VO2max ageing and all cause mortality in a global cohort of multi marathoners

Abstract

This study evaluated whether sustained participation in multi-marathon running is associated with elevated aerobic capacity, more gradual age-associated differences in VO₂max, and reduced estimated all-cause mortality risk. A global cross-sectional online survey collected self-reported data from 340 multi-marathoners (mean age = 52.2 years; 54.1% men) representing 24 countries, with an average of 121 lifetime marathon completions. VO₂max was estimated using wearable fitness device data and analysed by sex and age group. Estimates were compared against normative VO₂max trajectories from the FRIEND (Fitness Registry and the Importance of Exercise: A National Database) cohort. Mortality risk was modelled using published meta-analytic data indicating a 3.7% reduction in all-cause mortality per 1 mL/kg/min increase in VO₂max. Participants exhibited substantially higher VO₂max values than population norms across all age groups, with multiple age bands exceeding the 95th percentile threshold. Compared to normative data, multi-marathoners showed a more gradual age-associated pattern in VO₂max, particularly among older adults. Although men displayed higher absolute VO₂max, older women showed greater relative advantage. Estimated mortality risk was notably lower across the cohort. These findings suggest multi-marathoners maintain superior aerobic fitness into later life, supporting a model of healthy ageing; however, potential cohort effects warrant longitudinal research to clarify causality.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-52475-x.

Introduction

Multi-marathoning, the practice of completing multiple marathons, has emerged as a distinctive sub-culture within endurance sports¹. Unlike typical recreational marathon runners, who may participate in one or two events annually, multi-marathoners often complete hundreds of marathons across their athletic careers. This pattern of sustained, high-volume participation may lead to distinct physiological adaptations that prioritise efficiency at submaximal intensities, such as improved cardiovascular stability (i.e., the ability to maintain a steady heart rate and blood pressure during prolonged exertion), enhanced metabolic economy, and resistance to fatigue, over peak aerobic capacity alone. Supporting this, Milic et al. demonstrated that trained individuals maintain superior mechanical efficiency across a range of intensities, which may contribute to sustained aerobic performance in endurance athletes².While elite endurance performance is often associated with high VO₂max values, the demands of multi-marathoning are arguably more reflective of these submaximal adaptations than of peak aerobic capacity alone^3,4. Despite the growing popularity of this cohort, empirical research on the long-term physiological profiles and health implications of multi-marathon participation remains unexplored.

VO₂max, defined as the maximal rate at which the body can consume oxygen during intense exercise, reflects the integrated capacity of the cardiorespiratory system to deliver oxygen to the muscles and the muscles’ ability to extract and utilise it⁵. It remains a cornerstone measure of cardiorespiratory fitness and a well-established predictor of endurance performance and all-cause mortality^6,7.

While VO₂max does not directly reflect submaximal efficiency or running economy, it remains a valuable indicator of aerobic capacity and cardiovascular health, particularly in large-scale, observational studies like the present one, where direct assessments of metabolic cost are not feasible. In the context of multi-marathoners, VO₂max estimates offer a practical proxy for assessing aerobic fitness, even if they do not fully capture submaximal efficiency, mechanical economy, and fatigue resistance that likely underpinning sustained performance in this group. This study explores how the aerobic profiles of multi-marathoners differ from those of the general population, with particular attention to how these differences evolve across the ageing spectrum. By examining this unique endurance subgroup, the study provides a basis for exploring how long-term participation in high-volume aerobic training may relate to physiological ageing.

Extensive literature has established the physiological benefits of endurance training, including increased stroke volume, greater mitochondrial density in skeletal muscle, enhanced capillarisation, and increased skeletal muscle mass, all of which contribute to higher VO₂max^3,8. These adaptations not only enhance endurance performance but may also mitigate age-related declines in cardiovascular function, such as reductions in cardiac output or maximal heart rate^9,10. However, much of this research has focused on elite or recreational endurance athletes rather than high-volume, long-duration participants such as multi-marathoners. Whether these individuals experience enhanced or divergent physiological adaptations due to their unusual physical loads remains an open question.

Additionally, the relationship between training modalities and performance outcomes is complex and likely extends beyond purely physiological mechanisms. Hraste et al. demonstrated that integrated motor and cognitive approaches can enhance learning and physical performance in children, suggesting that cognitively engaged physical activity may foster neuromotor adaptations. While not specific to endurance athletes, this model may help explain potential cognitive benefits associated with sustained, goal-directed training such as multi-marathoning¹¹. In the context of multi-marathoners, who engage in repetitive, goal-directed activity across varied environments and conditions, such mechanisms may contribute to sustained participation and performance, particularly in older age. Research also suggests that physical training can bolster cognitive resilience, including reduced susceptibility to mental fatigue, which may further support consistent endurance performance over time¹². This perspective broadens the theoretical lens for interpreting the physiological patterns observed in multi-marathon participants.

To provide context for evaluating multi-marathoners, this study utilises the Fitness Registry and the Importance of Exercise: A National Database (FRIEND), a widely cited reference standard for normative VO₂max values across age and sex¹³. FRIEND provides a robust baseline for comparison and enables evaluation of whether multi-marathoners exhibit VO₂max profiles that deviate meaningfully from population norms. Furthermore, given the well-established gender differences in VO₂max, whereby men typically present higher absolute and relative values due to greater haemoglobin levels, cardiac output, and lean body mass¹⁴, a gender-stratified analysis is necessary. Some evidence also suggests that women may preserve VO₂max more effectively than men as they age, especially when training is maintained¹⁰, though findings remain inconsistent across endurance cohorts^9,15.

In addition to descriptive comparisons, understanding the health implications of elevated VO₂max in multi-marathoners is vital. Meta-analytic and large cohort studies consistently demonstrate that each 1 mL/kg/min increase in VO₂max is associated with an approximate 3.7% reduction in all-cause mortality risk^7,16,17. These associations provide a clinically meaningful interpretation of differences in cardiorespiratory fitness and offer a population-level perspective on the health benefits of multi-marathon participation.

The theoretical foundation for this study is supported by three distinct frameworks. First, the Cardiovascular Adaptation Theory proposes that repeated aerobic training leads to both structural and functional changes in the heart and vasculature, including greater skeletal muscle, cardiac efficiency and output, and enhanced oxygen transport capacity, which collectively raise VO₂max and preserve cardiovascular function over time^8,18. This theory suggests that multi-marathoners, due to their prolonged and repeated exposure to endurance stressors, such as high training volumes, sustained submaximal intensities, and cumulative fatigue, may exhibit superior cardiovascular adaptation compared to recreational athletes or sedentary individuals.

