Print version ISSN 1517-8692
Rev Bras Med Esporte vol.9 no.5 Niterói Sept./Oct. 2003
Ecuaciones de predicción de la aptitud cardiorespiratoria sin test de ejercicios y su aplicabilidad a los estudios epidemiológicos: revisión descriptiva y análisis de estudios
Geraldo de Albuquerque Maranhão NetoI, III, IV; Paulo de Tarso Veras FarinattiII
de Pós-Graduação em Saúde Coletiva, Departamento
de Epidemiologia, Instituto de Medicina Social, Universidade do Estado do Rio
IILaboratório de Atividade Física e Promoção da Saúde, Instituto de Educação Física e Desportos, Universidade do Estado do Rio de Janeiro, Rio de Janeiro, RJ
IIIEscola de Educação Física, Centro Universitário da Cidade (UniverCidade), Rio de Janeiro, RJ.
IVDepartamento de Educação Física, Escola de Reabilitação, Universidade Católica de Petrópolis (UCP), Petrópolis, RJ
A low cardiorespiratory fitness is an independent risk factor for mortality from all causes, but mainly for coronary heart disease. Nevertheless, there are many difficulties to evaluate it by exercise testing in the epidemiological context. Alternative forms of evaluation have therefore been suggested using non-exercise regression models. This study aimed to review and critically analyze these models and their applicability in epidemiological studies. A systematic review was conducted considering papers published between 1966 and 2002. There were selected 24 studies attending the inclusion criteria. Only five of them related the standard error of estimation (SEE), the equation fully reported, present a higher sample size and made the cross-validation. These studies presented a higher adjusted R2, what mean the quality and the prediction power of them. The authors conclude that cardiorespiratory evaluation by non-exercise models in epidemiological studies could be feasible. However, few models seem to fulfill the minimum external validation requirements to provide data that could be generalized for large populations.
Key words: Physical fitness. Regression analysis. Epidemiology.
Una reducida aptitud cardiorrespiratoria es considerada como factor de riesgo independiente para el óbito por todas sus causas, pero principalmente por enfermedad coronaria. Debido a esa importancia y a la dificultad de evaluarla a través de tests de ejercicios, fueron sugeridas formas de evaluación alternativas, envolviendo ecuaciones de predicción sin la necesidad de realizar ejercicios físicos. El presente estudio se centró en describir y analizar de forma crítica esos modelos y, principalmente, en su aplicabilidad en estudios epidemiológicos. Se realizó una revisión sistemática de los artículos publicados entre 1996 y 2003. En total, fueron seleccionados 24 de ellos, obedeciendo a criterios de inclusión. Solamente cinco artículos reportaron el error padrón de estimación (EPE), la ecuación completa, presentando un número mayor de muestras y principalmente realizaron una verificación cruzada; además de eso, figuran entre los que presentan mayor valor de R2 ajustado, lo que ratifica la calidad y la fuerza de predicción de los mismos. Concluyéndose en primer lugar, que los modelos sin ejercicios pueden constituirse en una alternativa viable para la evaluación de la aptitud cardiorrespiratoria en estudios epidemiológicos. A pesar de esto, las ecuaciones disponibles, cuya validez permita un grado aceptable de generalización, son pocas.
Palabras clave: Aptitud física. Análisis de regresión. Epidemiología.
Cardiorespiratory fitness is considered a health-related fitness component that indicates the capability of cardiovascular and respiratory systems in providing oxygen during a continuous physical activity1,2. Morbidity and mortality risks of chronic-degenerative diseases, among them coronary artery disease, systemic high blood pressure, diabetes mellitus and some types of cancer have been associated to low cardiorespiratory fitness and physical activity3-9. It would be important to assess cardiorespiratory capability on a general population.
The use of cardiorespiratory fitness as an exposure variable in epidemiological studies is limited by the high costs, by technical operational difficulties, and by the time spent to measure it10,11. These facts have fostered the development of more simple methods, where the maximal and submaximal exercise tests have been replaced by multiple linear regression models to predict cardiorespiratory fitness from physical features and living habits11-14. This type of techniques, more simple, less costly, and easy to apply, favors the use of cardiorespiratory fitness as an exposure variable in epidemiologic studies, particularly in low-infrastructure sites12,15. Thus, the purpose of this investigation was to assess studies on non-exercise cardiorespiratory fitness predictive models, to describe the evolution of this type of technique, and to assess the developed models, particularly in regard to their quality.
The potentially useful articles were retrieved from references of published articles and books (manually) and by research in databanks Medlars online (Medline) Silver Platter, and Lilacs ("Latin America and the Caribbean Literature on Health Sciences Information"). The review was limited from January 1966 until December 2002 in the Medline; and in Lilacs from 1982 until 2002.
All potentially useful articles retrieved from electronic media had their abstracts downloaded and were independently assessed by two reviewers, one, a health-applied physical fitness specialist, knowledgeable on the theme under study, and the other an epidemiologist experienced in systematic reviews. The main inclusion criteria was collecting studies that focused non-exercise prediction of cardiorespiratory fitness, based on easy-to-measure useful variables for populational studies, such as weight, height, anthropometric measures, and fitness status. From the collection and reading of the articles, their references were tracked in search of other potentially useful articles. This task was repeated as many times as necessary until one believed that none of the references presented studies was not yet identified.
