{"paper_id":"049cf182-9ded-4e06-a99b-45c60cbb5e1d","body_text":"1 \n \nThe Female Biomarker Challenge: Sex-\nSpecific Network Robustness Constrains \nBiological Age Estimation and \nGeroscience Trial Design \nAngus Silas Harding*, Jim Coward, Tianhai Tian \n \nDr Angus Silas Harding, PhD (corresponding author) \nBH Biotech Pty Ltd \nmailto:angus.harding@bhbiotech.com.au \n \nDr Jim Coward, MD-PhD \nAssociate Professor \nICON Cancer Care \nmailto:drjcoward@icon.team \n \nDr Tianhai Tian, PhD \nAssociate Professor \nSchool of Mathematics \nMonash University \nTianhai.Tian@monash.edu \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n2 \n \nHighlights \n- Sparse biomarker panels drawn from standard clinical pathology tests estimate \nbiological age that outperforms chronological age in predicting mortality and all \nfourteen age-related diseases examined — supporting their use as cost-effective \nsurrogate endpoints for anti-aging clinical trials. \n- Males and females require different biomarker panels, and male biological age is \nsubstantially more sensitive to both mortality prediction and lifestyle interventions — \nmaking older males the optimal proof-of-concept cohort for geroscience trials seeking \nmaximum signal-to-noise at minimum cost. \n- The reduced sensitivity of female biological age is consistent with greater female \nphysiological network robustness, and represents an urgent, solvable measurement \nproblem: developing validated female-specific biomarker panels should be treated as \nan independent research priority to enable mixed-sex trial designs to be both \nadequately powered and cost-effective. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n3 \n \nAbstract \nA current impediment to bringing anti-aging therapies to market is the lack of accepted \nclinical endpoints that fit within reasonable trial time horizons and budgets. Recent theoretical \nmodels predict that sparse sampling of interconnected physiological subsystems can capture \nthe essential dynamics of aging, suggesting that sparse biomarker panels could serve as \nsurrogate endpoints for geroscience clinical trials. Here, we test this prediction using \nNHANES 1999–2018 data linked to the National Death Index. To overcome variable dropout \ncaused by between-subsystem collinearity, we developed a two-stage dimensionality \nreduction architecture: Generalized Additive Models first compress each multi-variable \nsubsystem into a single non-linear mortality risk score, which is then integrated via Levine's \nbiological age algorithm. The resulting biological age estimates outperformed chronological \nage in predicting mortality and all fourteen age-related diseases examined, and detected the \neffects of diet, sleep, and physical activity on biological aging. Sex-stratified analysis revealed \nthat the mortality sex gap penetrates to every physiological subsystem measured, with males \nand females requiring different biomarker panels — consistent with sex-specific differences \nin physiological network topology. Critically, male biological age was substantially more \nsensitive to both mortality prediction and lifestyle interventions than female biological age, a \nrobustness–sensitivity trade-off predicted by network resilience theory. These findings carry \ndirect implications for trial design: older males currently offer the most favourable signal-to-\nnoise ratio for proof-of-concept geroscience trials using standard pathology tests, while the \ndevelopment of validated female-specific biomarker panels — capable of resolving the more \ndistributed aging signal imposed by greater female physiological robustness — should be \ntreated as an urgent and independent research priority. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n4 \n \nKey Words \nNetwork Physiology, Biological Age, Sparse Sampling, Sex-Specific Aging, Physiological \nRobustness, Geroscience Clinical Trials, Biomarker Panel Design \nGlossary \n• NHANES — National Health and Nutrition Examination Survey. \n• GAM / GAMs — Generalized Additive Model(s). \n• GLM — Generalized Linear Model. \n• HR — Hazard Ratio (from Cox proportional hazards). \n• OR — Odds Ratio (from logistic regression). \n• AUC-PR — Area Under the Precision-Recall Curve. \n• AIC — Akaike Information Criterion. \n• BIC — Bayesian Information Criterion. \n• BioAge — Biological age estimate based on routine tests (Levine algorithm). \n• BioAge Advance — BioAge minus chronological age. \n• NLR — Neutrophil-to-Lymphocyte Ratio. \n• MLR — Monocyte-to-Lymphocyte Ratio. \n• LMR — Lymphocyte-to-Monocyte Ratio. \n• SIRI — Systemic Inflammatory Response Index. \n• PIR — Poverty Income Ratio \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n5 \n \nIntroduction \nA central challenge in systems biology is determining whether the global state of a complex, \ninterconnected system can be inferred from sparse measurements of its subsystems. This \nquestion is particularly acute for human aging, where the organism functions as a tightly \ncoupled network of physiological systems 1,2, each aging asynchronously and influencing the \ntrajectory of the others 3,4. Different organs accumulate damage at different rates, yet the \nfailure of one system accelerates the decline of others 4, while lifestyle, environmental factors, \nand chronic disease uniquely perturb the biological age of individual organs 4. The \nobservation that organ-specific biological age predicts mortality better than chronological age \n4 confirms that aging is fundamentally a multi-system network phenomenon — but raises the \nthorny question of whether this distributed process can be captured without measuring the \nentire physiological state of the organism. \nRecent theoretical work suggests, remarkably, that it can. Three independent mathematical \nmodels of aging, each built on different biological assumptions, have converged on the same \nresult: sparse representations of interconnected subsystems are sufficient to reproduce \nGompertzian mortality dynamics. Nielsen and colleagues 5 modelled organisms as a \ncollection of connected subsystems where the failure of any subsystem can trigger cascading \nfailures in others; this sparse model not only reproduced Gompertzian mortality but also \nrealized the accelerating accumulation of failed subsystems that mirrors the exponential rise \nof non-communicable disease with age 6. Independently, Rutenberg and colleagues 7 showed \nthat random damage propagating through a scale-free biological network — where a few \nhighly connected hubs coexist with sparsely connected peripheral nodes — reproduced both \nGompertzian mortality and the age-dependent increase in frailty. A third model by Karin and \ncolleagues 8, in which senescent cells accumulate with age and suppress their own removal \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n6 \n \nthrough negative feedback, again quantitatively recapitulated the Gompertz law in both mice \nand humans. \nThis theoretical convergence is striking. Despite differing in their biological mechanisms, all \nthree models predict that the essential dynamics of aging can be captured by monitoring a \nrelatively small number of interacting subsystems. Moreover, Cohen and colleagues 9 \ndemonstrated empirically that robust physiological metrics can be derived from sparsely \nsampled networks, providing a direct methodological bridge between theoretical sparse \nmodels and practical biomarker measurement. Together, these results generate a testable \nhypothesis: sparse biomarker panels that sample across core physiological subsystems \nshould provide a robust estimate of biological age. If this hypothesis were true, then such a \nbiomarker panel could serve as a surrogate clinical endpoint to replace age and disability, \ngreatly simplifying trial design, shortening trial duration, and (crucially) greatly reducing the \ncost of anti-aging clinical trials 10,11. \nHowever, if aging dynamics are shaped by network topology, then organisms with different \nphysiological network architectures should exhibit different aging phenotypes — even if their \nrate of physiological aging is roughly equivalent. This prediction is directly relevant to \nhumans, specifically to the divergent mortality risk observed between men and women. \nWomen have outlived men consistently across populations and centuries, with documented \nsex mortality gaps in Sweden since 1751, Denmark since 1835, and England and Wales since \n1841 12,13. Recent work in network physiology has revealed a potential structural explanation: \nmale and female physiological networks differ in fundamental topological properties, with \nmale systems displaying higher small-world indices and greater modularity, while female \nnetworks are more densely connected overall and significantly more resistant to directed \nattack 14. Network resilience theory predicts that such topological differences should manifest \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n7 \n \nas differential robustness to age-related damage 15,16 — and, consequently, that males and \nfemales may require different biomarker panels to accurately estimate biological age. \nThis paper has two aims. Our primary aim is to test whether sparse biomarker panels selected \nacross physiological subsystems can estimate biological age to a standard suitable for use in \naging research — that is, whether they predict mortality, age-related disease, and respond \nappropriately to lifestyle interventions known to impact healthspan. Our secondary aim is to \ndetermine whether the sex differences predicted by differential network topology are \nempirically observable in aging biomarker selection, biological age estimation, and \nintervention sensitivity. If so, this carries immediate practical implications: clinical studies of \naging must account for sex-specific differences in network architecture when designing \nbiomarker panels, selecting interventions, and powering their analyses. \nResults \nSex-Specific Mortality and Disease Risk: Evidence for Distinct Aging \nPhenotypes \nGiven the well-established sex mortality gap 17-19, we began our analysis by assessing whether \nage-related mortality and disease risk differ between men and women within the NHANES \ndata set. We first compared the male and female samples to determine whether there were any \nobvious differences between the sexes. Apart from mortality risk during follow-up, where \nmales have the predicted elevated mortality risk compared to females, other known mortality \nrisk factors including age, education, poverty-to-income ratio, and ethnicity were not \nmeaningfully different between the male and female samples used in this study (Table 1). We \nthen performed Cox Regression to more precisely estimate the effect of biological sex on \nmortality risk. Crucially, biological sex altered the mortality risk profile of