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

Mounting evidence over the past 3 decades has firmly established that low levels of cardiorespiratory fitness (CRF) are associated with a high risk of cardiovascular disease (CVD) and all-cause mortality, as well as mortality rates attributable to various cancers, especially of the breast and colon/digestive tract.^1–4 Importantly, improvements in CRF are associated with reduced mortality risk.⁵ Although CRF is now recognized as an important marker of cardiovascular health, it is currently the only major risk factor not routinely assessed in clinical practice.

In 2013, the American Heart Association and the American College of Cardiology jointly released new guidelines for the prevention and treatment of coronary artery disease.⁶ Although CRF is the fourth-leading risk factor for CVD and has long been established as a significant prognostic marker,⁷ it was excluded from the risk calculator. The authors of the guidelines noted that the evidence that CRF would enhance risk classification was inconclusive, and thus, the added contribution of CRF to determine CVD risk was uncertain. There is, however, a large body of epidemiological and clinical evidence demonstrating not only that CRF is a potentially stronger predictor of mortality than established risk factors such as smoking, hypertension, high cholesterol, and type 2 diabetes mellitus (T2DM), but that the addition of CRF to traditional risk factors significantly improves the reclassification of risk for adverse outcomes.

The purpose of this statement is to review current knowledge related to the association between CRF and health outcomes, increase awareness of the added value of CRF to improve risk prediction, and suggest future directions in research. Although the statement is not intended to be a comprehensive review, critical references that address important advances in the field are highlighted. The underlying premise of this statement is that the addition of CRF for risk classification presents health professionals with unique opportunities to improve patient management and to encourage lifestyle-based strategies designed to reduce cardiovascular risk. These opportunities must be realized to optimize the prevention and treatment of CVD and hence meet the American Heart Association’s 2020 goals.⁸

CRF reflects the integrated ability to transport oxygen from the atmosphere to the mitochondria to perform physical work. It therefore quantifies the functional capacity of an individual and is dependent on a linked chain of processes that include pulmonary ventilation and diffusion, right and left ventricular function (both systole and diastole), ventricular-arterial coupling, the ability of the vasculature to accommodate and efficiently transport blood from the heart to precisely match oxygen requirements, and the ability of the muscle cells to receive and use the oxygen and nutrients delivered by the blood, as well as to communicate these metabolic demands to the cardiovascular control center. Clearly, CRF is directly related to the integrated function of numerous systems, and it is thus considered a reflection of total body health. About half of the variance in CRF is considered to be attributable to heritable factors⁹; similarly, the contribution of inherited factors to the response of CRF to physical activity approximates 45% to 50%.¹⁰ It is noteworthy that these heritability estimates are similar in magnitude to other CVD risk factors, including, for example, insulin, glucose, lipoproteins, blood pressure, and high-sensitivity C-reactive protein.¹¹

CRF can be measured directly, expressed as maximal oxygen consumption (Vo₂max), or estimated from the peak work rate achieved on a treadmill or a cycle ergometer or from nonexercise algorithms. Measured Vo₂ is more objective and precise, but because it is easier to obtain, estimated CRF derived from the peak work rate is the more common expression of fitness, particularly in epidemiological studies involving large populations. Numerous studies have reported that both measured and estimated CRF strongly predict health outcomes; in the following overview of these studies, CRF refers to estimated fitness unless otherwise stated.

Since the late 1950s, numerous scientific reports have examined the separate relationships between physical activity, CRF, and all-cause mortality. The past 2 decades in particular have seen an exponential growth in the number of studies assessing the association between measures of CRF, mortality, and other health outcomes.^12–16 A consistent finding in these studies was that after adjustment for age and other risk factors, CRF was a strong and independent marker of risk for cardiovascular and all-cause mortality. This observation has been made in healthy men and women, those with suspected or known CVD, and those with comorbid conditions, including obesity, T2DM, hypertension, and lipid abnormalities.^12–23 In a growing number of studies, CRF has been demonstrated to be a more powerful predictor of mortality risk than traditional risk factors such as hypertension, smoking, obesity, hyperlipidemia, and T2DM. In addition, CRF has been shown to be a more powerful predictor of risk than other exercise test variables, including ST-segment depression, symptoms, and hemodynamic responses.^{13,16,18,23–30} Moreover, numerous recent studies have expressed CRF in the context of survival benefit per metabolic equivalent (MET; a multiple of the resting metabolic rate approximating 3.5 mL·kg⁻¹·min⁻¹); selected studies are presented in Table 1. These studies are noteworthy in that each 1-MET higher CRF (a small increment achievable by most individuals) was associated with considerable (10%–25%) improvement in survival.

