Do women reduce the gap to men in ultra-marathon running?

Ethics

All procedures used in the study were approved by the Institutional Review Board of Kanton St. Gallen, Switzerland, with a waiver of the requirement for informed consent of the participants given the fact that the study involved the analysis of publicly available data.

Data sampling and data analysis

The data set for this study was obtained from the race website of the ‘Deutsche Ultramarathon-Vereinigung’ (DUV) (www.ultra-marathon.org). This website records all race results of all ultra-marathons held worldwide. Data of all competitors who ever participated in a 6, 12, 24, 48, 72 h, 6 days (144 h) and 10 days (240 h) ultra-marathon held worldwide between 1975 and 2013 were analysed. In time-limited races, athletes perform laps which are counted by lap counters or electronically. Any competitor is listed in the rankings as soon as she/he has completed one lap as minimum distance.

Statistical analysis

We used a multiple linear regression to analyse the gap between men and women (Table 1). To explore which variables may be accounted for, regression distances were graphically displayed against the variables (Fig. 1a, b). Due to the large amount of observations we used smoothing methods (i.e. loess if number of observations <1000 otherwise splines both implemented in the statistical software). The 95 % confidence regions are displayed and the polynomial fit for each ultra-marathon (UM) was added. We included the following variables in the model: sex, age at performance and calendar year of performance. To consider finishers who performed several races we included finisher as random variable in the model, although 48.9 % of the finishers in the data have only one finish. We justified including finisher as random variable by comparing the graphics of distance against age of the finisher with only one known finish with the finishers who have at least five finishes. Both graphs showed similar tendencies (graphs not shown). Visual inspection of Fig. 1a, b suggests using a cubic, quadratic and a cubic relation for age and year. To study the effect of sex we included also interactions. We also considered the heterogeneous variance of each UM-level. The final method was selected by Akaike information criterion (AIC) and visual inspection of the fitted values (Fig. 2a, b).

