Introduction

Much focus, anecdotal and evidential, has been presented on the role of testosterone in muscular hypertrophy and strength development (^7,37). The concept generally proposed is that testosterone increases anabolism, usually defined as increased protein synthesis into muscle growth, and exogenous testosterone does clearly increase myocytes into mature muscle structures (³⁵). A considerable amount of evidence is, however, based on data from anabolic steroid users or from testosterone replacement therapies in the treatment of hypogonadism and andropause (^4,21). Neither is necessarily likely to represent normal physiological ranges or responses found in healthy young athletes.

Recent literature suggests that training gains can be made in both strength and hypertrophy, seemingly without overt involvement of testosterone (^38,39). However, because the suppression of endogenous testosterone can prevent these changes (³¹), it is likely that testosterone has some permissive roles (not necessarily dose dependent within a normal physiological range) in the training process. Testosterone does appear to be better linked to speed and power in short timeframes, such as jumps or short sprints (^11,12,17). Similarly, certain aspects of physical performance (e.g., maximal squat strength) correlate well with power movements such as sprints or jumps (⁴⁰) and as such may be more relevant in trying to ascertain any hormonal relationships than hypertrophy-related measures lacking this overt relationship.

Throughout the literature, there are numerous examples of correlations between testosterone or cortisol and the training and performance outcomes (^{9,10,12,16,17,32}); however, the predictive ability of these hormones, and even the direction, of these correlations differs markedly. This apparent disparity may be because of several factors. Testosterone is an important stress hormone (^13,33), and its correlation to strength and hypertrophy may reflect its biomarker potential for stress (and hence be situational stress dependent), rather than necessarily any other direct anabolic effect. Testosterone can also affect behavior (^1,28) and influence (positively) calcium handling and muscle contractility (^18,20). These traits may themselves result in a higher quality and quantity of work performed, which could then promote strength and hypertrophy using other anabolic functions.

Most studies have also examined athletes of varying strength levels, and the training background of subjects is often highly mixed. The importance of training is confirmed by different correlations (i.e., their magnitude) as a function of player position, playing time in sport, and the time of the competitive season (^17,30). Training background and existing strength may also influence the expression of those neuromuscular structures (i.e., type 2 fibers, steroid receptors) that can actualize the steroid effects (^3,22). Therefore, it is possible that individuals with different strength profiles (e.g., good squatters vs. average squatters) may exhibit different patterns of hormone use, although hormone availability may be the same. Few correlational studies have performed repeated assessments on the same individuals to examine the questions outlined.

This study had 2 main aims: first, to establish whether a predictive relationship exists between salivary free testosterone and cortisol concentrations and back squat and sprinting performance in elite athletes and, second, to determine the effects of baseline strength on these relationships. It was hypothesized that individual testosterone would correlate to performance over a repeat assessment situation and that the magnitude of these correlations would differ between individuals of good and average squatting abilities.

Methods

Experimental Approach to the Problem

A mixed group of athletes were selected on the basis of their 1 repetition maximum (1RM) strength during a free weight squat. Over 10 sessions, each athlete was assessed for a 1RM back squat and a timed 10-m sprint. The free concentrations of testosterone and cortisol were measured in saliva, thereby representing the biologically active hormone at target tissue (²). Saliva was collected before testing to assess the relationship between preworkout testosterone and cortisol concentrations and performance. Data from 2 subgroups were compared; good squatters who could lift >2.0 × their body weight and average squatters who could lift <1.9 × their body weight. This was an arbitrary split designed to achieve 2 separate groups of a similar size.

Subjects

Ten elite male athletes were recruited for this project, and each provided written informed consent before the study commenced. All subjects were in excellent health, injury-free, and engaged in professional sport (e.g., rugby union, athletics) at a similar level. Each subject had a weight training history of >3 years, but at the time of this study, they were not performing any programmed weight training, although they were undertaking other forms of training on a regular basis (i.e., 4–5 sessions per week of running type activities). The free squatting ability of the subjects ranged from 181 to 228 kg. The subject's body weight was assessed before and after the study using electronic scales (see Table 1 for these details). A university ethics committee provided ethical approval for this study.

