Numerous longitudinal studies have shown that exercise intensity(as well as frequency, duration, and fitness) is an important training variable for improving maximal oxygen consumption (e.g.,31,32). Relatively few longitudinal endurance training studies have compared the influence of different exercise training intensities upon lactate and/or ventilatory thresholds, and the results have been equivocal (14,24,30,38). However, many experiments have reported changes in lactate and/or ventilatory thresholds with endurance training, and different treatment intensities were used by the various investigators. The time course of changes in lactate and/or ventilatory thresholds in response to exercise training has appeared to be asymptotic with most of the benefits occurring early in the training program (7,34,38). To date, no systematic analysis of the literature on the importance of training intensity upon lactate and ventilatory thresholds has been published. The purpose of this investigation was to merge the results of these independent experiments in a systematic and rigorous manner (meta-analysis) to determine the importance of training intensity for producing changes in the oxygen consumption associated with lactate and/or ventilatory thresholds in subjects with different levels of conditioning.
The literature was searched for published longitudinal training studies which measured either lactate threshold and/or ventilatory threshold. The author has maintained a file of studies on this topic for the past 24 years. In addition, a Medline search was conducted using the following terms: lactate threshold, ventilatory threshold, anaerobic threshold, aerobic threshold, exercise training, exercise, and lactic acid. References cited in these papers were located as well. The review covered the years from 1967 to 1994. Williams et al. (40) were the first to measure the effect of exercise training on a lactate threshold type variable (breakpoint in excess lactate). Prior to 1967, the training studies measured lactate response to a standard steady-state exercise test. Several studies were not included because of inadequate information or because data were reported in more than one publication.
The definitions of lactate and/or ventilatory thresholds varied somewhat among studies, but as long as the assessment procedures were the same at the starting, intermediate (if available), and ending times, the study was included in the pool. Data presented by Weltman et al.(38) suggested that various lactate threshold related variables (e.g., lactate threshold and fixed blood lactate concentrations) responded similarly to exercise training. In other words, the specific manner of measuring lactate threshold probably is not important if the procedures are consistent within a study. Therefore, all lactate threshold type data were merged for the data analyses. Conversely, Poole and Gaesser(23) reported that the lactate and ventilatory thresholds responded differently to a particular training stimulus; therefore, ventilatory threshold type variables (break in ventilatory volume, respiratory exchange ratio, volume of carbon dioxide, ventilatory equivalent for oxygen or carbon dioxide) were merged but analyzed separately from the lactate threshold data.
The preliminary dependent variables were oxygen consumption mean and SD in liters per minute at the threshold. When the data were reported in other units(e.g., mL·kg-1·min-1,%˙VO2max, SE) the data were converted via appropriate calculations. Other data that were recorded included: number of subjects, age, gender, fitness (sedentary, 0; conditioned, 1), maximal oxygen consumption, training intensity (see below), frequency of training, duration of training in weeks, and comparative statistics. Assignment to training intensity was made on the basis of descriptions provided in each study as follows: -1 or detraining: the subjects stopped training or became less active (conditioning vs downhill skiing); 0 or control: maintained current activity level; 1: trained below lactate or ventilatory threshold; 2: trained at the lactate or ventilatory threshold(intensity just prior to a nonlinear increase in lactate or ventilatory variables, respectively); 3: trained at the onset of blood lactate accumulation (OBLA - 4 mM·L-1) or respiratory compensation threshold; and 4: trained between intensity 3 and up to maximal. In some cases the intensity was not provided relative to the various thresholds; this required judgment in coding. For example, 40% ˙VO2max was coded as 1 (below threshold), 80% ˙VO2max was coded as 3 (OBLA or respiratory compensation threshold), and anything above 80% ˙VO2max was coded 4. When a study reported data on more than one group, each group was analyzed separately (study groups). When testing was conducted at intermediate points (at least 8 wk apart) during the study, the data were split into separate study groups for each sequential pair of tests, e.g., test 1 to test 2 was a study group, test 2 to test 3 was a study group, etc. When sequential testing occurred in studies that lasted more than 8 wk, subjects who were classified as sedentary for the first study group were classified as trained for subsequent study groups; it appeared that most of the training effect occurred during the first 8-10 wk. Sequential testing generally was performed in studies that were longer in duration, so splitting reduced the range of study durations. Treatment study groups' duration ranged from 3 to 18 wk (in contrast to 3-52 wk before splitting) with most of them lasting from 8 to 12 wk. In the 3-wk studies subjects trained six times per week in contrast to three times per week in many studies.
