Few physiological phenomena have a more robust basis in the literature than the general decline in physical performance with increasing age. However, accurate quantification of the specific impact of age on physical performance is confounded by several factors. Cohort differences, differences in performance conditions during field measurements, scarcity of physically active subjects in older decades, and changes in the quality and quantity of training with increasing age all contribute to the difficulty in clarifying the specific impact of age on physical performance. Although laboratory-based determinations of specific physiological qualities such as muscular strength or maximal oxygen consumption offer the advantage of control and accuracy, they are expensive and difficult to acquire in large numbers. Laboratory tests of work capacity and muscular endurance also may fail to elicit true maximal efforts such as those observed during competitive performances(1). Alternatively, the performance times of masters athletes in national and international competitions have been used to determine indirectly the effect of aging on aerobic capacity, muscular strength or power, and endurance by analyzing available archival data. This type of analysis has been reported for “lifetime” sports with strong masters programs, specifically, running and swimming(6,7,9,12,13).
Masters running and swimming share the advantage of attracting large numbers of individuals who perform under well organized, competitive conditions. However, disadvantages also exist. Running performances are strongly influenced by variability in weather and course conditions, making the accumulation of data from multiple competitions or across years confounded by non-age-related effects. The orthopedic stress of running is often poorly tolerated by older runners and may become a limiting factor in the training of these individuals. In swimming, differences in pool design impact wave damping characteristics, which in turn influence the wave drag experienced by the swimmers. In addition, the evolution of more optimal stroke technique and efficiency create age cohort differences that are of particular concern since mechanical efficiency in swimming appears to be more variable and a stronger predictor of performance than in other exercise modalities(7).
The sport of rowing also meets the criteria of large-age group participation and a well controlled competitive environment. Unfortunately, data collected from on-water events are often contaminated by wind, current, and water temperature variations. The relatively new discipline of competitive indoor rowing is not complicated by these variables. Indoor rowing is a unique modality for examining the functional impact of global changes in the cardiovascular and skeletal muscle systems during aging. Like swimming and running, Masters rowing has a large and growing participation in the United States and the world, with national and world veteran championships contested annually. Most rowers perform at least some of their training on rowing ergometers, which reproduce the basic biomechanical and physiological demands of on-water rowing (4,5). Rowing ergometry is widely used by local, elite, and national rowing teams during training and evaluation. Its popularity has resulted in the emergence of competitive indoor rowing. Because the athletes are competing indoors on an instrumented ergometer, this sport affords researchers the opportunity to examine physical performance in an almost laboratory-like test with the motivating environment of a competitive race.
The purpose of this study was two-pronged: 1) to quantify the differences in the standard 2500-m rowing ergometer race times and power of men and women over a wide age range and 2) to compare the regressions of these rowing performance curves on age in men and women.
Data and subject population. In this study, male and female individual race times reported in the annual World Concept II Rowing Ergometer Rankings (1990-1994 annual editions) served as the basis for cross-sectional data analyses. The performance data reported here, contributed by competitors from 29 countries, is based on the time to the nearest tenth of a second to row a standard distance of 2500 m. These data were collected from almost as many different rowing ergometers as there were subjects. Laboratory-based calibration or reliability determination is therefore impossible. However, the precision of the Concept II rowing machine as a laboratory quality ergometric instrument is supported by data comparing ergometer data with biomechanical transducer data acquired in parallel (16). The most convincing indication of the reliability of the CII rowing ergometer is its acceptance by the world-wide competitive rowing community, especially among world-class rowers and coaches. This reliability argument has been recently discussed in a similar context (15). In the large majority of cases, these performances were achieved during organized regional, collegiate, national, and international indoor rowing competitions. However, some of the times included in the ranking are also from individual time trials since the only inclusion criteria imposed by the rankings list organizers is that any time that is to be considered for a world record must have been performed during an organized competition. Annual rankings are composed of all submitted results received by March 1 of the listing year. Most indoor rowing regattas are held in January and February. The results of all of these regattas are then submitted to Concept II Inc. and published in full as an annual ranking. Participant age is the self-reported age of the competitor at the time of performance. Physical characteristics of the subjects are not provided in the ranking list.
Data analyses were restricted to individuals aged 24 and above. For men, a complete performance listing for 1993 (2) served as the basis for cross-sectional analyses. To increase the number of observations in this study for the older age groups, performances for men aged 60 and above from the 1991, 1992, and 1994 listings were combined with those from 1993. In cases where competitors were represented in more than one annual ranking, only the best performance was included in the final data set. This final data set consisted of 2,487 male performance times for individuals age 24-93.
