Construct validation of a non-exercise measure of cardiorespiratory fitness in older adults

Cardiorespiratory fitness (CRF) is associated with a decreased risk of all-cause mortality and chronic illnesses, especially cardiovascular disease [1, 2]. Additionally, cross-sectional studies and randomized clinical trials suggest that CRF is associated with brain structure and function [3] and that improvements in CRF brought about by exercise training are implicated in the restoration of neural and cognitive functioning in older adults [4–6]. Whereas the consideration of common risk factors such as physical inactivity and overweight status can provide information relative to the health status of an individual, CRF is a multifaceted construct that reflects the overall efficiency of the cardiovascular system and takes these other common health indicators into account.

Although being able to measure CRF expediently during a typical office visit health assessment or a population-based study would be valuable, it is a considerable challenge to adequately measure CRF in office medical settings or large-scale research investigations. Indeed, the increasing constraints of time, resources, and potential risk associated with assessing CRF by the "gold standard" method of graded exercise testing (GXT) make it all but impossible. Alternatives to a GXT are typically submaximal field tests. Examples of these include the Rockport 1-mile walk test [7] for estimating CRF, the Long Distance Corridor Walk (LDCW) [8, 9], and the UKK Walk Test [10]. Although these field tests provide reliable estimates of peak oxygen consumption, they still carry some burden of time, resources, and risk. Because CRF is an important indicator of the ability of older adults to carry out essential activities of daily living [11] and potentially an important precursor to functional limitation and disability [8, 12], identifying a relatively simple, low-cost, and low-risk measure of CRF could be of considerable benefit for use in clinical and research settings (e.g., for guiding exercise prescriptions).

Several research groups have developed equations that use variables associated with CRF such as age, sex, and physical activity level to approximate CRF in the absence of fitness testing [13–15]. Having a valid estimate of CRF could prove valuable in both healthcare and population-based research settings. Recently, Jurca et al. [16] have extended earlier attempts to develop equations to predict CRF using three very large samples (total N = 38,137). They expressed CRF in terms of metabolic equivalents (METs) and used several large data sets to establish the validity of a regression equation with the following components: age, sex, body mass index (BMI), resting heart rate (RHR), and self-reported physical activity (SRPA) level. Multiple correlation coefficients were similar across all data sets (r = 0.81, 0.77, 0.76) [16]. Moreover, the correlations between equation-estimated CRF in METs and METs measured by either maximal or submaximal graded exercise testing ranged between 0.74 and 0.80 for each sample. The original validation sample was characterized by participants who were primarily middle-aged (M age for men = 43.6 yrs; M age for women = 41.1 yrs) and relatively high fit (M measured METs for men = 11.72; M measured METs for women = 9.44).

The purpose of the present study was to test the construct validity of the regression equation developed by Jurca et al. [16] in a sample of older community-dwelling adults. The original equation was developed using primarily middle-age adults. However, given the rapidly increasing older adult population is it important to cross-validate the equation in this as well as other populations to maximize its applicability. To examine the construct validity, we performed a hierarchical regression analysis to determine the contribution of each of the individual equation components to measured METs. Additionally, we were interested in the extent to which the equation-estimated CRF value correlated with CRF measured by GXT, and how this relationship compared with the relationship between GXT-measured CRF and CRF estimated by a common field test. Finally, because CRF is often associated with cardiovascular conditions [17], we examined the association between self-reported conditions and each method of assessing CRF.

Cardiorespiratory fitness testing is costly, time-consuming, and, in the case of older adults, can necessitate some risk. These factors are prohibitive to routine clinical assessments of CRF and preclude use in larger population samples, in spite of the importance of CRF to numerous health outcomes [1]. The present study was designed to validate an equation developed by Jurca et al. [16] that combines a simple self-report physical activity index with age, sex, BMI, and RHR to estimate CRF in MET values [16] in a sample of community dwelling older adults. Initial development of this measure was conducted in several very large samples of community dwelling adults.