Second, the Cardiovascular Ageing Theory proposes that while VO₂max naturally declines with age, typically by 5–10% per decade beyond age 30, consistent aerobic training slows this decline by preserving maximal cardiac output and peripheral oxygen utilisation^9,19. According to this model multi-marathoners may preserve a higher VO₂max baseline and exhibit a more gradual age-associated decline in VO₂max, offering both performance and longevity advantages. While this theory is supported by longitudinal research, it remains unclear whether similar patterns are observable in cross-sectional comparisons among multi-marathoners.

Building on this, the Training Load and Longevity Hypothesis proposes that sustained high-volume endurance training not only elevates VO₂max but may also contribute to a decoupling of biological from chronological ageing in elite or ultra-endurance populations^20,21. Under this hypothesis, the physiological benefits of multi-marathon participation, particularly in older athletes, extend beyond performance maintenance to offer protection against age-related comorbidities and premature mortality. While some studies have raised concerns about cardiac strain associated with extreme endurance volumes, longitudinal evidence increasingly suggests that consistent high-volume training, such as that seen in multi-marathoners, confers durable cardiovascular benefits^22,23.

However, it is important to note that observed age-related patterns in aerobic fitness from cross-sectional data may also reflect cohort effects or survivor bias. For example, Berthelot et al. demonstrated that age-related trends in elite performance follow a biphasic curve of growth and decline that is observable at both the individual and population level²⁴. This pattern likely results from a combination of biological ageing and selection mechanisms, whereby older high performers represent a self-selected, resilient subgroup rather than a generalised effect of training on ageing trajectories. These insights are essential for interpreting the aerobic profiles of older multi-marathoners, who may differ systematically from the broader ageing population.

The present study was designed to investigate whether multi-marathon participation confers a sustained cardiorespiratory advantage over normative ageing patterns. Given the scale and real-world nature of the sample, the analysis relied on VO₂max estimates derived from self-reported wearable fitness devices, offering a practical, though indirect, estimate of aerobic fitness. While such estimates carry a typical error margin of approximately ± 10% compared to laboratory testing^25–27, this was accounted for by interpreting results in the context of that expected variability. These estimates were benchmarked against the FRIEND database and stratified by gender and age group. In addition to physiological comparisons, the study also modelled the association between VO₂max advantage and estimated all-cause mortality risk using established predictive models^7,16.

Based on previous literature and theoretical frameworks, we proposed the following cross-sectional hypotheses:

• H1: Multi-marathoners demonstrate significantly higher estimated VO₂max than age- and sex-matched population norms (50th percentile FRIEND data), reflecting group-level differences in cardiorespiratory fitness.

• H2: The association between age and VO₂max in multi-marathoners is more gradual than in normative datasets, suggesting a shallower age-associated trajectory.

• H3: Sex-based differences in VO₂max are observed, with men exhibiting higher absolute VO₂max than women, but older female multi-marathoners demonstrating relative greater advantage compared to normative values than their male counterparts.

• H4: VO₂max levels observed in multi-marathoners, relative to normative data, correspond to lower modelled all-cause mortality risk based on established VO₂max-risk prediction models.

Methods

Study design and participants

This cross-sectional study assessed estimated VO₂max in multi-marathon participants, with data collected between 1 January 2023 and 31 March 2024 via a global online survey hosted on Qualtrics. The survey captured demographic information (gender, age, and marathon completion count) and VO₂max estimates derived from wearable fitness devices.

Participants were recruited through collaboration with national and international multi-marathon clubs, running organisations and multi-marathon event organisers. These organisations acted as recruitment gatekeepers, distributing study invitations through annual general meetings, newsletters, websites, and social media platforms. Study invitations included a standardised description of the research and a direct link to the online survey, enabling participants to complete the survey voluntarily.

While key partner organisations were based in the United Kingdom and Ireland, recruitment outreach extended internationally through affiliated 100 Marathon Clubs and broader global multi-marathon networks. This approach facilitated access to a geographically dispersed population for which no central registry exists.

Eligibility criteria required participants to be adults (18 + years) of any gender who had completed, or were actively committed to completing, at least 100 marathons. This criterion ensured inclusion of highly experienced runners. The recruitment strategy aimed to capture diversity across age, gender, and nationality, reflecting the international nature of the multi-marathon community.

Informed consent was obtained from all participants prior to inclusion. Following data collection, responses were screened for plausibility based on established physiological thresholds for VO₂max. Predefined exclusion criteria were applied, and entries with missing or implausible VO₂max values, or incomplete key variables required for analysis, were removed to minimise the influence of device misreadings, reporting errors, or outliers unrepresentative of physiological capacity.

The study adhered to the Declaration of Helsinki and received ethical approval from Trinity College Dublin’s Faculty of Health Sciences Research Ethics Committee (FREC) (Reference: 231005).

Measures and data sources

VO₂max estimates were derived from participant-reported values obtained via commercially available wearable devices such as Garmin, Polar, and COROS. These devices typically use proprietary algorithms that combine heart rate data, GPS-derived pace, and user profile information (e.g., age, sex, weight, and training history) to estimate maximal oxygen uptake. Although the specific algorithmic details remain proprietary, validation studies and manufacturer documentation suggest that these estimates are based on models derived from submaximal exercise heart rate responses, typically during steady state running, and are calibrated against laboratory VO₂max testing in healthy adults^28–30.

For instance, Garmin devices employ Firstbeat analytics, which estimate VO₂max by analysing heart rate dynamics and running performance during outdoor sessions exceeding 10 min at moderate intensity³¹. Polar devices apply similar models using fitness tests or steady state running data validated through indirect calorimetry comparisons³⁰. A growing body of validation studies has reported reasonable levels of agreement between wearable-derived and laboratory VO₂max estimates, with typical deviations in the range of ± 5–10% under optimal conditions^27,32. This aligns with broader findings in the exercise science literature which emphasise that non-laboratory fitness assessments can yield meaningful and reliable data when properly validated. For instance, Padulo et al. highlighted the importance of rigorous reliability testing in isometric assessments, reinforcing that well-validated field methods can serve as reasonable proxies when laboratory testing is not feasible³³. However, accuracy can vary based on device brand, sensor quality, user compliance (e.g., tightness of wrist strap), and environmental conditions (e.g., GPS signal quality, terrain).