Although a variety of indirect estimates or surrogates for CRF have been associated with health outcomes dating back to the 1950s, Blair and colleagues² published a seminal study in 1989 in which fitness was estimated using maximal treadmill testing in >13 000 asymptomatic men and women. Participants were followed up for 110 482 person-years (averaging >8 years) for all-cause mortality. Key results from this analysis are presented in the Figure. Age-adjusted mortality rates were lowest (18.6 per 10 000 person-years) among the most fit and highest (64.0 per 10 000 person-years) among the least fit men; the corresponding mortality rates among women were 8.5 and 39.5 per 10 000 person-years, respectively. These findings closely parallel an earlier report among asymptomatic men from the Lipid Research Clinics (LRC) Mortality Follow-up,³⁹ in which each 2–standard deviation decrement in CRF (roughly 2–3 METs) was associated with a 2- to 5-fold higher coronary heart disease (CHD) or all-cause death rate. Numerous research groups worldwide have reported similar findings over the past 2 decades. These follow-up studies included subjects with and without CVD, T2DM, obesity, and lipid abnormalities and of varying ethnicities, as well as women who were apparently healthy at the time of their fitness evaluation.^14–30 Gulati et al²³ suggested that the strength of exercise capacity in predicting risk of mortality was even greater among women than men, demonstrating a 17% lower risk for every 1-MET increase in CRF. Similarly, Nes et al³⁸ reported a 21% lower risk for every 1-MET increase in CRF for both sexes in a large healthy population followed up for an average of 24 years. Furthermore, in the LRC Mortality Follow-up trial, nearly 3000 asymptomatic women underwent maximal exercise testing and were followed up for up to 20 years.²⁸ A 20% lower survival was observed for every 1-MET decrement in CRF. This study also highlighted the relative limitations of ischemic electrocardiographic responses in predicting cardiovascular and all-cause mortality among women.

In recent years, the association between CRF and a wide range of health outcomes has also been addressed in varied populations, for example, patients referred for exercise testing for clinical reasons.^{16,18,19,22,25–27,40,41} In a study performed among US veterans, 6213 men underwent maximal exercise testing for clinical reasons and were followed up for a mean of 6.2 years.¹⁸ Subjects were classified into 5 categories by quintiles of CRF. After adjusting for age, the largest gains in survival were noted when comparing the lowest to the next lowest CRF groups. Among apparently healthy subjects and those with CVD, the least fit individuals (<5 and <6 METs for subjects with and without CVD, respectively) had >4-fold increased risk of all-cause mortality compared with the most fit. Importantly, an individual’s CRF level was a stronger predictor of mortality than established risk factors such as smoking, hypertension, high cholesterol, and T2DM. Over the past several years, other cohorts, such as those from the Cleveland Clinic^25,26,40 and the Mayo Clinic,^29,41 as well as numerous ongoing follow-up trials in the United States and Europe,^{12,15,16,43,44} have documented the importance of CRF as a predictor of mortality among clinically referred populations. These clinically based studies confirm the early observations of Blair et al,² Framingham,⁴⁴ and the LRC Trial^28,39 among asymptomatic populations, underscoring the fact that the CRF level has a strong inverse association with the incidence of all-cause mortality. The strength of the association between CRF and mortality was further reinforced in a meta-analysis by Kodama et al.¹⁶ Data were extracted from 33 studies, including nearly 103 000 participants. Compared with subjects in the most fit tertile, those with low CRF had a 70% and 56% higher risk for all-cause and cardiovascular mortality, respectively. Across all studies, 13% and 15% reductions in cardiovascular and all-cause mortality, respectively, were observed per 1-MET increase in exercise capacity. This meta-analysis also confirmed the previous finding that the greatest mortality benefits occur when progressing from the least fit and the next least fit group; lesser improvements in health outcomes were noted when individuals in the moderate- to high-fit groups were compared.

Many recent studies have also demonstrated that low CRF is a stronger predictor of risk for adverse cardiovascular outcomes than traditional risk factors, including lipid abnormalities, hypertension, insulin resistance, obesity, and smoking.^{12,13,16–21,24} Despite these observations, the importance of CRF in the risk paradigm has historically received less attention from health professionals and CVD specialists. Moreover, when an exercise test is performed, there has been the tendency to focus on ischemic ST-segment displacement and the potential need for coronary revascularization without considering the prognostic value of CRF.^24,29,30 Reasons for the inverse association between CRF and mortality are not fully understood. Possible explanations include the fact that fitter people tend to have more cardioprotective cardiovascular risk profiles (mediated in part through higher activity levels), autonomic tone (potentially reducing arrhythmogenic risk), lower risk for thrombotic events, and improved indices of endothelial function. Numerous studies have documented that biological mechanisms for disease are favorably influenced by CRF. For example, in a cohort aged 20 to 90 years (n=4631), in which directly measured CRF was determined, women and men below the sex-specific median for CRF (women <35.1 mL·kg⁻¹·min⁻¹; men <44.2 mL·kg⁻¹·min⁻¹) were 5 and 8 times, respectively, more likely to have a cluster of cardiovascular risk factors than those in the highest quartile of CRF.⁴⁵ Additionally, each 5- mL·kg⁻¹·min⁻¹ lower level of CRF corresponded to a 56% higher odds of cardiovascular risk factors. Similarly, Arsenault et al,⁴⁶ in a cross-sectional evaluation of 169 healthy men without T2DM, noted that those in the lowest tertile of CRF had higher triglyceride levels, higher apolipoprotein B, and higher total cholesterol/high-density lipoprotein ratios than those in the highest tertile of CRF. Others have shown that higher CRF is associated with lower visceral adiposity, improved insulin sensitivity, lower levels of inflammation, more favorable lipid and lipoprotein profiles, and lower blood pressure.^20,46-48 Kawano et al⁴⁸ performed a randomized trial of exercise training and dietary intervention in 217 at-risk men and women and reported that lipid profiles improved with increases in CRF. Numerous recent studies have observed that C-reactive protein and other inflammatory markers are also lower among more fit individuals than among those who are less fit.^15,47

