According to the International Diabetes Federation, the number of individuals with diabetes in the world has increased to 415 million in 2015 (14). In the future, it will increase to 642 million before 2040 if strategies for diabetes prevention are absent (14). Physical inactivity is a major risk factor for the onset of type 2 diabetes mellitus (T2DM) (15), and physical activity is effective in preventing T2DM (6).
Currently, the benefit of maintaining high cardiorespiratory fitness (CRF), an objective marker of physical activity level (5,12,13), for the prevention of T2DM is widely accepted. This knowledge is mainly based on the evidence that higher compared with lower CRF at baseline is associated with lower risk of T2DM during follow-up (3). In a clinical setting, health care professionals often suggest their clients to improve or keep their CRF level by regular physical activity for the risk reduction of T2DM. However, they may not be able to indicate how long he/she must keep their CRF level higher, because whether consistently high CRF is necessary or transiently high CRF is sufficient for the risk reduction of T2DM is not been determined.
We hypothesized that consistently high CRF over time may be more beneficial than transiently high CRF for the prevention of T2DM. The purpose of this study was to examine the associations of consistently high CRF and transiently high CRF with future T2DM development using a long-term cohort design among Japanese men with multipoint measurement of CRF.
This is a cohort study of the relationship between CRF and health outcomes among Japanese men (20–22). The study participants were working for a gas supplier in Tokyo, and 13,009 workers had an annual health checkup between April 1986 and March 1987. Of these, 9221 workers underwent CRF exercise test. To obtain the total level of CRF over time, 7806 participants who underwent the exercise test for four times or more from April 1979 until March 1987 were selected. We excluded those who could not continue the CRF exercise test for over 4 min due to the appearance of abnormal electrocardiographic findings (n = 23). Women were not analyzed because the sample size (n = 384) was not large enough. Men with a history of cardiovascular disease (n = 6), stroke (n = 3), or diabetes (n = 44) before 1986 were excluded. Of the remaining 7346 participants, 152 who resigned at the end of year 1987 were excluded. Additionally, 36 men with a lack of data on potential confounders were excluded. Thus, a total of 7158 participants age 20 to 60 were followed-up until March 2010.
In our observational study, the clinical examinations were done under the Industrial Safety and Health Act and related laws in Japan. Therefore, we did not need to get written informed consent. This study was approved by the ethics committee of the National Institutes of Biomedical Innovation, Health and Nutrition.
Height, body weight, and resting blood pressure were measured in 1986. Body mass index (BMI) was calculated from height and body weight. The resting blood pressure was measured using an automatic blood pressure meter in a sitting position. In addition, we evaluated alcohol, smoking, desk work, and a family history of diabetes as potential confounders using a questionnaire.
Estimated maximal aerobic power was used as a measure of CRF using a submaximal exercise test with a cycle ergometer. This test consists of three steps (4 min per step). The initial loads for the participants in the 19–29, 30–39, 40–49, and over 50-yr age brackets were 98, 86, 74, and 61 W, respectively. The HR was measured based on the R-R interval on the electrocardiogram. The target HR was established as 85% of their age-predicted maximum HR (220 − age in years). The load was increased by 37 W per step until the target HR was reached. Based on the HR in each subject, we estimated maximal aerobic power using the Åstrand–Ryhming nomogram and Åstrand’s nomogram correction factors (1,2). The estimated maximal aerobic power was highly correlated with the direct method in a previous study (r = 0.92) (27).
To evaluate the total level of CRF over time, we calculated the area under the curve with respect to ground (AUCG) for CRF measurements between April 1979 and March 1987 in individuals (see Appendix Figure 1, Supplemental Digital Content 1, which represent the scheme for calculating the AUC for CRF, https://links.lww.com/MSS/A937) (19). Because the AUCG for CRF represents an integrated CRF level during the period, those with higher level of AUCG for CRF can be regarded as being able to keep their CRF level higher during the period. Because the frequency of CRF measurement during the period differed among the participants (four times, 538 men; five times, 817; six times, 1500; seven times, 2363; and eight times, 1951), the AUCG for CRF was standardized through dividing the total area by the years of measurement period. We calculated the AUCG for CRF without any complementation method to the missing CRF.