Testing Protocols

Over a period of 40 days, all subjects performed 10 testing sessions each (not less than 3 days apart) of the following protocol: The subjects arrived at the testing facility at 0900 hours and performed a 15-minute warm-up on a stationary cycle ergometer (5 minutes each at 60, 70, and 60 rpm unloaded). After the warm-up, a free squat routine was performed within 3–4 minutes as follows: 5 repetitions × 20 kg (Olympic bar only), 5 repetitions × 100 kg, 2 repetitions × 140 kg, then single repetition trials in steps of 2.5, 5, 10, or 20 kg based on their known 1RM, until they had reached their maximal 1RM for that day. A successful squat was said to occur only if they broke parallel at the bottom of the squat position (judged by a spotter) and returned to the start position without assistance. They were verbally encouraged by a spotter during all lifting attempts and across all testing sessions. In the case of a failed attempt, they were allowed 1 further trial at a weight 2.5–5 kg below the failure point, if they had not already achieved a repetition at this lesser weight, or if they had achieved this, then 1 more attempt was made at the failed weight.

Five minutes after the squat testing, all subjects performed a short running warm-up on an indoor track as follows: 1 × 10 m at 50% of self-perceived maximal pace, 1 × 20 m at 50% pace, 1 × 10 m at 80% pace with walk back recovery. Next, 3 × 10-m timed sprints were completed at maximal effort with walk back recovery between each trial. The best time of the 3 trials was recorded using electronic timing gates (Brower Timing System, Salt Lake City, UT, USA), as described (⁶). Timing started when the first timing gate beam was broken. The subjects were familiar with the testing procedures as part of their normal training and assessment for their sport. No attempts were made to modify the activity levels of the subjects because the study aim was to examine the relationships between the outcome variables under normal training conditions. Training and dietary information were recorded in training diaries. Each subject was instructed to replicate his normal dietary intake 24 hours before each session. This was verified by examination of the training diaries at the end of this study. In addition, all the subjects consumed at least 500 mL of fluid in the 3 hours before testing. Water was also allowed (300 mL per session) before and during the testing but was not allowed 5 minutes before sample collection to prevent dilution of the saliva samples.

Saliva Sampling

Saliva samples were taken immediately after the warm-up, but before the squats, to ensure a sample as close as possible to that of performance testing. Saliva was collected into sterile tubes by timed passive drool (2 mL over 2 minutes) and stored at −80°C until the time of the assay, as previously described (¹⁶). The saliva samples were analyzed in duplicate using commercial enzyme immunoassay kits (Salimetrics LLC, State College, PA, USA) and the manufacturer's guidelines. The minimum detection limit for the testosterone assay was 1.8 pg·mL⁻¹ with intra and interassay coefficients of variation (CVs) of 1.7–10.3%, based on high and low control samples. The cortisol assay had a detection limit of 0.3 ng·mL⁻¹ with an intra and interassay CV of 3.4–9.8%. Samples for each subject were run within the same assay to eliminate interassay variance.

Statistical Analyses

Subject characteristics were assessed before and after the study using paired and unpaired t-tests. Temporal changes in the hormonal and performance variables were assessed within and between the 2 squatting groups using a 2-way (group × session) analysis of variance with repeated measures. For each subject, the relationships between the outcome variables were examined using Pearson product moment correlations and averaged (using Fisher's z transformation) to provide a pooled correlation for each group (³⁴). Before analysis, the hormonal and performance variables were log transformed to normalize data distribution and reduce nonuniformity of error. All data are back transformed in their original units. The significance was set at an alpha level of p ≤ 0.05.

Results

The subjects were clearly definable into 2 groups on the basis of 1RM squat and 1RM squat performance divided by body weight (Table 1). In both cases, the good squatters were significantly stronger than the average squatters (p < 0.05–0.01). There were no other significant differences between the 2 groups. The body weight of subjects in each group did not change significantly from the start to the end of the study (data not shown).