NHANES \nparticipants, with both pre- and post-menopausal females showing a significantly lower \nmortality risk compared to males (Table 2 and Table 3 ). \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n8 \n \nVariable Category Overall Female Male Effect \nSize \nn — 29276 14215 15061 — \nAge (years), mean (SD) — 48.3 (16.6) 48.5 (16.5) 48.1 (16.7) Negligible \nDeath during follow-up, n (%) \nNo 25852 (88.3) 12836 (90.3) 13016 (86.4) \nSmall \nYes 3424 (11.7) 1379 \n(9.7) 2045 (13.6) \nEducation, n (%) \n 20 \ns(Serum_Lactate_Dehydrogenase_ \ntransformed) 9 4.04 0.91 0.085 \nFemale Liver Serum_Alkaline_Phosphatase NA 1 NA NA Linear term \n(no smooth) \nMale \nLiver \ns(Serum_Albumin) 19 2.63 0.92 0.122 k -> 20 \ns(Serum_Lactate_Dehydrogenase) 9 1.98 0.93 0.332 \nMetabolic \ns(Glycated_Hemoglobin_ \ntransformed) 9 3.89 0.93 0.198 \ns(WHR_transformed) 9 4.89 0.94 0.622 \nFemale Rbc \ns(Mean_Cell_Volume_ \ntransformed) 9 4.32 0.9 0.112 \ns(RBC_Folate) 9 3.11 0.93 0.858 \nMale Rbc \ns(RBC_Folate) 9 2.46 0.91 0.015 \ns(Red_Cell_Count) 9 4.3 0.91 0.06 \ns(Red_Cell_Width) 9 5.03 0.93 0.335 \n \nTable 6. Basis dimension adequacy for GAM smooth terms. Each row summarises one sex-specific \nsubsystem GAM model. n-smooth = number of penalised smooth terms; k' = effective basis dimension \n(maximum flexibility permitted); edf range = range of effective degrees of freedom across smooth terms within \neach model (flexibility actually used after penalisation); max edf/k' = highest basis utilisation for any smooth \nterm in the model; k-index range = range of basis adequacy indices across smooth terms (values approaching 1.0 \nindicate the basis captures the underlying relationship without systematic residual patterning); Basis a dequacy p \n= number of smooth terms with p < 0.05 on the k-index test. Low p-values in the absence of high basis \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n15 \n \nutilisation indicate minor residual patterning that the smoothing penalty correctly chose not to fit, rather than \nbasis insufficiency. Default basis dimension was k = 10 (k' = 9) for all smooth terms except where noted. \nCardiovascular (both sexes), female metabolic, and male platelet subsystems were fitted as single-predictor \ngeneralised linear models (GLMs) and are excluded from this table. \n \nTo explore whether the sex mortality gap is confined to population-level statistics or \npenetrates to individual physiological subsystems, we visually compared the mortality risk \nscores of males and females for each subsystem included in their respective BioAge \nassessments (Figure 2). Strikingly, males displayed a higher mortality risk score for every \nphysiological subsystem compared to women. This subsystem-level sex difference is \nconsistent with the hypothesis that female physiological networks are intrinsically more \nrobust to age-related perturbation, rather than being protected by a single organ system or \nhormonal mechanism. \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n16 \n \nFigure 2. Risk scores across physiological systems by chronological age for males and females. Logistic \nRegression (for single variable subsystems) and Multivariable Generalized Additive Models (GAMs: for multi -\nvariable subsystems) were used to calculate mortality risk scores for the following physiological systems: (a) \ncardiovascular, (b) electrolyte, (c) immune, (d) kidney, (e) liver, (f) metabolic, (g) red blood cell, and (h) platelet \n(males only). Risk scores were modeled as a function of system-specific biomarkers, stratified by sex, with \nmortality as the outcome variable. Mean risk scores are plotted across chronological age for males (blue) and \nfemales (orange), with shaded regions representing ±1 standard deviation. Notable differences in risk trajectories \nwere observed, with males generally exhibiting higher mortality risk scores across all systems. Note that the \nplatelet subsystem was not predictive of mortality in the female sample, therefore only the male curve is \ndisplayed. \nIn the second step, we combined the individual system risk scores to estimate Biological Age \nof the participants. Here, we combined the subsystem risk scores — now decorrelated by \nconstruction — into Levine's BioAge algorithm 28,29 without further variable dropout. Crucial \nto our goal of predictive biomarker preservation, compressing the multi-variable \nphysiological subsystems into a single mortality risk score maintained the positive predictive \npower of each subsystem in the final model, for both age and mortality, at a high-to-very-high \nstatistical significance (Table 7). Finally, mortality risk scores were inputted into Levine’s \nBioAge algorithm for estimating the biological age using male and female training and test \ndata sets 28,29. \nOutcome Physiological \nSystem \nMale Female \nEstimate p-value Estimate p-value \nMortality \nCardiovascular 4.34 2.08E-05*** 5.48 2.83E-08*** \nElectrolyte 3.62 0.00165** 5.49 2.16E-06*** \nImmune 4.06 2.86E-11*** 7.77 7.07E-06*** \nKidney 4.02 1.45E-12*** 5.20 1.82E-13*** \nLiver 3.03 3.20E-05*** 7.89 2.24E-07*** \nMetabolic 4.85 8.29E-08*** 5.87 0.000451*** \nRed Blood Cell 4.14 7.30E-13*** 10.23 3.04E-13*** \nPlatelet 8.41 1.10E-08*** --- --- \nAge \nCardiovascular 85 3.06E-72*** 118 6.17E-146*** \nElectrolyte 52 7.86E-27*** 80 1.25E-59*** \nImmune 8 0.00974** 40 7.68E-09*** \nKidney 20 2.68E-14*** 23 5.68E-11*** \nLiver 25 1.31E-14*** 64 6.26E-31*** \nMetabolic 102 4.93E-151*** 162 9.21E-142*** \nRed Blood Cell 66 2.64E-119*** 132 3.47E-98*** \nPlatelet 89 3.21E-49*** --- --- \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n17 \n \nTable 7. Logistic regression analysing the relationship between with subsystem risk scores (as the predictor \nvariables) and mortality or age (as the outcome variables). The strength of the interaction is indicated using the \nEstimate value, with statistical significance is shown by p-value: ***p < 0.001, **p < 0.01, *p< 0.05, , nsp ≥ 0.05 \n(not significant). The platelet subsystem is not predictive of female mortality (indicated by ---) and are therefore \nnot used in the female biological age estimate. \n \nWe used two complementary methods to assess BioAge performance relative to chronological \nage. First, we used a battery of performance metrics to compare BioAge to chronological age \nin predicting mortality. For both sexes, BioAge outperformed chronological age across \ndiscrimination, explained variance, and calibration metrics on both training and held-out test \ndata (Table 8). On the male test set, BioAge achieved an AUC-ROC of 0.85 versus 0.84 for \nchronological age, and an AUC-PR of 0.33 versus 0.30. Female results followed the same \npattern (AUC-ROC: 0.84 vs 0.83; AUC-PR: 0.26 vs 0.22). Notably, chronological age \nproduced a Matthews Correlation Coefficient of zero in all partitions, indicating it assigned all \nindividuals to the majority (survived) class at the 0.5 probability threshold, whereas BioAge \nachieved non-trivial classification in both sexes (male MCC = 0.25; female MCC = 0.14). \nMinimal degradation from training to test performance indicates the models generalise \nwithout overfitting. \nSex Data Test Metric Chronological Age BioAge \nMale \nTrain \nAUC_ROC 0.828 0.853 \nAUC_PR 0.314 0.372 \nMcFadden_R2 0.201 0.247 \nNagelkerke_R2 0.249 0.301 \nBrier_Score 0.064 0.061 \nLogLoss 0.225 0.212 \nF1_Score 0.958 0.959 \nMCC 0.000 0.236 \nTest \nAUC_ROC 0.831 0.846 \nAUC_PR 0.311 0.395 \nMcFadden_R2 0.206 0.241 \nNagelkerke_R2 0.257 0.297 \nBrier_Score 0.068 0.064 \nLogLoss 0.234 0.224 \nF1_Score 0.955 0.958 \nMCC 0.000 0.282 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n18 \n \nFemale \nTrain \nAUC_ROC 0.845 0.859 \nAUC_PR 0.237 0.269 \nMcFadden_R2 0.211 0.232 \nNagelkerke_R2 0.248 0.271 \nBrier_Score 0.046 0.045 \nLogLoss 0.167 0.162 \nF1_Score 0.972 0.972 \nMCC 0.000 0.130 \nTest \nAUC_ROC 0.830 0.846 \nAUC_PR 0.199 0.237 \nMcFadden_R2 0.181 0.206 \nNagelkerke_R2 0.212 0.239 \nBrier_Score 0.043 0.042 \nLogLoss 0.162 0.158 \nF1_Score 0.974 0.974 \nMCC 0.000 0.000 \n \nTable 8. Out-of-sample predictive performance of chronological age versus sparse-panel biological age \n(BioAge) for all-cause mortality. Logistic regression models were fitted on a 70% stratified training partition \nand evaluated on the held-out 30% test set without re-estimation. Discrimination was assessed via area under the \nreceiver operating characteristic curve (AUC_ROC) and precision-recall curve (AUC_PR). Explained variance \nwas quantified using McFadden's and Nagelkerke's pseudo-R². Calibration was assessed via Brier Score and log-\nloss, where lower values indicate better performance. Classification accuracy at a 0.5 probability threshold was \nevaluated using the F1 Score and Matthews Correlation Coefficient (MCC). An MCC of zero indicates the model \nassigns all observations to a single class. BioAge was derived from sex-specific sparse biomarker panels \ncompressed via generalised additive models into subsystem risk scores, then integrated using Levine's PhenoAge \nalgorithm. \n \nWe then estimated mortality hazard ratios using Cox proportional hazards regression with \nboth chronological age and BioAge as covariates (Table 9). When both predictors were \nincluded in the same model, BioAge displayed a significant hazard ratio in both males (HR = \n1.09, 95% CI: 1.08–1.10, p < 0.001) and females (HR = 1.10, 95% CI: 1.08–1.12, p < 0.001), \nwhile chronological age was rendered non-significant (males: HR = 1.00, p = 1.00; females: \nHR = 1.00, p = 0.99). This indicates that BioAge captures the age-associated mortality signal \nand provides independent predictive information beyond chronological age alone. \n \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n19 \n \nSex Variable Hazard \nRatio \nLower 95% \nCI \nUpper 95% \nCI \np-value \nMale \nChrolonlogical Age 1.00 0.99 1.01 1.00ns \nBioAge 1.09 1.08 1.10 < 0.001*** \nFemale \nChrolonlogical Age 1.00 0.98 1.02 0.99ns \nBioAge 1.10 1.08 1.12 < 0.001*** \n \nTable 9. Mortality hazard ratios from Cox proportional hazards regression including both chronological age and \nBioAge as covariates. Hazard ratios represent the change in mortality risk per unit increase in each predictor, \nadjusted for the other. 