The observations cited above highlight the fact that exceptionally high CRF levels are not necessary to provide significant health benefits. Individuals with a CRF level <5 METs tend to have a particularly high risk for mortality, whereas many epidemiological studies have observed that CRF levels >8 to 10 METs are associated with relative protection.^{14,16,18–20,42} However, a consistent observation is that the largest benefits occur between the least fit and the next least fit group of individuals studied. Stated differently, health benefits are most apparent at the low end of the CRF continuum. Although studies vary, this is generally the case for both all-cause and CVD mortality. This is an often misunderstood but important public health message, because one need not be athletic to gain substantial health benefits from improvements in CRF. This is illustrated in the Figure, in which more than half the reduction in all-cause mortality occurs between the least fit group and the next least fit group. Relatively less is gained by increasing CRF between moderately fit and highly fit individuals. Moreover, evidence suggests that even among subjects within the low-fit⁴⁹ or low-risk⁵⁰ groups, higher CRF is associated with reduced risk. This has implications for physical activity counselling, given that considerable benefits are likely to occur by encouraging the most sedentary or low-fit people to engage in modest activity levels. Simply stated, every effort should be undertaken to increase physical activity levels in sedentary adults (ie, some exercise is better than none). This might initially equate to increasing physical activity and exercise habits in previously sedentary individuals who continue to fall below current recommendations (guidelines). Over time, an individual can be “up-titrated,” approaching or exceeding the physical activity recommendations. This pragmatic approach could serve to enhance compliance with increased physical activity and exercise training, because a sedentary person who has been accustomed to minimal levels of physical activity for decades might perceive the immediate adoption of an active lifestyle to the level of current recommendations as unattainable. When counselling sedentary people, it is important to emphasize that substantive gains in health can be achieved with relatively modest increases in physical activity.

Risk prediction in the general population is challenging, because most people are at low risk. Nes and colleagues³⁸ found that CRF in healthy men (n=18 348) and women (n=18 764) <60 years of age at baseline was inversely associated with CVD mortality. Mean follow-up time was 24 years. Men <60 years of age at baseline and with a CRF below 85% of the age-expected value had an approximate 2-fold risk of dying of CVD compared with those at or above the age-predicted value. For each 1-MET increase in exercise capacity, the risk of CVD mortality was 21% lower in both men and women. Women below 85% of the age-predicted CRF had a 24% higher risk of CVD mortality. Similarly, Artero et al⁵¹ evaluated 43 356 adults (21% women) aged 20 to 84 years who were free of baseline history of CVD or cancer and followed them up for a median of 14.5 years. Both measured and estimated CRF were inversely associated with risk for fatal and nonfatal CVD events in men and for nonfatal CVD events in women. The risk reduction per 1-MET increase in measured CRF was 17% for fatal CVD and 10% for nonfatal CVD in men and 5% for fatal CVD and 23% for nonfatal CVD in women. Comparable findings have been reported in ongoing follow-up studies from cohorts including the Veterans Exercise Testing Study,^17–19,22 the Aerobics Center Longitudinal Study (ACLS),^2,5,13,20,51 and the Henry Ford Exercise Testing Project.^52–54 Importantly, the overall discriminative ability of CRF in these studies is comparable to that normally obtained in widely used risk models, such as the Framingham risk score and European SCORE (Systematic Coronary Risk Evaluation) algorithm.^55,56 For example, Laukkanen et al⁵⁷ reported that a 1-MET increment in CRF and a 1% increment in European risk score were associated with 16% and 15% changes, respectively, in risk for all-cause mortality. Subjects with high European or Framingham score and low peak Vo₂ represented the highest risk group.

Beyond cardiovascular and all-cause mortality, habitual physical activity and CRF have been linked to both cardiovascular and noncardiovascular surgical outcomes, including the timing of cardiac transplantation in ambulatory patients with heart failure (HF), their risk stratification, and the likelihood of HF hospitalization in later life, as well as the incidence of stroke in older adults. The prognostic value of CRF, including peak Vo₂, the ventilatory threshold, and other indices, has reinvigorated the clinical value of cardiopulmonary exercise testing (CPX), which has been used less frequently in recent years in favor of more advanced diagnostic imaging procedures (eg, exercise stress echocardiography, exercise myocardial perfusion imaging, pharmacological testing).⁵⁸ This section reviews clinically relevant epidemiological and observational studies, with specific reference to possible biological mechanisms underlying these associations, or the lack thereof.

