- Trainability may depend on the training stimulus or dose of exercise.
- The gene variants associated with V˙O2max trainability identified to date are remote from key physiological pathways in the O2 transport system like stroke volume and total body hemoglobin.
- We suggest a multidose exercise training study in twins might be able to help resolve a number of conceptual issues related to trainability and the exercise dose-response duration relation for V˙O2max.
Anyone who has ever spent much time in a gym or at sports practice will tell you that the rate at which a given individual within a group “gets in shape” or acquires new skills in response to training can vary dramatically. Trainability is the general term used to describe either the rate of change or the magnitude of change in fitness or skill in response to training. An individual who shows rapid and marked improvement to a given “dose” of training is said to be highly trainable. A slower responder or an individual who improves less over time is said to be less trainable. However, the rate of change to a training stimulus is not always uniform over time. So an initial slow or fast response may not always be predictive of the total magnitude of the response after a period of months or longer. In addition, because the duration of a vast majority of training studies is measured in months, it is not clear how years of training might influence ideas about trainability and nonresponders.
It also is important to distinguish trainability from intrinsic (or “natural”) ability. Some individuals might enter a training program faster, stronger, or more skilled than others. However, high levels of initial fitness are generally unrelated to the magnitude of change seen in response to standard training programs. In the case of endurance or strength-related sports, it also is possible that those who perform well on either laboratory or field tests with minimal formal training might simply have been more active during other facets of life. For example, in sedentary middle-aged obese individuals, there is a strong positive correlation between the intensity of incidental physical activity and maximal oxygen consumption (V˙O2max) (1). However, there are clearly inherent biological factors related to oxygen transport or muscular strength that are independent of physical activity. Thus, in terms of baseline fitness in free-living humans, it can be difficult to discern the role of intrinsic biological factors from acquired responses to the activities of daily living. There is speculation, for example, that average baseline V˙O2max in untrained subjects is typically higher in cultures where active transportation via cycling is more common.
In the context of these comments, trainability for the purposes of this article will be defined as the increase in V˙O2max in response to a specific training program. However, we recognize the limitations of this definition because there are many potential exercise training dose-response and duration interactions possible for a given individual. Thus, the determination of trainability for a given individual in most studies is a snapshot based on his or her response to a specific program. The concept of trainability also might be better served if more attention were paid to the range of values observed in a cohort in response to a specific program versus a more narrow focus on nonresponders to programs based on current physical activity guidelines.
In some studies, cardiorespiratory fitness (CRF) is reported. This measure is frequently used in large population studies of fitness and health or mortality when graded exercise testing without direct measurements of gas exchange has been performed. Although CRF and V˙O2max are highly correlated, CRF is estimated from peak METs levels and expressed as metabolic equivalents or METs. Each MET equaling 3.5 ml−1·kg−1·min−1 (e.g., resting metabolic rate), with increments of this unit being used to describe CRF. For example, CRF in an individual with a V˙O2max of 35 ml−1·kg−1·min−1 would equal 10 METs. In these cases, equations are used to link peak workload with a “MET equivalent.” There also are validated questionnaire-based approaches that can be used (2). Because CRF and V˙O2max are highly correlated, and because we are interested in the biology of oxygen transport, the remainder of the article focuses on V˙O2max.
Trainability and Why it Matters
Beyond its obvious importance to coaches and individuals interested in talent identification for sport, trainability has potential individual and public health implications. This is especially true for V˙O2max. The reasoning goes something like this: fitness is a powerful predictor of all-cause and cardiovascular mortality, and endurance exercise training is recommended to improve cardiovascular fitness. In population studies, each 1-MET (3.5 ml−1·kg−1·min−1) increase in V˙O2max is associated with a roughly 15% reduction in all-cause mortality (3). Importantly, such an increase in V˙O2max is generally viewed as achievable by most humans with modest levels of training.
In this context, if there are individuals who do not respond to training with an increase in V˙O2max, then do they receive a mortality benefit from training? Do such individuals need a different training program? Are the current physical activity and exercise guidelines sufficient to ensure most people who follow them increase their fitness? Finally, high levels of fitness may be more protective against cardiovascular and all-cause mortality than physical activity (4,5).