To quantify the degree of a transient temporal increase in CRF during the period, the peak AUC (AUCP) was calculated using the peak CRF value alone in individuals: AUCP = peak CRF–measurement period, and then differences (ΔAUCP) between the AUCG and AUCP were calculated (Appendix Figure 1B, Supplemental Digital Content 1, which represent the scheme for calculating the AUC for CRF, https://links.lww.com/MSS/A937). ΔAUCP is zero (minimum) when all CRF values during the measurement period are equal, whereas it increases when the peak CRF value deviates from the average of CRF values excluding the peak CRF value. Therefore, this parameter reflects the presence and the size of a “spike” in CRF during the measurement period, and may be thought of as an index of transiently high CRF. Similar to the AUCG for CRF, we standardized ΔAUCP by the measurement period.
On health checkups conducted between April 1986 and March 2009, we evaluated the year of T2DM development. The criteria for the diagnosis of T2DM were mainly based on the diagnostic guidelines of the American Diabetes Association and the Japan Diabetes Society (7,11). Because fasting blood tests were basically conducted for 35-yr-olds and those age 40 yr or older from 1986 to 1993, and for 25, 30, 35 yr, and those age 40 yrs or older from 1994 to 2009, respectively, in accordance with the Industrial Safety and Health Act and related laws in Japan, we used two criteria for the determination of T2DM based on whether the participants received blood tests or not. For the participants with the blood tests, participants with a fasting blood glucose level exceeding 7.0 mmol·L−1 (126 mg·dL−1) were regarded as having diabetes. The fasting condition was confirmed by a verbal confirmation. For the participants without blood tests, the results of an oral glucose test, which was conducted when nonfasting urinary glucose test showed a positive result, was adopted for the diagnosis of T2DM from April 1986 to March 1995. The year when the serum glucose level 2 h after 75-g glucose loading was 11.1 mmol·L−1 (200 mg·dL−1) or higher was regarded as the year of T2DM development. Although urinary tests are not recommended in the latest guidelines (7,11), in 1980s, the test followed by an oral glucose tolerance test for those found to have glycosuria had been the most common approach for screening diabetes in various settings, such as at military recruitment, in preemployment examination, and as a requirement for life insurance (31). Moreover, because the urinary tests were conducted for all participants, we used the information of urinary glucose and the subsequent oral glucose test upon positive urinary glucose to determine T2DM. Although participants with a blood test received a urinary test under fasting condition, they were determined as T2DM based on the results of fasting blood tests. In addition to these chemical tests, we asked all participants whether a diagnosis of T2DM was made by a physician using a questionnaire administered during health checkups. If more than one of the above criteria was met, the year of the earliest episode was regarded as the year of T2DM development.
We grouped men into four groups (GQ1, GQ2, GQ3, and GQ4) by totaling the participants in each quartile after dividing them into quartiles with respect to age (≤30, 31–35, 36–40, 41–45, 46–50, and ≥51 yr) based on the AUCG for CRF. For descriptive data, continuous variables were expressed as the median (interquartile range [IQR]), and category variables as percentages. Moreover, we compared the baseline characteristics between the participants included in and excluded from analysis to assess potential selection bias.
We used Cox proportional hazards models to estimate the relationship between the AUCG for CRF (quartiles) and incidence of T2DM. The hazard ratios (HR) were calculated. We adjusted for potential confounders including age (continuous variable), BMI (continuous variable), systolic blood pressure (continuous variable), smoking (never-smokers, past smokers, current smokers of 1–20 cigarettes per day, and 21 or more cigarettes per day), alcohol (none, 1–40 g·d−1, and 41 g or more per day), desk work (yes or no) at baseline (1986), family history of diabetes (present or not), and frequency of measuring CRF (continuous variable) (model 1). In addition to model 1, model 2 included CRF (continuous variable) at baseline, and model 3 included the peak CRF (continuous variable) during the measurement period, respectively. Furthermore, we examined interactions between the AUCG for CRF and age or BMI at baseline on the incidence of T2DM.