95% confidence intervals and p-values are reported. Models were fitted on the full cohort. \nns = not significant; *** p < 0.001. \n \nNext, we visually assessed BioAge performance (Figure 3). BioAge increased monotonically \nwith chronological age in both sexes (Figure 3a), with broadly overlapping trajectories. The \nfemale BioAge curve was marginally steeper, consistent with the sex-specific differences in \nsubsystem composition identified by the variable selection pipeline (Figure 2a). However, and \nin striking contrast to chronological age, for males, BioAge generates a sigmoidal mortality \nrisk curve with an asymptote approaching one (Figure 3b). Compare this to the chronological \nage, where male mortality probability did not show an asymptote, and the highest mortality \nprobability was slightly over 0.5 (Figure 1a). This suggests that BioAge, unlike chronological \nage, better captures the full mortality risk within the male – and to a lesser extent female – \nsample populations. Moreover, when plotted using the classic survival format (probability of \nsurvival vs log10BioAge) (Figure 3c), BioAge realizes the Type I survivorship curve that is \ntypical of human populations within developed countries 30,31. Again, this supports the \nhypothesis that our BioAge estimate better captures the full mortality dynamics of the \npopulation, despite our sample being cut-off at 79 years of age. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n20 \n \nFigure 3. Standardized biological age (BioAge) and its relationship to mortality risk and survival \nprobability. (a) Biological age (BioAge) vs chronological age: Standardized BioAge demonstrates a strong \nlinear relationship with chronological age across the lifespan, with no significant differences between males \n(blue) and females (orange). Mean BioAge values are shown with shaded regions representing ±1 standard \ndeviation. (b) Mortality probability vs Biological Age (BioAge): Logistic regression models reveal a sigmoidal \nrelationship between BioAge and mortality probability for males and females. Mortality risk asymptotically \napproaches 1 for males, indicating that BioAge captures the full mortality dynamics of the population, unlike \nchronological age, which showed a maximum mortality probability slightly above 0.5 (see Figure 1a). Shaded \nregions represent 95% confidence intervals. Crucially, males retain a higher mortality risk compared to females \nacross the entire Biological Age range. (c) Survival probability vs log10(BioAge): When plotted using the classic \nlog10 survival curve format, BioAge produces a Type I survivorship curve, with males and females exhibiting \nsimilar patterns. Type I survivorship curves describe populations with (i) high survivorship throughout the life \ncycle, (ii) few offspring, and (iii) a high investment in progeny, which is typical for humans in developed \ncountries such as the United States. Shaded regions represent 95% confidence intervals. Again, males have an \noverall lower survival probability compared to females across the range of Biological Age. \n \nCollectively, these data show that (i) physiological subsystem mortality risk scores can be \nused to generate BioAge estimates that significantly outperform chronological age in \npredicting mortality in both male and female populations; (ii) our BioAge estimates perform \nsimilarly on unseen data, arguing against over-fitting; (iii) our BioAge captures more of the \nmortality risk present within the male and female sample populations compared to \nchronological age ; (iv) GAMs-derived BioAge reveals the classic Type I survivorship curve \nexpected for human populations within developed countries, and finally (v) BioAge shows \nthat males are at higher risk of mortality compared to women, with an elevated male mortality \nrisk observed across all physiological subsystems measured. \nDoes Relative Biological Age Predict Mortality? \nWe now turn our attention to the suitability of using biological age to monitor the efficacy of \nanti-aging interventions. For this task, we focus on the concept of BioAge Advance 28,29: \nBioAge Advance = Biological Age – Chronological Age \nA positive BioAge Advance score indicates that the participant has experienced accelerated \naging and is biologically older than their chronological age, whereas a negative BioAge \nAdvance reveals that the participant has aged more slowly and is biologically younger than \ntheir chronological age. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n21 \n \nSex Model AUC-ROC AUC-PR Nagelkerke R2 \nMale \nChronological Age 0.83 0.31 0.25 \nBioAge Advance 0.58 0.19 0.05 \nChronological Age + BioAge Advance 0.85 0.38 0.30 \nFemale \nChronological Age 0.84 0.23 0.24 \nBioAge Advance 0.49 0.09 0.01 \nChronological Age + BioAge Advance 0.85 0.26 0.26 \n \nTable 10. Incremental predictive value of BioAge Advance beyond chronological age for all-cause mortality. \nLogistic regression models were fitted for chronological age alone, BioAge Advance alone (BioAge minus \nchronological age), and their combination. BioAge Advance alone performs poorly as expected — as a residual \nit captures deviation from age-expected biological state without the baseline age signal. However, the combined \nmodel outperforms chronological age alone across all metrics in both sexes, indicating that BioAge Advance \ncontributes independent mortality-relevant information not captured by chronological age. Discrimination was \nassessed via area under the receiver operating characteristic curve (AUC-ROC) and precision-recall curve \n(AUC-PR). Explained variance was quantified using Nagelkerke's pseudo-R². \n \nFirst, we investigated whether including BioAge Advance increases model performance in \npredicting mortality. For both males and females, chronological age plus BioAge Advance \nwas the best performing model for predicting mortality (Table 10). Next, we assessed the \nutility of BioAge Advance using mortality Hazard Ratios estimated by Cox Regression. \nCrucially, for both males and females, BioAge Advance showed comparable mortality Hazard \nRatios to chronological age for both male and female participants, with high statistical \nsignificance (Table 11). \nSex Variable Hazard \nRatio \nLower 95% \nCI \nUpper 95% \nCI p-value \nMale \nChronological Age 1.09 1.08 1.10 4.06E-130*** \nBioAge Advance 1.09 1.08 1.10 9.25E-65*** \nFemale \nChronological Age 1.10 1.09 1.11 1.91E-87*** \nBioAge Advance 1.10 1.08 1.12 7.92E-22*** \n \nTable 11. Both chronological Age and higher BioAge Advance increase the mortality risk in male and \nfemale participants. Statistical significance is shown by p-value: ***p < 0.001, **p < 0.01, *p< 0.05, , nsp ≥ \n0.05 (not significant). \n \nTo visually inspect the predictive power of BioAge on mortality, we plotted the survival \ncurves for men and women (Figure 4). For men, survival analysis of males based on whether \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n22 \n \ntheir BioAge-Advance is either positive (Biologically Older) or negative (Biologically \nYounger), reveals a clear survival advantage for the biologically younger male (Figure 4a). \nFor women, however, the story is more nuanced. While women who are biologically older do \nhave a higher risk of mortality, female participants must be +/- 4 years older or younger than \ntheir chronological age before a clear survival advantage between the older versus younger \ngroups emerges (Figure 4b). \n \nFigure 4. Survival probabilities and age-related disease outcomes are determined by biological age. \nKaplan-Meier survival curves for males (a) and females (b): participants were stratified by BioAge Advance, \nwith individuals classified as \"biologically older\" (red) having higher BioAge Advance than chronological age, \nwhile \"biologically younger\" (blue) have lower BioAge Advance compared to chronological age. Sh aded regions \nrepresent 95% confidence intervals. For males, biological age separation did not require adju stment to the \nclassification thresholds, while females required a ±4-year adjustment to differentiate confidence intervals. \nSurvival probabilities are consistently higher for biologically younger individuals, with the effect more \npronounced in males. \n \nTaken together, these data support the hypothesis that accelerated aging, measured using \nBioAge Advance, is predictive of mortality for both males and females. However, the BioAge \nAdvance estimate is a stronger predictor of mortality for males compared to females. \nDoes Relative Biological Age Predict Age-Related Disease? \nBecause researchers are developing healthy aging interventions to prevent age-related disease \nand disability, estimates of biological age must also measure the risk of age-related disease if \nthey are to be broadly useful in healthy aging clinical trials. A range of non-communicable \ndiseases are included within the NHANES data sets. However, for this study we selected \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n23 \n \ndiseases that showed significant age-association to focus on age-related pathologies, \nconsistent with the goal of measuring the human healthspan. \nTo test whether BioAge Advance does estimate the risk of non-communicable diseases, we \nassessed whether BioAge Advance predicts age-related diseases over-and-above \nchronological age. The power of BioAge Advance to predict age-related disease was \nestimated by calculating the Odds Ratio of BioAge Advance for each disease using logistic \nregression, with the disease as the outcome variable and both chronological age and BioAge \nAdvance as the predictor variables. As summarized in Table 12 and Figure 5, elevated BioAge \nAdvance increased the Odds Ratio for each disease in both Male and Female sample \npopulations, and this was highly significant for all age-related diseases. \n \n \nFigure 5. Forest plot summarizing the odds ratios (ORs) for the association between accelerated biological \nage advancement (positive BioAge Advance) and various disease outcomes in (c) biological males and (d) \nbiological females. Each row represents a disease, with the odds ratio (black square) and 95% confidence \nintervals (horizontal lines). Odds ratios greater than 1 indicate an increased likelihood of the disease in \nindividuals with a positive BioAge Advance (i.e., individuals who are biologically older than their chronological \nage). The reference line (OR = 1) indicates no association. Diseases with significant associations are highlighted \nwith asterisks, denoting levels of significance (*p < 0.05, **p < 0.01, ***p < 0.001). Diseases analyzed include \narthritis, anaemia (blood transfusion), coronary heart disease, stroke, and others, as detailed on the y -axis. \nAccelerated biological aging increases the risk for all diseases analyzed. \n \nSex Disease Odds Ratio Lower 95% \nCI \nHigher 95% \nCI \np-value \nMale arthritis 1.066 1.061 1.072 2.39E-153*** \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n24 \n \nblood transfusion 1.051 1.044 1.059 4.33E-45*** \ncongestive heart failure 1.089 1.074 1.105 2.91E-31*** \ncoronary heart disease 1.103 1.090 1.116 8.27E-58*** \nangina 1.086 1.070 1.101 2.44E-29*** \nheart attack 1.085 1.074 1.097 8.71E-51*** \nstroke 1.060 1.047 1.072 4.07E-22*** \nemphysema 1.074 1.058 1.090 2.26E-21*** \nthyroid problem 1.042 1.033 1.051 4.37E-19*** \nchronic bronchitis 1.030 1.021 1.039 7.77E-11*** \nliver condition 1.025 1.017 1.034 1.68E-09*** \ncancer 