Numerous studies have reported that adding CRF to a single or several established risk factors for CVD substantially improves the precision of risk prediction for CVD morbidity or mortality.^{8,12–19,82–84} However, although the evidence that CRF is inversely associated with mortality is strong and convincing, it does not necessarily mean that CRF directly enhances CVD mortality risk prediction. For CRF to truly be a novel risk marker, it must improve risk prediction beyond traditional markers.⁸⁵ There exists no single statistical test that provides all the information necessary to evaluate a new biomarker, and a combination of ≥2 statistical approaches has been suggested.^85–89 Recent studies suggest that the net reclassification improvement (NRI) and the integrated discrimination improvement can provide important insights beyond traditional statistical tools (eg, hazard ratios, odds ratios, C-index) when estimating risk for adverse outcomes. These tools more directly address the extent to which a given risk marker adds to existing markers to predict adverse outcomes. NRI indicates whether the addition of a biomarker correctly and significantly alters risk classification; it is defined as the net change in risk among those who do and do not experience an event.^86,88 Integrated discrimination improvement determines whether the addition of a new biomarker significantly improves risk discrimination, reflecting the improvement in true-positive rates minus the worsening of false-positive rates.^86–89 Several recent studies have used these metrics to help determine the additive value of CRF to traditional risk markers (Table 3).

Wickramasinghe et al⁹⁵ reported that the addition of CRF to a traditional risk prediction model (including age, body mass index, systolic blood pressure, T2DM, total cholesterol, and smoking) improved 30-year risk prediction in 13 627 men and 2906 women without known CVD at baseline. A low level of CRF (defined as an estimated peak Vo₂ <28 mL·kg⁻¹·min⁻¹ for men and <21 mL·kg⁻¹·min⁻¹ for women) was associated with a greater 30-year risk of dying of CVD in all risk factor strata. Importantly, CRF in particular added to long-term risk prediction. For example, a significantly higher 30-year risk for CVD mortality was noted among people with hypertension (stage II) and low versus high CRF (18.4% versus 10.1%) despite a similar risk at 10-year follow-up (2.3% versus 1.2 %). Laukkanen et al⁵⁷ studied a random population-based sample of 1639 men without known T2DM or atherosclerotic CVD at baseline. During the 16-year follow-up period, those with high Framingham or European risk scores and low CRF represented the group at highest risk of death of CVD and all causes. These results clearly demonstrated that the addition of CRF to established risk scores further improved risk prediction.

Gupta et al⁹¹ evaluated whether CRF improved risk classification when added to traditional risk factors in 49 307 men and 17 064 women examined in the ACLS between 1970 and 2006. Their traditional risk factor model included age, sex, systolic blood pressure, T2DM, total cholesterol, and smoking. Risk estimates were evaluated with and without CRF after 10 and 25 years of follow-up in men and after 25 years in women. In men, at 10 and 25 years of follow-up, the addition of CRF to the traditional risk model resulted in NRIs for CVD mortality of 12.1% and 4.1%, respectively. This suggests that 12.1% and 4.1% of subjects were correctly reclassified for CVD mortality beyond traditional risk factors at these time points. The corresponding relative integrated discrimination improvements were 29% and 11.1% at 10 and 25 years. The addition of CRF to the traditional risk model in women resulted in an NRI of 13.1% and relative integrated discrimination improvement of 13.5% at the 25-year follow-up, whereas too few women had died at the 10-year follow-up to conduct the analyses. Stamatakis and colleagues⁹⁰ evaluated whether CRF improved CVD mortality risk prediction in 17 669 women and 14 650 men aged 35 to 70 years who took part in health surveys in England and Scotland between 1994 and 2003. During a mean follow-up of 9 years, NRIs for CVD mortality were 27.2% and 21.0% for men and women, respectively. Myers et al⁹² followed up ≈7000 men referred for exercise testing for clinical reasons for a mean of 10 years and observed that the addition of CRF to a model that included traditional risk factors resulted in an NRI of 42.8%. Holtermann et al⁹⁴ reported an NRI of 30.5% for CVD mortality and an NRI of 24.5% by adding self-reported CRF to traditional risk factors among 8936 men and women in the Copenhagen City Heart Study.

Despite the aforementioned evidence linking CRF to longevity, it is not included in any of the currently used CVD risk prediction models from health authorities or health organizations throughout the world. A principal argument against the use of CRF in CVD risk-score models could be its precise quantification or the lack of evidence from randomized clinical trials, which would need to include all age groups and both sexes and use hard end points such as CVD morbidity and mortality. Although this limitation also applies to cigarette smoking, few people would dispute that smoking increases CVD risk. Nevertheless, data from large population-based studies and small-scale randomized clinical trials in selected populations suggest that CRF should be included in future CVD risk prediction models.

