Cardiopulmonary exercise test (CPET) presents indisputable clinical applicability for the diagnosis of exercise intolerance and its causes, as well to quantify cardiorespiratory fitness (CRF) [1, 2]. Therefore, the maximal pulmonary oxygen uptake (V˙O2max) obtained at the end of a CPET with maximum effort is widely used. Obese individuals present decreased CRF, mainly caused by low pulmonary complacency and functional residual capacity [3] combined with a less pronounced chronotropic response during sustained effort [4].

The total duration of CPET, as well as V˙O2max and maximum heart rate (HR_max), are inversely proportional to body mass index (BMI) [5]. Considering possible difficulties in reaching the maximum effort leading to interpretative limitation of the maximal incremental exercise variables (e.g., V˙O2max, HR_max, metabolic equivalent of task (MET), gas exchange rate (R) and the other variables registered at peak of effort), the CPET interpretation can be significantly improved by an evaluation of the tendency of submaximal relationships (e.g., anaerobic threshold (AT) values and lowest V˙E/V˙CO2, while OUES and the aforementioned ratios and intercepts) obtained before the peak of the exercise.

Previously identified as determinants for prognosis and survival in several clinical situations [6], these submaximal relationships have been recommended in the evaluation of chronic diseases patients, mainly cardiorespiratory [7]. Also known as dynamic physiological responses, the main advantage of these variables is its linear behavior during the increment or workload, regardless achieving maximum effort [8, 9]. In addition, the evaluation of these submaximal variables is especially interesting in obese population, which present difficulties to achieve maximum effort [10, 11].

The higher slope of the ΔHR/ΔV˙O2 relationship, for example, is an important indirect indicator of cardiovascular inefficiency and peripheral muscle disability. Similarly, ventilatory inefficiency can be measured either by the greater slope of the ratio between minute ventilation (V˙E) per unit of carbon dioxide (V˙CO2) released (ΔV˙E/ΔV˙CO2) or by the lowest point (known as nadir) of the quotient of ΔV˙E/ΔV˙CO2 (lowest ΔV˙E/ΔV˙CO2), which indicates increased respiratory dead space and/or hyperventilation. In addition, the slope of the relationship between tidal volume (Vt) and V˙E linearized by logarithmic transformation (ΔVt/ΔlnV˙E) translates the tachycardic pattern during exercise [1, 8]. The relationship between V˙O2 and V˙E logarithm in base 10 (logV˙E) (ΔV˙O2/ΔlogV˙E) translates the oxygen uptake efficiency slope (OUES) [1, 8].

Although several variables obtained in maximal effort identify worse values in obese, there is considerable controversy over dynamic physiological responses. Limited evidence shows that both OUES and V˙E/V˙CO2 present comparable values between obese and non-obese individuals [12, 13]. However, such evidence was obtained in studies with insufficient samples and inadequate statistical power and, hence, the association of obesity with these variables remains unknown. Therefore, we hypothesized that the main dynamic physiological responses of clinical interest do not suffer significant influence of obesity, maintaining their diagnostic and prognostic values unchanged. We aimed to evaluate the influence of adulthood obesity on maximal and submaximal physiological responses during CPET, emphasizing the analysis of submaximal dynamic variables.

From 1,724 CPETs, 1,594 were eligible for analysis (839 and 755 were non-obese and obese participants, respectively), as can be seen in Fig 2.

We considered 442 normal weight and 397 overweight participants as non-obese group. The obese group was composed by 292, 247 and 216, respectively, classes I, II and III. As expected, we found significant differences for anthropometric and cardiovascular risk variables when compared obese and non-obese participants. On average, obese group was mainly classified as obesity class II. Except of current smoking, obese participants had a higher proportion of cardiovascular disease risk factors in comparison with non-obese group (Table 1).

Obese participants presented the worst performance in most of the variables evaluated, when the indices were not adjusted by main confounders (t-test). Only few measures (lowest value of V˙E/V˙CO2 and ΔV˙E/ΔV˙CO2) were not significantly different between groups (Table 2).

Similarly, after we performed a multivariate analysis, obesity remained negatively influencing most of the 40 physiological variables investigated. When higher BMI, lesser CPET duration and increment (treadmill speed and grade). Cardiovascular variables, for instance, HR_max, HR_{% pred}, HR_reserve and ΔHR/ΔV˙O2, and the metabolic variables V˙O2max, V˙O2%pred, MET and R presented the similar behavior (Table 3).