Given this variability, a ± 10% correction band was applied when interpreting and visualising VO₂max estimates relative to normative standards. This adjustment reflects the upper boundary of published error margins and was chosen to conservatively account for potential over- or underestimation due to algorithmic or behavioural inconsistencies.

Normative VO₂max reference values were drawn from the FRIEND dataset, which is widely recognised as a leading source of cardiorespiratory fitness benchmarks for use in academic and clinical research¹³. The FRIEND dataset provides sex-specific and age-stratified VO₂max percentiles based on direct measurement via cardiopulmonary exercise testing across a large population of healthy individuals.

To ensure methodological consistency, participants in the present study were categorised into decade-based age groups that mirrored those used in the FRIEND reference data (e.g., 30–39, 40–49, and so on). This facilitated valid cross-group comparisons between self-reported wearable-based VO₂max estimates from the study sample and corresponding population norms.

Two comparative thresholds were used in the analysis. The 50th percentile of the FRIEND data was taken as the normative average, representing a typical cardiorespiratory fitness level for each age and sex category. In contrast, the 95th percentile served as the benchmark for top performers (individuals with exceptionally high aerobic capacity relative to their age and sex). Comparing the multi-marathon cohort against both the average and the top-tier normative thresholds enabled assessment of whether the sample exhibited superior cardiorespiratory fitness relative to the general population.

Statistical analysis

All statistical analyses were conducted in R³⁴. Descriptive statistics, including means, standard deviations, and ranges, were calculated for participant characteristics such as age, gender, marathon count, and estimated VO₂max. Analyses were stratified by gender throughout.

Prior to inferential testing, normality assumptions were assessed using the Shapiro–Wilk test alongside evaluations of skewness and kurtosis. Variables failing to meet parametric assumptions were log-transformed and reassessed. If violations persisted, non-parametric tests were used. Levene’s test was applied to examine homogeneity of variance before any parametric group comparisons were conducted. Only complete cases were included in the final dataset.

To assess group-level aerobic capacity, one-sample, one-tailed Welch’s t-tests³⁴ were conducted against the 50th and 95th percentile values from the FRIEND dataset. Tests were initially planned across all decade-based age groups (20–79 years); however, groups with small sample sizes (n ≤ 10) were excluded from inferential analysis due to limited statistical power and increased risk of Type II error. As a result, inferential comparisons were conducted only for participants aged 30–69 years. However, all age groups were retained in visualisations and descriptive summaries to illustrate broader trends across the full age range. A Bonferroni correction was applied to account for multiple comparisons, yielding an adjusted significance threshold of α = 0.00625. To verify results without relying on parametric assumptions, permutation-based p-values (10,000 iterations) were also computed and reported for each group comparison.

To quantify the magnitude of observed differences, effect sizes were calculated using Glass’s ∆, which uses the standard deviation from the normative dataset (FRIEND) as the denominator. This metric expresses how many standard deviations above (or below) the general population mean the multi-marathoner cohort lies, allowing for standardised interpretation across age and gender groups.

A post hoc sample size calculation was performed to assess whether the available sample was sufficient for the planned inferential analyses. Power analysis was conducted using standardised mean differences (Glass’s Δ; range = 1.07–4.16), calculated as the difference between multi-marathoner group means and FRIEND normative means, divided by the FRIEND standard deviations. Using a Bonferroni-adjusted α = 0.00625 and one-tailed testing with 95% power, the minimum required sample sizes per group ranged from 5 to 36. Seven of the eight retained subgroups exceeded these thresholds (actual N = 16–68), confirming adequate power to detect large effects. The exception was the Men 30–39 group (N = 22), which fell below the required sample size for 95% power (N = 36). Power calculations were performed in R³⁴ using the pwr.t.test() function from the pwr package (version 1.3-0). Overall, these results indicate that the available sample size was sufficient to support the planned inferential analyses, while acknowledging that some subgroup comparisons may be more limited in statistical power and that findings should be interpreted within the constraints of a cross-sectional design.

To evaluate age-related VO₂max trends, linear regression models were fitted for men and women separately, with age predicting VO₂max. The goal was to test whether multi-marathoners exhibited a less steep age-associated trajectory in aerobic capacity than population norms. Resulting slopes were extrapolated from age 20 to 90 and compared against 50th and 95th percentile FRIEND regression trajectories. The slope coefficients represented cross-sectional estimates of age-related change, and visual comparison highlighted differences in patterns across the ageing spectrum.

To test the statistical significance of these trajectory differences, a Monte Carlo simulation approach (10,000 iterations) to estimate whether the age-related decline in VO₂max among multi-marathoners was significantly more gradual than that observed in the FRIEND dataset. This simulation-based method follows established practices for comparing regression parameters under uncertainty³⁵. Simulated slopes were drawn from synthetic datasets using the FRIEND slope as the null, with Gaussian noise added based on the residual standard deviation observed in the multi-marathoner models. P-values were defined as the proportion of simulated slopes exhibiting a more gradual decline than or equal to the observed multi-marathoner slope. A Bonferroni-adjusted alpha level was applied to determine significance.

To evaluate model robustness, a leave-one-decade-out cross-validation (LODCV) procedure was applied to the multi-marathoner regression models. Standard individual-level cross-validation techniques (e.g., k-fold or leave-one-out) were not applicable due to the dataset structure, in which each observation represented an age-decade group mean rather than individual-level data. LODCV preserves internal consistency by excluding one age group (e.g., 40–49 years) at a time, refitting the regression model on the remaining groups, and recording the resulting slope and R². This method allowed for assessment of model generalisability and sensitivity to subgroup inclusion, while respecting the aggregated nature of the input data.

To assess the robustness of these findings to potential measurement variability, a sensitivity analysis was conducted across assumed wearable error margins (± 5%, ± 10%, ± 15% and ± 20%). For consistency, the ± 10% threshold was selected as the primary standard, as it reflects conservative published estimates of wearable VO₂max validity^26,27,36. The Monte Carlo simulation procedure was repeated under each condition, and statistical significance was reassessed using a Bonferroni-corrected threshold to account for multiple comparisons.

