The Age-Related Performance Decline in Marathon Running: The Paradigm of the Berlin Marathon

variation of marathon race time by age group has been used recently to model the decline of endurance with aging; however, paradigms of races (i.e., marathon running) examined so far have mostly been from the United States. Therefore, the aim of the present study was to examine the age of peak performance (APP) in a European race, the “Berlin Marathon”. Race times of 387,222 nishers (women, n=93,022; men, n=294,200) in this marathon race from 2008 to 2018 were examined. Men were faster by+1.10 km.h 1 (10.74 1.84 km.h 1 versus9.64 1.46 km.h 1 ,p<0.001, 2=0.065, medium e ect size) and older by+2.1 years (43.1 10.0 yearsversus41.0 9.8 years , p<0.001, 2=0.008, trivial e ect size) than women. APP was 32 years in women and 34 years in men using 1-year age groups, and 30–34 years in women and 35–39 years in men using 5-year age groups. Women's and men's performance at 60–64 and 55–59 age groups, respectively, corresponded to ~90% of the running speed at APP. Based on these ndings, it was concluded that although APP occurred earlier in women than men, the observed age-related di erences indicated that the decline of endurance with aging might di er by sex. Keywords:aerobic capacity; ageing; age of peak performance; exercise; gender 1. Introduction Recently, a dramatic increase has been observed in the number of outdoors running races—such as marathons—and the number of those

was concluded that although APP occurred earlier in women than men, the observed age-related di erences indicated that the decline of endurance with aging might di er by sex. Keywords:aerobic capacity; ageing; age of peak performance; exercise; gender 1. Introduction Recently, a dramatic increase has been observed in the number of outdoors running races—such as marathons—and the number of those participating in them [1,2]. For marathon running, mainly races and events held in the United States have been investigated more deeply [3–5]. An important performance characteristic in marathons, similarly to other sporting events, has been the age of peak performance (APP), i.e., the age of the best performance during the human lifetime [6]. The information about when APP occurs would be bene cial for coaches and athletes to set long-term training goals. In addition, the knowledge of APP might assist exercise physiologists and gerontologists in the study of the decline of endurance performance across the human life. APP has been well studied in marathon running using di erent sampling approaches (e.g., top athletes, all nishers) and statistical methods (e.g., multiple linear regression models, non-linear regression analyses, mixed-e ects regression analyses) [7–11]. Independent of the methodological approaches, APP in this endurance sport has been estimated ~25–35 years; however, the precise APP might vary by sex [7–11]. With regards to the role of sex, it has been suggested that APP was older in women than in men [7,9–11], with an exception [8] that showed the opposite trend. Int. J. Environ. Res. Public Health2019,16, 2022; doi:10.3390 /ijerph16112022 /journal/ijerph

Int. J. Environ. Res. Public Health2019,16, 2022 2 of 12 It should be acknowledged that these studies have enhanced our comprehension of APP in marathon running; however, it should be highlighted that most of the existing research has been conducted by using data from races held in the United States [6,7,11], and limited information was available with regards to European marathons [12]. Filling this gap in the existing literature might be of practical importance considering that the world record was achieved recently in a European marathon (Berlin). The course in Berlin seems to be the fastest marathon course in the world, since seven of the 10 fastest male marathon times were set in the Berlin Marathon, including the actual world record of 2:01:39 h:min:s set by Eliud Kipchoge in 2018 [13]. Therefore, the aim of the present study was to assess APP in the Berlin Marathon. We hypothesized that the age of peak marathon performance would be di erent in women and men based on data analyzed from marathon races held in the United States. 2. Materials and Methods 2.1. Ethics Approval The institutional review board of St Gallen, Switzerland, approved this study. Since the study involved analysis of publicly available data, the requirement for informed consent was waived. 2.2. Methodology Data (i.e., sex, age, calendar year, and running speed) on nishers in the Berlin Marathon from 2008 to 2018 were examined. Initially, 389,958 nishers were considered. These data were acquired from the o cial website of the race [13]. Race time in h:min:s was converted to running speed in km/h. Cases with missing age (n=133) or race time slower than the o cial time limit of 6:15 h:min (n=2603) were excluded, resulting in a nal sample of 387,222 that was entered in the analysis (women, n=93,022; men, n=294,200). Considering their age, nishers were classi ed in 1-year (e.g., 42 years, 43 years) and 5-year age groups (e.g., 40–44 years, 45–49 years). Both classi cations had practical applications; 1-year intervals would aid examining age-related decline in performance and study of the e ect of age with more detail, whereas 5-year

