During the early season (April to May 2024), before athletes had participated in any races, participants completed four time trials (TT) spread over two weekends. The first two TT were conducted in a fresh state, whereas the subsequent two TTs were performed after a fatiguing protocol (see Figure 1), as detailed below.

Fourteen endurance-trained cyclists took part in the study. Body mass was measured before the first TT under fresh conditions. The VO₂max data were obtained from the last standardized cardiopulmonary exercise test performed no longer than four months before the start of the study.

Inclusion criteria were as follows. –

Habitual training volume exceeding 7 h·week⁻¹.

Tests were conducted on two consecutive weekends. Participants were instructed to consume a standardized breakfast containing approximately 2 g·kg⁻¹ of carbohydrates and ∼800 ml of water one hour before testing, adhering to general pre-exercise nutritional guidelines (11). They were advised to avoid caffeine, vigorous exercise, and alcohol for 24 h prior to each test and to standardize their food intake in the evening before. During tests carbohydrate intake was standardized at 60 g·h⁻¹ using a 1:1 ratio of glucose and fructose provided through energy drinks and gels. The participants refrained from consuming caffeine during the fatiguing protocol and were allowed to drink water ad libitum.

After a 20 min standardized warm-up phase, the first test was carried out by means of 5 min and 20 min performance tests under fresh conditions [5′-Power output (PO)_fresh and 20′-PO_fresh]. The second test, which was performed according to a fatigue protocol (see below), was used to derive the 5 min and 20 min performance under fatigue.

The 5 min and 20 min bouts of fresh and fatigued exercise were separated by 40 min of active recovery, during which participants maintained a predetermined subjective effort that did not exceed 2 out of 10 on Borg's 10-point scale, equivalent to a “light effort” (12). Participants were allowed to freely choose their cadence during the fatigue protocol. Before each effort, participants were encouraged to achieve a maximum workload while maintaining a cadence between 80 and 100 revolutions per minute. The power values of the 5 min and 20 min maximal effort TT were plotted against 1/time (t), where t is the corresponding duration, to linearize the relationship between power and duration (13).

Prior to the second set of 5 min and 20 min tests, participants completed a fatiguing protocol consisting of: –

20 min at 70% of their initial 20′-PO_fresh as warm up.

All tests were performed using the participants' own road bicycles mounted on direct-drive, electromagnetically braked indoor cycling trainers (Kickr v5, Wahoo Fitness, Atlanta, USA; Tacx, Tacx, Wassenaar, The Netherlands). These indoor trainers were calibrated according to the manufacturers' specifications to ensure valid and reliable power output measurements (14). All participants used fans for cooling and conducted the tests in a well-ventilated room. Power output was measured using two brands of power meters: Garmin Rally (Garmin Ltd., Olathe, Kansas, USA) and Assioma (Favero Electronics, Treviso, Italy), recorded on portable head units (Garmin Edge 520 or Garmin Edge 1,040, Garmin Ltd., Olathe, Kansas, USA). Participants performed a “zero offset” calibration before each test, following manufacturer guidelines. Data files with irregularities, such as missing or erroneous power readings (e.g., sudden spikes or drops in power output that were not associated with corresponding changes in cadence or heart rate, as well as prolonged zero values despite ongoing activity) or incomplete data due to technical issues (e.g., battery failure), were excluded from analysis. Power output data were analyzed and visually inspected for errors by two independent researchers using commercial software (WKO 5 build 590).

Heart rate responses were assessed during both the 5 min and 20 min time trials to evaluate changes between fresh and fatigued states. Heart rate data were continuously recorded throughout each test using a heart rate chest strap (Garmin HRM-Pro, Garmin Ltd., Olathe, Kansas, USA), paired with portable head units. Comparisons were made between the fresh (first test) and fatigued (second test) conditions. These values were compared between successful and less successful athletes to determine whether fatigue influenced heart rate responses differently across groups.

The sample size was calculated a priori using G*Power (15). To achieve a statistical power of 80% at an alpha level of 0.05 and assuming a large effect size (Cohen's d = 0.4 for ANCOVA), the required total sample size of 26 participants was suggested. However, due to only 14 available athletes without exclusion criteria during the study period, the final sample comprised only 14 participants. The statistical analysis was performed using R (R Foundation for Statistical Computing, Vienna, Austria) on the basis of the available test subjects. Data are presented as mean ± standard deviation unless stated otherwise. The changes in power outputs were shown as a percentage. The Analysis of Covariance (ANCOVA) was used to analyze the differences in power output against the background of possible confounders. To fulfill the statistical requirements of ANCOVA the percentage changes were transformed into logits (16). T-tests were used for bivariate mean comparisons. Statistical significance was set at p < 0.05. Asterisks indicate statistical significance, with more asterisks representing lower p-values (e.g., *p < 0.05, **p < 0.01, *p < 0.001).