We observed differences mainly between non-obese and obese for gas exchange variables (PETO₂, PETCO₂, V˙E/V˙CO2, V˙E/V˙O2). Among the obese groups, PETO₂ and V˙E/V˙O2 were different only between the classes I and III (Table 3).

ΔVt/ΔlnV˙E and HR recovery were only altered when comparing normal weight and obese class III. Although PuO2 and OUES differed among the extreme groups (normal weight and obesity class III), we also found differences among the other groups in comparison to them. We found differences in AT when comparing obesity class III with the other groups, for both non-obese and obese. In contrast, all groups presented differences in SBP when compared to normal weight group (Table 3).

Regarding the ventilatory (RF, V˙E max, V˙E/MVV, V˙E/MVV) and some cardiovascular (DBP max and DP) variables, we observed inconsistent results (Table 3).

Lastly, Vt _max, Vt/VE, Vt/IC, Vt/FVC, Lowest V˙E/V˙CO2, ΔV˙E/ΔV˙CO2, and the y-intercepts of V˙E/V˙CO2, did not significantly differ among obese and non-obese participants, as can be seen in Table 3 and Fig 3.

The purpose of present study was to evaluate the influence of adulthood obesity on maximal and submaximal physiological responses during CPET, emphasizing the analysis of submaximal dynamic variables. We observed that both peak and submaximal responses are influenced by obesity, to a greater or lesser degree. Maximal cardiovascular and metabolic variables were the most affected responses regarding obesity, but, for some ventilatory and cardiovascular responses, our findings were inconclusive due to inconsistent results. For other variables, mainly gas exchange responses, we found differences when comparing non-obese and obese or related to severity of obesity. However, our main finding was the non-significant influence of obesity on ventilatory efficiency variables (Lowest V˙E/V˙CO2, ΔV˙E/ΔV˙CO2 and the y-intercepts of V˙E/V˙CO2), which may suggest best prognostic value regardless of overweight and obesity. To our knowledge, this is the first study to investigate the influence of obesity, according to severity and adjusted by the main cardiovascular risk factors, on the dynamic physiological responses obtained in the CPET in a robust sample.

As expected, maximal cardiovascular responses (HR_max, HR% _pred, HR _reserve) and cardiovascular efficiency (ΔHR/ΔV˙O2) were altered along with overweight and obesity. Our findings are corroborated by previously studies that showed rapidly increased cardiac output and unfavorable biomechanics leading to cardiovascular inefficiency.

In contrast, HR recovery and PuO₂ seems to be related to the severity of obesity. Similarly, previous studies have observed a slow HR_recovery in obese individuals [20, 21] and its improvement after weight loss [22]. In addition to relevant methodological differences, Carnethon et al. [20] characterized the sample according to normal weight, overweight and obese participants and presented proportionally less obese subjects when compared to our study. Thus, our study adds to literature by showing that this difference occurs in morbidly obese participants. Our findings also suggest that HR_recovery ≥ 22 bpm, as a parasympathetic normal response, can be applied to interpret results from normal weight, overweight and obesity classes I and II. Therefore, we recommend the use of HR_recovery in CPET analysis and interpretation since it is a simple, reproducible, and easy-to-access variable, besides having a high mortality predictive value, association with increased cardiovascular risk [23, 24] and additional value to the CRF prognosis [25].

Regarding PuO₂, literature remains scarce related to obesity and there is a lack of value to interpret this response. In the present study, morbidly obese had higher peak values of PuO₂, which may be possible explained by the high consumption of V˙O2 per workload unit presented by obese individuals. Similar results were observed by Serés et al. [26], who attributed these finds to a greater stroke volume related to an increased weight bearing in obese individuals. Previous study suggests that PuO₂ kinetics must be altered in cases of less cardiovascular efficiency, but its analysis requires visual inspection. Further studies interested in analyzing PuO₂ must include the additional analysis of its kinetics.

As expected, obesity influenced maximal V˙O2 in agreement with previous studies. The average energy expenditure during the CPET, represented by the MET, is higher in obese individuals due to a high energy expenditure to perform activities related to the excess of weight, which, according to Delany et al. [27], could mask the lower levels of physical activity performed by these subjects.

Regarding submaximal responses, AT was only significantly different in morbid obesity. Similarly, previous study found that AT was not different among obese and normal weight women [28]. However, we observed that AT in morbidly obese represents a high % V˙O2peak, instead of % V˙O2peak predicted, which may be explained by two main factors. First, the obese subjects are less physically active and conditioned, leading to the proximity of AT and peak V˙O2, due to an early CPET interruption. Secondarily, we used Hansen’s equation to estimate the V˙O2max [29], which can overestimate this value and, hence, underestimate the subjects by providing an inadequate ramp that also contributes to a possible early CPET interruption.