was entered in the analysis (women, n=93,022; men, n=294,200). Considering their age, nishers were classi ed in 1-year (e.g., 42 years, 43 years) and 5-year age groups (e.g., 40–44 years, 45–49 years). Both classi cations had practical applications; 1-year intervals would aid examining age-related decline in performance and study of the e ect of age with more detail, whereas 5-year intervals followed the o cial system of the race. 2.3. Statistical and Data Analysis The statistical package IBM Statistical Package for the Social Sciences (SPSS) v.20.0 (SPSS, Chicago, USA) and GraphPad Prism (Version 5, GraphPad Software, La Jolla, USA) were used to analyze the data. Mean standard deviation described age and speed. Normality of age and speed was tested by the Kolmogorov-Smirno test and visual inspection of normal Q-Q plots. The main e ects of sex, age group, and calendar year and their interactions on running speed were tested by a two-way analysis of variance (ANOVA) followed by a post-hoc Bonferroni test for di erences among calendar years or age groups. Eta square ( 2) examined the magnitude of the di erences among age groups or calendar years with the following criteria: small (0.010< 2 0.059), moderate (0.059< 2 0.138), and large ( 2 >0.138) [14]. The men-to-women ratio (MWR) was calculated as the ratio of men versus women nishers. The associations between calendar year and sex, and between age group (with at least 10 cases per sex) and sex were tested using chi-square test ( 2), and their magnitude was tested by Cramer's phi ('). APP (i.e., age of fastest running speed)—considering age groups in 1-year and 5-year intervals—was calculated by a non-linear regression model with a second order (quadratic) polynomial function (y=ax 2 +bx+c) that tted the data. The vertex of the quadratic function was calculated as p(x|y) = b/2a|c (b 2 /4a). Non-linear regression analysis was used instead of linear regression analysis, since it was previously observed that maximal oxygen uptake—a main correlate of running speed—varied in an inverse U trend across life-time [15]. Alpha level was set at 0.05.

The vertex of the quadratic function was calculated as p(x|y) = b/2a|c (b 2 /4a). Non-linear regression analysis was used instead of linear regression analysis, since it was previously observed that maximal oxygen uptake—a main correlate of running speed—varied in an inverse U trend across life-time [15]. Alpha level was set at 0.05.

Int. J. Environ. Res. Public Health2019,16, 2022 3 of 12 3. Results The overall MWR was 3.16. Men (race speed 10.74 1.84 km.h 1 , race time 4:02:41 0:41:42 h:min:s, age 43.1 10.0 years) were faster by+1.10 km.h 1 (p<0.001, 2=0.065, medium e ect) and older by+2.1 years (p<0.001, 2=0.008, trivial ES) than women (9.64 1.46 km.h 1 , 4:28:37 0:39:41 h:min:s, 41.0 9.8 years). 3.1. Trends of Participation, Running Speed, and Age across Calendar Years A sex calendar year association was observed ( 2=1813.69,p<0.001,'=0.068; Figure), where the men-to-women ratio was the smallest in 2018 (2.36) and the largest in 2009 (3.97). Compared to 2009, the number of women nishers increased by+69.9% in 2018, whereas the respective change in men was+0.9%. Accordingly, the number of women nishers increased across calendar years, MWR decreased, and the number of male nishers remained stable. A trivial e ect of calendar year on running speed was observed (p<0.001, 2=0.003), with 2018 being the slowest (10.21 1.94 km.h 1 ) and 2013 the fastest (10.62 1.75) (Figure) years. A trivial sex calendar year interaction on running speed was shown (p<0.001, 2<0.001), with the smallest sex di erence in 2009 (+1.02 km.h 1 ) and the largest in 2010 (+1.17 km.h 1 ). The overall trend in women, men, and sex di erence across calendar years was that the running speed remained stable during this period. A trivial e ect of calendar year on age was found (p<0.001, 2=0.001), with the youngest nishers in 2009 (41.74 9.87 years) and the oldest nishers in 2014 (42.94 10.02 years) (Figure). A trivial sex calendar year interaction on running speed was observed (p<0.001, 2=0.001), with the smallest sex di erence in 2013 (+1.4 years) and the largest in 2017 (+2.7 years). During the calendar years 2008–2018, the age of men and the sex di erence in age increased, whereas the age of women remained stable.Int. J. Environ. Res. Public Health 2019, 16, x 3 of 12 The overall MWR was 3.16. Men (race speed 10.74 ± 1.84 km.h -1 , race time 4:02:41 ± 0:41:42 h:min:s, age 43.1 ±