The analysis of the socio-demographic characteristics of the successful and less successful athletes reveals significant differences in weight, VO₂max and weekly training whereby age differences could not be determined (Table 1).

In the absence of age differences, the group of successful athletes consistently shows parameters that are conducive to higher performance. To rule out the possibility that the core results of this study can be explained solely by these parameters, a univariate model including the parameters shown in Table 1 was carried out for the comprehensive static analysis.

Table 2 shows the mean absolute changes of cycling performance and heart rate response and reveals that successful amateur cyclists have had a significantly smaller decrease in power output during the 20 min time trial under fatigue.

Figure 2 shows the mean absolute changes in cycling performance and heart rate response comparing the successful and less successful athletes in fresh and fatigued states. This figure with direct bivariate comparisons illustrates how the two groups differ in their ability to maintain performance under fatigue. However, due to the potential confounding effects of significantly different body weight, VO₂max and weekly training volume, these differences need to be tested for validity in the following analysis of covariance.

The ANCOVA shows that successful amateur athletes have a significantly lower drop in performance in the 20 min interval (p = 0.033) than their less successful counterparts (see Table 3). Based on the adjusted mean values, a mean performance drop of 6.5% can be calculated for successful athletes and 12.5% for less successful ones, which corresponds to a performance drop that is twice as high for the less successful athletes. In contrast to this finding, no significant difference was found for the 5 min interval.

Figure 3 shows the marginal plot of VO₂max and 20 min power decrement, comparing successful and less successful amateur athletes. The y-axis represents the drop in power over 20 min, expressed in logit transformed values, where higher logit values correspond to a greater drop in performance and the x-axis shows VO₂max.

The plot shows a significant interaction (p = 0.03), indicating that less successful athletes (blue line) show a steeper decline in performance as VO₂max decreases compared to successful athletes (red line). The shaded areas represent the 95% confidence intervals and illustrate the variability within each group.

Figure 4 represents the mean heart rate in beats per minute (bpm), and error bars indicate ± standard deviation. There were no significant differences in heart rate response between the two study groups (Table 4).

The most important practical finding from this study is that more successful amateur cyclists demonstrate greater durability, experiencing a smaller drop in performance over time compared to their less successful counterparts. This reinforces the crucial role of durability in endurance performance, particularly after substantial energy expenditure.

While the specific trainability of durability remains an area for further research, existing evidence suggests that its development is closely linked to long-term training consistency, high training volumes, and the inclusion of prolonged exercise sessions incorporating high-intensity efforts or progressive intensity increases. Coaches and athletes can apply these insights by integrating race-pace efforts late in training sessions, employing nutritional strategies to delay glycogen depletion, and monitoring heart rate and perceived exertion to manage fatigue effectively. Additionally, the ability to sustain power output under fatigue may serve as an indicator of an athlete's potential for endurance performance. This information can aid in talent identification and help guide training strategies tailored to improving resilience.

Since the a priori power analysis, assuming a large effect size (Cohen's d = 0.4), suggested a larger number of subjects (n = 26) than was available for the study presented here, the size of the study population must be discussed as a major limiting factor. However, van Erp and colleagues (8), have been able to show validity of small study populations in this still young field of research about durability in cycling. For this reason, this study can be seen as a valuable contribution to the scientific world in the field of sports medicine and a corresponding template for further studies with a larger sample.

Moreover, the use of different brands of power meters to collect power data, which can lead to variations in measurements (26). Because study participants were not continuously observed before the testing sessions including the periods in between, we cannot completely guarantee the absence of protocol violations regarding adherence to the personal pre-test instructions. While our study did not assess cardiovascular drift in the conventional sense, we examined heart rate responses in 5- and 20 min time trials to evaluate transient changes between fresh and fatigued states, which may provide insights into short-duration submaximal effort-induced heart rate dynamics. Future studies with larger sample sizes of amateur athletes should incorporate analysis of the relationship between VO₂max and durability to better quantify the variance explained by VO₂max. Another limitation is that most of the data was collected from male athletes. It remains unclear whether there is a similarly strong relationship between endurance and race performance in female cyclists, and further research is needed to investigate this aspect.