Numerous studies have assessed CRF in the context of established risk prediction models such as the Framingham risk score. Gander⁹⁶ examined the association of CRF with 10-year risk of CHD while controlling for Framingham risk score in 29 854 men from the ACLS who were examined between 1979 and 2002. At baseline, all participants were free of CVD or cancer and between 30 and 74 years of age. Men who developed CHD during the follow-up were older and had an estimated CRF ≤38 mL·kg⁻¹·min⁻¹. Risk of CHD was 20% lower for each 1-MET-higher increment in CRF. In addition, being categorized as having a high CRF (defined as the highest 40% of CRF in the entire ACLS population [mean CRF 48±7 mL·kg⁻¹·min⁻¹]) was associated with a 33% lower risk compared with men who had low CRF (defined as the lowest 20%; mean CRF 30±4 mL·kg⁻¹·min⁻¹). The study also stratified subjects into low, moderate, or high Framingham risk score groups at baseline and found that CRF was significantly protective across the range of Framingham risk scores.

The impact of CRF as a biomarker is valuable not only to determine a person’s risk for future adverse clinical outcomes, but also to optimize treatment strategies. Determining CRF on a serial basis is valuable in gauging the effectiveness of treatment strategies, including recommendations for participation in physical activity. Blair et al³¹ studied 9777 men given 2 preventive medical examinations, each of which included assessment of CRF by maximal exercise testing, a mean of 5.1 years apart. The highest age-adjusted all-cause death rate was observed in men who were unfit at both examinations (122.0/10 000 man-years); the lowest death rate was observed in those who were physically fit at both examinations (39.6/10 000 man-years). Men who improved from unfit to fit between the first and second examination had a reduction in mortality risk of 44% relative to men who remained unfit at both examinations. Lee et al⁹⁷ reported that in relatively fit men (n=14 345, average estimated CRF 41.7 mL·kg⁻¹·min⁻¹), maintaining or improving CRF from baseline to a second examination 6 years later was associated with 27% and 42% reduced risks for CVD and all-cause mortality, respectively, during an 11.4-year follow-up period compared with those whose CRF decreased over the same period. Importantly, men who had a reduction in CRF between examinations were at increased risk of dying of CVD regardless of changes in body mass index. Every 1-MET increase in CRF was associated with a 19% lower risk of CVD mortality. Similarly, Kokkinos et al⁹⁸ reported that unfit individuals whose CRF improved had a 35% lower mortality risk during a median follow-up period of 8.1 years compared with those who remained unfit. The largest randomized trial of exercise training in HF patients, HF-ACTION (Heart Failure and a Controlled Trial Investigating Outcomes of Exercise Training), reported that every 6% increase in CRF (measured peak Vo₂) over 3 months was associated with a 4% lower risk of cardiovascular mortality or cardiovascular hospitalization and an 8% lower risk of cardiovascular mortality or HF hospitalization after adjustment for potential confounding variables.⁹⁹

Although is it well documented that higher levels of CRF are associated with lower CVD risk, over the past 2 decades numerous other health benefits have been linked to higher levels of CRF.

Non–exercise-based equations or models are available to conveniently estimate CRF without performing a maximal or submaximal exercise test.¹⁷¹ This approach uses variables commonly assessed in clinical settings to provide a rapid and inexpensive way of estimating CRF in public health and clinical settings.

One of the first nonexercise prediction equations was developed by Jackson et al¹⁷² in 1990 using 1393 male and 150 female employees from the National Aeronautics and Space Administration (NASA)/Johnson Space Center, aged 20 to 70 years. Regression models were used to estimate CRF from age, sex, body mass index or percentage body fat, and self-reported physical activity, with an SEE of ≈5.5 mL·kg⁻¹·min⁻¹.¹⁷² This equation has been cross-validated with independent samples^173–176 and used to link estimated CRF with disease outcomes.^177,178 Other researchers have developed nonexercise equations to estimate CRF in populations that differed in age, sex, and ethnicity. The accuracy of the predicted CRF values was improved by incorporating other lifestyle and health indicators.

One systematic review¹⁷¹ of 13 nonexercise equations is presented in Table 6.^{172,179–190} These equations were developed with cross-sectional data using age, sex, body weight (or body mass index, percentage of body fat, waist circumference), physical activity/exercise/training (self-reported or measured), smoking, resting HR, or perceived functional ability as predictors of CRF. The R² and SEEs ranged from 0.50 to 0.86 and 2.98 to 6.90 mL·kg⁻¹·min⁻¹, respectively. The nonexercise CRF estimates were similar in accuracy to submaximal exercise prediction models.^{172,174,181,191} A limitation of these equations is that they tend to underestimate and overestimate CRF at the upper and lower ends of the distribution, respectively.^{172,179,180,182,185,190} The underestimation is unlikely to affect highly fit individuals, who will still be correctly classified into the higher CRF categories; however, the overestimation for people with low CRF could be a concern because of the associated heightened risk among these men and women.^{38,172,182,192} Despite this, most models derived from large studies correctly classify individuals into low-fitness categories. For example, in a study by Nes et al¹⁹⁰ that included 2067 men and 2193 women, 90.2% of women and 92.5% of men in the 2 lowest quartiles of fitness were correctly classified into 1 of the 2 lowest measured quartiles when using their CRF prediction algorithm. That study also correctly classified a high percentage of men (93.6%) and women (91.2%) within the closest measured “high-fit” quartile.