To visualise the physiological advantage across the lifespan, percentage improvement plots were generated by comparing predicted VO₂max trajectories for multi-marathoners against the FRIEND 50th and 95th percentile reference lines across ages 20 to 90. This illustrated the relative physiological benefit retained by multi-marathoners over time.

Finally, to contextualise the findings in clinical terms, estimated VO₂max values were translated into predicted reductions in all-cause mortality risk. This was achieved by applying a linear conversion model derived from seminal meta-analyses, which estimate an approximate 3.7% decrease in mortality risk per 1 mL/kg/min increase in VO₂max^7,16,17. The model was applied to the mean VO₂max differences between multi-marathoners and the FRIEND 50th and 95th percentile reference values, stratified by age and gender. This allowed the observed fitness advantage to be interpreted in clinically meaningful terms, contextualising cardiorespiratory fitness improvements in relation to long-term health outcomes. However, the model assumes a constant linear relationship across the VO₂max range and does not account for potential nonlinearities at extreme values or in advanced age. As such, estimates (particularly in older adults) should be interpreted as directional rather than absolute and may partially reflect extrapolated VO₂max differences rather than a strictly proportional physiological benefit.

Analytical rigor and reproducibility

To ensure transparency and reproducibility, all data processing and statistical analyses were conducted using reproducible R scripts, with version control and session information archived. Analytical decisions, including the use of resampling techniques and conservative correction methods, were made to minimise bias and overestimation of significance. While VO₂max estimates derived from wearable devices are subject to inherent variability, the integration of error-modelling and sensitivity analyses strengthens the reliability of the conclusions.

To ensure consistency in self-reported data, a ± 1 year resolution was applied to age and a ± 1 marathon completion count was applied, accounting for minor reporting variances while preserving analytical precision.

This rigorous analytical approach provides a robust foundation for identifying superior aerobic capacity and a more gradual cross-sectional VO₂max decline across age, relative to normative population data.

Results

Participant characteristics

A total of 341 participants were recruited, representing 24 countries. One participant self-identified as a gender other than male or female and was excluded from gender-stratified analyses, yielding a final analytic sample of 340 participants (184 men [54.1%], 156 women [45.9%]). Summary statistics for age, marathon experience, and estimated VO₂max are presented in Table 1. VO₂max estimates were self-reported from wearable fitness devices (e.g., Garmin, Polar, COROS).

Table 1.

Participant characteristics (Mean/SD).

Characteristic	Total (N = 340)	Men (n = 184)	Women (n = 156)
Age (years)	52.22 / 9.92	52.24 / 9.95	52.23 / 9.89
Marathon count	120.00 / 149.27	118.42 / 145.82	120.65 / 149.39
VO2max (mL/kg/min)	47.57 / 6.98	47.63 / 6.94	47.51 / 6.94

Descriptive statistics

Table 2 presents the descriptive statistics of the study sample, stratified by gender and age group. Metrics include the number of participants (N), mean and standard deviation of VO₂max estimates, chronological age, and total marathon completions. These figures provide a foundational understanding of the aerobic fitness and running history of the cohort prior to comparative and inferential analyses.

Table 2.

Participant characteristics by gender and age group.

Gender	Age group	N	VO2MAX (Mean ± SD)	Age (Mean ± SD)	Marathon count (Mean ± SD)
Men	20–29	2	60.5 ± 9.2	27.5 ± 2.12	14.0 ± 1.41
Men	30–39	22	53.5 ± 6.0	35.1 ± 2.9	114.5 ± 144.4
Men	40–49	44	52.8 ± 5.0	45.1 ± 2.9	122.8 ± 149.9
Men	50–59	68	49.9 ± 5.5	55.1 ± 2.9	129.9 ± 161.1
Men	60–69	44	47.4 ± 6.2	65.1 ± 2.9	121.7 ± 159.4
Men	70–79	4	47.2 ± 9.3	72.8 ± 2.4	232.0 ± 166.0
Women	30–39	16	46.6 ± 8.4	35.1 ± 2.9	114.5 ± 144.4
Women	40–49	45	45.7 ± 5.6	45.1 ± 2.9	122.8 ± 149.9
Women	50–59	59	44.3 ± 5.7	55.1 ± 2.9	129.9 ± 161.1
Women	60–69	32	41.5 ± 4.5	65.1 ± 2.9	121.7 ± 159.4
Women	70–79	4	44.2 ± 7.3	73.8 ± 4.4	251.0 ± 296.0

Values are presented as Mean ± SD. N = number of participants.

Normality and homogeneity of variance

Normality was assessed for age, estimated VO₂max, and marathon completion count using the Shapiro–Wilk test, skewness (± 2), and kurtosis (< 3). Both VO₂max and age met the assumptions for normality, with Shapiro–Wilk p-values of 0.3561 and 0.7411, respectively, and skewness and kurtosis within acceptable ranges. However, marathon completion count did not meet normality assumptions (p = 7.05 × 10⁻²⁵), showing moderate skewness (− 0.44) and kurtosis (2.36). A log transformation was applied to the marathon count, and the transformed data subsequently passed the normality test and was used in analysis.

Homoscedasticity was evaluated using Levene’s test. No significant differences in variance were detected for VO₂max across gender (p = .78), or age group (p = .62), supporting the use of parametric statistical tests in subsequent analyses.

Comparison of multi-marathoners’ VO₂max to FRIEND norms

To evaluate aerobic fitness in multi-marathoners, estimated VO₂max values were compared against age- and gender-specific norms from the FRIEND dataset. Analyses focused on age groups 30–69 with adequate sample sizes (n > 10). Across all eight age-by-gender subgroups, multi-marathoners significantly exceeded the 50th percentile VO₂max norms (perm p < .001). In contrast, comparisons to the 95th percentile varied: men aged 60–69 and women aged 40–69 demonstrated significantly higher VO₂max than the 95th percentile (perm p < .001), while other groups were not significantly different.

Full results are presented in Table 3, including sample sizes, VO₂max means and standard deviations, normative reference values from the FRIEND dataset (50th and 95th percentiles), effect sizes calculated using Glass’s Δ, and permutation-based p-values based on 10,000 iterations.

Table 3.

Comparison of multi-marathoners’ estimated VO₂max with FRIEND 50th and 95th percentile norms by gender and age group.