1.096 1.086 1.106 9.55E-87*** \nkidney failure 1.043 1.030 1.057 1.40E-10*** \ndiabetes 1.065 1.058 1.071 4.77E-93*** \nFemale \narthritis 1.083 1.078 1.088 3.22E-238*** \nblood transfusion 1.049 1.043 1.054 2.58E-70*** \ncongestive heart failure 1.076 1.058 1.095 1.71E-17*** \ncoronary heart disease 1.094 1.075 1.114 2.63E-23*** \nangina 1.066 1.051 1.082 2.41E-18*** \nheart attack 1.062 1.047 1.077 1.14E-17*** \nstroke 1.059 1.047 1.071 2.60E-24*** \nemphysema 1.066 1.048 1.084 1.54E-13*** \nthyroid problem 1.043 1.038 1.048 5.29E-67*** \nchronic bronchitis 1.022 1.015 1.029 9.64E-11*** \nliver condition 1.028 1.019 1.037 2.08E-09*** \ncancer 1.053 1.046 1.059 1.76E-55*** \nkidney failure 1.031 1.020 1.042 7.19E-08*** \ndiabetes 1.065 1.058 1.072 2.20E-88*** \nTable 12. Association between BioAge Advance and self-reported physician-diagnosed disease prevalence. \nOdds ratios were estimated from logistic regression models predicting each disease outcome from BioAge \nAdvance adjusted for chronological age, fitted separately by sex. Odds ratios represent the increase in odds of \ndisease per one-unit increase in BioAge Advance — the deviation of an individual's biological age from their \nchronological age. All associations remained significant at p < 0.001. \n \nBioAge Advance was significantly associated with all 14 physician-diagnosed conditions \nexamined in both sexes. These results indicate that accelerated biological aging, as captured \nby the deviation of BioAge from chronological age, is a consistent and robust risk factor for \nage-related disease across cardiovascular, metabolic, renal, hepatic, pulmonary, and neoplastic \nconditions. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n25 \n \nDoes Relative Biological Age Respond to Lifestyle Interventions? \nSleep \nCircadian rhythm disruptions in the form of insomnia are a known risk factor for accelerated \naging 32, and low-risk interventions for improving sleep, such as cognitive behavioural \ntherapy (CBT), have been extensively tested in clinical trials (for example, see 33). Cox \nproportional hazards regression revealed that insomnia was associated with a greater than \ntwofold increase in mortality risk independent of chronological age in both males and females \n(Table 13). \nSex Variable hazard_ratio lower_95 upper_95 p_value \nMale \nage 1.09 1.08 1.10 3.31E-67*** \ninsomnia 2.36 1.61 3.45 1.02E-05*** \nFemale \nage 1.10 1.08 1.11 9.33E-37*** \ninsomnia 2.99 1.84 4.85 9.11E-06*** \n \nTable 13. Mortality hazard ratios from Cox proportional hazards regression with chronological age and \ninsomnia status as covariates. Insomnia was coded as a binary variable (1 = insomnia, 0 = healthy sleep \npattern) derived from self-reported sleep questionnaire data. Hazard ratios represent the change in mortality risk \nper unit increase in each predictor, adjusted for the other. 95% confidence intervals and p-values are reported. \nModels were fitted separately by sex on the full cohort. *** p < 0.001. \n \nTo further analyze the link between circadian rhythm disruption and mortality risk, we \ngenerated survival curves for males and females comparing participants with insomnia to \nthose with healthy sleeping patterns. Males with insomnia showed a clear increase in \nmortality risk, with non-overlapping 95% confidence intervals (Figure 6a). Females with \ninsomnia also showed an elevated mortality risk compared to healthy sleepers, however the \nconfidence-intervals between female insomniacs and healthy-sleepers overlap, revealing that \nfemales may be less sensitive to the impact of insomnia than males, at least with respect to \nshort-term mortality risk (Figure 6b). \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n26 \n \n \nFigure 6. Association between insomnia and mortality risk (a, b) and BioAge Advance (c, d) in male and \nfemale participants. (a, b) Age-adjusted survival curves from Cox proportional hazards models fitted separately \nfor individuals with insomnia (blue) and healthy sleep patterns (red), with 95% confidence intervals shaded. (c, \nd) BioAge Advance (BioAge minus chronological age) by insomnia status. Group differences were assessed \nusing Wilcoxon rank-sum tests with Cliff's Delta effect size. Males with insomnia displayed significantly \nelevated BioAge Advance (Cliff's Delta = −0.25, small effect, p < 0.001), while the female difference was not \nstatistically significant (Cliff's Delta = -0.1, negligible effect, p = 0.07). \n \nWe then asked whether insomnia impacts biological age. For male participants, insomnia \nincreases biological age with high statistical significance (p < 0.001), albeit with a small \neffect size (Figure 6c). In contrast, female participants with insomnia did not display an \nelevated BioAge compared to healthy sleepers (Figure 6d). \nIn males, insomnia was associated with both elevated mortality risk (HR = 2.36) and \nsignificantly accelerated biological aging (Cliff's Delta = −0.25, p < 0.001), suggesting that \nthe male sparse biomarker panel captures pathways through which sleep disruption \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n27 \n \naccelerates physiological decline. In females, insomnia conferred a comparable mortality risk \n(HR = 2.99) but was not reflected in BioAge Advance (Cliff's Delta = −0.09, p = 0.17). This \ndissociation indicates that insomnia-driven mortality risk in females may operate through \nphysiological pathways not captured by the female sparse panel — consistent with the sex-\nspecific network topology identified by our variable selection pipeline, suggesting that the \nrobust female physiology may require additional biomarkers to fully capture circadian-\nmediated accelerated aging. \nDiet \nCalorie restriction is a proven method to extend lifespan in a range of animals 34, and a \nhealthy diet strongly correlates with many aspects of healthy aging 35. To assess whether diet \nquality impacts biological age, we first examined how chronological age modulates the \nrelationship between diet and mortality risk. We stratified male and female participants by age \ngroup and visualised predicted mortality risk as a function of Healthy Eating Index (HEI) \nscore. For both sexes, the protective effect of diet quality on mortality risk was most \npronounced in older adults (Figure 7a and 7b), motivating a focused analysis of participants \naged 40 and above. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n28 \n \n \nFigure 7. Association between diet quality and mortality risk (a, b), BioAge Advance (c, d), and categorical \ndiet comparison (e, f) in male and female participants. (a, b) Predicted mortality risk from logistic regression \n(age + HEI score) plotted against Healthy Eating Index score, stratified by age group (18 –39, 40–59, 60–79), \nwith GAM trend lines. Mortality risk decreases with improving diet quality, with the effect most pronounced in \nolder age groups. (c, d) Linear regression of BioAge Advance against HEI score in participants aged 40 and \nabove, with 95% confidence intervals shaded. Higher diet quality was associated with reduced BioAge Advance \nin both males (β = −0.047, R² = 0.01109, p < 0.001) and females (β = −0.035, R² = 0.0172, p < 0.001). (e, f) \nBioAge Advance by diet category, comparing unhealthy (HEI ≤ 40) to healthy (HEI ≥ 75) diets. Group \ndifferences were assessed using Wilcoxon rank-sum tests with Cliff's Delta effect size. Both males (Cliff's Delta \n= −0.31, small effect, p < 0.001) and females (Cliff's Delta = −0.40, medium effect, p < 0.001) with healthy diets \ndisplayed significantly lower BioAge Advance. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n29 \n \nWe next asked whether diet quality is reflected in BioAge Advance among older adults. For \nboth men and women, higher HEI scores were associated with a statistically significant \nreduction in BioAge Advance (males: β = −0.047, R² = 0.011, p < 0.001; females: β = −0.034, \nR² = 0.017, p < 0.001; Figure 7c and 7d). Comparing the healthiest (HEI ≥ 75) to the least \nhealthy (HEI ≤ 40) diets, both sexes showed significantly lower BioAge Advance in the \nhealthy diet group (males: Cliff's Delta = −0.31, small effect, p < 0.001; females: Cliff's Delta \n= −0.40, medium effect, p < 0.001; Figure 6e and 6f). Cox proportional hazards regression \nconfirmed that each one-point increase in HEI was associated with a ~2% reduction in \nmortality risk independent of chronological age, with near-identical effects across sexes (HR \n~ 0.98 in both males and females, p < 0.001; Table 14). \n \nSex Variable hazard_ratio lower_95 upper_95 p_value \nMale \nage 1.095 1.084 1.106 7.34E-76*** \nscore 0.984 0.977 0.991 1.02E-05*** \nFemale \nage 1.091 1.079 1.104 3.47E-50*** \nscore 0.981 0.973 0.989 3.45E-06*** \n \nTable 14. Mortality hazard ratios from Cox proportional hazards regression with chronological age and \nHealthy Eating Index (HEI) score as covariates. HEI scores were derived from NHANES dietary recall data \nusing the heiscore R package, with higher scores indicating greater dietary quality. Models were fitted separately \nby sex on participants aged 40 and above. Hazard ratios represent the change in mort ality risk per unit increase \nin each predictor, adjusted for the other. 95% confidence intervals and p -values are reported. *** p < 0.001. Data \nsource: NHANES 2005–2018 linked to the National Death Index. \n \nThe convergence of these three independent analyses — age-stratified mortality modelling, \ncontinuous BioAge Advance regression, and categorical diet comparison — provides robust \nevidence that diet quality is captured by BioAge Advance and represents a modifiable \ndeterminant of biological aging in both sexes. \nExercise \nPhysical activity in the form of aerobic and anaerobic exercise has been shown to decrease \nage-related disease risk and increase healthy lifespan 36. Consistent with the broader literature, \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n30 \n \nthe impact of exercise on mortality was age-dependent, with older adults showing the greatest \nreduction in mortality risk relative to younger participants (Figure 8a, b). \nTo maximise sensitivity, subsequent analyses were restricted to older adults (aged 60 and \nabove), the cohort showing the strongest exercise-mortality relationship. Even within this \nresponsive subgroup, linear regression revealed only a weak association between weekly \nphysical activity and BioAge Advance. The association was statistically significant but \nnegligible in magnitude for both males (β = −2.43 × 10⁻⁴, p = 0.0014, R² = 0.007) and females \n(β = -7.337 × 10⁻⁵, p = 0.215, R² = 0.00034; Figure 8c, d). The near-zero R² values indicate \nthat continuous variation in physical activity explains essentially none of the variance in \nBioAge Advance, and the slight impact of physical activity as measured by questionnaire did \nnot reach statistical significance for females. \nDichotomising participants guided by the WHO-recommended physical activity threshold \nrevealed a somewhat clearer picture. Older males who met an activity threshold of 600 MET-\nminutes/week or above displayed a highly significant reduction in BioAge Advance compared \nto sedentary males, albeit with a small effect size (Wilcoxon W = 214,527, p = 6.04 × 10⁻⁶; \nCliff's δ = 0.152, 95% CI [0.086, 0.218]; Figure 8e). Active older females also showed a \nsignificant reduction relative to sedentary females, however the effect size was negligible (W \n= 486,750, p = 5.32 × 10⁻⁵; Cliff's δ = 0.108, 95% CI [0.056, 0.160]; Figure 8f). Critically, \nCox proportional hazards modelling confirmed that physical activity retains a statistically \nsignificant independent inverse association with 10-year mortality after adjusting for age in \nboth sexes (males: HR = 0.9999 per MET-minute/week, p = 0.001; females: HR = 0.9999, p = \n0.023; Table 15)— a biologically meaningful signal that BioAge Advance only captures in \nmales. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n31 \n \n \nFigure 8. Physical activity is associated with reduced biological age acceleration in older adults. (a, b) \nEstimated 10-year mortality risk (derived from PhenoAge) plotted against weekly physical activity (MET-\nminutes/week) for males (a) and females (b), stratified by age group (18 –39, gold; 40–59, blue; 60–79, green). \nRegression lines are shown for each age stratum. (c, d) BioAge advance (biological age − chronological age \nresidual) plotted against weekly MET-minutes for older males (c; aged 40–79) and older females (d; aged 40–\n79). Red lines indicate ordinary least-squares regression fits. Linear regression revealed a statistically significant \ninverse association in older males (β = −2.43 × 10⁻⁴, p = 0.0014, R² = 0.007), but a small positive association of \nuncertain biological interpretation in older females (β = +6.50 × 10⁻⁵, p = 0.0015, R² = 0.002), likely indicating \nno meaningful effect (see below). (e, f) BioAge advance compared between sedentary and active older males (e) \nand older females (f), dichotomised at the WHO recommended threshold of 500 MET-minutes/week. Wilcoxon \nrank-sum tests indicated significantly lower BioAge advance in active individuals (males: W = 214,527, p = 6.04 \n× 10⁻⁶; females: W = 486,750, p = 5.32 × 10⁻⁵). Effect sizes estimated by Cliff's delta were small for males (δ = \n0.152, 95% CI [0.086, 0.218]) and negligible for females (δ = 0.108, 95% CI [0.056, 0.160]). \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n32 \n \nSex Variable Hazard \nRatio \nLower 95% \nCI \nUpper 95% \nCI p-value \nMale \nage 1.0838 1.0629 1.1051 0.0000 \nphysical activity 0.9999 0.9998 1.0000 0.0024** \nFemale \nage 1.0963 1.0692 1.1240 0.0000 \nphysical activity 0.9999 0.9998 1.0000 0.0138* \n \nTable 15. Cox proportional hazards model estimates for age and physical activity as predictors of 10-year \nmortality risk in adults over 60. Hazard ratios (HR) and 95% confidence intervals are shown for age (years) \nand weekly physical activity (MET-minutes/week) as co-predictors in sex-stratified Cox proportional hazards \nmodels. Age was a strong independent predictor of mortality in both males (HR = 1.086 per year, 95% CI [1.077, \n1.096], p = 1.06 × 10⁻⁷⁴) and females (HR = 1.090 per year, 95% CI [1.078, 1.102], p = 7.86 × 10⁻⁵¹). Physical \nactivity retained a statistically significant independent inverse association with mortality after adjusting for age \nin both males (HR = 0.9999 per MET-minute/week, p = 0.0024) and females (HR = 0.9999 per MET-\nminute/week, p = 0.0138). \n \nThus, BioAge Advance detects the impact of physical activity in older males at the level of a \nsmall effect but lacks sufficient sensitivity to capture the female response — even though the \nmortality benefit of activity in females is independently validated by the Cox model. Across \nall three lifestyle domains examined — sleep, diet, and exercise — male biological age \nshowed consistently greater responsiveness than female biological age. This pattern is \npredicted by network robustness theory: systems with greater topological robustness resist \nperturbation in both directions, limiting measurable responsiveness to beneficial interventions \nas well as conferring protection against harmful exposures 15,16. The sex difference in BioAge \nsensitivity is therefore not a limitation of the measure per se, but a reflection of underlying \nphysiological architecture. \nDiscussion \nOur primary aim was to test the theoretical prediction that sparse sampling across \nphysiological subsystems can capture the essential dynamics of aging. Three independent \nmathematical models — built on cascading subsystem failure 5, damage propagation through \nscale-free networks 7, and senescent cell feedback dynamics 8 — converge on the prediction \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n33 \n \nthat sparse representations are sufficient to reproduce Gompertzian mortality. Our results \nprovide direct empirical support for this prediction. \nUsing a two-stage dimensionality reduction architecture — GAMs to compress subsystem \nvariables into non-linear mortality risk scores, followed by integration via Levine's algorithm \n— we found that biological age estimated from sparse biomarker panels i) more fully \ncaptured population mortality dynamics, including the expected sigmoidal mortality curve \nand the Type I survivorship pattern typical of industrialized populations 30,31; ii) outperformed \nchronological age in predicting mortality; iii) outperformed chronological age in predicting all \nfourteen age-related diseases analyzed; and iv) displayed appropriate sensitivity to lifestyle \ninterventions known to impact aging. Crucially, our two-stage architecture preserved all \npredictive biomarkers in the final panel by resolving between-subsystem collinearity at the \ncompression stage rather than through variable elimination — maintaining the comprehensive \nsubsystem coverage that the theoretical models identify as essential. \nSex-Specific Network Architecture and the Mortality Sex Gap \nOur secondary aim was to test whether the sex differences in aging predicted by differential \nphysiological network topology are empirically observable at the level of biomarker selection, \nbiological age estimation, and intervention sensitivity. The persistent sex mortality gap — \ndocumented across centuries and populations 12,13 — was recapitulated within the NHANES \ndata, with adult males experiencing elevated mortality risk and an increased burden of non-\ncommunicable disease. Critically, this sex difference was not confined to population-level \nstatistics: males displayed elevated mortality risk scores across every individual physiological \nsubsystem analyzed, revealing that the sex mortality gap is embedded at the subsystem level \nof the physiological network. \nThe existence of a robust sex mortality gap raised the specter that males and females may \nrequire different biomarker panels for estimating biological age. Alas, our analysis, combined \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n34 \n \nwith historic studies 23-25, confirms this fear. For example, alterations in the platelet subsystem \npredicts mortality in males, but not females. Similarly, excess body fat is a risk factor for \nmales, but again, not for females. However, the most important difference identified in our \nstudy is the relative insensitivity of female biological age estimates in predicting mortality \nand the impacts of lifestyle interventions compared to male biological age. Separating males \naccording to those who are biologically older or younger than their chronological age is \nsufficient to see a clear difference in mortality risk. In contrast, females require an eight-year \nseparation in relative biological age (i.e., +/- 4 years) before a similar discrimination in \nmortality risk is observable. Physical activity, a known protective lifestyle intervention, has a \npositive impact on biological age in older males, but no measurable impact on the biological \nage for older females. Similarly, insomnia had a significant negative impact on male \nbiological age, but a comparatively subdued effect on female biological age. \nSince males and females age at equivalent rates yet experience markedly different mortality \noutcomes, the explanation must lie in how male and female physiology responds to \naccumulating age-related damage — that is, in the robustness of their respective physiological \nnetworks. \nAs outlined in the Introduction, male and female physiological networks differ in fundamental \ntopological properties: male systems display higher small-world indices and greater \nmodularity, while female networks are more densely connected and significantly more \nresistant to directed attack 14. Our data are consistent with the prediction that these \narchitectural differences produce divergent aging phenotypes. The observation that males \nshow elevated mortality risk across every physiological subsystem — rather than in one or \ntwo specific organs — points to a system-level vulnerability rather than organ-specific \npathology. Conversely, the reduced sensitivity of female biological age to both mortality \nprediction and lifestyle interventions is precisely what network theory predicts for systems \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n35 \n \nwith greater topological robustness: such systems resist perturbation in both directions, \nconferring protection against damage while simultaneously limiting measurable \nresponsiveness to beneficial interventions 15,16. \nTwo additional mechanisms likely modulate this network-level explanation. The first is \ngenetic: across tetrapod species, the sex bearing the shorter sex chromosome consistently \nshows reduced longevity 37-39. Whether this reflects increased vulnerability to recessive \nmutations on the homologous long chromosome 38 or deleterious accumulation on the short \nchromosome itself — the 'toxic Y hypothesis' 39 — remains unresolved. The second \nmodulating mechanism is hormonal. Male sex hormones appear to increase mortality risk \nwhile female sex hormones confer protection 40,41, as strikingly illustrated by the observation \nthat prepubescent castration in mice equalizes male and female lifespans 42. However, our \nage-stratified analysis revealed that the female survival advantage remained consistent across \nthe menopausal divide (HR<60 = 0.65 vs HR≥60 = 0.64), suggesting that fundamental sex \ndifferences in mortality risk are not directly coupled to menopause status. Nevertheless, \novarian hormones — particularly the protective effects of estradiol and the increased \ncardiovascular and metabolic risks following hormone depletion — contribute to sex-specific \naging trajectories, particularly frailty, and warrant investigation with direct menopausal status \nmeasurement and longitudinal hormone data. \nThe hypothesis that female physiological networks are intrinsically more robust than male \nnetworks receive strong independent support from across the lifespan. For one, males are at \nhigher mortality risk in utero than females 43-46, a phenotype which continues throughout early \nchildhood 47,48, and into old age 49. Second, women are more resistant to infections from \nparasites and viruses compared to males 50,51. Furthermore, women mount significantly \nstronger immune responses to vaccinations 50 and are far more likely to survive ischemic heart \ndisease (IHD), heart failure, and cancer than men 49. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n36 \n \nHowever, the physiological robustness of women is most convincingly illustrated in the \nfascinating study of Zarulli and Colleagues, who demonstrated that the female survival \nadvantage persists under conditions of extreme mortality such as famines, epidemics, and \nslavery 52. Strikingly, the female survival advantage is most evident in infants, with baby girls \nbetter able to survive harsh conditions than baby boys 52. \nGiven that our panels specifically measure physiological parameters rather than genetic or \nhormonal markers, differential network topology provides the most parsimonious explanation \nfor our observed sex differences in biomarker predictive power and intervention sensitivity. \nThe robustness-sensitivity trade-off we observed — where female biological age resists both \ndamage and beneficial perturbation — is a fundamental property of robust networks, not an \nartefact of panel design. However, this physiological robustness must be contextualized with \nthe well-documented impacts of menopause on musculoskeletal function 53. A striking \ndichotomy emerges: the aging male tends to remain physically robust while becoming \nphysiologically fragile, whereas the aging female remains physiologically robust while \nbecoming frail. Clearly this dichotomy is an oversimplification. Precisely how male and \nfemale aging trajectories are shaped by genetics, hormones, and other sex-specific variables \nremains a crucial but unanswered question. However, the existence of clear sex-specific \ndifferences in aging phenotypes strongly suggest that sex-specific dosing and/or intervention \nstrategies will be required to increase the healthspan of both males and females. \nIn conclusion, our results empirically validate the theoretical prediction that sparse sampling \nacross physiological subsystems can capture the essential dynamics of aging, while revealing \nthat sex-specific network architecture imposes fundamental constraints on biological age \nestimation. The robustness-sensitivity trade-off between male and female physiological \nnetworks — predicted by network resilience theory and confirmed across mortality, disease, \nand intervention analyses — represents a systems-level property that any aging biomarker \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n37 \n \nframework must accommodate. Practically, we advocate for the immediate use of standard \nclinical tests to estimate biological age, with the critical caveat that sex-specific differences in \naging dynamics, biomarker panel composition, intervention sensitivity, and statistical power \nmust be accounted for in study design. \nImplications for Clinical Trials \nOur primary motivation for undertaking this work was to develop cheap, reliable biological \nage endpoints for our own Phase 1 trials – trials that include older females and female cancer \nsurvivors. The discovery that female biological age is substantially less sensitive to both \nmortality prediction and lifestyle interventions was a most unwelcome revelation. \nWithin the context of facilitating cost-effective geroscience clinical trials, our findings suggest \nthe following practical path forward. \nOlder males as the proof-of-concept cohort. Male biological age, estimated from standard \nclinical pathology tests, shows clean separation between biologically younger and older \nindividuals at BioAge Advance greater or less-than zero, detects the effects of diet (Cliff's \nDelta = 0.31, small), sleep disruption (Cliff's Delta = 0.29, small), and physical activity \n(Cliff's Delta = 0.15, small), and predicts all fourteen age-related diseases examined. \nCrucially, all of this is achievable using tests available at any standard pathology laboratory, at \na cost accessible to small research groups. For a Phase 1 geroscience trial with the primary \ngoal of detecting a biological age signal — establishing proof-of-concept that an intervention \nmeasurably slows or reverses biological aging — older males currently offer the most \nfavourable signal-to-noise ratio with the smallest required sample at the lowest cost. To be \nclear, we are not arguing for excluding women from aging research. We are, however, \nintending to capitalize on the patient cohort where signal detection for anti-aging intervention \nefficacy is most reliable (i.e., older males), while in parallel developing cost-effective \nbiomarker panels for women. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n38 \n \nThe female panel problem is solvable but requires a focused and dedicated research \neffort. The current female panel detects dietary effects at a medium effect size (Cliff's Delta = \n0.40), confirming that female biological aging is measurable in principle. The problem is \nprecision and breadth: the panel cannot resolve the survival advantage conferred by BioAge \nAdvance unless a ±4-year separation is imposed, and it fails to detect exercise and sleep \neffects that are clearly present in the Cox mortality data. This gap is most parsimoniously \nexplained by the network robustness argument developed throughout this paper — female \nphysiology is more robust, therefore physiological aging is more distributed across \nsubsystems, with smaller per-subsystem effect sizes that our current sparse panel under \nsamples. We posit that closing this gap likely requires three categories of expansion. First, \nphysical performance measures — grip strength, gait speed, and the short physical \nperformance battery — capture the musculoskeletal decline that is a hallmark of the female \naging phenotype. Second, the inclusion of more physiological subsystems. Several spring \nimmediately to mind. For example, immune subsystem expansion beyond the inflammatory \nmarkers currently included is warranted, given that women mount stronger and more complex \nimmune responses than men, and the current panel likely undershoots this dimension of \nfemale physiology. In addition, hormonal and stress markers — at minimum estradiol, FSH, \ncortisol and others — are necessary to properly characterise the menopausal transition and its \ncontribution to accelerating biological aging trajectories in mid-to-late life, as well as the \nknown female susceptibility to anxiety and depression. Third, algorithmic improvement. We \ndeliberately undertook a less-is-more approach to biomarker selection to drive-down costs. \nThis worked for males but unfortunately failed for females. An alternative approach, and one \nthat we strongly advocate for, is the use of more sophisticated algorithms that can \naccommodate interacting variables, thereby widening the effective search space of \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n39 \n \nestablished pathology tests and thereby keeping clinical trials cost-effective and accessible to \nsmaller research groups like ours. \nThe bottom line is that the development and validation of female biological age panels that \nare sensitive to anti-aging interventions should be treated as an urgent and independent \nresearch priority. \nSample size implications. Investigators designing geroscience trials with biological age \nendpoints should account explicitly for the sex-specific effect sizes reported here and \nelsewhere. The sleep data provides a stark warning: the male Cliff's delta for the impact of \ninsomnia on BioAge (0.29) was statistically significant, while the female delta (0.09) was not. \nAchieving 80% power to detect a female-equivalent effect at that magnitude would require \napproximately four to eight times the male sample size, depending on the specific intervention \nand outcome. Until validated female-specific biomarker panels are available, we recommend \nthat mixed-sex geroscience trials should analyse males and females separately, while treating \nthe female biological age result as exploratory pending panel validation. Treating a trial with \nmixed-sex enrolment without accounting for the more robust female physiology risks \nsystematically underestimating the efficacy of anti-aging and healthspan-improving \ninterventions and potentially discarding therapies that have clinical utility. \nLimitations \nSeveral limitations of our study should be noted. First, key physiological systems known to \nplay crucial roles in aging were not included in our analysis due to insufficient data. Second, \nwe could not include several highly informative clinical markers (such as established immune \ncytokines and metabolic markers) due to the low number of patient data points for these \nhighly desirable biomarkers. Thus, there is considerable room for improving on our panels if a \nmore comprehensive set of systems and biomarkers are included. Third, while the NHANES \ndataset provides robust population-level data, it is cross-sectional rather than longitudinal, \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n40 \n \nlimiting our ability to track individual aging trajectories and preventing us from identifying \npotential mechanisms explaining the difference between male and female aging trajectories. \nFourth, our lifestyle intervention analyses relied on self-reported data, which do not fully \ncapture intervention effects. This is particularly relevant to physical activity, which is \nespecially challenging to capture using a questionnaire format. Finally, our study population \nwas limited to ages 18-79, potentially missing important aging dynamics in the oldest-old \npopulation. For example, we have likely underestimated the impact of biological sex on aging \nas these are most apparent in the oldest-old. \nAdditionally, while we interpret the differential sensitivity of male and female biological age \nthrough the lens of network resilience theory, our study does not directly measure network \ntopology. Direct confirmation would require simultaneous measurement of physiological \ncoupling structure and aging biomarkers within the same cohort. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n41 \n \nMaterial and Methods \nDetailed mathematical descriptions of the methods and libraries used, as well as the data \ncleaning, transformation, and variable selection strategy deployed, are provided in the \nsupplementary methods section. Code used to clean NHANES data and generate the figures \nand tables presented in the manuscript are publicly available at \nhttps://github.com/AngusHarding/harding-et-al-2026-biological-age . \nData Sets Used \nNHANES records ‘gender’; here we analyze male/female as biological sex based on \navailable fields and use ‘sex’ throughout. We reserve ‘gender’ for identity constructs not \nmeasured in NHANES. Biomarkers were selected the following standard pathology and \nclinical tests from the publicly available Continuous National Health and Nutrition \nExamination Survey (NHANES) survey: Blood Pressure (BPX), Body Measures (BMX), \nUrine Albumin & Creatinine (ALB_CR), Complete blood count with 5-part differential \n(CBC), Folate (FOLATE), Glycohemoglobin (GHB), High-Sensitivity C-Reactive Protein \n(HSCRP), Standard Biochemistry Profile (BIOPRO). The Medical Conditions questionnaire \n(MCQ), the Diabetes questionnaire (DIQ), and the kidney Kidney Conditions – Urology \nquestionnaire (KIQ_U), were used to assess the presence of non-communicable, age-related \ndisease. The Demographics (DEMO) questionnaire was used to identify participant sex, age, \nand pregnancy status (pregnant participants were excluded from analysis), while the Current \nHealth Status questionnaire (HSQ) was used to identify acutely ill participants (acutely ill \nparticipants were excluded from analysis). The publicly available NHANES mortality data \nsets provided the mortality and time-to-death data. \nThe impact of disrupted sleep was measured using data extracted from the Sleep Disorders \n(SLQ_J) Questionnaire from 2005-2014. Diet was assessed using the Healthy Eating Index \n(HEI) scores calculated using the publicly available package ‘heiscore’ 54 from National \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n42 \n \nHealth and Nutrition Examination Survey 24-hour dietary recall data. Physical Activity was \ndetermined using the Physical Activity Questionnaire (PAQ) from the years 2007-2018. \nMetabolic Minutes per Week was calculated following the recommendations outlined in \nAppendix 1: Suggested MET Scores Table, associated with the Physical Activity \nQuestionnaire. \nHazard Ratio \nTo investigate the effect of predictor variables (for example, age and sex) on mortality risk, \nwe calculated the Cox proportional hazards regression model using the coxph function from \nthe survival package in R. \nOdds Ratio \nWhere time-to-event data was absent, we estimated the Odds Ratio using logistic regression \nanalyses using the glm() function in R. \nRisk Score Estimation Using Generalized Additive Models (GAMs) \nGuided by approaches pioneered using network physiology 9, we first allocated biomarkers \ninto their respective physiological subsystems and then selected the variables from each \nsubsystem. However, variable drop-out occurred when subsystem variables were combined \ninto a single model. This was likely due to interactions between biomarkers, as expected in \ncomplex physiological systems 1. In the context of developing biomarker panels for clinical \ntrials, we believe that preserving all predictive variables in the final biomarker panel is \noptimal because it maximizes the search space for identifying efficacious interventions. Here, \nwe chose Generalized Additive Models (GAMs) as our variable reduction method for two \nreasons. First, many biomarkers have non-linear relationships with age and mortality, and \nunlike traditional regression, GAMs capture non-linear relationships 26,27. Second, by using \nGAMs, we can directly link variable selection and reduction to what we consider to be the \nmost important aging clinical outcome, mortality. Any anti-aging intervention should, in our \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n43 \n \nview, have the foundational goal of reducing the risk of premature death. Furthermore, within \nclinical trials, it is standard practice to exclude patients who are at a high risk of death. For \nthese reasons, we argue that aging biomarkers must directly inform clinicians about mortality \nrisk. Fortunately, this can readily be achieved when using GAMS by explicitly using mortality \nas the outcome variable. \nTo assess physiological subsystem-specific contributions to mortality risk, we employed \nlogistic models for subsystems containing a single biomarker using the base R glm() function, \nand Generalized Additive Models (GAMs) for subsystems containing two-or-more variables \nusing the mgcv package in R. Basis dimension adequacy was assessed for all GAM smooth \nterms via k-index testing 55 (Table 6). Maximum basis utilisation across all models was 76%, \nwith k-index values ranging from 0.89 to 0.97, confirming adequate basis dimensions \nthroughout. Where the default basis dimension (k = 10) proved insufficient (Serum Albumin \nin both liver models), k was increased to 20, resolving the limitation. Serum Alkaline \nPhosphatase in the female liver model was entered as a linear term based on the observed \nlinear mortality relationship. Subsystems with single predictors (cardiovascular, female \nmetabolic, and male platelet) were fitted as standard generalised linear models rather than \nGAMs. To visualize age-related trends in subsystem-specific risk, we plotted the mean risk \nscores with shaded areas representing ±1 SD for males and females using the R package \nggplot2. \nEstimating Biological Age (BioAge) \nThe BioAge algorithm is an algorithm designed to estimate biological age by incorporating (i) \nclinical biomarkers that reflect physiological function across various organ systems, plus (ii) \nchronological age 28. The Levine BioAge algorithm was implemented in R using the BioAge \npackage 29. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n44 \n \nBioAge Advance \nThe difference between an individual's BioAge and their chronological age, herein termed \n\"BioAge Advance\" serves as an indicator of accelerated or decelerated aging 29. \nModel Performance Evaluation \nTo evaluate the performance of logistic regression models across multiple datasets, we \ncalculated five key metrics: Area Under the Precision-Recall Curve (AUC-PR) 56, Akaike \nInformation Criterion (AIC) 57, Bayesian Information Criterion (BIC) 58, McFadden’s R-\nsquared 59, and Nagelkerke’s R-squared 60. The above metrics were computed using standard \nfunctions from the following R libraries (caret, pROC, PRROC, and fmsb). Results for each \nmetric were summarized in a tabular format for all datasets. \nEffect Size Estimation \nTo summarize male – female sample baseline comparability we reported standardized effect \nsizes (no hypothesis tests). Continuous variables were summarized as mean ± SD; categorical \nvariables as column-wise n (%). Calculations were unweighted and include NHANES non-\nresponders as an explicit level for education and PIR. \n• Continuous (Age): standardized mean difference (SMD, Hedges’ g). \nLet 𝑥ˉ𝑚, 𝑥ˉ𝑓and 𝑠𝑚, 𝑠𝑓be group means and SDs with sizes 𝑛𝑚, 𝑛𝑓. \nThe pooled SD 𝑠𝑝 = √\n(𝑛𝑚−1)𝑠𝑚2 +(𝑛𝑓−1)𝑠𝑓\n2\n𝑛𝑚+𝑛𝑓−2 . \nCohen’s 𝑑 = (𝑥ˉ𝑚 − 𝑥ˉ𝑓)/𝑠𝑝; Hedges’ correction 𝐽 = 1 −\n3\n4(𝑛𝑚+𝑛𝑓)−9; \nwe report 𝑔 = 𝐽 ⋅ 𝑑. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n45 \n \n• Binary (Death during follow-up): standardized difference of proportions. \nWith event rates 𝑝𝑚, 𝑝𝑓and pooled 𝑝 =\n𝑛𝑚𝑝𝑚+𝑛𝑓𝑝𝑓\n𝑛𝑚+𝑛𝑓\n, \nΔ𝑝 =\n𝑝𝑚−𝑝𝑓\n√𝑝(1−𝑝). \n• Multi-category (Education, PIR, Race/Ethnicity): Cramér’s V from the \nsex×category contingency table. \nWith chi-square statistic 𝜒2, total 𝑛, and table dimensions 𝑟 × 𝑘, \n𝑉 = √\n𝜒2\n𝑛⋅min⁡(𝑟−1, 𝑘−1). \n(Non-responders are included as a category.) \nInterpretation followed common thresholds: negligible if ∣ SMD ∣< 0.10⁡or ∣ Δ𝑝 ∣< 0.10, and \nnegligible for Cramér’s 𝑉 < 0.10(0.10 – 0.20 = small; 0.20 – 0.30 = moderate; ≥ 0.30 = \nlarge). These metrics are for descriptive balance only; substantive sex differences in \nmortality are evaluated with time-to-event models in the main analysis. \nCliff’s delta is a non-parametric effect size measure that provides an indication of the \nmagnitude of the difference between two independent groups 61. Cliff’s delta was calculated \nusing the cliff.delta() function in the R package effsize and classified as large, moderate, small \nor negligible based on the absolute value (abs) of the Cliff’s delta values as follows: \nabs(cliffs_delta) < 0.147 ~ \"negligible\", abs(cliffs_delta) >= 0.147 & abs(cliffs_delta) < 0.33 \n~ \"small\", abs(cliffs_delta) >= 0.33 & abs(cliffs_delta) < 0.474 ~ \"medium\", abs(cliffs_delta) \n>= 0.474 ~ \"large\". \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n46 \n \nAuthor Contributions \nAngus Harding (corresponding Author) \nStudy conception and design, data acquisition, data cleaning and analysis, data interpretation, \ncode generation, creation of figures and tables, manuscript writing, manuscript submission. \nThe data and analyses included herein were originally submitted as a Masters Thesis by \nAngus Silas Harding (Specialisation: Applied and computational mathematics) (Conferred: \n05-Feb-2025) Cost-Effective Biomarker Panels for Aging Clinical Trials, Monash University \nJim Coward \nStudy conception and design. \nTianhai Tian \nStudy conception and design, manuscript writing and review. \nClaude (Anthropic) assisted with for reformatting the article for bioRxiv Systems Biology, \nincluding drafting and editing text, and making insightful editorial suggestions. All scientific \nhypotheses, literature analysis, data interpretation, content, and conclusions are the sole \nresponsibility of the carbon-based authors. \nData Availability \nCode available at https://github.com/AngusHarding/harding-et-al-2026-biological-age \nAll data used in this study is publicly available. The publicly available NHANES data files \nwere downloaded from the NHANES web site \n(https://wwwn.cdc.gov/nchs/nhanes/Default.aspx). Publicly available NHANES mortality \ndata was downloaded from the CDC website (https://www.cdc.gov/nchs/data-\nlinkage/mortality-public.htm). The downloaded ascii files were converted to .csv data frames \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n47 \n \nusing the publicly available R package, available at (https://www.cdc.gov/nchs/data-\nlinkage/mortality-public.htm). All Python and R packages used for data cleaning, variable \nselection, and data analysis, are open-source and freely available. \nCompeting Interests \nThe author(s) declare no competing interests. \nReferences \n \n1. Ivanov PC. The New Field of Network Physiology: Building the Human Physiolome. \nFront Netw Physiol. 2021;1:711778. doi:10.3389/fnetp.2021.711778 \n2. Cohen AA, Ferrucci L, Fülöp T, et al. A complex systems approach to aging biology. \nNature Aging. 