THE HISTORY OF NON-EXERCISE MODELS
The reviewing stage allowed the finding of 24 articles that met inclusion criteria, all of them original studies published since 1967. The development of the definition of purposes and methodological features of the studies made evident the efforts toward more accurate cardiorespiratory capability models. Thus, the studies will be presented chronologically. In order to enhance discussion on them, during the study description process one saw fit to present other investigations that, at some time, had contributed to the knowledge on the issue or cross-validated original predictive models.
The first investigations to suggest assessment of maximum oxygen uptake through variables other than exercise tests were carried out at the end of the 60s. At that time, research focused on measuring the amount to potassium in the body through a radio-diagnostic technique of muscular tissue16,17. This technique accepts that potassium levels in fat-free mass to be constant. Thus, once the amount of potassium in the body is established, it is possible to make predictions regarding lean mass. The rationale to assess lean mass is that a physically active individual would present a positive relation between cardiorespiratory fitness and muscular mass.
Shephard et al.18 published, in 1971, the first study aiming to predict cardiorespiratory fitness through multiple regression and without the use of exercise tests. Thirty-seven anthropometric measures and body strength indices were collected from 46 children and adolescents of both genders, as part of a randomized sample of Toronto (Canada) students. The most promising and applicable models for other studies were those based on body area (calculated through weight and height measures), in addition to skinfold of the thigh and age. The authors concluded that, for children, cardiorespiratory fitness could be conveniently predicted according to the proposed method. Two years later, Bruce et al.19 established some cardiorespiratory fitness predictive models with and without exercise tests, demonstrating that it could be predicted through variables such as gender, age, weight and the habit of practicing physical exercise, by the use of stepwise multiple regression analysis. This was the first study to use adults and to demonstrate that cardiorespiratory fitness could be predicted not only through anthropometric data, but also from behavioral variables, such as the practice of daily physical activity.
Among the studies that used anthropometric variables following the line of Shephard et al.18, Mayhew and Gifford's20 in 1975, and Bonen's et al.21 in 1979 stand out. In the former20, 31 boys age 7 to 9 years were studies, and VO2max was estimated through a number of anthropometric measurements. Initially, just the simple correlation of VO2max with the measurements was performed. Next, the stepwise multiple regression analysis was used to select the most representative models. Again, the most significant measurements were related to lower limbs: volume and skinfold of leg and thigh presented higher coefficient of explanation (R2 = 0.64). In the later21, also with children and adolescents, the authors checked the predictive power of age, weight and height of 100 boys age 7 to 15 years. According to the authors, the high coefficient presented (R2 = 0.88) and the fact that children did not adapt well to most exercise tests would strengthen even more the idea that predictive models with easy-to-measure variables would be an excellent alternative to an indirect calculation of aerobic power. Finally, in 1978, Taylor et al. 22 tried to predict the total time spent in minutes during a treadmill stress test through the total of scores from the Minnesota Leisure Time Physical Activity (MLTPA) developed at the University of Minnesota. The MLTPA seeks to assess physical activities practiced over the past year. This model did not present a very strong association (R2 = 0.27), which could indicate that just the history of physical activity should not be used to predict.
In the 80s, only three studies carried out in India sought to predict cardiorespiratory fitness using anthropometric variables only. In the first23, 27 anthropometric measurements using the stepwise multiple regression analysis were performed, to check which variables could significantly predict VO2max of 120 women and men. Four variables remained in the final model: weight, height, elbow diameter and chest skinfold. In another study with 70 male subjects age 11 to 18 years, Verma et al.24 found a VO2 max relationship with age, weight and height, identifying a higher explanation power (R2 = 0.81) in a regression model based only on weight. Finally, in 1998, with a sample of 146 men, Verma et al.25 checked how cardiorespiratory fitness could be predicted from age, height and weight. A model including age and weight was then designed. The two Indian studies with adult subjects had similar coefficients of explanation in their models (respectively, 0.29 and 0.35), suggesting that the use of anthropometric variables was not as suitable for adults as it was for children and adolescents. One possibility to consider, which would justify the relative success of the use of only anthropometric variables in models for children and adolescents, is the fact that biological age in childhood is directly related to body proportions. However, we do not have data to confirm this hypothesis.
In 1981, Leon et al.26 predicted the time spent on a maximum treadmill test by 175 middle-age men, using both anthropometric and behavioral variables. This was the first study to suggest the use of cardiorespiratory fitness prediction without exercise tests for epidemiological studies, based on the increase in number of evidences of a low cardiorespiratory fitness and the risk of dying from coronary artery disease. Eleven predictive variables were selected: age, rate of intense activities according to MLTPA, body mass index (BMI), past or current smoking, typical performance on a sweat- or dyspnea-causing occupational activity, amount of coffee, tea, or cola-type soft drink drank a week, habit of smoking pipe or cigar, leisure activities that caused sweat or dyspnea, average hours of sleep, and heart rate at rest. The authors concluded that a good cardiorespiratory capability could be predicted from standardized questionnaires, along with simple physical measurements, in spite of the determination coefficient value be moderate (R2 = 0.53).