Female performance rankings from both 1992 and 1993 were merged and culled to include only the best times of repeat performers. As was the case for male data, older age group performances for 1990, 1991, and 1994 were added to the data set to increase the number of observations. The final female data set for s ages 24-84 consisted of 1,615 individual performance times. The age distribution of the sample population for both men and women is reported inTable 1.
From the complete data sets described above, top male and female performers in each age group were identified. The complete data sets were split into 2-yr age increments, beginning at age 24. Top performers were identified based on the 95th percentile performance time for each 2-yr age increment (using SPSS 6.0, standard percentile method). The complete data sets include a small number of performers between 75 and 93 yr of age. These oldest age groups were not included in the top performers' analysis because of the small sample size. A total of 119 men and 78 women with performances at or above 95th percentile were identified and analyzed from rowers between 24 and 74 yr of age.
From the 2500-m performance time, average performance power was calculated as described in the Concept II manual: Power (W) = 1.14 × 108/(pace per 500 m in seconds)2.75. This equation is a simplified version of the more complex third order polynomial that is truly representative of the power requirements. Nonetheless, it provides an accurate calculation of developed rowing power.
Analyses. Reliability of ergometer performance was assessed by analyses of performance times for individuals who participated in both the 1991 and 1992 indoor rowing world rankings (men, N = 362; women,N = 176). Specifically, Pearson product moment correlations were calculated for 1991 vs 1992. The correlation between 1991 and 1992 ergometer performances was r = 0.95 for men (Fig. 1) and r = 0.93(regression not shown) for women.
Linear and quadratic regression analyses were performed to determine the effects of age and gender on both 2500-m ergometer rowing time and on average power output. These analyses were performed on the data from all the rowers of each gender and separately on the data from the best performers for each gender, as specified earlier. Comparison of the linear and quadratic models was performed by a general linear F test:Equation where SSErrorReduced is the sum-of-square error term from the reduced (in this case linear) model, SSErrorFull is the SS error term from the full (in this case quadratic) model, dfErrorReduced is the degrees of freedom term for the error term of the reduced model, and dfErrorFull is the degrees of freedom term for the error term of the full model. Because of the large sample size used in correlation analysis, statistical significance was set at P < 0.01. Statistical analyses were performed using Microsoft Excel (Bellevue, WA) and SPSS for Windows 6.0 on a 586-based personal computer.
Rowing times and power of all men sampled.Figure 2a depicts the individual performance times versus age and regression equations for the entire male subject population(N = 2487). Moderate correlations were found between age and performance time for both linear (r = 0.56, P < 0.001) and quadratic models (r = 0.58, P < 0.001). Age was also moderately related to male performance power (Fig. 2b) both linearly(r = 0.56, P < 0.001) and curvilinearly (r = 0.56, P< 0.01). General linear F tests revealed that the quadratic model significantly improved the prediction of time (F1,2484 = 82.59, P = < 0.001) and power (F1,2484 = 9.71,P = < 0.002) over the linear model.
Rowing times and power for all women sampled.Figure 3a depicts the individual performance times versus age for the entire female data set (N = 1615). As with the men, age and performance time were moderately related with both linear (r = 0.46,P < 0.001) and quadratic models (r = 0.46, P < 0.001). Female power production was related moderately to age(Fig. 3b) in both linear (r = -0.45, P < 0.001) and quadratic models (r = 0.45, P < 0.001). The quadratic model did not provide a significantly better fit to the data for performance time (F1, 1612 = 2.69, P > 0.10), but it did for power (F1, 1612 = 6.20, P = 0.01).
Rowing times and power of best male performers. Age-performance relationships were increased substantially when only performances at or above the 95th percentile were analyzed (Fig. 4a). The correlations of the best male performance times with age were very high for both linear (r = 0.91, P < 0.001) and quadratic models (r = 0.94,P < 0.001). The age-power relationship in the best male performers were also significant and very high: linear (r = -0.93, P< 0.001) and quadratic (r = -0.93, P < 0.001). The quadratic model provided a significantly better fit to the data for male performance time (F1, 119 = 56.23, P < 0.001) and power(F1,119 = 11.83, P < 0.001).
Rowing times and power of best female performers. Similar relationships were found for the best female performers(Fig. 4a). Age and performance time were significantly correlated for both linear (r = 0.95, P < 0.001) and quadratic models (r = 0.96, P < 0.001). As was the case for men, the relationships between age and power in the best female rowers were very high: linear model, r = -0.95, P < 0.001, and quadratic model, (r =-0.95, P < 0.01; Fig. 4b). F tests revealed that the quadratic relationship was a significantly better predictor in the best female rowers for time (F1,79 = 9.78, P< 0.01) but not for power (F1,79 = 2.73, P > 0.1).