To investigate the association between transiently high CRF and T2DM, we estimated the relationship between quartiles (ΔPQ1, ΔPQ2, ΔPQ3, and ΔPQ4) of ΔAUCP and incidence of T2DM. Moreover, to confirm the association between the CRF level at a point and incidence of T2DM, we calculated HR using quartiles of CRF value at the baseline year (1986) and also using the highest and lowest CRF value during the measurement period. In these analyses, the AUCG for CRF was included in model 4 as well as model 1. The proportionality assumption of the models was tested using a log-minus-log plot; no evidence of violation was found.
For sensitivity analysis, we conducted three analyses in a different viewpoint. First, we excluded participants who developed T2DM within 3 yr to eliminate the influence of possible preexisting T2DM at baseline. Second, we used more conservative criteria of evaluating T2DM development in which T2DM was diagnosed when the participants met the criteria twice or more, because there were possibly some participants who did not report noncompliance or who forgot they had eaten. Third, we analyzed the association of quartiles of AUCG and ΔAUCP for CRF with the incidence of T2DM among participants with CRF measures for 8 times to assess the influence of missing values of CRF. In addition, we also used a multiple imputation with chained equations for missing data on CRF from 1979 to 1986. All variables at baseline described in Table 1 and CRF values from 1979 to 1986 were used to impute the missing data on CRF. Twenty imputed datasets were created, and we obtained pooled estimates for the analyses.
For statistical analysis, all analyses were done using Stata version 14.1 (Stata Corp, College Station, TX) and SPSS version 22 (IBM Japan, Tokyo, Japan). All statistical tests were two-sided.
The median (IQR) age of the participants at baseline was 37 (32–45) yr. The median and maximum follow-up periods were 18 and 23 yr, respectively. During follow-up involving a total of 115,084 man-years, 1495 participants developed T2DM. During follow-up before 2009, 3491 participants dropped out. The baseline characteristics of the participants according to the quartiles of the AUCG for CRF are shown in Table 1. BMI and blood pressure were lower in groups with a higher AUCG for CRF between 1979 and 1986. The proportions of heavy smokers and drinkers were lower in groups with a higher AUCG. The proportion of participants with a family history of T2DM was lower in groups with a higher AUCG for CRF. Furthermore, the initial, peak, and minimal CRF level between 1979 and 1986 and CRF at baseline were higher in groups with a higher AUCG.
When compared with men excluded from our analysis (see Appendix Table 1, Supplemental Digital Content 2, baseline characteristics of participants according to men included in and excluded from the analysis, https://links.lww.com/MSS/A938), the men included in the analysis were older at baseline. Moreover, they had lower BMI, lower systolic and diastolic blood pressure, and lower percentage of smokers and drinkers. On the other hand, the level of CRF and the percentage of desk-worker among the men included in the analyses were lower than the men excluded from the analyses.
AUCG for CRF and T2DM
Table 2 shows the HR of T2DM according to the quartiles of the AUCG for CRF. When comparing the first quartile group (GQ1) with the other groups (GQ2, GQ3, and GQ4), the HR were lower in groups with higher AUCG (P for trend < 0.001). This negative relationship remained after adjusting for confounders in model 1 (P for trend < 0.001). Although additional adjustment for the CRF at baseline (model 2, P for trend = 0.002) or the peak CRF between 1979 and 1986 (model 3, P for trend = 0.001) attenuated this relationship, the relationship between AUCG for CRF and incidence of T2DM was obtained. Figures 1 and 2 show the cumulative incidence curves and restricted cubic spline regression after adjustment with confounders for T2DM according to the quartiles of the AUCG for CRF (model 2), respectively. Although the interaction of AUCG and age on T2DM development was not significant (P for interaction = 0.74), the negative association between the quartile of the AUCG for CRF and the incidence of T2DM was stronger among participants 40 yr or older as compared with participants 39 yr or younger (see Appendix Table 2, Supplemental Digital Content 3, hazard ratios of the incidence of T2DM according to quartiles of an index of consistent CRF stratified by the age of 40 yr, https://links.lww.com/MSS/A939). On the other hand, the interaction of AUCG and BMI on T2DM development was significant (P for interaction = 0.021, see Appendix Table 3, Supplemental Digital Content 4, hazard ratios of the incidence of T2DM according to quartiles of an index of consistent CRF stratified by BMI, https://links.lww.com/MSS/A940).