Using self-reported physical activity to predict maximum oxygen uptake, Siconolfi et al.27 observed, in 36 men and 32 women, that the predictive power of the models was higher if they checked the intensity of the physical activity performed, rather than just checking whether subjects performed them or not. From this study on, the intensity of physical activity became a constant and very important variable in predictive studies. For instance, two years later Milesis28 estimated the time of performance in a maximum stress test of 126 men and 70 women, based on the variables gender, age, reciprocal weight index (height divided by the cubic root of weight), level of physical activities according to categories 1 to 5 (sedentary, little active, active, highly active, and athlete), background of smoking, according to categories 0 to 2 (never smoked, smoker of less than 20 cigarettes/day, and smoker of more than 20 cigarettes/day), and heart rate at rest. Kohl et al.15, through a questionnaire sent by mail, predicted the maximum performance time (in minutes) in a stress test applied to 375 subjects with mean age of 47.1 years. The predictive model included age and physical activity-related variables, such as a score for participating in activities such as walking and running, and the frequency these activities were performed under intensity enough to cause sweating.
A year later, in an important study because of the size of the sample, Blair et al.12 developed a model to predict the time of a maximum stress test on a treadmill, with 15.627 men (42.5 ± 9.5 years) and 3.943 women (42.1 ± 10.7 years). The subjects were divided in five groups according to age range, from those 20 to 29 years until those over 60 years, and got predictive models with explication coefficients ranging from 0.49 to 0.60 for males and 0.20 to 0.49 for females. The models included the following variables: BMI, heart rate at rest, rate of physical activity and leisure at the past month (being 1 equal to no physical activity practiced at the past month, and 5 to walking, running or jogging more than 32 km a week), and smoking (whether the subject smoked or not). In this study there is an additional evidence: alterations in BMI and rest were accountable for 14 to 19% of changes treadmill time.
Studies to predict cardiorespiratory fitness in subjects with heart condition were pioneered by Lee et al.29, at the end of the 80s, through the Specific Activity Scale - SAS30. Lee et al.29 demonstrated, in 36 heart-condition patients and healthy subjects that the self-reported ability in performing daily-life activities (such as putting on clothes, taking a shower or going up a flight of stairs) could add to the stress test in predicting cardiorespiratory capability. This could even enable the health team to decide whether or not the subject should undertake the test, depending on the reported limitations. Soon after, in order to use longitudinal epidemiological studies, Hlatky et al.31 validated a cardiorespiratory fitness predictive model without the use of exercise test in heart-condition patients. Initially, maximum oxygen uptake was correlated to functional capability of 50 patients according to the Duke University's Duke Activity Status Index - DASI). This index included 12 items, comprising activities related to personal care, home-making activities, sex, and recreational activities, weighted according to their individual metabolic expenditure measured in METs. Spearman's correlation was high (0.80). However, this first group was interviewed, but other 50 subjects filled out a questionnaire, and correlation was lower (0.58). At the end of the process, a simple regression model was generated from data of the first and second groups. According to the authors, further studies are necessary to check whether DASI is sensitive to detect longitudinal changes. Moreover, they do not believe that the questionnaire may replace the stress test, even being a good tool to assess the autonomy of coronary artery disease patients.
In 1990, Jackson et al.13 developed two models to predict cardiorespiratory fitness using variables gender, age, body composition and self-reported physical activity practice (from 0 to 7 according to intensity, being 0 for the person who did not take part in any physical activity or sports over the past month, and 7 for ones who run more than 10 miles or spent more than 3 hours per week practicing a physical activity similar to running). One of the models used BMI as measure of body composition, and the other the amount of fat (%F) predicted by skinfold measurements. Both, the %F (R2 = 0.66) and the BMI (R2 = 0.62) models showed, according to the authors, good predictive values for 1.393 males and 150 females age ranging from 20 and 70 years. The model's accuracy was confirmed when it was applied in the cross-validation sample with 423 males and 43 females, healthy and with high blood pressure. Pearson's correlation coefficients between predicted and observed values in the model including %F and BMI were of 0.82 and 0.79, respectively. Only in subjects with high level of fitness (VO2 max ³ 55 ml.kg-1.min-1) the models tended to underestimate fitness. However, this type of people is high above the average of the population, and does not affect applicability of the models for a large sample. These models showed to be more accurate than Åstrand's and Ryhming's32 predictive treadmill model, that used heart rate measured at submaximal exertion on the treadmill. These results confirmed the ideas advocated by Shephard et al.18 that non-exercise models could be more accurate that submaximal physical tests. Moreover, this was the first study the investigators had special concern with cross-validation procedures.
The interest for the models proposed by Jackson et al.13 lead to the carrying out of two studies, whose purpose was to check the accuracy of the proposed models in two samples of different features: Kolhorst and Dolgener33 checked model validity in 69 physically active university students. The study included 28 men and 41 women, mean age of 21 ± 2 years. Upon applying Pearson's correlation to compare results of measured and predicted cardiorespiratory fitness, the authors observed that the two models of Jackson et al.13 did not present good correlation (r = 0.72), confirming the conclusion of the original study that their applicability was limited to highly fit subjects. In 1996, Williford et al.34 checked cross-validation of non-exercise models13 in a sample of 165 women, as the cross-validation sample in the original study was small (n = 43). Both, the BMI and the %F models showed good correlation (r = 0.81 and 0.86 respectively), confirming accuracy of these models also in women aged 18 to 45 years. The model was able to predict fitness of 87% of the women with VO2max < 32 ml.kg-1.min-1, a value with higher association to mortality risk from all causes12, suggesting its use in epidemiological investigations.