All rowers. The linear regression model for the men's 2500 in time as a function of age was time = 444 + 3.1 (age). The slope (3.11) indicates an increase of 3 s·yr-1. Based on the men's average time of 578 s, this represents an increase of 0.54% per year. Similarly the regression model for the women's 2500 m: time = 548 + 4(age), indicating an increase of 4 s·yr-1. Because the average time for all women was 699 s, this represents a nearly identical increase of 0.57% per year. Using the linear regression equation for power output decline, male mean power output decreased 3.25 W, or about 1.25% per year, and female power decreased approximately 2.2 W, or 1.4% per year.
Top rowers. Gender differences were also apparent even when performance were considered relative to the very best youthful performance. Relative performance times and power were expressed for the 95th percentile performers in each age group by dividing each time for men and women by the best male and female scores in the data set, respectively. Thus, age group scores for men were divided by the scores of a man in the 24- to 29 yr-old age group, and those for women were divided by the scores of a woman in the 30- to 34-yr-old age group. Relative performance times and power are shown inFigure 5. Men in the age range 40 to 60 clearly maintained performance time and power scores relative to the best men and better than the women. Many men in their 40s and 50s were less than 11% slower than the fastest young rower. Indeed, the direction of the quadratic curves was slightly opposite for men and women. Whereas middle-aged women either lost power linearly or more rapidly in their 40s and 60s, middle-aged men maintained power nonlinearly.
In the present study, cross-sectional rowing performance data spanning five decades were compared among age groups and between genders. When regression of all subjects' ergometer race times on chronological age was performed, age alone accounted for only one-third of the performance variation in men and even less in women (21%). Differences in physical stature, inherent endurance capacity, training habits, competitive desire, and a host of other factors are a greater source of performance variation than age alone, even in a competition that tends to select for very specific physical and endurance characteristics. These differences appear to be greater in female rowers than in male rowers. In marked contrast, when the regression analysis is restricted to those performers in the best 5% of all times in a specific 2-yr age group who are more likely to be homogeneous in adherence to a rigorous training program, performance differences across a 50-yr span are almost completely accounted for by age. About 90% of the variability of both men and women can be attributed to age in these elite performers.
Visual inspection of age-related performance time differences for the best men and women across five decades (Fig. 4a) suggests different age-performance associations in men and women. In the best performing men performance differences are not a strict linear function of age. Using the nonlinear regression curve, between the ages of 24 and 50, performance time increases only 3% per decade. From age 50 to 74, the rate of performance time change more than doubles to 7% per decade. If only performances at ages 24 and 74 are compared, performance time increases 26% over five decades or 5.2% per decade. In the best performing female rowers, average performance time change approximates 7.5% per decade. However, in women the pattern of performance change is more linear across the entire 50-yr span analyzed. Even though the F tests indicated that the quadratic function significantly improved the curve fit for female scores, this is more a tribute to the large numbers employed in this study. Except for the male performance time regression, quadratic regression results in only negligible improvement in correlation. However the sensitivity of the general linearF test to the change in SS error makes these small differences significant.
Comparisons of the cross-sectional decline in rowing performance with age-related declines in running (8) and cycling(14) record performances reveal both similarities and key differences. In all three sports, rowing, running, and cycling, women appear to decline linearly across decades, whereas men decline at a slower rate from ages 30-55 and at a faster rate after the mid-50s. Despite a small difference in the pattern of decline, the overall rate of performance change in the 800-m and 10,000-m runs is very similar, approximately 13% per decade for men and 19% per decade for women.
Hartley and Hartley (6) reported the impact of age, race distance, and gender on the cross-sectional pattern of performance decline in swimming. In that study, swimming performance decline with age was highly curvilinear in men performing the shorter sprint races and became more linear as race distance increased. In contrast to men, performance decline in women racing short or long events from age 30 to 70 was linear (≈18% per decade). The data indicate that endurance performance decline is age, gender, and event specific, with values ranging from a decrease of only 3% per decade in cycling to 20% per decade in older swimmers.