For the 1RM squats, a significant main effect was identified by group (p < 0.001) and session (p < 0.001), and a group × session interaction (p = 0.003). The good squatters were significantly stronger than the average squatters across all sessions (p = 0.006, Figure 1). The stronger group also showed a significant improvement in squat performance in session 10 from session 1 (p < 0.001), with the positive changes in sessions 3 and 4 approaching significance (p = 0.070–0.079). No significant group and group × session interactions were seen for the 10-m sprints (Figure 2), but a significant session effect was identified (p < 0.001). Post hoc analysis revealed a decrease in sprinting time (p = 0.05) in session 8 (vs. in session 1).

For both hormones, a significant main effect was seen across the testing sessions (p < 0.001). As seen in Figure 3, free testosterone concentrations increased slightly in session 4 (p = 0.019), compared with session 1 data and more so in sessions 5–10 (p < 0.001). Free cortisol followed a similar trend (Figure 4), being significantly elevated during sessions 4 (p = 0.051), 7 (p = 0.026), and 9 (p = 0.033) from session 1. There were no significant effects for the hormonal variables when the group and group × session interactions were examined.

As seen in Table 2, the pooled correlations between free testosterone and the performance variables were found to be very strong and significant in the good squatters (r = 0.92 for squats, r = −0.87 for sprints, p < 0.01) but weak and nonsignificant for the average squatters (r = 0.35 for squats, r = −0.18 for sprints, p > 0.05). To further highlight the differences between the good and average squatting groups, the testosterone scatter plots for the subjects in each group are respectively presented in Figure 5A, B (squats) and in Figure 6A, B (sprints).

In both squatting groups, only weak and nonsignificant correlations (r = −0.23 to 0.35) were identified between free cortisol and subsequent 1RM squat and 10-m sprinting performance (Table 2), but no major group differences were identified in the magnitude of these relationships.

Discussion

In support of our initial hypotheses, the good squatters revealed a strong predictive relationship of pretesting free testosterone concentrations to subsequent squatting and sprinting performance, whereas these relationships were much weaker in the average squatters. This finding also casts doubt on the generalized use of hormonal data to predict performance in athletes of varying strength levels.

We found a significant effect of baseline strength on the pooled correlative relationships. In the good squatters, the temporal changes in free testosterone concentrations and the 1RM squats and 10-m sprints were strongly related on an individual level, but these measures were not related in the average squatters. To our knowledge, this is the first time this has been documented in elite athletes during repeated observations. Several studies on athletic populations have quoted weak to moderate correlations between testosterone concentrations and different exercise and training parameters (^{11,12,14,17,30}). These weaker relationships may be explained by the mixing of athletes of varying strength levels and different training backgrounds. Given the strong and consistent individual relationships in this study, we speculate that free testosterone is a strong individual predictor of squatting and sprinting performance in individuals with relatively high strength levels.

A central focus of testosterone studies on athletes has been the concept of anabolism, more generally speaking around protein synthesis and muscle growth. This focus is perhaps too narrow and as such uninformative to the training role of testosterone. Much of the nonsport data on this androgen focuses on broader roles such as stress responses (^13,33), along with mood and behavior in neuromuscular processing (^1,28). Testosterone has also been linked to calcium processing in muscle and contractile function (^18,20) and motor cortex outputs (⁸). In fact, many of these testosterone actions can occur on a very short-term timescale (i.e., seconds, minutes to hours). These factors may outrank any anabolic effect linked to muscle growth and subsequent performance. Such a conceptual approach also helps to align studies by West et al. (^38,39) and Kvorning et al. (³¹). In other words, testosterone may well be permissive to muscle strength and hypertrophy but may not need a dose-response role in anabolism to be so.