ΔAUCP for CRF and T2DM
Table 3 shows the HR of T2DM according to the quartiles of ΔAUCP as a parameter of transiently high CRF (see Appendix Table 4, Supplemental Digital Content 5, which shows the baseline characteristics of study participants according to quartiles of ΔAUCP, https://links.lww.com/MSS/A941). Before adjustment for confounders, there was a negative relationship between the quartiles of ΔAUCP and incidence of T2DM (P for trend = 0.037). However, when considering confounders (model 1, P for trend = 0.82) and CRF at baseline (model 2, P for trend = 0.09), there was no relationship between ΔAUCP and T2DM development. Figure 3 shows restricted cubic spline regression for the association between the quartiles of ΔAUCP and the incidence of T2DM.
Baseline, peak, and minimal CRF and T2DM
We also analyzed the association between the baseline, peak, and minimal CRF levels during the measurement period and incidence of T2DM, considering potential confounders. The baseline (see Appendix Table 5, Supplemental Digital Content 6, hazard ratios of the incidence of T2DM according to quartiles of the CRF at baseline, https://links.lww.com/MSS/A942), peak (see Appendix Table 6, Supplemental Digital Content 7, hazard ratios of the incidence of T2DM according to quartiles of the peak of CRF during 1979–1986, https://links.lww.com/MSS/A943), and minimal (see Appendix Table 7, Supplemental Digital Content 8, hazard ratios of the incidence of T2DM according to quartiles of the minimal level of CRF during 1979–1986, https://links.lww.com/MSS/A944) CRF level were negatively associated with the incidence of T2DM. However, after the adjustment for AUCG for CRF, the association between baseline CRF level and the incidence of T2DM was markedly attenuated (P for trend = 0.03). On the other hand, there was no association between the peak (P for trend = 0.57) and minimal (P for trend = 0.67) CRF during the measurement period and the incidence of T2DM.
To minimize the influence of preexisting diabetes at baseline on the relationship between the AUCG for CRF and the incidence of T2DM, we performed a sensitivity analysis which excluded participants who developed T2DM within 3 yr after the start of follow-up (see Appendix Table 8, Supplemental Digital Content 9, hazard ratios of the incidence of T2DM according to quartiles of an index of consistent CRF excluding those who developed T2DM within 3 yr after the onset of follow-up, https://links.lww.com/MSS/A945). The relationship between the AUCG for CRF and the incidence of T2DM was attenuated, but there remained a negative relationship. In addition, even when more conservative criteria for T2DM was used, we obtained a similar result on the association between the AUCG for CRF and the incidence of T2DM (see Appendix Table 9, Supplemental Digital Content 10, hazard ratios of the incidence of T2DM according to quartiles of an index of consistent CRF using more conservative criteria during follow-up, https://links.lww.com/MSS/A946). Finally, the analyses among 1951 men with CRF measures for eight times showed comparable results from analyses with all participants (see Appendix Table 10, Supplemental Digital Content 11, hazard ratios of the incidence of T2DM according to quartiles of an index of consistent CRF during 1979 to 1986 among 1951 men with eight times of the measures of CRF, https://links.lww.com/MSS/A947; see Appendix Table 11, Supplemental Digital Content 12, hazard ratios of the incidence of T2DM according to quartiles of an index of transiently high CRF during 1979–1986 among 1951 men with eight times of the measures of CRF, https://links.lww.com/MSS/A948). Moreover, the result of multiple imputation for missing data on CRF during the 8-yr measurement period was closely similar to that in the main analysis (data not shown).