	Coefficient	Standard error	p value
6 h	55.7	0.13	<0.0001
Ultramarathon
12 h	34.4	0.24	<0.0001
24 h	87.0	0.35	<0.0001
48 h	166.3	1.33	<0.0001
72 h	219.6	6.23	<0.0001
144 h	463.5	3.99	<0.0001
240 h	776.9	18.36	<0.0001
Sex (female)	−5.6	0.31	<0.0001
Age
Age centered linear	−0.25	0.01	<0.0001
Age centered squared	−0.01	0.00	<0.0001
Age centered cube	0.00	0.00	<0.0001
Sex (female) × age
Age centered linear	0.00	0.03	0.902
Age centered squared	0.00	0.00	0.007
Age centered cube	0.00	0.00	0.907
Year
Year linear	−0.30	0.02	<0.0001
Year squared	0.00	0.00	0.357
Sex (female) × year
Year linear	0.12	0.05	0.007
Year squared	−0.01	0.01	0.122
Sex (female) × ultramarathon
12 h	−1.98	0.53	<0.0001
24 h	1.33	0.81	0.099
48 h	5.04	2.76	0.068
72 h	16.39	14.25	0.250
144 h	−19.44	8.41	0.021
240 h	−71.72	31.78	0.024
Ultramarathon 12 h × age
Age centered linear	−0.06	0.02	0.020
Age centered squared	−0.02	0.00	<0.0001
Age centered cube	0.00	0.00	0.046
Sex (female) × ultramarathon 12 h × age
Age centered linear	−0.02	0.05	0.678
Age centered squared	0.00	0.00	0.096
Age centered cube	0.00	0.00	0.401
Ultramarathon 24 h × age
Age centered linear	0.19	0.04	<0.0001
Age centered squared	−0.04	0.00	<0.0001
Age centered cube	0.00	0.00	0.031
Sex (female) × ultramarathon 24 h × age
Age centered linear	−0.13	0.09	0.131
Age centered squared	−0.01	0.00	0.002
Age centered cube	0.00	0.00	0.803
Ultramarathon 48 h × age
Age centered linear	0.63	0.14	<0.0001
Age centered squared	−0.07	0.01	<0.0001
Age centered cube	0.00	0.00	0.749
Sex (female) × ultramarathon 48 h × age
Age centered linear	−0.05	0.35	0.898
Age centered squared	−0.01	0.01	0.711
Age centered cube	0.00	0.00	0.484
Ultramarathon 72 h × age
Age centered linear	−0.71	0.54	0.194
Age centered squared	−0.09	0.02	<0.0001
Age centered cube	0.00	0.00	0.033
Sex (female) × ultramarathon 72 h × age
Age centered linear	−1.18	1.40	0.398
Age centered squared	−0.12	0.07	0.088
Age centered cube	0.00	0.00	0.717
Ultramarathon 144 h × age
Age centered linear	0.29	0.41	0.482
Age centered squared	−0.18	0.02	<0.0001
Age centered cube	0.00	0.00	0.354
Sex (female) × ultramarathon 144 h × age
Age centered linear	−0.10	0.99	0.923
Age centered squared	−0.01	0.04	0.727
Age centered cube	0.00	0.00	0.276
Ultramarathon 240 h × year
Year centered linear	2.54	1.87	0.173
Year centered squared	−0.22	0.07	0.002
Year centered cube	−0.01	0.01	0.011
Sex (female) × ultramarathon 240 h × age
Year centered linear	−4.16	3.72	0.264
Year centered squared	0.21	0.12	0.069
Year centered cube	0.01	0.01	0.420
Ultramarathon 12 h × year
Year centered linear	−0.54	0.04	<0.0001
Year centered squared	0.00	0.00	0.324
Sex (female) × ultramarathon 12 h × year
Year centered linear	−0.23	0.10	0.018
Year centered squared	−0.01	0.01	0.222
Ultramarathon 24 h × year
Year centered linear	−1.32	0.07	<0.0001
Year centered squared	−0.02	0.01	<0.0001
Sex (female) × ultramarathon 24 h × year
Year centered linear	−1.01	0.17	<0.0001
Year centered squared	−0.04	0.02	0.009
Ultramarathon 48 h × year
Year centered linear	−2.73	0.25	<0.0001
Year centered squared	−0.04	0.02	0.016
Sex (female) × ultramarathon 48 h × year
Year centered linear	−2.54	0.53	<0.0001
Year centered squared	−0.14	0.04	0.001
Ultramarathon 72 h × year
Year centered linear	−1.92	1.08	0.076
Year centered squared	−0.26	0.11	0.020
Sex (female) × ultramarathon 72 h × year
Year centered linear	−1.91	2.61	0.465
Year centered squared	0.04	0.24	0.864
Ultramarathon 144 h × year
Year centered linear	0.75	0.84	0.372
Year centered squared	0.36	0.05	<0.0001
Sex (female) × ultramarathon 144 h × year
Year centered linear	−1.47	1.91	0.441
Year centered squared	0.02	0.13	0.896
Ultramarathon 240 h × year
Year centered linear	1.37	3.46	0.692
Year centered squared	0.48	0.68	0.480
Sex (female) × ultramarathon 240 h × year
Year centered linear	5.52	5.88	0.348
Year centered squared	1.56	1.21	0.196

Variable were centered: age = 46, calendar year = 2007

Women were not able to narrow the gap to men with increasing race duration

A first important finding was that men were faster than women. The differences between women and men were between 0.2 and 10.0 % for all durations for calendar year 2007 (Table 2). However, these differences were lower than the general sex difference of 11–12 % reported for endurance and ultra-endurance performance (Cheuvront et al. 2005; Coast et al. 2004; Lepers and Cattagni 2012).

An approach of supporting the assumption of women outrunning men was reported by Speechly et al. (1996) comparing performances of both sexes in 90 km events, while matching marathon times of female and male runners. They found that women performed better than men in a 90 km event. Addressing the assumption of Speechly et al. (1996), Hoffman (2008) matched both sexes for running times in 50, 80 and 161 km in the same year and found no difference in running speed. It is important to mention that the runners investigated by Speechly et al. (1996) and Hoffman (2008) were matched for the running speed in shorter races. Therefore, the conclusion that women were as fast as men in ultra-marathon running is only partly true as no women exist who can be matched with the fastest men in the shorter running distances.