The present data help to explain this disparity in the apparent impact of age on performance. Conversion of ergometer time to average applied power reveals that average power for the best male and female rowers declines linearly at the same absolute rate of ≈4 W·yr-1. The incongruence between the age-power and age-performance time relationships is a mechanical issue, not a physiological one. The relationship between a change in muscular power output and the corresponding change in movement velocity of an object moving through air or water is not linear because of the exponential relationship between movement velocity and the resulting drag force acting on the object (10,11). Power is a third-order polynomial function of velocity whether one is moving through air or water. This relationship is also true for the rowing ergometer used in this study because the resistance is created by a circular wind vane moving through air. Interpretation of performance data from events such as time-trial cycling, flat-water kayaking, on-water and ergometer rowing, and swimming is complicated by this nonlinearity between applied power and velocity. In all of these events, body weight is supported, and air and/or water drag is the primary resistance to movement. In contrast, the impact of wind drag on running is relatively less important compared with the resistance caused by gravity at typical running velocities in endurance events (16-24 km·h-1).
Gender differences in the pattern of performance decline are also created by the different peak average powers produced by elite men and women. Young men produce higher power outputs than women, but lose power at a similar absolute rate (4 W·yr-1). Thus, their relative performance power is better maintained (Fig. 5b). Cross-sectionally, the relative power decline in elite men is ≈0.9% per year versus 1.2% per year in elite women. At young ages where the gender performance difference is most dramatic, the mean external power output sustained by elite men is nearly 50% higher than same-age elite women. Young elite men average 460 W compared to 325 W for their female counterparts. These power values are almost identical to the average power outputs of the 1992 male and female Olympic rowing teams in a 2000-m race simulation (instead of the 2500-m distance used in the present data) on the CII ergometer (467 W, N = 35 men and 310 W,N = 25 women, (4). The 50% gender difference in ergometer power output in these national team members was paralleled by a 45% difference in absolute ˙VO2max (6.25 L·min-1 vs 4.37 L·min-1, (4). This substantial gender difference in external power output is associated with only a 15% greater mean velocity on the ergometer and an 8-10% greater velocity in the boat (where increased body weight in men results in increased boat wetted surface area and attenuates the benefit of increased power). At high ergometer (or boat) velocities, large changes in mechanical power applied to the oar will elicit smaller absolute changes in velocity compared with the same power increment applied at low velocities. Consequently, although mean power declines in parallel in men and women between the mid-20s and mid-50s(Fig. 4b), the impact on rowing ergometer velocity is less substantial in elite young to middle-aged men because they perform on a steeper portion of the exponential power-velocity curve. After this age, men perform on a shallower portion of this curve and performance times decrease more for the same decrement in power. On the basis of the power-velocity relationship, the conversion of performance time to mean power yields a more meaningful quantification of physiological decline across age. This conversion can be made accurately on rowing ergometer performance, but it has relevance to the interpretation of performance data from several other sports where such a conversion is more problematic.
Age was a modest predictor of rowing performance in a large heterogenous population of competitors participating in the world indoor rowing rankings. Differences in stature, training volume, and technique create large variability independent of age. However, when only performers at or above the 95th percentile of each 2-yr age category were employed in the regression analysis, age predicted ≈90% of the difference in performance time and average power. Averaged over five decades, the rowing power declines for top men and women were nearly identical in absolute terms (≈4 W·yr-1). However, relative power declines are greater in top women (1.2% per year vs 0.9% per year). Viewed cross sectionally, the rate of power decline in the upper 5% of male or female performers is somewhat lower than the rate of decline for the entire sample (0.9% per year vs 1.2% year for men, and 1.2% year vs 1.4% per year for women).
An important observation emerging from these analyses is that the time course of performance change is different between men and women. In the best performing men, performance was a highly curvilinear function of age. Between ages 24 and 50, performance time increased only 3% per decade. From ages 50-74, the rate of performance decline more than doubled to 7% per decade. For women, the pattern of performance decline was essentially linear across the entire 50-yr span analyzed. We have shown that gender differences can be explained by converting time to power output; the best men and the best women at any age perform at different positions on the power-velocity curve. Because the best and youngest men perform on a steeper portion of the exponential power-velocity curve, their decline is relatively linear. After ages 50-55, when men are performing on a shallower portion of the power-velocity curve(more similar to 24-yr-old elite women), performance times decrease more for the same decrement in power. These data may help to explain the disparity in the apparent effect of age on performance in different endurance sports that share similar basic physiological constraints.
This study was supported in part by a grant from Concept II Inc., R.R.1, Box 1100, Morrisville, VT 05661-9727.
Address for correspondence: Stephen Seiler, Ph.D., Faculty of Health and Sport Agder College, Kongsgård alle 20, N-4604 Kristiansand, Norway. E-mail: Stephen.Seiler@hia.no.