Gender	Age group	N	VO2max (Mean ± SD)	FRIEND VO2max (50th / 95th)	Glass’s ∆ vs. 50th / 95th	Perm p (50th / 95th)
Men	30–39	22	53.5 ± 6.0	42.4 / 59.8	1.84 / -1.03	< 0.001 / 1.000
Men	40–49	44	52.8 ± 5.0	37.8 / 55.6	2.97 / -0.57	< 0.001 / 1.000
Men	50–59	68	49.9 ± 5.5	32.6 / 50.7	3.12 / -0.15	< 0.001 / 0.898
Men	60–69	44	47.4 ± 6.2	28.2 / 43.0	3.09 / 0.71	< 0.001 / <0.001
Women	30–39	16	46.6 ± 8.4	30.2 / 45.8	1.95 / 0.10	< 0.001 / 0.351
Women	40–49	45	45.7 ± 5.6	26.7 / 41.7	3.42 / 0.72	< 0.001 / <0.001
Women	50–59	59	44.3 ± 5.7	23.4 / 35.9	3.68 / 1.47	< 0.001 / <0.001
Women	60–69	32	41.5 ± 4.5	20.0 / 29.4	4.74 / 2.67	< 0.001 / <0.001

Groups with sample sizes ≤ 10 were excluded from statistical inference due to limited power and increased risk of Type II error. Effect sizes are reported as Glass’s ∆, calculated as the difference between the multi-marathoner group mean and the normative percentile, divided by the normative standard deviation.

While comparisons to the 50th percentile illustrate the magnitude of deviation from general population norms, the inclusion of the 95th percentile offers a more stringent and physiologically relevant benchmark. Notably, in three of the eight analysed subgroups (men 60–69, women 40–49, and women 50–69), multi-marathoners exceeded the 95th percentile. In others, their values closely approximated it. This pattern suggests that the observed effect sizes, though large, reflect alignment with the highest levels of age- and gender-matched fitness rather than statistical artefacts.

These effect sizes are notable when compared to improvements observed in structured physical intervention trials. For instance, a recent cluster-randomised study by Stojanović et al. demonstrated that even among younger participants, supervised training interventions led to significant but comparatively modest gains in fitness and body composition³⁷. This provides important context for interpreting the substantial fitness advantage observed in this highly trained older cohort.

Figure 1 provides a visual overview of VO₂max trends for multi-marathoners compared with the FRIEND 50th and 95th percentile lines. These plots depict age-associated VO₂max trends for men and women, illustrating how multi-marathoners maintain notably higher VO₂max across the lifespan. For completeness, VO₂max estimates for the 20–29 and 70–79 age groups are included in Fig. 1. However, due to limited sample sizes in these categories (n ≤ 10), no inferential statistics were performed for these groups. Their inclusion is intended for descriptive and illustrative purposes only.

Fig. 1.

Comparison of multi-marathoners’ VO₂max with FRIEND’s percentile norms across age groups and genders.

Age-related VO₂max trends: multi-marathoners vs. FRIEND norms by gender

LRM model fit was strong across all groups (R² = 0.81–1.00), with particularly high explanatory power for the normative datasets (R² ≥ 0.95) and multi-marathoner men (R² = 0.93). These models enable direct comparison of age-related VO₂max trends between multi-marathoners and population norms. The fitted regression lines reflect associations observed at a single time point rather than individual changes over time. Figure 2 illustrates these relationships, with projections extended to age 90 for illustrative purposes only.

Fig. 2.

Cross-sectional age-associated VO₂max trends.

Table 4 provides the corresponding regression equations and R² values for each group, enabling direct comparison of model strength and trend trajectories.

Table 4.

LRMs for VO₂max trends in multi-marathoners and FRIEND norms by gender.

Group	Men - LRM equation	Men - R2	Women -LRM equation	Women - R2
Multi-marathoners	y = -0.23x + 61.31	0.93	y = -0.15x + 51.58	0.81
95th percentile (Top performers)	y = -0.54x + 76.73	0.99	y = -0.61x + 66.4	0.99
50th percentile (Average performers)	y = -0.47x + 56.88	1	y = -0.37x + 42.8	0.95

LRM = Linear Regression Model.

The linear models showed that multi-marathoners had higher intercepts and more gradual negative slopes than the corresponding FRIEND regression lines, indicating higher estimated VO₂max across the age spectrum. In both men and women, the observed cross-sectional slopes were significantly less steep than those of the 50th and 95th percentile normative trajectories (p < .001 in all comparisons). These results were confirmed through Monte Carlo simulations and remained robust across all sensitivity analyses incorporating wearable VO₂max estimation error (± 5%, ± 10%, ± 15%, and ± 20%). While these findings do not imply longitudinal preservation, they reinforce the consistency of between-group differences in age-associated VO₂max patterns, even under substantial measurement uncertainty.

To assess potential overfitting in the multi-marathoner regression models, leave-one-group-out cross-validation (LOGOCV) was conducted by iteratively excluding each age group and refitting the models. This approach is more appropriate than individual-level k-fold or LOOCV methods in this context, as the models operate on group-level mean VO₂max values (i.e., one row per age group), limiting the applicability of standard cross-validation techniques. The LOGOCV procedure yielded consistent slopes and high R² values across iterations (R² = 0.88–0.96 for men; R² = 0.75–0.93 for women), supporting the stability and generalisability of the fitted regression lines despite the limited number of age-level data points.

Percentage VO₂max advantage of multi-marathoners over FRIEND norms

To quantify the relative fitness advantage of multi-marathoners, VO₂max improvement was calculated as a percentage difference from age- and gender-matched values in the FRIEND dataset. Figure 3 presents these comparisons separately for men and women, benchmarked against both the 50th (average performers) and 95th (top performers) percentiles.

Fig. 3.

Percentage VO₂max difference in multi-marathoners relative to FRIEND norms (Log Scale). The plots show the estimated percentage difference in VO₂max for multi-marathoners compared to the FRIEND 50th percentile (orange line) and 95th percentile (red dashed line), stratified by gender.

Relative to FRIEND normative values, multi-marathoners demonstrated up to a 300% higher estimated VO₂max in later age groups. Consequently, even modest absolute differences can translate into disproportionately large relative gains. This pattern should be interpreted with caution, as it may overstate the physiological advantage of multi-marathoners in later decades. Comparisons to the 95th percentile also revealed consistent advantages across all age groups, with the largest relative differences observed among older adults. These estimates remained robust within a ± 10% variability band, reflecting wearable-derived uncertainty.