None of the included studies reported a within-subjects SD although it is the appropriate measure for determining the SE of the difference for testing the significance of the changes over time.1 However, all studies reported SD for each testing period except one that did not include a value for the posttest. For the missing posttest value, the pretest value was used. The SD of the differences could be calculated from:Equation
where s2p is the pooled pretest/posttest variance and rxy is the correlation between the pretest and posttest scores(11). The reliability of the threshold type tests has been reported to be approximately 0.90(2,25,39). Estimated rxy values were determined for several of the current studies based on statistics reported, e.g., pretest-posttest correlation or t-values and these calculated values were consistent with the reported reliabilities. Therefore, it was assumed that rxy was 0.90 for all of the studies.
The data were analyzed using methods described by Hedges and Olkin(13). Effect sizes (ES) were calculated from the mean pretest to posttest differences divided by the within-subjects SD. These effect sizes were corrected for small sample sizes. Weighted effect sizes were determined by dividing the corrected effect size by the corresponding variance of the effect size. The final derived value for each study group was determined by dividing the squared corrected effect size by the variance of the effect size.
Subsequent analyses involved testing of all effect sizes, the effect sizes within classes, and the effect sizes among classes. The inverse of the variance of the individual study group effect sizes was used as the weighting factor. The resulting Q values were tested with Chi square. Significant-among-classes results were followed with Bonferroni tests.
The independent variables were type of threshold, intensity, and fitness, and the dependent variable was weighted effect size of oxygen consumption at the specified lactate or ventilatory threshold. Because comparable data were not available for all levels of these independent variables, separate analyses were performed. In addition, analyses were conducted on combinations of the independent variables where comparable data were available.
Figure 1 presents the data in a scattergram format. When all data were pooled and analyzed, the only significance was that the detraining group was different than the trained groups. When the data were split into separate analyses on the basis of type of threshold and fitness level, several effects were evident. However, there still was considerable variability in treatment effects. Therefore, all study groups that deviated from the mean of similar study groups by more than two SD's were analyzed for extenuating factors that may have contributed to their deviant results. Five deviant studies (3,4,18,28,29) were discarded because of the age of the subjects, disease, medications, the use of walking as the treatment, the fact that treatment was running but testing was performed on a cycle ergometer, and/or interval training; all but one had two or more of these factors. One deviant walking study group was retained because the subjects were competitive race walkers(42). The analyses reported here included the remaining 29 studies and 69 study groups.
The original data included 34 studies with 85 study groups (seeTable 1). Significant treatment effects were found in 29 studies. Nine of 12 ventilatory threshold studies and 23 of 25 lactate threshold studies found treatment effects. Fourteen of 16 studies lacking comparison groups demonstrated a treatment effect, while 15 of 18 studies with comparison groups showed a treatment effect. In four of the five studies that reported no treatment effect, comparisons were performed usingt-tests (the author ran paired t-tests on two studies); conversely, 11 of 15 studies in which t-tests were used did report treatment effects. Twelve of 18 studies which included comparison groups equated these groups randomly or by matching. No patterns were evident regarding the effect size versus gender, frequency of sessions, or duration of sessions.
Summary data for the lactate threshold studies are presented inTable 2 and Figure 2. The change in lactate threshold (ΔTHLa) and the corrected effect size(EScorr) values are presented for points of reference only, and the statistical analyses were performed on the weighted effect sizes. For sedentary subjects each training intensity group improved in comparison with the control group (P < 0.01), but there were no significant differences among the training intensities (P > 0.13). In the conditioned subjects, detraining reduced lactate threshold in comparison with training groups (P < 0.05), but there were no other significant differences (the control group vs the detraining group (P < 0.06), the control group and training groups (P > 0.10) or among the training groups (P > 0.36)). All three low fitness groups improved more than the two lower intensity, higher fitness groups (P< 0.003 - 0.013). A two-way analysis of three training intensities (2, 3, and 4) by fitness showed that sedentary groups improved more than the conditioned groups (P < 0.05), and there was a significant intensity by fitness interaction (P < 0.02) in that only the highest intensity appeared to be effective in the conditioned groups.
Summary data for the ventilatory threshold studies are presented inTable 3 and Figure 3. For the sedentary comparisons the lowest (P < 0.002), second lowest(P < 0.009), and highest (P < 0.007) training intensities produced significant improvements over the control group, but there were no significant differences among the training intensities(P > 0.219). These same sedentary groups improved more than the high fitness groups (P < 0.001-0.006). For the conditioned groups the only significant differences were that each treatment intensity was greater than the detraining group (P < 0.001). A two-way analysis of two levels of training intensities (2 and 3) and fitness status produced no significant differences (P > 0.05).