In 1992, Ainsworth et al.13 developed a model to predict cardiorespiratory fitness by asking the frequency a subject would perform intense physical activities for over 15 minutes, in addition to other easy-to-assess variables, such as age, gender and BMI. For that, they had a somewhat small sample of 27 men and 47 women, age between 21 and 59 years. An interesting feature of this study is that for the authors to reach the most suitable question on the physical activity, they applied a number of physical activity questionnaires used in epidemiological investigations35-39. At the end, just one question on the regular practice of more intense physical activities remained in the model37, strengthening the idea that the variable physical activity could be assessed in a simple way, to inform on cardiorespiratory fitness, as long as the intensity of the activity was taken into consideration, along with other variables. Two years later, Myers et al.40 developed a model to predict performance in a maximum stress test on a treadmill, in 207 men and 5 women (62 ± 8 years) with heart condition, through the Veteran Specific Activity Questionnaire - VSAQ and age. The study subjects informed in the questionnaire which physical activity they were able to perform without exertion limiting symptoms (fatigue, uneasiness in the chest, dyspnea). Through the multiple regression model generated, the authors developed a nomogram in which, from the number of METs defined at the questionnaire, and the age of the subject, his/her performance on the treadmill was predicted. According to the authors, the model did not have the purpose of replacing the ergometric test, but would enable the health team to have an idea of the subject's physical fitness, and would adjust the test to this status.
Whaley et al.41 developed another fitness prediction model with variables gender, age, heart rate at rest, weight, proportion of fat, smoking (from 1 to 8 according to the frequency and number of cigarettes smoked, 1 being for non-smokers, and 8 to smoking more than two packs a day), and self-reported physical activity (from 1 to 6 according to the intensity, 1 being for the sedentary subject, and 6 for the highly fit subject, who runs, cycles or swims more than 20 miles a week). Seven hundred and two males and 473 females took part in the study, and the predictive model presented good accuracy (R2 = 0.72). Like Jackson's et al.13 study, this one also carried out a cross-validation of the model. Pearson's correlation between predicted and measured values (r = 0.85) led the authors to consider the model valid. Still in that year, Heil et al.14 validated a non-exercise model with variables gender, age and age2, proportion of fat, and the score of Jackson's et al.13 physical activity levels in 229 women and 210 men aged 20 to 79 years (R2 = 0.77). Cross-validation was carried out in 65 subjects with features similar to the group to which the model was validated. According to the authors, Pearson's correlation was good (r = 0.85); however, the small sample somewhat limits the results of the cross-validation. Notwithstanding, the generated model reinforces the idea that it is actually possible to predict cardiorespiratory fitness from some variables suggested by Jackson et al.13.
In 1996, another predictive questionnaire was validated, based on functional impairment of heart-condition patients: the Specific Activity Questionnaire - SAQ, with 13 questions related to daily-life42. Ninety-seven patients (being 12 females) had their fitness predicted through SAQ score, height, age and weight (R2 = 0.50). Pearson's correlation was calculated between SAQ and other questionnaires in regard to cardiorespiratory fitness, and the following results were found: SAQ (r = 0.71), SAS29 (r = 0.35), DASI31 (r = 0.62) and VSAQ40 (r = 0.66). For the authors, this evidenced the potential use of SAQ in studies with heart-condition subjects if the stress test was costly or unfeasible. In that same year, Cardinal43 published a study in which he checked whether the models proposed by Jackson et al.13 and Ainsworth et al.10 were associated between themselves and with other physical activity indices, in 123 healthy women (age = 38.8 ± 8.4 years). The conclusion was that both, the models (r = 0.80) and the physical activity indices (0.26 a 0.74), had an overall good correlation between themselves, and followed similar classification criteria. A year later, George et al.44 established a predictive model adjusted for young, physically active students, in a sample of 50 males and females aged 18 to 29 years. To increase accuracy of the final model (R2 = 0.72), as attempts to predict fitness of highly fit individuals had failed so far, new variables were added. Among them, there was a question on the perceived fitness to perform activities such as walking and running, in which people should inform at what pace they could move without become extremely tired. Another question was related to the history of physical activity practice, ranging from 0 to 10 over a six-month period, rather than from 0 to 7 over one month, as proposed in Jackson's et al.13 study. The authors considered this to be the first model with no need for any measuring, as weight and height to calculate BMI were self-reported. In this study, the process of cross-validation was different than the prior ones. Instead of using a sub-sample of the whole group under investigation, which, according to the authors, would limit the sample, it was used the method of adding the square of the predicted residues (PRESS). This method allows the use of all subjects in the sample, in both validation and cross-validation. For this purpose, it is based on the calculation of the predicted residues for each subject, while he/she is excluded from the original model45. From adding the square of these residues it is possible to calculate R2 (0.71) and the standard error of the estimate, evidencing the good accuracy of the model.
In 1999, Mathews et al.11 proposed a model and examined its sensitivity to rate cardiorespiratory fitness. The authors considered that not doing this would limit the application of the models in epidemiological studies. The rating would enable disease-risk estimates to be compared among different fitness levels. Following the example of George et al.44, only self-reported variables were included in the model: age, age2, gender, reported physical activity (as proposed by Jackson et al.13), height and weight, (R2 = 0.74). Rating accuracy of the model was assessed by tabulating data into age and gender categories, and distributing them in fifths of measured and predicted cardiorespiratory fitness. The overall accuracy rating of the model was modest (36%). However, 83% of all subjects were appropriately classified, or in the closest fifth. The extreme error in classifying from the lowest to the highest fifth was seldom observed (0.13%), leading to the conclusion that the predicted fitness values could be used as an exposure variable in epidemiological studies when the stress test was not a feasible option. For the process of cross-validation, the PRESS method was also used, confirming the validity of the model (R2 = 0.74).