One group who developed nonexercise equations using objective measures of physical activity reported more accurate prediction of CRF in Japanese men and women than with more traditional models using self-reported physical activity.^187–189 More recently, a new longitudinal nonexercise algorithm has been developed using data from the ACLS that addressed 2 limitations from previous cross-sectional studies.¹⁹¹ First, the longitudinal equations used quadratic modeling to account for the well-documented nonlinear relationship between age and CRF. Second, the longitudinal nonexercise models provided valid estimates of changes in CRF over time. Nevertheless, this longitudinal equation¹⁹¹ should be cross-validated in other populations.

Recently, several studies have sought to determine the validity of nonexercise estimated CRF and long-term health risk, including mortality.^38,51,90,193 Pooled data from 8 British cohorts included 32 319 people aged 35 to 70 years, with a 9-year follow-up.⁹⁰ In this study, the 2005 nonexercise model proposed by Jurca et al¹⁸⁵ was used to estimate CRF. A 9.4% and 7.4% lower risk of all-cause death and a 15.6% and 16.9% lower risk of CVD death per 1-MET increase was observed in men and women, respectively. Nes et al³⁸ examined the predictive validity of nonexercise CRF using a cross-sectional model derived previously.¹⁹⁰ The sample included 37 112 individuals who were followed up for a mean of 24 years in the HUNT study (Nord-Trøndelag Health Study). After adjustment for potential confounders, each 1-MET higher CRF was associated with 21% lower CVD mortality for both men and women who were <60 years of age at baseline, and the corresponding lower risks for all-cause mortality were 15% for men and 8% for women. Artero et al⁵¹ investigated the association between nonexercise estimated CRF using a longitudinal 2012 model¹⁹¹ that examined all-cause mortality and nonfatal CVD events among 43 356 adults (21% women, aged 20–84 years) from the ACLS. With a median follow-up of 14.5 years, estimated CRF among men was associated with a 15% and 13% lower risk of all-cause death and nonfatal CVD events per MET, whereas in women the values were 11% and 13%, respectively. Martinez-Gomez et al¹⁹³ also explored the impact of this new longitudinal model¹⁹¹ on all-cause mortality among 2930 adults >60 years of age during an average follow-up of 9.4 years. The investigators reported a 20% lower risk of death with each 1-MET increment only in older women. The aforementioned studies demonstrate that the risk reduction associated with each 1 MET increase in nonexercise estimated CRF ranges from 7.4% to 21% and from 8% to 16.9% for all-cause mortality and CVD mortality, respectively. These results are consistent with the risk reduction as reported from a meta-analysis of 33 studies (13% and 15% for all-cause and CVD mortality, respectively).¹⁶

In summary, nonexercise estimated CRF provides an alternative approach for large epidemiological research and routine clinical practice with the goal to identify individuals with low CRF who are at increased health risk. Researchers or practitioners should select the equations that are most suitable for the population being evaluated. The 13 nonexercise equations given in Table 6 have all been cross-validated. Among these equations, the estimated CRF values in the 2005 model developed by Jurca et al¹⁸⁵ and the 2011 model developed by Nes et al¹⁹⁰ predicted long-term mortality^90,98 and showed a comparable risk reduction to measured CRF.¹⁶ A step-by-step procedure for using the Nes equation¹⁹⁰ to estimate CRF with routine clinical measures can be readily accessed by both the practitioner and the patient.¹⁹⁴ Estimation of CRF provides the practitioner with a platform for counselling the patient regarding the importance of physical activity. However, in most clinical patient subsets, nonexercise estimated CRF should not be viewed as a replacement for objective assessment of CRF.

For a given age, men generally demonstrate higher CRF levels than women, which is largely attributed to their higher peak cardiac output, hemoglobin levels, and skeletal muscle mass.^195,196 Also, a recent report showed those with moderate or high CRF had blunted age-related declines in maximal HR.¹⁴⁸ Although it is widely accepted that CRF decreases with age, the rate and causes of the decrease in aerobic capacity remain poorly understood. Using men (n=435) and women (n=375) from the Baltimore Longitudinal Study of Aging, researchers found a decline in peak Vo₂ of 3% to 6% per decade for the third and fourth decades, but after age 70, the rate accelerated to >20% per decade.¹⁹⁶ Using the much larger ACLS data set (3429 women and 16 889 men), others have confirmed that the longitudinal decline in CRF of women and men is not linear, noting an increase in the rate of decline starting at approximately age 45.¹⁹⁵ Jackson et al¹⁹¹ observed the rate of decline in CRF was steeper for men than women, but when the rate of decline was expressed as a percentage of peak CRF, men and women were almost identical.