Comparable data across the intensity × fitness × threshold matrix were available only for intensities 2 and 3(Tables 2 and 3). This analysis showed significant results for the fitness by intensity interaction (P < 0.05), fitness by threshold interaction (P < 0.05), and fitness × intensity × threshold interaction (P < 0.05). The fitness by intensity interaction appeared to be a result of sedentary subjects showing a slightly greater response on average to the higher training intensity, while the conditioned subjects responded marginally less to the higher training intensity. The fitness by threshold interaction appeared to be a result of divergent threshold differences in the effect for sedentary study groups versus similar effects for the fit study groups. The fitness × intensity× threshold interaction occurred because sedentary subjects responded less on ventilatory threshold to the moderate intensity than the low intensity, while the response of lactate threshold was reversed. None of the other factors were significant: intensity (P > 0.50), fitness(P > 0.05), type of threshold (P > 0.90), and intensity by threshold interaction (P > 0.30).
Meta-analysis attempts to combine the data from many studies so that the overall effects can be analyzed. Several problems arise during such analyses. In this case it was assumed that merging threshold data that were collected in different ways was reasonable. Within-subjects variability was not reported in any of the studies but was calculated assuming a correlation of 0.90 between pre- and post-test scores. Training intensities were coded from descriptions provided by the study authors. In some cases these intensities were not provided relative to the various thresholds; this required a judgment for coding. Training frequencies, training time per session, study durations, and subject characteristics varied among the studies. The data from various studies were provided in different formats and required calculations to convert them to a common format. Some study results were outliers, and subjective judgments were required to determine if there were valid reasons for discarding them. All of the problems were dealt with in a rigorous manner, and it was felt that these analyses were as valid as possible under the circumstances.
The results showed that endurance training at a lactate threshold intensity or higher improved lactate threshold in sedentary subjects. No studies were available for the effect of below threshold training on lactate threshold. The effect on ventilatory threshold was not as clear, but three of four training intensities (1, 2, and 4) produced significant effects, while intensity 3 was not quite significant (ρ < 0.056). The lack of significance probably was caused by rather small effects in several of the ventilatory threshold study groups (9,10,12). In addition, two studies which measured both lactate and ventilatory threshold concluded that the latter was less responsive to training than the former(9,23). These results suggest that metabolic adaptations at the muscle level occur more readily than ventilatory control changes. However, the lactate and ventilatory threshold effect sizes were of similar magnitude overall and the fitness × intensity × threshold analysis showed that the differences between these effect sizes were not significant.
Conditioned subjects clearly showed smaller changes in the threshold values than sedentary subjects. Although the improvement in conditioned subjects was not statistically significant, it was close to being so. This asymptotic trend also was evident in the long-term studies that reported measurements at intermediate times (7,34,39) and probably reflects fitness by training intensity limitations, i.e., a point of diminishing returns. Although not significant, it appeared that in the conditioned subjects a higher training intensity (intensity 4) was required to produce a treatment effect on the lactate variables. No ventilatory threshold studies with conditioned subjects used such high intensity training, so it is not known how general this phenomenon may be.
The primary purpose of the current analyses was to investigate the effect of training intensity on the thresholds. In general, intensity did not appear to be a major factor. All the one-way analyses showed that there were no significant differences among the training intensities (2, 3, and 4 for the lactate threshold studies and 1-4 for the ventilation threshold studies). The more complex statistical procedure identified a fitness by intensity interaction; the conditioned subjects responded better to the highest intensity on the lactate threshold variables. However, only one conditioned study group was available for the highest intensity for the lactate threshold. Since comparable ventilatory threshold data were not available, it may be prudent to withhold judgment on the fitness by intensity interaction effect.
At the other end of the intensity spectrum, one would expect a minimum intensity (training intensity threshold) to produce a training effect in sedentary subjects. It was clear that an intensity corresponding to the lactate and/or ventilatory threshold was as effective as higher intensities. Only one study group trained at lower than this intensity (at 40% of˙VO2max) (30) and in the current analysis showed a significant improvement. However, in the original study this effect was not significant but probably was close, while a group that trained at 80% of ˙VO2max did improve. The authors concluded that it was necessary to train at an intensity above the ventilatory threshold to produce a training effect. Apparently the pretest-posttest correlation was less than the assumed r = 0.90 used in the current investigation (thereby leading to a deflated within subjects standard deviation and inflated effect size). Conversely, the detraining study groups clearly showed a decrease in the threshold values. In fact, Karvonen (17) reported a decline in lactate threshold with 13 wk of reduced training intensity (Alpine skiing) in previously endurance trained subjects. Unpublished data from our laboratory suggested that the minimum necessary intensity to produce a training effect is slightly below the lactate threshold.