Anecdotally, a number of observations support the contention that baseline strength may determine certain responsive properties, postactivation potentiation after heavy load weights being one (^25,29), but little documented evidence has been presented from a hormonal perspective. In this study, it appears that group differences in testosterone concentrations (i.e., between-individual variances) may be less important in determining the performance outcomes than the concentration variation within individuals, because this did not differ between the good and average squatters. Strength training is known to contribute to the development and recruitment of the larger type 2 fibers (²³). The training of these fibers may also help to actualize the steroid effects (^3,22), and resistance training has been shown to regulate androgen receptor content in type 2 muscle fibers (¹⁹). There is also evidence that training can influence the correlative relationships (^17,30). Whether in stronger athletes there is simply a better link between properties such as mood, confidence to perform, neuromuscular function, and testosterone on one side and performance outcome on the other, and whether this is inherent or trained remains to be seen. It would be informative to explore the possibility that a crossover point in strength exists, at which point the correlations become strong or if this is some sort of threshold effect.

In the good squatters, the temporal changes in testosterone and squat performance tended to covary (all increasing) across the middle and latter sessions. As such, it is possible that the sequential change in testosterone concentrations was merely a training response that accompanied the increase in squat performance, although only the 1RM changes in session 10 were significant. Adding to the difficulty with interpretation, each subject maintained his own training regimes, and these were not controlled, or at least normalized, in any manner. Testosterone and cortisol responsiveness (i.e., changes from pre to posttraining) is also indicative of training outcomes in weight-trained athletes (^24,26,27), but this was not examined in this study. Even if this were the case, it does not necessitate an argument promoting an anabolic effect for these hormones because it could simply reflect their biomarker potential of stress (^13,33), with stress being indicative of load or numerous other cofactors.

Free cortisol concentrations displayed nonsignificant correlations to squatting and sprinting performance, irrespective of groupings or individuals. This finding could be explained by the greater variability in pretesting cortisol concentrations across the experimental period. The secretory patterns of cortisol can change before maximal strength testing and more so than testosterone, especially during competition (³²). Highly trained athletes might also possess an adrenal system that responds more readily to competitive situations than do lesser-trained individuals (^32,36). Overall, our work is the first to show a clear separation in the predictive ability of testosterone between high-level strength athletes and athletes who, although are still strong, possess a lesser ability in the squat exercise. This would seem to us to be important in suggesting a number of levels of ongoing questioning: firstly, why there should be such a marked difference relative to strength and, secondly, what this difference means in terms of predicting performance variation over time.

We contend that monitoring hormonal changes with individual performance over time can better inform research on elite athletes. The importance of a case study approach for understanding individual responses and trends in elite sport has been highlighted conceptually in a recent review by Crewther et al. (¹⁵) and confirmed by experimental research using testosterone as an individual biomarker of training (⁵). Although interpretation of this study is limited by the number of subjects tested, the pool of elite athletes meeting the inclusion criteria for this study is limited. The descriptive nature of this project is another limiting factor when trying to ascertain a causative role for testosterone. Nevertheless, it was meaningful to see the different hormonal and performance correlations in each squatting group. This work has also highlighted the possibility that testosterone may elicit very rapid effects and also a much wider range of effects on adaptive physiology, than previously thought (^15,37).

Our key conclusion is that the temporal variance in squat and sprinting performance in an elite athletic group with strong squatting ability may be dependent, to a reasonable extent, on pretesting testosterone concentrations. The current results also confirm that the grouping of athletes of mixed strength ability may bias predictive results in a manner not reflective of a more elite population or indeed an elite individual.

Practical Applications

The results of this research suggest that temporal variances in free testosterone concentrations may help to predict the performance outcomes during normal training in stronger weight-trained athletes. This information can assist coaches, trainers, and performance scientists in making informed training decisions that could subsequently influence athlete development. For example, the preworkout measurement of testosterone could provide a possible indicator of training outcomes over time or a readiness to train at a certain intensity level, especially if real-time hormonal measurements are made. Our results also highlight the need to separate group and individual hormonal data during the repeated testing of athletes with variable strength levels.

Acknowledgments

We acknowledge with gratitude the professional athletes who contributed to this study. Part of this study was supported by grants (Elite Sport Performance Research in Training with Pervasive Sensing programme - EP/H0097441/10) from the Engineering and Physical Sciences Research Council United Kingdom and by the United Kingdom Sport Council.