The aforementioned questions also highlight the contrasts and potential overlap between the terms training and physical activity. Although exercise training is a structured program designed to improve physical capacity, there also can be adaptive responses to incidental, leisure time, or occupational physical activity. As previously noted, individuals who engage in more high-intensity incidental physical activity have higher V˙O2max values (1). In this context, an advantage of measuring responses to training versus estimates of other forms of physical activity is that better quantification and control of the input (training) stimulus is possible.
Before delving into the physiological and biological determinants of V˙O2max and trainability, it is important to note that some of the epidemiological and outcomes literature on exercise relate fitness to health or mortality outcomes, whereas other studies relate measures of physical activity to outcomes. As discussed earlier, there is some overlap in these variables; in general fitness is more protective than physical activity (3–5). In either case, the protection afforded by either high levels of CRF or physical activity are generally greater than might be predicted based on their effects on traditional risk factors like hypertension, diabetes, and blood lipids when compared with drug therapy (3,6).
In summary, trainability is important when considering the relation between V˙O2max and health outcomes. Because some individuals show bigger changes in V˙O2max in response to training, these variable responses raise questions about how uniform the health benefits of training are. They also raise questions about the extent to which health benefits are related to increases in V˙O2max and improvements in other risk factors linked to health outcomes.
Determinants of V˙O2max
V˙O2max describes the maximum ability of a whole organism to transport oxygen from the air to the tissues and especially the exercising skeletal muscles (7). It is thus dependent on and potentially limited by numerous sites in the so-called O2 transport cascade. This cascade includes the following: 1) pulmonary ventilation, 2) diffusion of oxygen across the pulmonary capillary membrane to the blood, 3) the bulk “flux” of O2 away from the lungs via a combination of cardiac output and arterial O2 content, 4) increased blood flow to contracting muscles, and 5) diffusion of O2 from blood to tissue and ultimately oxidative metabolism in the mitochondria.
Although each of the five factors previously outlined can influence V˙O2max, in large groups of healthy humans with a 2- to 3-fold range in V˙O2max, peak or maximum cardiac output and total body hemoglobin mass seem to predominate as determinants of V˙O2max (Fig. 1). The other factors only become potentially rate limiting in either specific patient groups and in some cases (e.g., pulmonary ventilation), elite endurance athletes (8).
Current public health guidelines typically recommend adults engage in 150–300 min·wk−1 of moderate to vigorous aerobic physical activity, with lesser amounts suggested for those engaging in more vigorous exercise (10). When large numbers of previously untrained healthy young and middle-aged humans engage in guideline-based training for 10–20 wk, the increase in V˙O2max ranges from 0 to approximately 40%–60% (11–13) with average values of around 15%–20% usually reported and higher mean increases (approximately 25%–30%) seen with more frequent or intense training (11,14). Importantly, the increase in V˙O2max with training is not related to baseline V˙O2max and is not influenced by race, sex, or age (up to middle age) (12,13). In this context, the iconic HERITAGE study reported that roughly 10%–20% of subjects responded only minimally to training with an increase in V˙O2max and thus demonstrated limited or no trainability in response to a public health guideline-informed program (12,13).
Along similar lines, when the effects of such guideline-based training programs on blood pressure, blood lipids, and blood glucose are evaluated, roughly 10%–30% of subjects show no improvement or perhaps a worsening of values (15). However, the risk factor adverse responder concept has been challenged on a number of grounds (16). In addition, many of the protective effects of fitness and physical activity on health likely operate via mechanisms not captured via traditional risk factors (6).
In contrast to guideline-based training programs, studies that have used more intense levels of exercise have typically shown that the vast majority of young healthy subjects are trainable (14). In addition, one of us (see Montero and Lundby) recently showed that when healthy young men performed intense cycle training 4 or 5 d·wk−1, all of the studied subjects showed an increase in V˙O2max (11). By contrast, some subjects who were only training 1, 2, or 3 d·wk−1 showed limited trainability. However, when two additional training sessions per week were added to the limited-trainability subjects, they all became trainable. In this study, trainability was defined as an increase greater than the typical measurement error for maximal attained workload. This is the only study where individual study participant “improvements” versus “no-improvements” is based on an objective measure. Figure 2 shows key data from this study. It also seems that all subjects in training studies using repeated intervals of 3–5 min at exercise intensities greater than 90% of V˙O2max show an increase in V˙O2max (17). Finally, only a few studies have followed people during prolonged periods, and Howden et al. (18) found that with 1 yr of high-volume training that included high-intensity exercise, male subjects responded with 22% increases in V˙O2max, whereas the number was 15% for female participants. In addition, this increase was somewhat higher than that observed by Scharhag-Rosenberger et al. (19) (average 16%, range 9%–20%), who followed subjects over the course of 1 yr of guideline-based (3 d·wk−1) training.