Estimated reductions in all-cause mortality risk

To provide a clinically meaningful interpretation of the VO₂max advantage observed in multi-marathoners, values were converted into estimated reductions in all-cause mortality risk using a published linear model (3.7% per 1 mL/kg/min VO₂max). Figure 4 presents the modelled reductions across age groups for men and women, benchmarked against both the 50th and 95th percentiles from the FRIEND dataset. Predicted reductions increased with age, reflecting the widening gap between multi-marathoners and population norms. However, especially in older adults, these reductions should be interpreted as directional rather than literal, as the linear model may overestimate effects when extrapolated to the extremes of age or VO₂max. Even when compared with the FRIEND 95th percentile, notable predicted advantages were retained, and findings were robust across wearable VO₂max variability margins (± 10%).

Fig. 4.

Estimated reduction in all-cause mortality risk in multi-marathoners compared to FRIEND 50th percentile norms.

Discussion

Key discussion points

This study presents novel data on estimated VO₂max in a large, international cohort of multi-marathon runners. Compared to age- and gender-matched normative data from the FRIEND study, multi-marathoners demonstrate significantly greater aerobic capacity across a wide age range, with several subgroups exceeding the 95th percentile of population norms. Regression models revealed that multi-marathoners exhibit a less steep cross-sectional association between age and estimated VO₂max compared to normative datasets. While this may reflect training-related differences, it could also result from cohort differences or selection bias, particularly among older participants. As noted by Berthelot et al., performance trajectories in ageing athletes are often shaped by survivor effects, with only the most physiologically resilient individuals remaining active into later life²⁴. As such, these findings should not be interpreted as evidence of slowed physiological decline. Rather, they likely reflect a combination of training-related adaptations and long-term selection dynamics, wherein only the most resilient, genetically advantaged, or highly motivated individuals remain active at older ages. As Güllich and Emrich note, sustained elite performance often emerges from systemic selection pressures over time, including attrition, dropout, and differential survival³⁸. These processes likely contribute to the exceptional fitness profiles observed in this cohort. When translated into health outcomes, these VO₂max differences corresponded to substantial modelled reductions in all-cause mortality risk. Collectively, these results highlight the potential of multi-marathon participation as a model for high-functioning ageing and sustained cardiorespiratory fitness, while underscoring the need for longitudinal research to confirm underlying mechanisms.

Interpreting the results

These findings support H1, confirming that multi-marathoners exhibit significantly elevated VO₂max compared to age- and gender-matched population norms. Mean VO₂max estimates exceeded the 50th percentile of the FRIEND dataset in all age and gender subgroups and, in three age-groups also surpassed the 95th percentile, reflecting exceptional aerobic capacity in parts of the cohort. This supports previous literature indicating that endurance training induces beneficial cardiovascular adaptations^3,8, and extends this observation to the under-researched group of multi-marathoners. These results are consistent with the Cardiovascular Adaptation Theory, which proposes that long-term endurance training leads to both structural and functional cardiovascular benefits^8,18.

Notably, the study found that multi-marathoners exhibit a more gradual cross-sectional decline in age-associated VO₂max estimates compared to population-based slopes from the FRIEND dataset (H2). This pattern supports the Cardiovascular Ageing Theory^9,19, which suggests that sustained aerobic training mitigates the typical age-related deterioration in VO₂max by preserving cardiac output and peripheral oxygen utilisation. Additionally, research by Milic et al. has shown that trained individuals maintain superior mechanical efficiency across a range of exercise intensities, including uphill walking at iso-efficiency speeds². Such findings offer a mechanistic explanation for how long-term endurance training may support not only cardiovascular preservation but also more efficient movement patterns that contribute to sustained aerobic capacity in older age. While these findings are based on cross-sectional data and not direct longitudinal measurements, the observed slopes in multi-marathoners suggest preservation of aerobic capacity well into later life. Although longitudinal data is required to confirm causality, this preliminary evidence reinforces prior assertions that multi-marathon participation can help maintain high aerobic capacity into later life and may attenuate biological markers of cardiovascular ageing. It also should be noted that although VO₂max provides a useful benchmark of aerobic fitness, it does not account for submaximal efficiency or running economy, which likely contribute to the sustainability of multi-marathon participation, especially in older individuals.

Although the multi-marathoner regression models demonstrated strong fit, concerns about potential overfitting were addressed using leave-one-age-group-out cross-validation. This analysis confirmed consistent slope estimates and high R² values across iterations, reinforcing the stability of the age–VO₂max trajectories. In contrast, the exceptionally high R² values observed for the FRIEND normative regressions (≥ 0.95) likely reflect the smoothing techniques used to generate centile curves rather than true model precision. Because these norms are based on fitted population percentiles rather than raw individual-level data, standard cross-validation techniques were not applicable in that context.

Gender-stratified results revealed that, as expected, men displayed higher absolute VO₂max values than women, reflecting known sex-based physiological differences in haemoglobin concentration, cardiac output, and lean body mass¹⁴(H3). However, older female multi-marathoners demonstrated relatively greater VO₂max advantage compared to normative values than their male counterparts. This aligns with previous evidence suggesting that women may experience a slower decline in VO₂max with age when training is sustained^10,14, although other studies have reported inconsistent findings across endurance cohorts^9,15. These findings highlight the need for further research into sex-specific trajectories in multi-marathoners.

In addition to physiological adaptations, it is also possible that sustained multi-marathon participation supports broader cognitive and neuromotor functions. Previous research suggests that integrated motor and cognitive engagement during physical activity can enhance learning and performance¹¹. Given the repetitive, goal-directed, and often mentally demanding nature of multi-marathon training and competition, such adaptations may contribute to the resilience and functional preservation observed in older participants. Although these mechanisms were not directly assessed in the present study, they offer a valuable theoretical perspective on the interplay between cognitive and physical pathways in lifelong endurance athletes.

This interplay between physical training and cognitive resilience is further supported by findings from Migliaccio et al.¹², demonstrated that mental fatigue impairs reaction time in sportsmen and proposed that sustained physical training may confer resilience to such effects through improved attentional control.