In a study that seems to be, until the writing of this text, the last on non-exercise models to predict cardiorespiratory fitness, Wu e Wang46 established a model from the observation of 24 workers of both genders living in Taiwan. The significant variables in the regression model were gender, age and BMI (R2 = 0.77), confirmed by the cross-validation process in a small sample (N = 6). The authors believed that the model could suit an occupationally active population. However, extrapolation of the results is obviously impaired by the very small sample.
CRITICAL ANALYSIS OF THE REVIEWED MODELS
Tables 1 and 2 present the studies, their country of origin, sample, gender, age group, and predictive models with adjusted R2 and standard error of estimation (SEE). Table 1 presents all studies that used VO2 max as a dependent variable, both in relative (ml.kg-1.min-1) and absolute (l.min-1) terms. In table 2, the dependent variable was the time spent performing a maximum stress test on treadmill (in minutes) or its maximum intensity (in METs).
Multiple linear regression has been the statistical analysis most often used to predict cardiorespiratory fitness without exercise test. It is to be used when the investigator intends to explain which variables add to the prediction of the dependent variable (cardiorespiratory fitness), and the magnitude of their role47. In some studies, only simple linear regression was studied24,27,29,31. All articles presented the equation's R (multiple correlation coefficient) or R2 (explication coefficient) related to the explanatory capability of the model. The presentation of these data is according to the recommendations found in the literature48. The adjusted R2 can be easily calculated and is useful for a better analysis of the models, as it is not influenced by the number of independent variables. On the other hand, R2 tends to inflate as a function of the amount of variables included in the model. The adjusted R2 is calculated through the formula48
adjusted R2 = 1 - [(1 - R2) n-1/n-p],
where n is the number of subjects in the sample and p the number of parameters: the rationale for calculating R2 is to analyze and compare the quality of the adjustment of predictive models with different amounts of variables. Based on this calculation, one can state that models that present value higher than the adjusted R2 are those with higher explanatory capability in the sample for which they were validated. As to the SEE, it indicates the variation not explained by the regression line, being a discrepancy measure among the observed and predicted variables. Some authors consider that non indicating SEE lessens the quality of the study, particularly if they do not present the complete model48,49. The fact that some studies did not predict maximum oxygen uptake (table 2) implies that the models relate to a doubly-indirect independent variable, which would affect even more the quality of the models, for the time spent on a maximum test is already an indirect indicator of cardiorespiratory fitness. Mechanical efficiency is also a factor that interferes in the result, regardless of VO2. Furthermore, other studies21,23-25 have reported that the maximum oxygen intake was predicted rather than measured, which affect even more the quality of these predictions.
The five studies with the higher adjusted R2 value (table 1) and among the most recent are those that include SEE and the model, had higher number of subjects in the sample, and performed cross-validation. Some considerations must be made as to advantages and disadvantages of these studies: a tendency in the two most recent predictive studies11,44 was the use of only self-reported variables in the prediction model, in order to further decrease time of application. This tendency made models with measured predictive variables, such as heart rate at rest41, to be hampered. The reason being that, in spite of the measures seeming to be simple to assess, they may require some time to be assessed. For instance, it takes from 5 to 10 minutes to accurately assess heart rate, which is longer than the time required for some submaximal tests. Variables that need well-trained personnel to properly assess them, such as skinfold measures, may also be of disadvantage to be applied in models for the study of big samples. In some studies13,14,41 the skinfold method was used to predict the percentage of body fat through another model, which may include errors from restrictions proper to this type of prediction.
An interesting point that should be reviewed in George's et al.44 study is that, as the authors mentioned, it would be specific for well-fit individuals. If this is to be true, it can be argued that its application is for a small portion of the population. On the other hand, individuals with low cardiorespiratory fitness present higher risk to develop cardiovascular and metabolic diseases. This considered, this model could be regarded as detrimental for epidemiological research. Another important factor to be taken into consideration in any study aiming to predict non-exercise cardiorespiratory fitness is the influence of genetics in the level of fitness. Some studies showed that this could influence about 30 to 40% of the magnitude of results50. Notwithstanding, variables related to perceived exertion in daily activities, such as walking and running, used by George et al.44, may be an alternative for cardiorespiratory fitness not related to physical activity history, as a subject may present a good cardiorespiratory fitness without necessarily practicing exercises regularly.
The methodological progress of the studies and the high accuracy of the established models, most of them for healthy subjects, suggests this type of prediction to be a good alternative to rate cardiorespiratory fitness. However, these models are still little used in epidemiological investigations. The models proposed by Jackson et al.13 are the only ones used so far by other studies51,52. However, the testing of model applicability is still deficient. Some reasons may be pointed out: first, most models have been developed using sample of subjects of high or average social, economic and cultural levels. As this profile does not match the social features of most of the population, the extrapolation potential of the models is being limited, and they should not be broadly applied.