A key question that arises when guideline-based studies are compared with studies using more rigorous training programs is the effect on the range of V˙O2max responses. Is the range of responses similar but shifted to a higher value? Does the range increase, or conversely, does the range decrease? The limited data available on this topic suggest that the range of responses may be similar but shifted to a higher value (14). However, these data do not provide information about whether a non (or limited) responder to training also would show a limited response to a more rigorous or perhaps longer duration program. In terms of training volume, it can however be seen from Figure 2 that the most extreme responses (21%, 25%, and 24%) observed when training 1, 2, and 3 times per week, respectively, become the norm when training 4 (26%) or 5 (32%) times per week. Hence, comparing studies applying different training frequencies, intensities, and durations may be of limited utility.
Our discussion of the divergent results seen in response to differing training programs and the questions raised by this discussion highlight a number of related ambiguities in the current data and thinking on this topic: First, have reports of V˙O2max nonresponders to specific training programs been the focus of too much attention? Do nonresponders primarily represent the tail of a distribution of responses to a given dose of exercise applied for a given duration? Second, consistent with our use of the word context in the title of this article, uncertainty in the trainability and nonresponder discussion also can arise based on technical errors in the measurement of V˙O2max and also day-to-day variability in human biology and motivation. These potential confounds need clear definition when discussing the results of a given study, comparing results between studies, and when considering the topic as a whole. Third, clear metrics related to subject adherence to any program are critical. Fourth, there also are potential issues about the scaling of V˙O2max values that need to be considered. In general, we favor liters per minute but this is not ironclad, and there are numerous scaling issues that might need to be considered depending on the size and body composition range of the subjects involved (20). Fifth and finally, will the nature of the distribution of V˙O2max responses to training change if the dose and duration of exercise is changed?
Even the most demanding training programs used in controlled studies pale in comparison to the hours per day of training that elite athletes do for years (21). If this sort of elite training represents a near maximal dose applied for years, would all individuals exposed to it show marked increases in V˙O2max and would the distribution of responses be expanded or compressed? Our hypothesis is that a maximal dose exercise stimulus applied for a long duration would result in marked training responses in essentially all humans and compress the range of responses. However, we fully acknowledge that this proposition is likely “not testable” via a randomized controlled trial.
As mentioned previously, in young and middle-aged populations, sex and age do not seem to influence trainability in response to a guideline-based program, and this also includes both white and black subjects. However, there is little information about sex and age influences on trainability in response to more rigorous programs. This is especially true in adults aged 60 or older. There are some data (low subject number) suggesting that in response to 1 yr of rigorous training, cardiac hypertrophy is less in young women than men (18). There also are some data to suggest the loss of female reproductive hormones at menopause might influence training responses (22). Aging and sex differences are clearly areas that are ripe for additional studies on trainability that include a range of long-term programs with varying degrees of frequency, intensity and duration.
Biological Factors — Genetics
Data from a limited number of studies on twins indicate that V˙O2max values in the untrained state and the response to training are markedly influenced by genetics (23–25). Likewise, HERITAGE used a family study design to show that approximately 47% of the increase in V˙O2max to training was heritable (12,13). The HERITAGE team subsequently identified a number of genetic markers and developed a retrospective genetic panel based on 21 SNP gene variants that predicted approximately 50% of the training response among individuals (26).
However, several observations challenge the idea that genetic factors are the major determinant of trainability. First, we have argued that none of the gene variants identified in HERITAGE are clearly linked to cardiac output, stroke volume, blood volume, and red cell mass — the predominant physiological determinants of V˙O2max (7,8). In addition, a survey of approximately 3000 elite male endurance athletes who regularly compete at the international level has identified no common variants associated with unusually high V˙O2max values in these men (27). Of note, it seems reasonable to assume that these men’s training is likely sufficient to generate maximum biological adaptations in their O2 transport systems.