In summary, endurance training at an intensity near the ventilatory threshold (and probably the lactate threshold) is necessary to produce a training effect. Training at higher intensities has minimal benefit over the minimum intensity in sedentary subjects, but may be beneficial in conditioned subjects. Most of the changes that eventually will occur take place in the first 8-12 wk of training, but small changes may accrue beyond this period. Detraining will occur if training intensity is reduced.
Using an among-subjects variance as suggested by Looney et al. (20) did not appear appropriate for theoretical and practical reasons. The proper error variance for an individual repeated measures design is the appropriate within-subjects mean square, and many of the studies followed this approach in one form or another (dependent“t”, REANOVA, or ANCOVA). Most studies with sedentary subjects found statistically significant changes, but when the procedures of Looney et al. were used, there were no significant results.
1. Acevedo, E. O. and A. H. Goldfarb. Increased training intensity effects on plasma lactate, ventilatory threshold, and endurance.Med. Sci. Sports Exerc.
2. Aunola, S. and H. Rusko. Comparison of two methods for aerobic threshold determination. Eur. J. Appl. Physiol.
3. Belman, M. J. and G. A. Gaesser. Exercise training below and above the lactate threshold in the elderly. Med. Sci. Sports Exerc.
4. Blumenthal, J. A., C. F. Emery, D. J. Madden, et al. Effects of exercise training on cardiorespiratory function in men and women>60 years of age. Am. J. Cardiol.
5. Coyle, E. F., W. H. Martin, III, S. A. Bloomfield, O. H. Lowry, and J. O. Holloszy. Effects of detraining on responses to submaximal exercise. J. Appl. Physiol.
6. Davis, J. A., M. H. Frank, B. J. Whipp, and K. Wasserman. Anaerobic threshold alterations caused by endurance training in middle-aged men. J. Appl. Physiol.
7. Denis, C., R. Fouquet, P. Poty, A. Geyssant, and J. R. Lacour. Effect of 40 weeks of endurance training on the anaerobic threshold.Int. J. Sports Med.
8. Ekblom, B., P-O. Åstrand, B. Saltin, J. Stenberg, and B. Wallstrom. Effect of training on circulatory response to exercise.J. Appl. Physiol.
9. Gaesser, G. A. and D. C. Poole. Lactate and ventilatory thresholds: disparity in time course of adaptations to training. J. Appl. Physiol.
10. Gaesser, G. A., D. C. Poole, and B. P. Gardner. Dissociation between ˙VO2max
and ventilatory threshold responses to endurance training. Eur. J. Appl. Physiol.
11. Glass, G. V., B. McGaw, and M. L. Smith.Meta-Analysis in Social Research
. London: Sage: 1981, p. 117.
12. Golden, H. P. and P. Vaccaro. The effects of endurance training intensity on the anaerobic threshold. J. Sports Med. Phys. Fit.
13. Hedges, L. V. and I. Olkin. Statistical Methods for Meta-Analysis
. New York: Academic Press, 1985, pp. 147-165.
14. Henritze, J., A. Weltman, R. L. Schurrer, and K. Barlow. Effects of training at and above the lactate threshold on the lactate threshold and maximal oxygen uptake. Eur. J. Appl. Physiol.
15. Hill, D. W., K. J. Cureton, S. C. Grisham, and M. A. Collins. Effect of training on the rating of perceived exertion at the ventilatory threshold. Eur. J. Appl. Physiol.
16. Hurley, B. F., J. M. Hagberg, W. K. Allen, et al. Effect of training on blood lactate levels during submaximal exercise.J. Appl. Physiol.
17. Karvonen, J., E. Rauhala, and J. Chwalbinska-Moneta. The effect of three months' slalom training on physical performance capacity.J. Sports Med. Phys. Fit.
18. Keyser, R. E., K. Delafuente, and J. McGee. Arm and leg cycle cross-training effect on anaerobic threshold and heart rate in patients with coronary heart disease. Arch. Phys. Med. Rehabil.