Sex differences in running performance have been shown to vary by race’s distance. For instance, Cheuvront et al. (2005) reported a sex difference of 8–14 % for running distances from 1500 m to 42 km, Lepers and Cattagni (2012) a sex difference of ~11 % in the ‘New York City Marathon’ from 1980 to 2009 and Coast et al. (2004) a sex difference of ~12.4 % in running distances from 100 m to 200 km. Across all these distances the sex difference in performance seemed rather to increase than to decrease with increasing race distance. The 240 h races belong to the longest races held worldwide (www.ultra-marathon.org) and therefore serve well for the statement that women will not outrun men in ultra-running distances.

The most important differences between women and men regarding running performance are differences in physiology and anthropometry. Women have more body fat than men in both elite (Vernillo et al. 2013) and recreational (Hoffman et al. 2010a, b) athletes. In elite runners both sexes are considerably leaner than recreational runners (Hetland et al. 1999). In both elite and recreational runners the percentage of body fat is higher in women compared to men (Blaak 2001). It could be argued that fatty tissue may be used as an energy reserve and this could be an advantage for ultra-distances since runners tend to lose body fat during multi-hours running competitions (Karstoft et al. 2013; Schütz et al. 2013). Women might benefit from their higher percentage of body fat since both sexes lose a similar amount of fat during an ultra-endurance performance such as a 100 km ultra-marathon (Knechtle et al. 2010a, b, 2012a, b). Another sex difference in anthropometry is the percentage of skeletal muscle mass (Holden 2004). In ultra-marathoners, both sexes have a lower body fat percentage and the percentage of skeletal muscle tissue is higher (Knechtle et al. 2010a, b, 2012a, b) than in recreational runners. However, body fat and training characteristics, not skeletal muscle mass, were associated with running times in half-marathoners, marathoners, and ultra-marathoners (Knechtle et al. 2012a, b).

Considering physiological aspects, maximum oxygen uptake (VO₂max) was considered as the most significant predictor of athletic performance (Bassett and Howley 2000). While elite male athletes reach a VO₂max of ~85 ml min⁻¹ kg⁻¹ (Saltin and Astrand 1967), VO₂max is lower in elite women with a maximum of ~70 ml min⁻¹ kg⁻¹ (Ridout et al. 2010). VO₂max is mainly dependent from the heart’s performance and the lung capacity (Steding et al. 2010). The maximal cardiac output (Fomin et al. 2012) and the maximal lung capacity (Guenette et al. 2007) are higher in elite male compared to elite female athletes. VO₂max depends directly from both maximal cardiac output and lung capacity and is therefore larger in men than in women (Steding et al. 2010).

Another important aspect for running performance is running economy (Anderson 1996; Piacentini et al. 2013). Running economy is defined as the necessary effort to transport 1 kg of weight for 1 m (Morgan et al. 1989). Although there is a significant difference in running economy between elite and recreational runners, sexes show no difference (Morgan et al. 1989). Bassett and Howley (2000) found VO₂max, body fat and running economy as the major three factors contributing and predicting running performance. Therefore, women are disadvantaged in two out of three factors and have no chance to outrun men.

The sex gap for all race durations increased with increasing age

A second important finding was that the sex differences in performance were larger in the older runners. This discrepancy among age groups should be attributed to the men-to-women ratio in each age group. This ratio increased consistently with increasing age for most of the race durations, i.e. a relatively lower number of women participated in the older age groups compared to men. An increase in sex difference in age group athletes has already been reported for athletes competing in shorter race distances. For age group pool swimmers and marathon runners, the sex difference increased with age. However, the increase in sex difference was lower in running compared to swimming (Senefeld et al. 2016). The increase in sex difference in these ultra-marathoners was due to the lower number of women in older age groups. This finding has already been reported for runners in short distances. For marathoners, the increase in sex difference with increasing age was explained by the lower number of women compared to men (Hunter and Stevens 2013).