The elevated VO₂max levels observed in multi-marathoners translated into substantial estimated reductions in all-cause mortality risk, particularly in older age groups (H4). These outcomes reflect the sustained aerobic fitness advantage of this cohort and align with meta-analytic models demonstrating a strong, linear association between VO₂max and longevity. They also support the Training Load and Longevity Hypothesis^20,21 which proposes that multi-marathon participation may modulate the ageing process and reduce risk of age-related disease. While the magnitude of the estimated risk reductions was most pronounced in older individuals, this trend underscores the increasing health impact of aerobic fitness with age, rather than suggesting absolute immunity. Taken together, the findings reinforce the potential of multi-marathon participation not just as a model for sustained athletic engagement, but as a vehicle for promoting long-term health and healthy ageing.

However, the magnitude of estimated reductions at older ages should be interpreted with caution. As normative VO₂max declines steeply with age, modest absolute differences can yield disproportionately large percentage gains and risk reductions, reflecting mathematical artefact more than substantial physiological enhancement. Additionally, our mortality risk model, while based on established dose–response relationships in general populations, may not extrapolate reliably to highly trained individuals. As Lear et al. note, the benefits of physical activity may plateau at high fitness levels, especially among those already selected for cardiovascular health³⁹. Consequently, our projections should be viewed as directional group-level indicators, not precise individual-level risk estimates.

An important consideration when interpreting these findings is the extent to which the observed elevated VO₂max values in multi-marathoners reflect physiological adaptations induced by sustained endurance training, versus the possibility that this cohort represents a unique phenotype predisposed to ultra-endurance activities. Our study lacked data on participants’ baseline fitness prior to multi-marathon participation, training history, injury incidence, or dropout rates, which limits causal inference on the mechanisms underlying these physiological differences. It remains unclear whether the exceptional aerobic capacity is primarily a consequence of long-term, high-volume training or selection bias favouring individuals with inherently superior cardiovascular profiles.

Practical implications

These findings suggest that multi-marathon participation may serve as a practical model for sustaining cardiovascular health into later life, with some participants exceeding the 95th percentile of normative VO₂max values. This reinforces the potential of high-volume endurance activity (outside elite sport) as a meaningful contributor to long-term health and fitness.

Crucially, this presents a compelling recruitment narrative for older adults seeking purposeful physical activity. By showcasing the physiological benefits of sustained aerobic training, multi-marathoning can help reframe ageing as a period of continued athletic and health potential.

To ensure accessibility and safety, especially for ageing participants, strategies should include age-appropriate training plans and adequate recovery protocols. Moreover, gender imbalances persist, with men still comprising most multi-marathon participants. Expanding participation among women, particularly in older age groups, may require inclusive event design, targeted outreach, and community-based support.

With these considerations in place, multi-marathoning holds promise as a scalable, inclusive model for promoting cardiovascular fitness, healthy ageing, and lifelong physical engagement.

Future research

To clarify the observed trends, future research should prioritise longitudinal designs to determine whether the more gradual age-related decline in VO₂max observed among multi-marathoners reflects genuine physiological preservation or is influenced by cohort effects or selection bias. While the current findings are based on estimated VO₂max, direct physiological assessments are also essential to corroborate and expand upon these estimates. Previous research has demonstrated discrepancies between wearable-derived and laboratory-based VO₂max values^26,27,36, underscoring the need for validation. Moreover, VO₂max alone does not capture other key determinants of endurance performance, such as running economy and submaximal efficiency, which are likely to underpin sustained multi-marathon participation. Laboratory-based evaluations of oxygen cost during steady-state exercise, biomechanical efficiency, and lactate threshold would provide a more comprehensive understanding of these adaptations. Additionally, sex-specific trends, particularly the relatively greater VO₂max advantage observed in older female participants, warrant further investigation. Psychosocial and motivational factors should also be explored to better understand the traits supporting long-term engagement in multi-marathoning.

Future studies should aim to include individuals who have discontinued multi-marathon participation, to better understand attrition patterns, barriers to sustained endurance training, and health outcomes in less-selective populations. Longitudinal designs tracking both active and former participants would offer greater insight into the full spectrum of physiological ageing trajectories and help address survivorship bias inherent in studies of currently active athletes such as this.

While the current findings are robust across multiple simulated error margins (± 5% to ± 20%), future studies should incorporate direct laboratory-based assessments to validate the observed trajectories. Wearable-derived VO₂max estimates, while scalable and practical for large samples, are subject to variability and may not capture finer-grained physiological adaptations. Integrating gold-standard testing protocols would enhance measurement precision and help clarify the true extent of cardiorespiratory advantage in ageing endurance athletes.

Future studies should aim to characterise baseline health, socioeconomic factors, and attrition patterns to disentangle training effects from survivorship or self-selection, as emphasised by Güllich and Emrich³⁸.

Beyond cardiorespiratory fitness, long-term endurance training may also confer cognitive benefits. Migliaccio et al. demonstrated that mental fatigue adversely affects reaction time in trained sportsmen, highlighting the interplay between physical exertion and cognitive control mechanisms¹². Given the sustained demands of pacing, environmental adaptation, and psychological regulation in multi-marathoning, it is plausible that these athletes develop enhanced tolerance to mental fatigue over time. This may contribute not only to sustained endurance performance but also to preserved executive function and broader cognitive resilience, particularly in older age. Although mental fatigue was not directly measured in this study, its potential role as both a confounder and an adaptation warrant further investigation. Future research should explore the neurocognitive implications of high-volume endurance activity, including its potential to mitigate age-related cognitive decline.

Prospective, longitudinal studies with richer phenotyping are essential to unravel causal mechanisms underlying the exceptional aerobic profiles observed in multi-marathoners. Incorporating detailed training histories, injury and dropout tracking, comprehensive physiological testing, and psychosocial assessments will deepen understanding of how sustained endurance activity shapes ageing trajectories and health outcomes. Such approaches will help distinguish true training adaptations from inherent phenotypic traits and inform targeted interventions for healthy ageing through endurance exercise.

Limitations

While the study aimed to maximise data accuracy and ecological validity, several limitations should be acknowledged. Its cross-sectional design captures physiological data at a single time point, limiting the ability to infer causal relationships or assess longitudinal changes in VO₂max. A longitudinal design would be required to determine how multi-marathon participation influences cardiovascular health over time.

VO₂max estimates were derived from wearable fitness devices. In analyses where statistical significance was assessed, a ± 10% correction was applied to improve comparability with laboratory-based standards, in line with published validation studies. However, this correction was not applied across the entire dataset. Device variability and usage conditions may have introduced measurement error, and results should be interpreted with appropriate caution.

