, or individual factors unrelated to these variables. We need both deeper measurement within individuals and broader sampling across demographics, applied together. Implementation faces real barriers that warrant acknowledgment. Increasing sample sizes to include diverse populations requires additional recruitment effort and funding. Analysing demographic subgroups demands statistical power that small studies cannot achieve. Training the scientific workforce to recognize and investigate diversity-related questions requires changes in graduate curricula and mentorship practices. Importantly, achieving meaningful diversity in samples requires diversity in the scientific workforce itself – investigators from underrepresented backgrounds are more likely to recognize and investigate diversity-related research questions (Hofstra et al. 2020). These barriers are genuine and require institutional commitment: funding agencies must allocate resources for larger, more diverse samples; journals must value demographic analysis as rigorously as technical innovation; and universities must prioritize recruitment and retention of scientists from diverse backgrounds. The infrastructure exists, but implementation requires redistribution of resources and reconceptualization of scientific priorities. Funding agencies and journals should require justification for demographic restriction and treat diversity deficits as validity concerns requiring response during review, not afterthoughts for discussion sections. NIH formalized this expectation in 2014 with policies requiring consideration of sex as a biological variable (Collins & Tabak 2014), later expanding these requirements through comprehensive rigor and reproducibility guidance (National Institutes of Health 2015). This guidance mandates that grant applicants address relevant biological variables – including sex, age and other demographic factors – in experimental design, analysis and reporting. A recent Government Accountability Office assessment of NIH-supported animal research identified incomplete reporting of biological variables as a persistent barrier to reproducibility and translatability (U.S. Government Accountability Office 2024). NIH now provides extensive training resources to support implementation, including ORWH educational modules on integrating sex as a biological variable throughout the research process (Office of Research on Women's Health 2025). Demographic reporting should be as routine as methodological reporting, with the same expectation of transparency and justification. Physiology stands at a crossroads. Our capacity to measure biological phenomena has advanced dramatically, revealing substantial individual heterogeneity across motor, metabolic, muscular and cardiovascular systems. Yet our sampling practices ensure we cannot understand this heterogeneity. The dual diversity crisis – demographic homogeneity preventing investigation of biological heterogeneity – represents a field-wide validity threat. The examples presented span 2024–2025 publications, demonstrating this is not a historical problem but a current crisis demanding immediate attention. The path forward requires no technical innovations, only different priorities: treating demographic diversity as essential to validity rather than peripheral to generalizability. Our most sophisticated methods reveal that biology is heterogeneous. Our sampling practices should enable us to understand why. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. The author has no conflicts of interest. Sole author. AI was used to improve English language clarity and readability. All scientific content, conceptual development, and intellectual contributions are solely the work of the author. None.