CRF is directly influenced by the hemodynamic determinants of the Fick equation: Vo₂=Qc×a-v O₂ difference (oxygen uptake = cardiac output times the arteriovenous difference for oxygen) (see Levine¹⁹⁷ and Heinonen et al¹⁹⁸ for discussion). Cardiac output is determined by the product of HR and stroke volume. Because virtually every exercise training study, regardless of length or intensity, has reported no change or even a slight decline in HR max, increases in CRF occur primarily via increases in stroke volume, arteriovenous O₂ difference, or both. Although total blood volume and hemoglobin increase with training, hemoglobin concentration remains stable or declines slightly, so that arterial oxygen content remains unchanged. Therefore, the 2 major adaptations that occur with exercise training are an increase in maximal stroke volume and a decrease in venous oxygen content caused by increased O₂ extraction.

Generally, stroke volume increases via an increase in end-diastolic volume as a function of 3 key adaptations: an increase in total blood volume¹⁹⁹; an improvement in left ventricular distensibility (larger left ventricular end-diastolic volume for the same filling pressure)²⁰⁰; and improvement in diastolic function.²⁰¹ Stroke volume can also increase via a decrease in end-systolic volume with improved ventriculoarterial coupling, likely because of enhanced endothelial function.^202,203 Enlargement of the right ventricle appears to occur early in the course of exercise training and might be necessary to facilitate left ventricular adaptations.²⁰⁰

There are also significant changes in skeletal muscle that increase O₂ extraction. Probably the most important is an increase in muscle capillary density, which increases mean transit time for diffusion.²⁰⁴ There are also increases in the size and number of skeletal muscle mitochondria and oxidative enzymes after training,²⁰⁵ although the capacity for mitochondrial respiration and skeletal muscle blood flow far exceed that of the circulation to deliver blood and oxygen to the muscle, even in untrained individuals. For young people, most studies show a balanced increase in maximal cardiac output (from an increase in maximal stroke volume), and arteriovenous O₂ difference^200,204,206 with training. For older people, the training responses are more variable, although the final mechanisms of improvement largely depend on the duration and intensity of training.

The concept of peak/maximal Vo₂ Vwas established in 1923 by Hill and Lupton.²⁰⁷ Early on, it was reported to vary with age, sex, and endurance training status^208–210 and to be increased by regular physical activity.^211,212 Also, CRF was shown to be an excellent measure of cardiorespiratory integrity.^213,214 Subsequently, numerous scientists evaluated a wide variety of factors that impact a person’s peak/maximal Vo₂, with specific reference to the dose of physical activity needed to increase CRF. The key components of physical activity considered in determining the exercise dose include activity type, intensity, session frequency, time (session duration), program duration, activity pattern, and progression.²¹⁵ Although frequently considered separately, each of these components interact with one another to impact the training response. Collectively, these data were consolidated into recommendations by the American College of Sports Medicine (ACSM) in “Position Stands” published in 1978, 1990, 1998, and 2011.^215–218

In recent years, the major addition to the CRF dose-response literature was from the increasing number of reports evaluating the effects of HIT and sprint interval training. Interval training, the alternating of higher- and lower-intensity bouts of exercise during a single session, was originally used by endurance athletes and evaluated by sport medicine scientists in Europe >50 years ago.^282,283 More recently, moderate-intensity interval training (50%–75% HR reserve or Vo₂R) has been used in health-oriented fitness regimens for healthy adults and in cardiovascular and pulmonary rehabilitation.^272,284

In healthy adults, HIT regimens have been shown to be effective by inducing greater increases in CRF than moderate-intensity continuous training (MICT) regimens, especially when total amounts of energy expended in the different regimens are similar.^236,237,285 For example, Gormley et al studied 61 healthy young men and women who were randomized to a nonexercise control group or 1 of 3 exercise groups: MICT (60 minutes, 4 days per week at 50% Vo₂R), vigorous intensity (40 minutes, 4 days per week at 75% Vo₂R), or near-maximal effort (HIT 3 days per week, 5 minutes at 75% Vo₂R followed by 5 intervals of 5 minutes at 95% Vo₂R and 5 minutes at 50% Vo₂R cool-down).²³⁶ Total work over the 6-week program was similar for the 3 exercise groups. Mean baseline CRF for all participants was ≈10 METs. The net increase in peak/maximal Vo₂ in response to each of the physical activity programs was significant: MICT, 3.4±3.9 mL·kg⁻¹·min⁻¹ (9.4%); vigorous intensity, 4.8±3.2 mL·kg⁻¹·min⁻¹ (13.7%); near maximal, 7.2±4.3 mL·kg⁻¹·min⁻¹ (20.6%); and no-exercise control, 0.7±3.8 mL·kg⁻¹·min⁻¹ (0.6%). The increase in peak/maximal Vo₂ was significantly different between each of the exercise groups, demonstrating a dose response for exercise intensity at the high end of the intensity spectrum. HIT and MICT effects on CRF have also been compared in healthy obese men and women,²⁵⁸ people with metabolic syndrome,²⁶¹ hypertensive patients,²⁵⁰ and patients with T2DM,²³⁸ with HIT eliciting significantly greater increases in CRF than MICT despite similar energy expenditure.