2022;2(7):580-591. \n3. Schaum N, Lehallier B, Hahn O, et al. Ageing hallmarks exhibit organ-specific \ntemporal signatures. Nature. 2020;583(7817):596-602. \n4. Tian YE, Cropley V , Maier AB, Lautenschlager NT, Breakspear M, Zalesky A. \nHeterogeneous aging across multiple organ systems and prediction of chronic disease and \nmortality. Nature medicine. 2023;29(5):1221-1231. \n5. Nielsen PY , Jensen MK, Mitarai N, Bhatt S. The Gompertz Law emerges naturally \nfrom the inter-dependencies between sub-components in complex organisms. Sci Rep. Jan 12 \n2024;14(1):1196. doi:10.1038/s41598-024-51669-5 \n6. Belikov A V . Age-related diseases as vicious cycles. Ageing Res Rev. Jan 2019;49:11-\n26. doi:10.1016/j.arr.2018.11.002 \n7. Rutenberg AD, Mitnitski AB, Farrell SG, Rockwood K. Unifying aging and frailty \nthrough complex dynamical networks. Experimental gerontology. 2018;107:126-129. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n48 \n \n8. Karin O, Agrawal A, Porat Z, Krizhanovsky V , Alon U. Senescent cell turnover slows \nwith age providing an explanation for the Gompertz law. Nat Commun. Dec 2 \n2019;10(1):5495. doi:10.1038/s41467-019-13192-4 \n9. Cohen AA, Leblanc S, Roucou X. Robust Physiological Metrics From Sparsely \nSampled Networks. Front Physiol. 2021;12:624097. doi:10.3389/fphys.2021.624097 \n10. Newman JC, Milman S, Hashmi SK, et al. Strategies and Challenges in Clinical Trials \nTargeting Human Aging. J Gerontol A Biol Sci Med Sci. Nov 2016;71(11):1424-1434. \ndoi:10.1093/gerona/glw149 \n11. Rolland Y , Sierra F, Ferrucci L, et al. Challenges in developing Geroscience trials. Nat \nCommun. Aug 19 2023;14(1):5038. doi:10.1038/s41467-023-39786-7 \n12. Mascitelli L, Pezzetta F, Sullivan JL. Why women live longer than men: sex \ndifferences in longevity. Gend Med. Dec 2006;3(4):341; author reply 342. doi:10.1016/s1550-\n8579(06)80221-4 \n13. Dorling D. Life expectancy: women now on top everywhere. Unequal Health. Policy \nPress; 2013:272-275. \n14. Barajas-Martínez A, Ibarra-Coronado E, Fossion R, et al. Sex Differences in the \nPhysiological Network of Healthy Young Subjects. Front Physiol. 2021;12:678507. \ndoi:10.3389/fphys.2021.678507 \n15. Carlson JM, Doyle J. Complexity and robustness. Proceedings of the national \nacademy of sciences. 2002;99(suppl_1):2538-2545. \n16. Kitano H. Biological robustness. Nature Reviews Genetics. 2004;5(11):826-837. \n17. Organization WH. Health of the elderly: report of a WHO Expert Committee [meeting \nheld in Geneva from 3 to 9 November 1987]. 1989; \n18. Myers GC. World statistical trends and prospects. Principles and Practice of Geriatric \nPsychiatry. 2002:87-89. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n49 \n \n19. Grundy E. Gender and healthy aging. Longer life and healthy aging. Springer; \n2006:173-199. \n20. Appiah D, Nwabuo CC, Ebong IA, Wellons MF, Winters SJ. Trends in age at natural \nmenopause and reproductive life span among US women, 1959-2018. Jama. \n2021;325(13):1328-1330. \n21. Furnival GM, Wilson RW. Regressions by leaps and bounds. Technometrics. \n2000;42(1):69-79. \n22. Tibshirani R. Regression shrinkage and selection via the lasso. Journal of the Royal \nStatistical Society Series B: Statistical Methodology. 1996;58(1):267-288. \n23. Dallmeier D, Denkinger M, Peter R, et al. Sex-specific associations of established and \nemerging cardiac biomarkers with all-cause mortality in older adults: the ActiFE study. \nClinical chemistry. 2015;61(2):389-399. \n24. Nakamura E, Miyao K. Sex differences in human biological aging. The Journals of \nGerontology Series A: Biological Sciences and Medical Sciences. 2008;63(9):936-944. \n25. Yang Y , Kozloski M. Sex differences in age trajectories of physiological \ndysregulation: inflammation, metabolic syndrome, and allostatic load. Journals of \nGerontology Series A: Biomedical Sciences and Medical Sciences. 2011;66(5):493-500. \n26. Chen C. Generalized additive mixed models. Communications in Statistics - Theory \nand Methods. 2000/01/01 2000;29(5-6):1257-1271. doi:10.1080/03610920008832543 \n27. Hastie TJ. Generalized additive models. Statistical models in S. Routledge; 2017:249-\n307. \n28. Levine ME, Lu AT, Quach A, et al. An epigenetic biomarker of aging for lifespan and \nhealthspan. Aging (Albany NY). Apr 18 2018;10(4):573-591. doi:10.18632/aging.101414 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n50 \n \n29. Kwon D, Belsky DW. A toolkit for quantification of biological age from blood \nchemistry and organ function test data: BioAge. Geroscience. Dec 2021;43(6):2795-2808. \ndoi:10.1007/s11357-021-00480-5 \n30. Demetrius L. Adaptive value, entropy and survivorship curves. Nature. \n1978;275(5677):213-214. \n31. Ketz AC, Storm DJ, Barker RE, Apa AD, Oliva‐Aviles C, Walsh DP. Assimilating \necological theory with empiricism: Using constrained generalized additive models to enhance \nsurvival analyses. Methods in Ecology and Evolution. 2023;14(3):952-967. \n32. Carvalhas-Almeida C, Cavadas C, Álvaro AR. The impact of insomnia on frailty and \nthe hallmarks of aging. Aging Clin Exp Res. Feb 2023;35(2):253-269. doi:10.1007/s40520-\n022-02310-w \n33. Barroso D, Hespanhol L, Siegloch ML, et al. Effect of minimal cognitive behavioral \ntherapy for patients with acute insomnia: A systematic review and meta-analysis. Sleep Med. \nOct 2024;122:171-176. doi:10.1016/j.sleep.2024.08.003 \n34. Lim GM, Maharajan N, Cho GW. How calorie restriction slows aging: an epigenetic \nperspective. J Mol Med (Berl). May 2024;102(5):629-640. doi:10.1007/s00109-024-02430-y \n35. Surugiu R, Iancu MA, Vintilescu Ș B, et al. Molecular Mechanisms of Healthy Aging: \nThe Role of Caloric Restriction, Intermittent Fasting, Mediterranean Diet, and Ketogenic \nDiet-A Scoping Review. Nutrients. Aug 28 2024;16(17)doi:10.3390/nu16172878 \n36. Goh J, Wong E, Soh J, Maier AB, Kennedy BK. Targeting the molecular & cellular \npillars of human aging with exercise. Febs j. Feb 2023;290(3):649-668. \ndoi:10.1111/febs.16337 \n37. Marais GAB, Gaillard J-M, Vieira C, et al. Sex gap in aging and longevity: can sex \nchromosomes play a role? Biology of sex differences. 2018;9(1):33-14. doi:10.1186/s13293-\n018-0181-y \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n51 \n \n38. Connallon T, Beasley IJ, McDonough Y , Ruzicka F. How much does the unguarded X \ncontribute to sex differences in life span? Evolution Letters. 2022;6(4):319-329. \ndoi:10.1002/evl3.292 \n39. Sultanova Z, Downing PA, Carazo P. Genetic sex determination, sex chromosome size \nand sex‐specific lifespans across tetrapods. Journal of evolutionary biology. 2023;36(2):480-\n494. doi:10.1111/jeb.14130 \n40. Fritz García JHG, Keller Valsecchi CI, Basilicata MF. Sex as a biological variable in \nageing: insights and perspectives on the molecular and cellular hallmarks. Open Biol. Oct \n2024;14(10):240177. doi:10.1098/rsob.240177 \n41. Huang Y , Li H, Liang R, Chen J, Tang Q. The influence of sex-specific factors on \nbiological transformations and health outcomes in aging processes. Biogerontology. Oct \n2024;25(5):775-791. doi:10.1007/s10522-024-10121-x \n42. Jiang N, Cheng CJ, Gelfond J, Strong R, Diaz V , Nelson JF. Prepubertal castration \neliminates sex differences in lifespan and growth trajectories in genetically heterogeneous \nmice. Aging Cell. Aug 2023;22(8):e13891. doi:10.1111/acel.13891 \n43. Wyllie J. Sex differences in infant mortality. Canadian Public Health Journal. \n1933;24(4):177-185. \n44. Sheiner E, Levy A, Katz M, Hershkovitz R, Leron E, Mazor M. Gender does matter in \nperinatal medicine. Fetal Diagn Ther. Jul-Aug 2004;19(4):366-9. doi:10.1159/000077967 \n45. Mondal D, Galloway TS, Bailey TC, Mathews F. Elevated risk of stillbirth in males: \nsystematic review and meta-analysis of more than 30 million births. BMC Med. Nov 27 \n2014;12:220. doi:10.1186/s12916-014-0220-4 \n46. Baines KJ, West RC. Sex differences in innate and adaptive immunity impact fetal, \nplacental, and maternal health†. Biology of Reproduction. 2023;109(3):256-270. \ndoi:10.1093/biolre/ioad072 \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n52 \n \n47. Alkema L, Chao F, You D, Pedersen J, Sawyer CC. National, regional, and global sex \nratios of infant, child, and under-5 mortality and identification of countries with outlying \nratios: a systematic assessment. The Lancet Global Health. 2014;2(9):e521-e530. \n48. Townsel CD, Emmer SF, Campbell WA, Hussain N. Gender Differences in \nRespiratory Morbidity and Mortality of Preterm Neonates. Front Pediatr. 2017;5:6. \ndoi:10.3389/fped.2017.00006 \n49. Ostan R, Monti D, Gueresi P, Bussolotto M, Franceschi C, Baggio G. Gender, aging \nand longevity in humans: an update of an intriguing/neglected scenario paving the way to a \ngender-specific medicine. Clin Sci (Lond). Oct 1 2016;130(19):1711-25. \ndoi:10.1042/cs20160004 \n50. Flanagan KL, Fink AL, Plebanski M, Klein SL. Sex and Gender Differences in the \nOutcomes of Vaccination over the Life Course. Annu Rev Cell Dev Biol. Oct 6 2017;33:577-\n599. doi:10.1146/annurev-cellbio-100616-060718 \n51. Nasiri MJ, Haddadi S, Tahvildari A, et al. COVID-19 Clinical Characteristics, and \nSex-Specific Risk of Mortality: Systematic Review and Meta-Analysis. Front Med \n(Lausanne). 2020;7:459. doi:10.3389/fmed.2020.00459 \n52. Zarulli V , Barthold Jones JA, Oksuzyan A, Lindahl-Jacobsen R, Christensen K, Vaupel \nJW. Women live longer than men even during severe famines and epidemics. Proc Natl Acad \nSci U S A. Jan 23 2018;115(4):E832-e840. doi:10.1073/pnas.1701535115 \n53. Lu L, Tian L. Postmenopausal osteoporosis coexisting with sarcopenia: the role and \nmechanisms of estrogen. Journal of endocrinology. 2023;259(1) \n54. Sarkar VRBHA. Score and Plot the Healthy Eating Index from NHANES Data. 2024; \n55. Wood SN. Generalized additive models: an introduction with R. chapman and \nhall/CRC; 2017. \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint \n\n53 \n \n56. Keilwagen J, Grosse I, Grau J. Area under precision-recall curves for weighted and \nunweighted data. PloS one. 2014;9(3):e92209. \n57. Akaike H. A new look at the statistical model identification. IEEE transactions on \nautomatic control. 1974;19(6):716-723. \n58. Schwarz G. Estimating the dimension of a model. The annals of statistics. 1978:461-\n464. \n59. McFadden D. Conditional logit analysis of qualitative choice behavior. 1972; \n60. Nagelkerke NJ. A note on a general definition of the coefficient of determination. \nbiometrika. 1991;78(3):691-692. \n61. Cliff N. Dominance statistics: Ordinal analyses to answer ordinal questions. \nPsychological bulletin. 1993;114(3):494. \n \n.CC-BY 4.0 International licenseperpetuity. It is made available under a \npreprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in \nThe copyright holder for thisthis version posted March 16, 2026. ; https://doi.org/10.64898/2026.03.12.711462doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}