Stratifying participants by age, gender, and performance level reduced sample sizes within some subgroups, which may have limited statistical power and increased the risk of overfitting. The exceptionally high R² values observed in the regression models (0.81–1.00), particularly those fitted to FRIEND normative data (R² ≥ 0.95), may reflect overfitting or over smoothing due to the use of smoothed percentile curves rather than raw individual-level data. While this approach improves visual clarity and trend estimation, it may artificially inflate model fit. Future work should consider applying additional cross-validation techniques (e.g., k-fold) to assess model generalisability, particularly when working with raw data. While our current study applied a leave-one-age-group-out approach for multi-marathoner regressions, broader cross-validation methods may be appropriate in datasets with sufficient granularity.

Self-reported data on training volume, marathon count, and demographics introduced potential for reporting bias and inaccuracy. Although patterns aligned with expectations and internal consistency checks were conducted, objective verification was not possible within this study’s scope.

The use of voluntary online recruitment introduces the potential for self-selection bias, as individuals who chose to participate may differ systematically from those who did not. In the context of multi-marathoners, this may reflect differences in engagement, health status, or motivation. Recruitment through established clubs and international multi-marathon networks was used to facilitate broad dissemination across a geographically and demographically diverse population, rather than targeting specific performance levels. Eligibility criteria were based on participation thresholds rather than performance metrics, reducing the likelihood of over-representation of higher-performing individuals. Nevertheless, the sample may not be fully representative of all multi-marathoners, and findings should be interpreted as cross-sectional associations within a self-selected cohort.

While participants represented 24 countries, the sample was skewed toward regions with established multi-marathon cultures, potentially limiting the generalisability of findings. Cultural, environmental, and socioeconomic factors in underrepresented regions may influence participation and physiological adaptation in ways not captured here.

The study focused on VO₂max as a key indicator of aerobic capacity, but this does not account for running economy or submaximal efficiency, which may be especially relevant to multi-marathon performance. Direct measurement of oxygen cost during steady-state exercise, as well as biomechanical and lactate threshold assessments, would provide a more complete physiological profile.

Percentage-based VO₂max improvements in older age groups may overstate the aerobic advantage of multi-marathoners due to the steep decline in normative reference values with age. Modest absolute differences can yield disproportionately high percentage gains, which reflect mathematical artefact more than substantial physiological enhancement.

Predicted reductions in all-cause mortality risk, while based on established VO₂max–mortality associations, should be interpreted with caution. Linear extrapolation may exaggerate benefits at older ages due to the steep normative decline in VO₂max and the potential for mathematical artefact when comparing against low reference baselines. These estimates are directional rather than predictive and would benefit from future validation using longitudinal or event-based outcome data.

The study did not include individuals who had previously engaged in, but discontinued, multi-marathoning. This introduces survivorship bias, as the analytic sample likely overrepresents individuals with above-average health, resilience, or access to training resources. As a result, the findings may not be generalisable to all multi-marathon runners or to those with intermittent or short-lived participation.

While VO₂max provides a valuable index of aerobic capacity, it does not fully capture efficiency-based adaptations that are highly relevant to marathon performance. Long-term endurance training can improve running economy (the oxygen cost of steady-state exercise) through biomechanical, neuromuscular, and metabolic refinements. These changes may support superior performance and resilience in multi-marathoners, even when VO₂max gains plateau. As such, the high aerobic capacity observed in this cohort likely interacts with other physiological adaptations not captured in the present study.

Another important limitation lies in the use of the FRIEND dataset as a normative benchmark. While FRIEND is robust and widely adopted, it represents a clinical population referred for exercise testing, which may differ systematically from our self-selected sample of highly motivated multi-marathoners. Differences in health consciousness, training history, socioeconomic status, and baseline health may partly explain the group-level VO₂max differences observed. Future research should aim to use more closely aligned comparator datasets—ideally longitudinal cohorts of endurance athletes or stratified normative data that better match the demographic and training profiles of multi-marathoners.

This study did not include data on participants’ baseline fitness prior to multi-marathon participation, training history, injury incidence, or dropout rates. Consequently, it remains unclear whether the observed elevated VO₂max values represent physiological adaptations resulting from sustained training or reflect a unique, pre-existing phenotype that predisposes individuals to multi-marathon participation. The cross-sectional design limits causal inference, and these gaps highlight the need for prospective longitudinal studies to disentangle training effects from selection bias within this specialised endurance cohort.

Finally, the study did not control for unmeasured confounding factors such as socioeconomic status, access to training resources, coaching, or underlying health conditions, all of which could have influenced both participation and physiological outcomes.

Despite these limitations, the study offers a strong foundation for understanding the cardiovascular profiles and subgroup characteristics of multi-marathoners and provides clear directions for future research.

Conclusion

This study provides evidence that multi-marathon participation is associated with enhanced aerobic capacity and more gradual age-associated differences in VO₂max when compared to normative population data. While VO₂max estimates were derived from wearable devices, the magnitude of observed differences, combined with theoretical alignment and subgroup trends, supports the plausibility of the findings.

Multi-marathoners, as a distinct endurance cohort, appear to maintain exceptional cardiorespiratory fitness across the lifespan, with potential implications for longevity and long-term health. However, given the cross-sectional design, these results should not be interpreted as evidence of attenuated physiological decline. Instead, they reflect group-level patterns that may arise from training effects, selection bias, or both.

The study’s strength lies not in proving causation but in documenting the physiological characteristics of this unique population and establishing a framework for future research. These results affirm the credibility of multi-marathon participation not only as a performance pursuit but as a viable model for high-functioning ageing. As such, multi-marathoning warrants greater recognition in both endurance research and public health discourse, with future studies expanding upon its unique physiological and behavioural profile.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Leo Lundy was involved in all aspects of the study, including conceptualization, methodology, data collection, analysis, interpretation, and writing of the original draft. Richard B. Reilly and Neil Fleming contributed to the conceptualization and methodological design of the study, provided supervisory oversight, and were actively involved in reviewing and editing the manuscript. All authors approved the final version of the manuscript and agree to be accountable for its content.

Funding

The authors received no specific funding for this work.

Data availability

Results of Vo2max data gathering is presented in a supplementary file.

Competing interests

The authors declare no competing interests.