In contrast, a rare variant in at least one Olympic champion skier has been associated with high red cell mass and hemoglobin values have been associated with a very high V˙O2max value (28). In addition, both blood doping and exogenous erythropoietin (EPO) administration can increase V˙O2max in both trained and untrained subjects (9,29,30). The finding of a rare variant related to EPO in a single outlier human is important because emerging concepts in genetics research have argued that such findings can inform the search for more common variants that might affect the same biological systems. In this context, it is interesting that common gene variants in EPO-related signaling pathways have not emerged in studies of large numbers of elite endurance athletes with high V˙O2max values (31).
The previous discussion on genetic factors versus the well-established physiological determinants that explain V˙O2max is perhaps informed by recent theoretical work on complex traits (32). This work has advanced the idea that certain “core genes” or genetic pathways can be major players in a given phenotype with modifying contributions from many more pathways. The inability so far to identify specific genes and variants associated in a major way with the key physiological pathways central to exercise capacity leads us to question the extent to which this conceptual model applies to V˙O2max. The standard response to this critique is that larger sample sizes and more sophisticated modeling are needed to unravel this problem. A counterresponse is that if no obvious signal is seen in large groups of outlier phenotypes, then the odds of such a signal emerging via a larger sample size in the general population seems low. In either case, it should be remembered that as late as 2006, at least one leading proponent of genetic causation (Francis Collins) was quoted in the Wall Street Journal as suggesting that “… there are about 12 genes involved [in diabetes], and that all of them will be discovered in the next two years,” (33). Although this was not an exercise-specific comment, it highlights the earlier expectation that a limited number of common variants would explain complex human phenotypes.
Another response to the argument that the DNA variants reported to date are not linked to the known physiological determinants of V˙O2max is that they may be drivers of adaptability via regulation of replication, transcription, translation, autophagy, apoptosis, angiogenesis, miRNAs, and other molecular transducers of the adaptations to the repeated stress of exercise that in the end impact the determinants of the Fick equation. In this context, we note that substantial responses to exercise training still occur in genetic knockouts of PGC1α and other pathways proposed as master regulators of many exercise responses (34,35). We interpret these data as highlighting that the responses to exercise are incredibly redundant with many potential biological pathways possible to generate a phenotypic response. Along similar lines, lessons from experimental evolution studies in model organisms show predicable convergent phenotypes but unpredictable karyotypes in response to a given selective pressure. Although perhaps not directly related to short-term adaptations to exercise training in humans, these observations are another example of the concept that multiple pathways to a given phenotype are possible (36).
Back to the Future: Is a Next Wave of Twin Studies Needed?
So far, we have defined the concept of trainability as it applies to V˙O2max and we have reviewed why this is important from both an individual wellness and public health perspective. For healthy young and middle-aged people, we have concluded the following: 1) trainability varies among people, 2) if given a sufficient stimulus, the likelihood of a true nonresponder in these age groups seems low, 3) it is not known if extremely rigorous training over years will alter the range of V˙O2max increases with training or merely shift them upwards, 4) twin and family studies suggest a significant heritable (genetic) component to trainability, 5) the search for genetic explanations that might explain trainability has not identified key variants that intersect in a clear-cut way with the deterministic physiological pathways. In addition to these five main points, there are issues like sex differences and aging that might further confound concepts related to trainability, not to mention specific disease states.
When we consider the aforementioned five main points, it seems to us they can be collapsed into three central questions that could be addressed in comprehensive studies on mono and dizygotic twins: first, will rigorous training reduce the variability in V˙O2max observed between untrained dizygotic twins (24,25)? Second, in dizygotic twins, will rigorous training increase the correlation in V˙O2max between twin A and twin B so that it is similar to the more uniform V˙O2max values seen in monozygotic twins (23–25)? Third, will rigorous training in either dizygotic or monozygotic twins expand or reduce the range of increases in V˙O2max seen among groups of twins? In this context, it is important to note that the most informative aerobic exercise training study on twins that focused on V˙O2max was conducted in monozygotic twins, and we would like to see an intervention that included evaluation of expanded dose-response-duration relation in both mono and dizygotic twins (23).
A simple experimental design might include a period of 10 wk of guideline-based training and outcome measurements followed by 10 wk of a rigorous program similar to that pioneered by Hickson (37) in the late 1970s. In addition to V˙O2max, key outcome variables might include changes in stroke volume and cardiac size along with red cell mass. In addition, the interpretation of such data is likely to be straightforward. If the responses to our three questions converge in mono and dizygotic twins after rigorous training, we would argue that trainability can be more about the training stimulus than genetic factors in response to high-dose training. If the responses diverge, then the role of genetic factors as determinants of trainability would be emphasized. Figures 3A–C show idealized results demonstrating these possible categories of outcomes.