We conducted a cohort study based on the hypothesis that consistently high CRF, rather than transiently high CRF, may be more closely associated with lower incidence of T2DM. The results supported this hypothesis; there was a negative dose–response relationship between the AUCG for CRF, representing the integrated CRF level during the measurement period, and the incidence of T2DM, whereas there was no association between the ΔAUCP for CRF, representing the degree of a temporal increase in CRF during the period, and the incidence of T2DM.
Results in relation to other studies
Currently, the benefit of maintaining high CRF, as a result of regular moderate-to-vigorous intensity physical activities, for the prevention of T2DM is already widely accepted. However, this knowledge is mainly based on the evidence that higher CRF at baseline is associated with lower risk of T2DM during follow-up (3,8,16,20,24,25). Our previous study showed an inverse relationship between long-term trends in CRF and the incidence of T2DM, even when the initial CRF level was considered (21). Moreover, a joint analysis by combined trends in CRF and initial CRF level showed that the risk of T2DM increased as the CRF level decreased over the time and the initial CRF level was lower. Although these findings suggested that consistently high CRF over time was associated with the development of T2DM, the secular change of CRF can fluctuate but not consistently increase or decrease, and the trends in CRF were susceptible to an initial and last value of measured CRF. In addition, whether transiently high CRF is sufficient for the risk reduction of T2DM has not been determined. To address this question, in this study, total exposure to CRF over time was calculated using the AUCG based on CRF measurements during the 8-yr measurement period before the baseline. In line with our previous study, we found the AUCG for CRF was negatively associated with the incidence of T2DM over 20 yr. On the other hand, there was no association between the ΔAUCP, reflecting the presence and the size of a “spike” in CRF during the measurement period and the incidence of T2DM. Thus, consistently high CRF, but not transiently high CRF, was associated with a lower risk of T2DM.
Genetic factors may explain why consistently high CRF level is a strong predictor for T2DM. Previous studies reported that the heritability of CRF among inactive individuals ranges from 25% to 65% (28), and that the intrinsic CRF level may markedly influence glucose metabolism.(30) On the other hand, recent studies suggested that epigenetic modification including DNA demethylation through exposure to a specific amount of physical activity for a specific period could play a potential preventive role of T2DM development (4,17,18). This phenomenon is called “metabolic memory.” In fact, an animal experiment showed that the favorable effects of exercise during the prepubertal period on glucose uptake capacity and insulin resistance was maintained up to middle-age periods in rats (23). These results lend support to our findings for a role of consistently high CRF.
This study provides evidence for the first time to support the belief that continuously maintaining a high level of CRF is associated with lower risk of T2DM. In addition to the effect of consistently higher CRF level on T2DM, this study also provides a plausible answer to why CRF measured once at baseline is useful as a strong predictor for future T2DM development, despite possible changes in CRF during the follow-up period. Because the CRF at baseline was strongly correlated with AUCG (r = 0.80), but only moderately with ΔAUCP (r = 0.42), this suggests that those who had higher CRF level at baseline were likely to consistently maintain higher CRF level over many years.
In this study, the criteria of T2DM differed among participants. This is a critical limitation of this study. We could not use the information of fasting blood glucose as the criteria of T2DM among participants who did not undergo blood tests. For them, the result of an oral glucose tolerance test, which was conducted when a nonfasting urinary glucose test showed a positive reaction, was used along with a self-reported questionnaire. Generally, urinary glucose test has poorer sensitivity for the detection of undiagnosed T2DM as compared with blood glucose test (9). Therefore, there is a possibility that undiagnosed diabetic participants were failed to be excluded at baseline, especially among younger participants, leading to reverse causality of our findings. The sensitivity analysis, however, with the exclusion of those who developed T2DM within 3-yr after the start of follow up showed a negative dose–response relationship between the AUCG for CRF and the incidence of T2DM. Among participants 39 yr or younger, the negative relationships between the AUCG for CRF and T2DM was substantially attenuated when BMI was considered. This result might come from overlooking potentially diabetic participants, who were prone to have higher BMI, at baseline, because BMI is a strong factor that influences not only diabetes but also CRF. Given the criteria of T2DM differed among participants, therefore, the comparability of the results might be low.