Another aspect is that the self-reported physical activity is limited, in most studies, to leisure activities 11,13-15. Only Whaley's et al. study41 included occupational tasks in their model, in addition to leisure activities. In Ainsworth's et al.10 study, occupational activity was also assessed. However, as the sample subjects had sedentary or extremely light occupational activities, this variable did not add to the final predictive model. By the way, populations of low occupational-level activities are found in a number of studies11,13-15,34. Even the population of Blair's et al. study12, with 15,627 males and 3,943 females could not be considered as representative of the American population, as they performed low-intensity occupational activities, had high educational level, and were from average to high social-economic level. This limitation was acknowledged by the authors themselves12. Finally, Wu and Wang's study46, in spite of proposing to apply the model to subjects occupationally active, does not include any physical activity-related activity, which may be necessary in a higher, more heterogeneous group. As to the existing models to predict fitness in heart-condition patients, one may say that they provide additional information on the autonomy of the subjects. However, they typically do not dare to predict a cardiorespiratory fitness.
The non-exercise studies may be compared to some submaximal tests used in epidemiological investigations. If some submaximal tests showed an R2 value higher than those without exercise (0.81; 0.85), SEE values are comparable (± 4 ml.kg-1.min-1 )53, or are not reported54, which limits their use. Moreover, in some submaximal tests54, validation was performed in a very narrow age group, which hampers a comprehensive use of the models. Another interesting issue is related to the comparison of the models with the questionnaire of physical activity history, often used in epidemiological research. It is a fact that nature and intensity of daily activities may influence the subject's cardiorespiratory fitness56,57. Therefore, it should not be ignored as a dependent variable in the development of predictive models. On the other hand, the correlation between physical activity and cardiorespiratory fitness tends to be small when based on information from questionnaires. High correlation is connected to intense physical activities. In spite of physical activity be considered the main determinant of cardiorespiratory fitness, physical activity information given on an interview or self-reported, with average duration between 15 to 45 minutes, does not seem suitable to assess cardiorespiratory fitness39. However, this may be an interesting aspect to be studied, as many a time predictive models derive from simple questionnaires, whose predictive power is enhanced as one adds easy-to-measure variables in more complex models. The strategies used by Mathews et al.11 of considering self-reported variables only, even for weight and height, for not requiring long measuring time nor trained evaluators, seem to be quite appropriate to apply in epidemiological studies, and future studies should consider these principles.
Non-exercise predictive models are a subject of interest for investigators worldwide. In principle, non-exercise models may be a feasible alternative to assess cardiorespiratory fitness in epidemiological studies. The fact that only a few models exist whose validity allows for an acceptable degree of generalization shows that this area has been scarcely explored, and new investigations on the theme should be carried out.
Some prospective issues should be addressed. First, for further applicability in epidemiological studies, the new investigations should not limit themselves to validation and cross-analysis only, but should tackle longitudinal sensitivity of VO2max prediction. This means, so far, one does not know if changes in predictive variables over time (changes due to training) may be detected by non-exercise models. Moreover, one must acknowledge that there are few studies that focus the development of models to be applied to special groups, such as (particularly) the elderly, children, adolescents, women, or heart-condition patients. Those that exist have low generalization capability, due to the small sample used in their development. Another important aspect is the role of social, economic and cultural components: the designing of regression models to predict cardiorespiratory fitness should take into consideration the characteristics of the population under investigation, particularly in the selection of predictive variables. It is clear the need to include, in studies to be validated, subjects of average and low social-economic status, as well as the physical activity required by their occupation. These features tend to be specific for the observed region or population. This evidences the need for studies that develop models suited to the different Brazilian regions and social status of their population, or, at least, to check the validity of the existing models in the Brazilian scenario.
We thank Doctors Antonio Claudio Lucas da Nóbrega and Antonio Ponce de Leon for their important collaboration, and to undergraduate student Marcelo Moreira Antunes for the translation from the Chinese. This study was partially funded by the State of Rio de Janeiro Research Support Foundation (Fundação de Apoio à Pesquisa do Estado do Rio de Janeiro - Process E-26/151.802/1999) and by the National Council on Technical and Scientific Development (Conselho Nacional de Desenvolvimento Científico e Tecnológico Research Productivity modality, process 300754/2000-0).
All the authors declared there is not any potential conflict of interests regarding this article.