An obvious limitation to our paradigm is that it will be “small-N” and thus unlikely to shed light on any specific gene variants or pathways that might play a role in trainability. However, our paradigm will provide fundamental insights into the concepts related to trainability and its biological basis. It also will provide an important interventional study that will test the primacy of the physiological pathways that are largely seen as deterministic for V˙O2max. If the findings from our hypothetical study were to show increased heritability estimates for V˙O2max (and also cardiac output and red cell volume as key contributors to trainability), then a case to continue to search for additional gene variants and complex networks that might explain trainability would be strengthened. If the findings suggest that the training stimulus per se is a more critical factor, then perhaps reductionist searches for gene variants related to trainability and V˙O2max might be deemphasized.
1. McGuire KA, Ross R. Incidental physical activity is positively associated with cardiorespiratory fitness. Med. Sci. Sports Exerc
. 2011; 43(11):2189–94.
2. Nes BM, Janszky I, Vatten LJ, Nilsen TI, Aspenes ST, Wisløff U. Estimating V˙O2peak
from a nonexercise prediction model: the HUNT Study, Norway. Med. Sci. Sports Exerc
. 2011; 43(11):2024–30.
3. Ross R, Blair SN, Arena R, et al. Importance of Assessing Cardiorespiratory Fitness in Clinical Practice: A Case for Fitness as a Clinical Vital Sign: A Scientific Statement From the American Heart Association. Circulation
. 2016; 134(24):e653–99.
4. Lavie CJ, Arena R, Swift DL, et al. Exercise and the cardiovascular system: clinical science and cardiovascular outcomes. Circ. Res
. 2015; 117(2):207–19.
5. DeFina LF, Haskell WL, Willis BL, et al. Physical activity versus cardiorespiratory fitness: two (partly) distinct components of cardiovascular health? Prog. Cardiovasc. Dis
. 2015; 57(4):324–9.
6. Joyner MJ, Green DJ. Exercise protects the cardiovascular system: effects beyond traditional risk factors. J. Physiol
. 2009; 587(Pt 23):5551–8.
7. Joyner MJ, Casey DP. Regulation of increased blood flow (hyperemia) to muscles during exercise: a hierarchy of competing physiological needs. Physiol. Rev
. 2015; 95(2):549–601.
8. Lundby C, Montero D, Joyner M. Biology of V˙O2max: looking under the physiology lamp. Acta. Physiol. (Oxf.)
. 2017; 220(2):218–28.
9. Lundby C, Robach P, Saltin B. The evolving science of detection of 'blood doping'. Br. J. Pharmacol
. 2012; 165(5):1306–15.
10. CDC. Current physical activity guidelines CDC 2008. Available at: https://www.cdc.gov/cancer/dcpc/prevention/policies_practices/physical_activity/guidelines.htm
. Accessed April 6, 2018.
11. Montero D, Lundby C. Refuting the myth of non-response to exercise training: 'non-responders' do respond to higher dose of training. J. Physiol
. 2017; 595(11):3377–87.
12. Bouchard C, An P, Rice T, et al. Familial aggregation of V˙O2max
response to exercise training: results from the HERITAGE Family Study. J. Appl. Physiol. (1985)
. 1999; 87(3):1003–8.
13. Skinner JS, Jaskólski A, Jaskólska A, et al. Age, sex, race, initial fitness, and response to training: the HERITAGE Family Study. J. Appl. Physiol. (1985)
. 2001; 90(5):1770–6.
14. Ross R, de Lannoy L, Stotz PJ. Separate effects of intensity and amount of exercise on interindividual cardiorespiratory fitness response. Mayo Clin. Proc
. 2015; 90(11):1506–14.
15. Bouchard C, Blair SN, Church TS, et al. Adverse metabolic response to regular exercise: is it a rare or common occurrence? PLoS One
. 2012; 7(5):e37887.
16. Leifer ES, Mikus CR, Karavirta L, et al. Adverse cardiovascular response to aerobic exercise training: is this a concern? Med. Sci. Sports Exerc
. 2016; 48(1):20–5.