Our findings show the prediction equation yielded similar estimates of CRF as those reported in the original validation study and offer cautiously optimistic support for the use of this measure to estimate CRF in older adults. The correlation between equation-estimated CRF and CRF measured by GXT, the "gold standard," was relatively strong (r = 0.66) and almost identical to the correlation between CRF measured by graded exercise testing and CRF estimated from submaximal field testing. This latter finding alone suggests that assessment of CRF using this non-exercise method approximates the accuracy of using an established field test such as the Rockport one-mile walk to estimate fitness in older adults. It should be noted that the correlation between Rockport-estimated CRF and GXT-measured CRF in this sample (r = 0.68) is slightly lower than the correlation reported in the original validation paper for adults aged 60-69 (r = 0.74) [7]. In addition, other field-based measures such as the 400 meter walk of the LDCW correlate higher with measured CRF [8]. However, unlike the index validated herein, such measures still require participants to engage in exercise, and may not be practical for use with certain populations. Our findings deviated from those of Jurca et al. slightly in that resting heart rate was not a significant predictor of CRF in the present study. However, it must be noted that the regression coefficient for resting heart rate in the original validation paper was very small, suggesting that its statistical significance may have been driven by the large sample size.

We further tested the construct validity of the non-exercise CRF measure by examining the relationship of the MET values obtained from this measure and the maximal and submaximal exercise tests with self-reported cardiovascular conditions. As CRF is a reflection of efficient cardiovascular functioning, one would expect individuals with high CRF to have fewer cardiovascular conditions, and vice versa. All three measures of fitness were significantly associated with this measure, although the correlations with fitness measured by GXT were the strongest. This latter finding is to be expected, given that the GXT measures CRF most precisely. Although weaker, the correlations of cardiovascular conditions with estimated fitness obtained from the regression equation and submaximal methods were not significantly different from the association with the gold standard measure. More importantly, the correlation with the equation-predicted measure was almost identical to the correlation with the non-maximal measure, again suggesting that the non-exercise estimation of fitness is at least as strongly associated with cardiovascular conditions as a measure that requires significant resources in terms of medical supervision, time, cost, and exercise participation. If these findings can be replicated in other samples and when using other field and submaximal tests, this equation could be a very useful substitution to exercise testing, as it would significantly minimize the required resources.

The appeal of a non-exercise estimate of CRF is straightforward. The equation utilizes information routinely collected in a research laboratory, clinic, or physicians' practice to easily obtain an estimate of CRF and provide some initial indication of health status. For practitioners, the equation could be used as part of a fitness evaluation to provide more accurate and appropriate exercise recommendations that reflect initial fitness levels. For scientists conducting population-based research, such information would be particularly useful in examining associations between CRF and other important physical and mental health outcomes such as cognitive function, disability, and well-being. Although such potential outcomes are promising, continued validation evidence is necessary, particularly with older adults, other populations, and in larger samples. Moreover, it is of further interest to determine whether the equation-estimated CRF is predictive of change in fitness over time, either as a function of self-initiated behavior change or intervention participation. Finally, future studies should work towards determining a criterion estimated CRF level at which an individual is at greater risk for related health outcomes (CHD, mortality, etc.).

Our study is not without limitations. Our sample was composed largely of relatively healthy, white, female, well-educated, and economically advantaged community-dwelling older adults. However, the parent study from which the sample was drawn specifically recruited sedentary older adults and therefore the sample would appear to be representative of the older population. Replication in other adult populations is obviously warranted. Additionally, the equation does not consider other factors which may influence fitness levels such as genetic influences, prescription medications, or possible disease states. Although medications or chronic disease conditions could add to errors of prediction, the fact that participants were not disease- or medication-free also enhances the generalizability of our results to other community-dwelling older adult samples. One must also consider that Jurca et al. [16] used both maximal and submaximal exercise testing data to develop the original equation, and this likely introduced additional error to its predictive value. Finally, the self-reported physical activity level element, which is weighted heavily in the equation, is a subjective rating and, therefore, is open to the participant's interpretation. However, that self-reported physical activity makes a considerable contribution to the prediction of CRF adds further support to the validity of this aspect of the equation.

When performing a pre-exercise evaluation for at-risk populations or conducting research that necessitates precise measurement of CRF, it is recognized that there is no acceptable substitute for graded maximal exercise testing. There are several issues associated with using prediction equations, including determination of which is the "best" or most appropriate equation for a given population, potential lack of sensitivity to detect changes in CRF, and error associated with self-reported information. However, if a reasonable estimate of CRF is sufficient, and precise accuracy is not the primary concern, then the benefits of a non-exercise regression model such as that proposed by Jurca et al. [16] may be considerable.

This research provides preliminary evidence to support the use of a low-cost, low-risk, and relatively expedient way of estimating cardiorespiratory fitness in relatively healthy older adults [16]. This estimate of CRF does not involve any exercise testing, provides similar estimates to a popular field test estimate of CRF, and could be of utility in both clinical and research settings.