19. Kim, H. S., K. Tanaka, and K. Maeda. Effects of exercise training at an intensity relative to lactate threshold in mildly obese women. Ann. Physiol. Anthrop.
20. Looney, M. A., C. J. Feltz, and C. N. VanVleet. The reporting and analysis of research findings for within-subjects designs: Methodological issues for meta analysis. Res. Q. Exerc. Sport
21. Mayes, R., A. E. Hardman, and C. Williams. The influence of training on endurance and blood lactate concentration during submaximal exercise. Br. J. Sports Med.
22. Pierce, E. F., A. Weltman, R. L. Seip, and D Snead. Effects of training specificity on the lactate threshold and VO2peak
.Int. J. Sports Med.
23. Poole, D. C. and G. A. Gaesser. Response of ventilatory and lactate thresholds to continuous and interval training. J. Appl. Physiol.
24. Poole, D. C., S. A. Ward, and B. J. Whipp. The effects of training on the metabolic and respiratory profile of high-intensity cycle ergometer exercise. Eup. J. Appl. Physiol.
25. Posner, J. D., K. M. Gorman, H. S. Klein, and C. J. Cline. Ventilatory threshold: measurement and variation with age. J. Appl. Physiol.
26. Ramsbottom, R., C. Williams, N. Fleming, and M. L. G. Nute. Training induced physiological and metabolic changes associated with improvements in running performance. Br. J. Sports Med.
27. Ready, A. E. and H. A. Quinney. Alterations in anaerobic threshold as a result of endurance training and detraining.Med. Sci. Sports Exerc.
28. Rotstein, A., R. Dotan, O. Bar-Or, and G. Tenenbaum. Effect of training on anaerobic threshold, maximal aerobic power, and anaerobic performance of preadolescent boys. Int. J. Sports Med.
29. Rumley, A. G., R. Taylor, S. Grant, A. R. Pettigrew, I. Findlay, and H. Dargie. Effect of marathon training on the plasma lactate response to submaximal exercise in middle-aged men. Br. J. Sports Med.
30. Sady, S., V. Katch, P. Freedson, and A. Weltman. Changes in metabolic acidosis: evidence for an intensity threshold. J. Sports Med. Phys. Fit.
31. Sharkey, B. J. Intensity and duration of training and the development of cardiorespiratory endurance. Med. Sci. Sport
32. Shephard, R. J. Intensity, duration, and frequency of exercise as determinants of the response to a training regime. Int. Z. Angew. Physiol.
33. Sjodin, B., I. Jacobs, and J. Svedenhag. Changes in onset of blood lactate accumulation (OBLA) and muscle enzymes after training at OBLA. Eur. J. Appl. Physiol.
34. Smith, D. A. and T. V. O'Donnell. The time course during 36 week's endurance training of changes in ˙VO2max
and anaerobic threshold as determined with a new computerized method. Clin. Sci.
35. Tanaka, K., F. Nakadomo, H. Kitao, H. Watanabe, and S. Sumida. Physiologic status at 1 year follow-up of obese women engaged in a supervised conditioning program. Ann. Physiol. Anthrop.
36. Tanaka, K., H. Watanabe, Y. Konishi, et al. Longitudinal associations between anaerobic threshold and distance running performance. Eur. J. Appl. Physiol.
37. Thomas, S. G., D. A. Cunningham, J. Thompson, and P. A. Rechnitzer. Exercise, training, and “ventilation threshold” in elderly. J. Appl. Physiol.
38. Weltman, A., R. L. Seip, D. Snead, et al. Exercise training at and above the lactate threshold in previously untrained women.Int. J. Sports Med.
39. Weltman, A., D. Snead, P. Stein, et al. Reliability and validity of a continuous incremental treadmill protocol for determination of lactate threshold, fixed blood lactate concentration, and ˙VO2max
.Int. J. Sports Med.
40. Williams, C. G., C. H. Wyndham, R. Kok, and M. J. E. von Rahden. Effect of training on maximal oxygen intake and on anaerobic metabolism in man. Int. Z. Angew. Physiol.
41. Yoshida, T., Y. Suda, and N. Takeuchi. Endurance training regimen based upon arterial lactate: effects on anaerobic threshold.Eur. J. Appl. Physiol.
42. Yoshida, T., M. Udo, M. Chida, M. Ichioka, K. Makiguchi, and T. Yamaguchi. Specificity of physiological adaptation to endurance training in distance runners and competitive walkers. Eur. J. Appl. Physiol.