17. Bacon AP, Carter RE, Ogle EA, Joyner MJ. V˙O2max trainability
and high intensity interval training in humans: a meta-analysis. PLoS One
. 2013; 8(9):e73182.
18. Howden EJ, Perhonen M, Peshock RM, et al. Females have a blunted cardiovascular response to one year of intensive supervised endurance training. J. Appl. Physiol
. 2015; 119(1):37–46.
19. Scharhag-Rosenberger F, Meyer T, Walitzek S, Kindermann W. Time course of changes in endurance capacity: a 1-yr training study. Med. Sci. Sports Exerc
. 2009; 41(5):1130–7.
20. Jensen K, Johansen L, Secher NH. Influence of body mass on maximal oxygen uptake: effect of sample size. Eur. J. Appl. Physiol
. 2001; 84(3):201–5.
21. Tonnessen E, Sylta O, Haugen TA, Hem E, Svendsen IS, Seiler S. The road to gold: training and peaking characteristics in the year prior to a gold medal endurance performance. PLoS One
. 2014; 9(7):e101796.
22. Spina RJ, Ogawa T, Kohrt WM, Martin WH 3rd, Holloszy JO, Ehsani AA. Differences in cardiovascular adaptations to endurance exercise training between older men and women. J. Appl. Physiol
. 1993; 75(2):849–55.
23. Prud'homme D, Bouchard C, Leblanc C, Landry F, Fontaine E. Sensitivity of maximal aerobic power to training is genotype-dependent. Med. Sci. Sports Exerc
. 1984; 16(5):489–93.
24. Klissouras V. Heritability of adaptive variation. J. Appl. Physiol
. 1971; 31(3):338–44.
25. Klissouras V, Pirnay F, Petit JM. Adaptation to maximal effort: genetics and age. J. Appl. Physiol
. 1973; 35(2):288–93.
26. Bouchard C, Sarzynski MA, Rice TK, et al. Genomic predictors of the maximal O2
uptake response to standardized exercise training programs. J. Appl. Physiol
. 2011; 110(5):1160–70.
27. Rankinen T, Fuku N, Wolfarth B, et al. No evidence of a common DNA variant profile specific to world-class endurance athletes. PLoS One
. 2016; 11(1):e0147330.
28. de la Chapelle A, Traskelin AL, Juvonen E. Truncated erythropoietin receptor causes dominantly inherited benign human erythrocytosis. Proc. Natl. Acad. Sci. U. S. A
. 1993; 90(10):4495–9.
29. Ekblom BT. Blood boosting and sport. Baillieres Best Pract. Res. Clin. Endocrinol. Metab
. 2000; 14(1):89–98.
30. Jelkmann W, Lundby C. Blood doping and its detection. Blood
. 2011; 118(9):2395–404.
31. Bomba L, Walter K, Soranzo N. The impact of rare and low-frequency genetic variants in common disease. Genome Biol
. 2017; 18(1):77.
32. Boyle EA, Li YI, Pritchard JK. An expanded view of complex traits: from polygenic to omnigenic. Cell
. 2017; 169(7):1177–86.
33. Regalado A. New genetic tools may reveal roots of everyday ills: rapid DNA tests can search many variations at once; probing obesity, memory. The Wall Street Journal. Available at: https://www.wsj.com/articles/SB114497175117125656
. Accessed April 6, 2018.
34. Ballmann C, Tang Y, Bush Z, Rowe GC. Adult expression of PGC-1α and -1β in skeletal muscle is not required for endurance exercise-induced enhancement of exercise capacity. Am. J. Physiol. Endocrinol. Metab
. 2016; 311(6):E928–38.
35. Tanner CB, Madsen SR, Hallowell DM, et al. Mitochondrial and performance adaptations to exercise training in mice lacking skeletal muscle LKB1. Am. J. Physiol. Endocrinol. Metab
. 2013; 305(8):E1018–29.
36. Simões P, Fragata I, Seabra SG, et al. Predictable phenotypic, but not karyotypic, evolution of populations with contrasting initial history. Sci. Rep
. 2017; 7(1):913.
37. Hickson RC, Bomze HA, Holloszy JO. Linear increase in aerobic power induced by a strenuous program of endurance exercise. J. Appl. Physiol. Respir. Environ. Exerc. Physiol
. 1977; 42(3):372–6.