(+2.7 years). During the calendar years 2008–2018, the age of men and the sex di erence in age increased, whereas the age of women remained stable.Int. J. Environ. Res. Public Health 2019, 16, x 3 of 12 The overall MWR was 3.16. Men (race speed 10.74 ± 1.84 km.h -1 , race time 4:02:41 ± 0:41:42 h:min:s, age 43.1 ± 10.0 years) were faster by +1.10 km.h− 1 (p <0.001, η 2 = 0.065, medium effect) and older by +2.1 years (p <0.001, η 2 = 0.008, trivial ES) than women (9.64 ± 1.46 km.h− 1 , 4:28:37 ± 0:39:41 h:min:s, 41.0 ± 9.8 years). 3.1. Trends of Participation, Running Speed, and Age across Calendar Years A sex × calendar year association was observed (χ 2 = 1813.69, p <0.001, φ = 0.068; Figure 1), where the men-to-women ratio was the smallest in 2018 (2.36) and the largest in 2009 (3.97). Compared to 2009, the number of women finishers increased by +69.9% in 2018, whereas the respective change in men was +0.9%. Accordingly, the number of women finishers increased across calendar years, MWR decreased, and the number of male finishers remained stable. A trivial effect of calendar year on running speed was observed (p <0.001, η 2 = 0.003), with 2018 being the slowest (10.21 ± 1.94 km.h −1 ) and 2013 the fastest (10.62 ± 1.75) (Figure 2) years. A trivial sex × calendar year interaction on running speed was shown (p <0.001, η 2 <0.001), with the smallest sex difference in 2009 (+1.02 km.h− 1 ) and the largest in 2010 (+1.17 km.h −1 ). The overall trend in women, men, and sex difference across calendar years was that the running speed remained stable during this period. A trivial effect of calendar year on age was found (p <0.001, η 2 = 0.001), with the youngest finishers in 2009 (41.74 ± 9.87 years) and the oldest finishers in 2014 (42.94 ± 10.02 years) (Figure 3). A trivial sex × calendar year interaction on running speed was observed (p <0.001, η 2 = 0.001), with the smallest sex difference

trivial effect of calendar year on age was found (p <0.001, η 2 = 0.001), with the youngest finishers in 2009 (41.74 ± 9.87 years) and the oldest finishers in 2014 (42.94 ± 10.02 years) (Figure 3). A trivial sex × calendar year interaction on running speed was observed (p <0.001, η 2 = 0.001), with the smallest sex difference in 2013 (+1.4 years) and the largest in 2017 (+2.7 years). During the calendar years 2008–2018, the age of men and the sex difference in age increased, whereas the age of women remained stable. Figure 1. Number of finishers by sex and calendar year. Note: MWR = men-to-women ratio; shadowed areas denote 95% confidence intervals;  = women, ⚫ = men. Figure 1. Number of nishers by sex and calendar year. Note: MWR=men-to-women ratio; shadowed areas denote 95% con dence intervals;#=women, =men.

Int. J. Environ. Res. Public Health2019,16, 2022 4 of 12Int. J. Environ. Res. Public Health 2019, 16, x 4 of 12 Figure 2. Running speed of finishers by calendar year. Note: Δ = sex difference; error bars denote standard deviation; shadowed areas denote 95% confidence intervals;  = women, ⚫ = men. Figure 3. Age of finishers by calendar year. Note: Δ = sex difference; error bars denote standard deviation; shadowed areas denote 95% confidence intervals;  = women, ⚫ = men. 3.2. Trends in Participation by Age Group A sex × age association was shown when age groups were considered in 5-year groups (χ 2 = 3304.46, p <0.001, φ = 0.092; Figure 4), with the lowest MWR (2.06) in age group 25–29 years and the highest MWR (8.66) in the age group 75–79 years. Also, a sex × age association was found when age groups were considered in 1-year groups (χ 2 = 3483.08, p <0.001, φ = 0.095; Figure 5), with the lowest MWR (1.96) in the 26 years age group and the highest MWR (11.16) in the 71 years age group. Figure 2. Running speed of nishers by calendar year. Note:D=sex di erence; error bars denote standard deviation; shadowed areas denote 95% con dence intervals;#=women, =men.Int. J. Environ. Res. Public Health 2019, 16, x 4 of 12 Figure 2. Running speed of finishers by calendar year. Note: Δ = sex difference; error bars denote standard deviation; shadowed areas denote 95% confidence intervals;  = women, ⚫ = men. Figure 3. Age of finishers by calendar year. Note: Δ = sex difference; error bars denote standard deviation; shadowed areas denote 95% confidence intervals;  = women, ⚫ = men. 3.2. Trends in Participation by Age Group A sex × age association was shown when age groups were considered in 5-year groups (χ 2 = 3304.46, p <0.001, φ = 0.092; Figure 4), with the lowest MWR (2.06) in age group 25–29 years and the highest MWR (8.66) in the age group 75–79 years. Also, a sex × age association was found when age groups were considered in