In addition to the determination of T2DM, there are several limitations of this study. First, CRF was estimated by submaximal exercise test. However, given that our procedure was highly correlated with the direct method in previous study (r = 0.92) (27), we assume that the CRF used in this study are sufficiently accurate as a measure of whole-body endurance capacity. Moreover, we defined the AUCG based on multipoint measurement of CRF as an exposure in this study. The calculation of AUCG for CRF has mainly two advantages in multiple measurements; (1) the consistency and accuracy of the individual’s CRF level, and (2) the comparability of total CRF level during the period with consideration to serial changes in CRF. Second, our participants were limited to approximately half of total individuals who underwent the health check-ups at the baseline. Although we compared the difference of baseline characteristics between the included (7185 men) and excluded (1272 men) participants among those who underwent our CRF exercise test at baseline, we could not compare the difference between the included (7158 men) and excluded (5824 men) participants among total 13,009 participants. Therefore, there is a possibility that our findings is affected by selection bias. The generalizability of our findings might be limited to participants with relatively higher CRF. Moreover, our participants were limited to Japanese men. However, previous studies have also shown that CRF is a long-term predictor of T2DM development in races other than East Asians (8,16,24), and in women (26). Therefore, our hypothesis is likely applicable to other races and women. Third, some of the confounders was evaluated by a self-reported questionnaire. Additionally, some studies have shown that dietary factors are related to the risk of T2DM (10,29). However, we did not have information on diet and thus could not control for this. It is possible that high-fit persons may have had more healthy dietary habits than low-fit persons. Therefore, the relationship between consistently higher CRF and the incidence of T2DM may have been overestimated.
The results showed that the risk of T2DM development was lower in participants with consistently higher CRF level over time, whereas transiently high CRF did not influence the risk of T2DM. Thus, maintaining high CRF level over time is beneficial for risk reduction of T2DM.
The authors thank the study participants and the Tokyo Gas Health Promotion Center physicians and medical staff for assistance with data collection. The authors also thank Benjamin Howe for helpful comments.
This work was supported by the Japan Society for the Promotion of Science [JP16K16591 to H. M.] and the National Institutes of Biomedical Innovation, Health and Nutrition to S. S. S. The funders had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; or the decision to submit the manuscript for publication.
The authors have no conflicts of interest to disclose. The results of the present study do not constitute endorsement by the American College of Sports Medicine. The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.
1. Astrand I. Aerobic work capacity in men and women with special reference to age. Acta Physiol Scand Suppl
2. Astrand PO, Ryhming I. A nomogram for calculation of aerobic capacity (physical fitness) from pulse rate during sub-maximal work. J Appl Physiol
3. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis. Eur J Epidemiol
4. Barres R, Yan J, Egan B, et al. Acute exercise remodels promoter methylation in human skeletal muscle. Cell Metab
5. Church TS, Earnest CP, Skinner JS, Blair SN. Effects of different doses of physical activity on cardiorespiratory fitness among sedentary, overweight or obese postmenopausal women with elevated blood pressure: a randomized controlled trial. JAMA
6. Colberg SR, Sigal RJ, Fernhall B, et al. Exercise and type 2 diabetes: the American College of Sports Medicine and the American Diabetes Association: joint position statement. Diabetes Care
7. Committee of the Japan Diabetes Society on the Diagnostic Criteria of Diabetes Mellitus, Seino Y, Nanjo K, et al. Report of the committee on the classification and diagnostic criteria of diabetes mellitus. J Diabetes Investig
8. Crump C, Sundquist J, Winkleby MA, Sieh W, Sundquist K. Physical fitness among Swedish military conscripts and long-term risk for type 2 diabetes mellitus: a cohort study
. Ann Intern Med
9. Engelgau MM, Narayan KM, Herman WH. Screening for type 2 diabetes. Diabetes Care
10. Esposito K, Maiorino MI, Ceriello A, Giugliano D. Prevention and control of type 2 diabetes by Mediterranean diet: a systematic review. Diabetes Res Clin Pract
11. Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Report of the expert committee on the diagnosis and classification of diabetes mellitus. Diabetes Care
. 2003;26(1 Suppl):S5–20.