1. Caspersen CJ, Powell KE, Christensen GM. Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public Health Rep 1985;100:126-31. [ Links ]
2. American College of Sports Medicine. ACSM's guidelines for exercise testing and prescription. 6th ed. Baltimore: Williams and Wilkins, 2000. [ Links ]
3. Bouchard C, Shephard RJ, Stephens T. Physical activity, fitness, and health: international proceedings and consensus statement. Champaign, IL: Human Kinetics, 1994. [ Links ]
4. Pate RR, Pratt M, Blair SN, Haskell WL, Macera CA, Bouchard C, et al. Physical activity and public health. A recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 1995;273:402-7. [ Links ]
5. U.S. Department of Health and Human Services. Physical activity and health: a report of the surgeon general. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996. [ Links ]
6. Lee CD, Blair SN, Jackson AS. Cardiorespiratory fitness, body composition, and all-cause and cardiovascular disease mortality in men. Am J Clin Nutr 1999;69:373-80. [ Links ]
7. Wei M, Kampert JB, Barlow CE, Nichaman MZ, Gibbons LW, Paffenbarger RS, et al. Relationship between low cardiorespiratory fitness and mortality in normal-weight, overweight, and obese men. JAMA 1999;282:1547-53. [ Links ]
8. Delvaux K, Philippaerts R, Lysens R, Vanhees L, Thomis M, Claessens AL, et al. Evaluation of the influence of cardiorespiratory fitness on diverse health risk factors, independent of waist circumference, in 40-year-old Flemish males. Obes Res 2000;8:553-8. [ Links ]
9. Blair SN, Cheng Y, Holder JC. Is physical activity or physical fitness more important in defining health benefits? Med Sci Sports Exerc 2001;33:S379-99. [ Links ]
10. Ainsworth BE, Richardson MT, Jacobs DR, Leon AS. Prediction of cardiorespiratory fitness using physical activity questionnaire data. Med Exerc Nutr Health 1992;1:75-82. [ Links ]
11. Mathews CE, Heil DP, Freedson PS, Pastides H. Classification of cardiorespiratory fitness without exercise testing. Med Sci Sports Exerc 1999;31: 486-93. [ Links ]
12. Blair SN, Kannel WB, Kohl HW, Goodyear N, Wilson PWF. Surrogate measures of physical activity and physical fitness. Am J Epidemiol 1989; 129:1145-56. [ Links ]
13. Jackson AS, Blair SN, Mahar MT, Wier LT, Ross RM, Stuteville JE. Prediction of functional aerobic capacity exercise testing. Med Sci Sports Exerc 1990;22:863-70. [ Links ]
14. Heil DP, Freedson PS, Ahlquist LE, Price J, Rippe J. Nonexercise regression models to estimate peak oxygen consumption. Med Sci Sports Exerc 1995;27:599-606. [ Links ]
15. Kohl HW, Blair SN, Paffenbarger RS Jr, Macera CA, Kronenfeld JJ. A mail survey of physical activity habits as related to measured physical fitness. Am J Epidemiol 1988;127:1228-39. [ Links ]
16. Cotes JE, Davies CTM, Healy MJR. Factors relating to maximum oxygen intake in young adult male subjects. J Physiol 1967;189: 79-80. [ Links ]
17. Cotes JE. Relationships of oxygen consumption, ventilation and cardiac frequency to body weight during standardized submaximal exercise in normal subjects. Ergonomics 1969;12:415-27. [ Links ]
18. Shephard RJ, Weese CH, Merriman JE. Prediction of maximal oxygen intake from anthropometric data. Int Z Angew Physiol 1971;29:119-30. [ Links ]
19. Bruce RA, Kusumi F, Hosmer D. Maximal oxygen and nomographic assessment of functional aerobic impairment in cardiovascular disease. Am Heart J 1973;85:546-62. [ Links ]
20. Mayhew JL, Gifford PB. Prediction of maximal oxygen uptake in pre-adolescent boys from anthropometric parameters. Res Quart 1975;46: 302-11. [ Links ]
21. Bonen A, Heyward VH, Cureton KJ, Boileau RA, Massey BH. Prediction of maximal oxygen uptake in boys, ages 7-15 years. Med Sci Sports 1979;11:24-9. [ Links ]
22. Taylor HL, Jacobs DR, Schucker B, Knudsen J, Leon AS, Debacker G. A questionnaire for the assessment of leisure time physical activities. Journal of Chronic Diseases 1978;31:741-55. [ Links ]
23. Verma SS, Bharadwaj H, Malhotra MS. Prediction of maximal aerobic power in healthy Indian males from anthropometric measurements. Z Morphol Anthropol 1980;71:101-6. [ Links ]
24. Verma SS, Gupta RK, Kishore N, Gupta JS. A simple relationship between aerobic power and body weight in Indian adolescent boys. Indian J Med Sci 1986;40:93-6. [ Links ]
25. Verma SS, Sharma YK, Kishore N. Prediction of maximal aerobic power in healthy Indian males 21-58 years of age. Z Morphol Anthropol 1998;82:103-10. [ Links ]
26. Leon AS, Jacobs DR, DeBacker G, Taylor HL. Relationship of physical characteristics and life habits to treadmill capacity. Am J Epidemiol 1981; 113:653-60. [ Links ]
27. Siconolfi SF, Lasater TM, Snow RCK, Carleton RA. Self-reported physical activity compared with maximal oxygen uptake. Am J Epidemiol 1985;122: 101-5. [ Links ]
28. Milesis CA. Prediction of treadmill performance from clinical characteristics in healthy persons. J Cardiopulm Rehabil 1987;7:365-73. [ Links ]
29. Lee TH, Shammash JB, Ribeiro JP, Hartley LH, Sherwood J, Goldman L. Estimation of maximum oxygen uptake from clinical data: performance of the specific activity scale. Am Heart J 1988;115:203-4. [ Links ]