sex × age association was shown when age groups were considered in 5-year groups (χ 2 = 3304.46, p <0.001, φ = 0.092; Figure 4), with the lowest MWR (2.06) in age group 25–29 years and the highest MWR (8.66) in the age group 75–79 years. Also, a sex × age association was found when age groups were considered in 1-year groups (χ 2 = 3483.08, p <0.001, φ = 0.095; Figure 5), with the lowest MWR (1.96) in the 26 years age group and the highest MWR (11.16) in the 71 years age group. Figure 3. Age of nishers by calendar year. Note:D=sex di erence; error bars denote standard deviation; shadowed areas denote 95% con dence intervals;#=women, =men. 3.2. Trends in Participation by Age Group A sex age association was shown when age groups were considered in 5-year groups ( 2 =3304.46,p<0.001,'=0.092; Figure), with the lowest MWR (2.06) in age group 25–29 years and the highest MWR (8.66) in the age group 75–79 years. Also, a sex age association was found when age groups were considered in 1-year groups ( 2=3483.08,p<0.001,'=0.095; Figure), with the lowest MWR (1.96) in the 26 years age group and the highest MWR (11.16) in the 71 years age group.

Int. J. Environ. Res. Public Health2019,16, 2022 5 of 12Int. J. Environ. Res. Public Health 2019, 16, x 5 of 12 Figure 4. Finishers by 5-year age groups. MWR = men-to-women ratio. The horizontal dashed line shows the overall MWR. Figure 5. Finishers by 1-year age groups. MWR = men-to-women ratio. The horizontal dashed line shows the overall MWR. 3.3. Age of Peak Performance When the age was considered in 1-year age intervals, APP was observed at 32 years in women and 34 years in men (Table 1, Figure 6, Figure 7); a small main effect of age group on running speed was shown (p <0.001, η 2 = 0.027), with the fastest running speed at 33 years (10.79 ± 1.97 km.h− 1 ) and the slowest at 80–84 years (7.26 ± 1.82 km.h− 1 ). A trivial sex × age group interaction on running speed Figure 4. Finishers by 5-year age groups. MWR=men-to-women ratio. The horizontal dashed line shows the overall MWR.Int. J. Environ. Res. Public Health 2019, 16, x 5 of 12 Figure 4. Finishers by 5-year age groups. MWR = men-to-women ratio. The horizontal dashed line shows the overall MWR. Figure 5. Finishers by 1-year age groups. MWR = men-to-women ratio. The horizontal dashed line shows the overall MWR. 3.3. Age of Peak Performance When the age was considered in 1-year age intervals, APP was observed at 32 years in women and 34 years in men (Table 1, Figure 6, Figure 7); a small main effect of age group on running speed was shown (p <0.001, η 2 = 0.027), with the fastest running speed at 33 years (10.79 ± 1.97 km.h− 1 ) and the slowest at 80–84 years (7.26 ± 1.82 km.h− 1 ). A trivial sex × age group interaction on running speed Figure 5. Finishers by 1-year age groups. MWR=men-to-women ratio. The horizontal dashed line shows the overall MWR.

Int. J. Environ. Res. Public Health2019,16, 2022 6 of 12 3.3. Age of Peak Performance When the age was considered in 1-year age intervals, APP was observed at 32 years in women and 34 years in men (Table, Figures); a small main e ect of age group on running speed was shown (p<0.001, 2=0.027), with the fastest running speed at 33 years (10.79 1.97 km.h 1 ) and the slowest at 80–84 years (7.26 1.82 km.h 1 ). A trivial sex age group interaction on running speed was found (p<0.001, 2=0.001); considering 1-year age groups with at least 10 nishers in each sex, the smallest sex di erence was observed at 73 years (1.0%) and the largest at 38 years (13.7%). Table 1. Parameters in the second-order polynomial regression running speed (km.h 1 )=a+bx+cx 2 , using the running speed of all runners in 1-year age and 5-year age groups by sex. 1-Year Age Groups 5-Year Age Groups Parameter Women Men Women Men a (km.h 1 ) 0.00117 0.001928 0.02773 0.04594 b (km.h 1 .years 1 ) 0.07472 0.131 0.218 0.399 c (km.h 1 .years 2 ) 8.650 8.868 9.413 10.230 Age (years) 31.93 33.97 30–34 35–39 Running speed (km.h 1 ) 9.84 11.09 9.84 11.10 The parameters a, b, and c are coe cients in the polynomial equation that shows the relationship between running speed and age. Based on the regression analysis, we calculated the parameters Age (the age or age group of peak performance) and Running speed (the performance of the corresponding Age).Int. J. Environ. Res. Public Health 2019, 16, x 6 of 12 was found (p <0.001, η 2 = 0.001); considering 1-year age groups with at least 10 finishers in each sex, the smallest sex difference was observed at 73 years (1.0%) and the largest at 38 years (13.7%). Table 1. Parameters in the second-order polynomial regression running speed (km.h −1 ) = a + bx + cx 2 , using the running speed of all runners in 1-year age and 5-year age groups by sex. 1-year age groups 5-year age groups Parameter Women Men Women

This study analyzes age-related performance decline in marathon running.