12. Garber CE, Blissmer B, Deschenes MR, et al. American College of Sports Medicine position stand. Quantity and quality of exercise for developing and maintaining cardiorespiratory, musculoskeletal, and neuromotor fitness in apparently healthy adults: guidance for prescribing exercise. Med Sci Sports Exerc
13. Hainer V, Toplak H, Stich V. Fat or fit: what is more important? Diabetes Care
. 2009;32(Suppl 2):S392–7.
14. International Diabetes Federation. Diabetes Atlas
. 7th ed. Brussels, Belgium: International Diabetes Federation; 2015.
15. Lee IM, Shiroma EJ, Lobelo F, Puska P, Blair SN, Katzmarzyk PT. Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet
16. Lee DC, Sui X, Church TS, Lee IM, Blair SN. Associations of cardiorespiratory fitness and obesity with risks of impaired fasting glucose and type 2 diabetes in men. Diabetes Care
17. Lindholm ME, Marabita F, Gomez-Cabrero D, et al. An integrative analysis reveals coordinated reprogramming of the epigenome and the transcriptome in human skeletal muscle after training. Epigenetics
18. Nitert MD, Dayeh T, Volkov P, et al. Impact of an exercise intervention on DNA methylation in skeletal muscle from first-degree relatives of patients with type 2 diabetes. Diabetes
19. Pruessner JC, Kirschbaum C, Meinlschmid G, Hellhammer DH. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology
20. Sawada SS, Lee IM, Muto T, Matuszaki K, Blair SN. Cardiorespiratory fitness and the incidence of type 2 diabetes: prospective study of Japanese men. Diabetes Care
21. Sawada SS, Lee IM, Naito H, et al. Long-term trends in cardiorespiratory fitness and the incidence of type 2 diabetes. Diabetes Care
22. Sawada SS, Muto T, Tanaka H, et al. Cardiorespiratory fitness and cancer mortality in Japanese men: a prospective study. Med Sci Sports Exerc
23. Shindo D, Matsuura T, Suzuki M. Effects of prepubertal-onset exercise on body weight changes up to middle age in rats. J Appl Physiol (1985)
24. Sieverdes JC, Sui X, Lee DC, et al. Physical activity, cardiorespiratory fitness and the incidence of type 2 diabetes in a prospective study of men. Br J Sports Med
25. Someya Y, Kawai S, Kohmura Y, Aoki K, Daida H. Cardiorespiratory fitness and the incidence of type 2 diabetes: a cohort study
of Japanese male athletes. BMC Public Health
26. Sui X, Hooker SP, Lee IM, et al. A prospective study of cardiorespiratory fitness and risk of type 2 diabetes in women. Diabetes Care
27. Teräslinna P, Ismail A, MacLeod D. Nomogram by Astrand and Ryhming as a predictor of maximum oxygen intake. J Appl Physiol
28. Teran-Garcia M, Rankinen T, Bouchard C. Genes, exercise, growth, and the sedentary, obese child. J Appl Physiol (1985)
29. Wang PY, Fang JC, Gao ZH, Zhang C, Xie SY. Higher intake of fruits, vegetables or their fiber reduces the risk of type 2 diabetes: a meta-analysis. J Diabetes Investig
30. Wisloff U, Najjar SM, Ellingsen O, et al. Cardiovascular risk factors emerge after artificial selection for low aerobic capacity. Science
31. Diabetes mellitus. Report of a WHO Study Group. World Health Organ Tech Rep Ser