In several recent randomized controlled studies involving patients with CVD, including HF, that compared the effects of HIT versus MICT on CRF or physical working capacity, HIT was found to be superior in some^256,264,266 but not all studies.^230,258,262 SAINTEX-CAD (Study on Aerobic Interval Exercise Training in CAD) compared the effects of HIT versus MICT on CRF in 200 patients with coronary artery disease.²³⁰ After 12 weeks of training, investigators found no difference in the mean increase in peak/maximal Vo₂ (22.7% for HIT and 20.3% for MICT). However, in this study, the difference between training intensity was smaller than planned between the 2 groups (the MICT group trained at 80% versus planned 65%–75% of peak HR, whereas the HIT group trained at 88% versus the planned 90%–95% of peak HR), which made the training protocols nonisocaloric. Three small meta-analyses reported significant increases in CRF in response to selected HIT and MICT regimens in patients with various manifestations of CVD.^277,286,287 In another meta-analysis, 6 randomized clinical trials with a total of 153 patients were included, but only 4 randomized controlled trials with 111 patients had adequately reported peak/maximal Vo₂ data.²⁷⁷ Compared with MICT, HIT significantly improved peak/maximal Vo₂ (mean difference, 3.6 mL·kg⁻¹·min⁻¹; 95% confidence interval, 2.3–4.9). Similar results were shown in patients with HF with preserved ejection fraction.²¹⁹ Weston and colleagues²⁸⁶ included 10 studies and 273 patients in their meta-analysis and concluded that HIT significantly increased CRF by almost double that of MICT (HIT 5.4 mL·kg⁻¹·min⁻¹ versus MICT 2.6 mL·kg⁻¹·min⁻¹; mean difference, 3.03 mL·kg⁻¹·min⁻¹; 95% confidence interval, 2.00–4.07 mL·kg⁻¹·min⁻¹). On the basis of their meta-analysis of 6 randomized controlled trials comparing HIT and MICT in patients with coronary artery disease, Elliott and colleagues²⁸⁷ concluded that HIT was more effective than MICT for increasing CRF but also recommended that long-term studies assessing morbidity and mortality after HIT are required before this approach can be more widely adopted. Moholdt et al²⁴⁹ compared HIT and MICT regimens in patients after coronary artery bypass surgery and reported that for short-term training (4 weeks), peak/maximal Vo₂ increased significantly in both groups. However, with continued training up to 6 months, those patients performing HIT further increased their Vo₂ peak (P<0.001), whereas the MICT patients did not.²⁴⁹ Thus, the duration of the training program might influence which regimen is most effective for increasing CRF in selected patient populations.

Although there is now substantial evidence that low levels of CRF are associated with a heightened risk of cardiovascular and all-cause mortality, unanswered questions remain. Here, we provide recommendations for future research that although not exhaustive, offer direction to unravel some of the vagaries between CRF and selected health outcomes.

An underlying premise of this statement is that CRF should be measured in clinical practice if it can provide additional information that influences patient management. Indeed, decades of research have produced unequivocal evidence that CRF provides independent and additive morbidity and mortality data that when added to traditional risk factors significantly improves CVD risk prediction. On the basis of these observations alone, not including CRF measurement in routine clinical practice fails to provide an optimal approach for stratifying patients according to risk. As noted in numerous recent American Heart Association scientific statements, the measurement of CRF in clinical settings is both important and feasible.^{74,75,152,153,292} Additionally, estimates of CRF using nonexercise algorithms have pragmatic importance and provide values for CRF that enhance risk prediction when direct CRF measures are not feasible. In fact, routine estimation of CRF in clinical practice is no more difficult than measuring blood pressure, and procedures for incorporating CRF estimation into routine clinical assessments in a pragmatic manner are provided in Tables 9 and 10. Finally, of crucial importance is the repeated observation that one does not need to be highly fit to gain benefit from improvements in CRF. Indeed, numerous epidemiological studies have now demonstrated that more than half the reduction in all-cause and CVD mortality generally occurs when moving from the least fit group to the next least fit group. For many people, this can be achieved by routine, moderate-intensity exercise consistent with consensus guidelines; lower levels of physical activity may be all that is needed to derive a clinically significant benefit in habitually sedentary individuals. This has implications for physical activity counselling, given that considerable benefits are likely to occur by encouraging the most sedentary or low-fit individuals to engage in modest amounts of physical activity accumulated throughout the day. Although gaps in knowledge remain, and refinement of CRF targets for risk reduction across age and sex need further investigation, the evidence reviewed suggests that the measurement of CRF improves patient management and that its omission from routine clinical practice for the vast majority of patients is unacceptable. Accordingly, the inclusion of CRF measurement or estimation in routine practice affords clinicians with a vitally important opportunity to improve patient management and, more importantly, patient health.