30. Goldman L, Hashimoto B, Cook F, Loscalzo A. Comparative reproducibility and validity of systems for assessing cardiovascular functional class: advantages of a new specific activity scale. Circulation 1981;64:1227-34. [ Links ]
31. Hlatky MA, Boineau RE, Higginbotham MB, Lee KL, Mark DB, Califf RM, et al. A brief self-administered questionnaire to determine functional capacity (the duke activity status index). Am J Cardiol 1989;64:651-4. [ Links ]
32. Astrand PO, Ryhming I. A nomogram for calculation of aerobic capacity (physical fitness) from pulse rate during submaximal work. J Appl Physiol 1954;7:218-21. [ Links ]
33. Kolkhorst FW, Dolgener FA. Nonexercise model fails to predict aerobic capacity in college students with high aerobic VO2peak. Res Q Exerc Sport 1994;65:78-83. [ Links ]
34. Williford HN, Scharff-Olson M, Wang N, Blessing DL, Smith FH, Duey WJ. Cross-validation of non-exercise predictions of VO2peak in women. Med Sci Sports Exerc 1996;28:926-30. [ Links ]
35. Paffenbarger RS Jr, Hyde RT, Wing AL, Hsieh CC. Physical activity, all-cause mortality, and longevity of college alumni. N Engl J Med 1986;314: 605-13. [ Links ]
36. Haskell WL, Taylor HL, Wood PD, Schrott H, Heiss G. Strenuous physical activity, treadmill exercise test performance and plasma high-density lipoprotein cholesterol. The Lipid Research Clinics Program Prevalence Study. Circulation 1980;62:IV53-61. [ Links ]
37. Godin G, Shephard RJ. A simple method to assess exercise behavior in the community. Can J Appl Sport Sci 1985;10:141-6. [ Links ]
38. Sallis JF, Haskell WL, Wood PD, Fortmann SP, Rogers T, Blair SN, et al. Physical activity assessment methodology in the Five-City Project. Am J Epidemiol 1985;121:91-106. [ Links ]
39. Jacobs DR Jr, Ainsworth BE, Hartman TJ, Leon, AS. A simultaneous evaluation of 10 commonly used physical activity questionnaires. Med Sci Sports Exerc 1993;25:81-91. [ Links ]
40. Myers J, Do D, Herber W, Ribisl P, Froelicher VF. A nomogram to predict exercise capacity from a specific questionnaire and clinical data. Am J Cardiol 1994;73:591-96. [ Links ]
41. Whaley MH, Kaminsky LA, Dwyer GB, Getchell LH. Failure of predicted VO2máx to discriminate physical fitness in epidemiological studies. Med Sci Sports Exerc 1995;27:85-91. [ Links ]
42. Rankin SL, Briffa TG, Morton AR, Hung J. A specific activity questionnaire to measure the functional capacity of cardiac patients. Am J Cardiol 1996;77:1220-3. [ Links ]
43. Cardinal BJ. Predicting cardiorespiratory fitness without exercise testing in epidemiologic studies: a concurrent validity study. J Epidemiol 1996;6:31-5. [ Links ]
44. George JD, Stone WJ, Burkett, LN. Non-exercise VO2max estimation for physically active college students. Med Sci Sports Exerc 1997;22:415-23. [ Links ]
45. Holiday D, Ballard J, McKeown B. PRESS-related statistics: regression tools for cross-validation and case diagnostics. Med Sci Sports Exerc 1995;27:612-20. [ Links ]
46. Wu HC, Wang MJJ. Establishing a prediction model of maximal oxygen uptake for young adults. Journal of the Chinese Institute of Industrial Engineers 2002;19:1-7. [ Links ]
47. Greenhalgh T. How to read a paper: statistics for the non-statistician. II: "significant" relations and their pitfalls. BMJ 1997;315:422-5. [ Links ]
48. Altman, D. Practical statistics for medical research. London: Chapman and Hall, 1995. [ Links ]
49. Morrow JR, Jackson AW, Disch JG, Mood DP. Measurement and evaluation in human performance. Champaign, IL: Human Kinetics, 1995. [ Links ]
50. Perusse L, Tremblay A, Leblanc C, Bouchard C. Genetic and environmental influences on level of habitual physical activity and exercise participation. Am J Epidemiol 1989;129:1012-22. [ Links ]
51. Rossy LA, Thayer JF. Fitness and gender-related differences in heart period variability. Psychosom Med 1998;60:773-81. [ Links ]
52. Kelley GA, Lowing L, Kelley K. Gender differences in the aerobic fitness levels of young African-American adults. J Natl Med Assoc 1999;91:384-8. [ Links ]
53. Jette M, Campbell J, Mongeon J, Routhier R. The Canadian home fitness test as a predictor for aerobic capacity. CMAJ 1976;114:680-2. [ Links ]
54. Siconolfi SF, Garber CE, Lasater TM, Carleton RA. A simple, valid step test for estimating maximal oxygen uptake in epidemiologic studies. Am J Epidemiol 1985;121:382-90. [ Links ]
55. Salmon J, Owen N, Bauman A, Schmitz MK, Booth M. Leisure-time, occupational, and household physical activity among professional, skilled, and less-skilled workers and homemakers. Prev Med 2000;30: 191-9. [ Links ]
56. Smolander J, Blair SN, Kohl HW 3rd. Work ability, physical activity, and cardiorespiratory fitness: 2-year results from Project Active. J Occup Environ Med 2000;42:906-10. [ Links ]
Laboratório de Atividade Física e Promoção da Saúde
Instituto de Educação Física e Desportos
Universidade do Estado do Rio de Janeiro
Rua São Francisco Xavier, 524, sala 8.133-F
20550-013 - Rio de Janeiro, RJ, Brasil
Received in 19/6/03
2nd version received in 21/9/03
Approved in 27/9/03