The gap between women and men across calendar years

A third important finding was that the gap between the sexes decreased for certain ultra-marathons (i.e. 6, 72, 144 and 240 h) across years but increased for others (i.e. 24 and 48 h). Findings for a decrease in sex difference were reported over a large variety of distances as in 100 m sprints (Tatem et al. 2004), marathons (Whipp and Ward 1992) and ultra-marathons (Da Fonseca-Engelhardt et al. 2013; Eichenberger et al. 2012). Promoters of the theory that women would outrun men favoured linear models in performance to support their theory (Tatem et al. 2004; Whipp and Ward 1992). The use of linear models was, however, controversially discussed (Reinboud 2004) but mainly found to be worse than non-linear models.

A potential explanation for the increase in sex difference could be the participation in ultra-marathons. Several studies reported an increase in the percentage of female ultra-marathoners (Da Fonseca-Engelhardt et al. 2013; Hoffman et al. 2010a, b; Zingg et al. 2013). The first women officially ran a marathon in 1967 (www.baa.org). Forty-six years ago, the percentage of female overall finishers started to increase at the ‘Boston Marathon’ from less than 1 to 39.5 % (www.baa.org) as well as in other marathons (www.worldmarathonmajors.com). In ultra-marathons such as the ‘Western State 100 Mile Endurance Run’, the percentage of female finishers increased from virtually none in the late 1970s to nearly 20 % since 2004 (Hoffman et al. 2010a, b). Therefore, the density of both elite and recreational female finishers increased. Nevertheless, the density of the world’s fastest runners is still lower in women than in men (Deaner 2013). This leaves a possibility of a further decrease in the sex difference in ultra-running performance in case the number of female finishers will match with the number of male finishers.

The change in sex difference in performance differed between different distances (Bam et al. 1997; Tatem et al. 2004; Whipp and Ward 1992) as well as between elite and recreational runners (Hunter et al. 2011). As short and middle distance races up to 10,000 m have been held longer for both sexes in the Olympic Games, the first Olympic marathon for women was held in 1984 (www.olympic.org). Even later, women started to compete in ultra-marathons (www.ultra-marathon.org; Hoffman et al. 2010a, b). Therefore, the improvement of female performance would be faster in the first years than the improvement in men. While the sex difference in performance stabilized in running distances up to the marathon distance (Hunter et al. 2011), the sex difference in performance still decreased in ultra-marathons (Hoffman et al. 2010a, b).

Strength, weakness, limitations and implications for future research

The strength of the study is the inclusion of all athletes competing in ultra-marathons in duration between 6 h and 10 days. Furthermore, multiple finishes per athlete were included since the aspect of previous experience seems very important in ultra-marathon running (Hoffman and Parise 2015; Knechtle et al. 2009, 2011a, b). To the best of our knowledge, the data set is the most extensive for ultra-running in time-limited ultra-marathons so far. A possible weakness could be that some events from 6 h to 10 days were not recorded in the data base and therefore were not included in the data set. Furthermore, the study is limited since variables such as anthropometric characteristics (Knechtle et al. 2009, 2010a, b, 2011a, b), training data (Hagan et al. 1981), nutrition (Maughan and Shirreffs 2012; Rodriguez et al. 2009), fluid intake (Williams et al. 2012), exercise-associated hyponatremia (Hoffman et al. 2013), physiological parameters (Billat et al. 2001), and environmental conditions (Ely et al. 2007) were not considered. These variables may have had an influence on race outcome. Future studies may investigate the sex difference for all running distances from 60 m to 3100 miles for the world fastest women and men.

Practical applications

Despite these limitations, the findings of the present study would have important practical implications for both researchers and practitioners working with long-distance runners. Since the analysed data were the most extensive ever studied in time-limited ultra-marathons and covered a large period (~40 years), the findings might be used in future studies as reference. Moreover, runners and practitioners working with them (e.g. fitness trainers) should consider the identified sex differences in the present study in order to develop sex-tailored training programs.