We showed that OUES significantly differed in morbidly obese, but it should be carefully analyzed. Previous findings observed that obese individuals consume more O₂ at the same effort intensity [30]. Although a promising dynamic physiological response [31, 32], OUES is dependent to exercise intensity since it can underestimate V˙O2max at low intensity, but overestimate at high intensity effort [30]. Additionally, OUES present an important interindividual variability [33] that can compromise its use in clinical practice. Thus, OUES interpretation may be difficult [13] despite its potential role as prognostic value for mortality regardless BMI.

In our study, we observed that both the Vt/IC ratio, which can be used to help determine the ventilatory reserve, and the Vt/FVC ratio did not present significant differences between non-obese and obese participants. Our results agree with previous studies on obese individuals with COPD, which observed that obesity did not worsen dyspnea or exercise performance in these individuals [34, 35]. A possible explanation found are related to alterations in the elastic properties of the lungs, increased intra-abdominal pressure, decreased pulmonary hyperinflation, and preserved inspiratory capacity during effort [34, 35].

Related to ΔV˙E/ΔV˙CO2, Neder et al. [36] observed a greater ventilatory efficiency in individuals in the early stages of COPD and flattening of this relationship with the progression of disease severity. Supposedly, these finds indicate better ventilatory efficiency in these patients, but physiologically, this efficiency should be worsened in these individuals due to the impairment of limited ventilatory mechanics combined with increased CO₂ pressure (PETCO₂) [36]. Similarly, obese individuals also have limited ventilatory mechanics due to central obesity, which may reduce the compliance of the lungs and, therefore, may lead to a restrictive ventilatory pattern during effort [37]. However, we did not observe an altered response in ΔV˙E/ΔV˙CO2, which was an unexpected finding possibly associated with the use of BMI classification since each class can range from 2 to 3 times in visceral adiposity [38].

Concomitant with this paradoxical result, there was an increase in the value of the y-intercept of ΔV˙E/ΔV˙CO2 related to the severity of obesity. It is possible that the y-intercept of the ΔV˙E/ΔV˙CO2 is more sensitive than the relationship slope to indicate the progression of ventilatory inefficiency in patients with COPD during effort [36]. Both the y-intercept and the ΔV˙E/ΔV˙CO2 slope remained unchanged in our study. Therefore, these indexes have a promising role in prognostic and diagnostic related to ventilatory inefficiency.

In addition, the lowest V˙E/V˙CO2 was also not different across non-obese and obese groups. Considering the great clinical interest for early identification of respiratory disorders and incipient respiratory diseases, any change found in V˙E/V˙CO2 suggests a ventilatory inefficiency during effort and should be further investigated.

The fat deposits in the mediastinum and the abdominal cavities due to obesity, as previously discussed, reduce the compliance of the lungs and, therefore, may lead to a restrictive ventilatory pattern during effort [37]. However, ventilatory pattern (ΔVt/ ΔlnV˙E) showed no significant influence of obesity. In this context, lower values of this ratio, indicative of tachypnea respiratory pattern, would most likely suggest an incipient ventilatory disorder regardless of obesity severity. Recently, Chao et al. [39] compared spirometry and CPET to detect incipient ventilatory disorders in workers exposed or not to occupational particulate matter. In addition to best sensibility of CPET, the ΔVt/ΔlnV˙E was one of the most useful variables being able to differentiate exposed from unexposed workers. Therefore, our results reinforce the clinical usefulness and prognostic value of this submaximal ratio.

As expected, the increment (treadmill speed and grade) decreased with obesity severity. The treadmill is an extremely popular ergometer in cardiology and sports medicine. A disadvantage of this ergometer is the workload, which cannot be accurately measured, since it depends on other factors (sustained body mass, anthropometry, and biomechanical aspects of gait) [40, 41]. In comparison to cycle ergometer, using a treadmill leads to greater cardiorespiratory stress which occurs due to the greater muscle mass involved in the exercise, venous return, HR, cardiac output and V˙O2 for the same workload [40, 41]. In addition, these physiological responses enable a best diagnosis of electrocardiographic changes, especially at maximal effort. Finally, using a treadmill provides greater representativeness of daily living activities (walking and running) [40–42].

The present study has some limitations that must be addressed. It is important to consider that the use of the treadmill implies a lack of objective workload. However, we registered the speed and grade during effort, which minimized this limitation. Our main limitation was the estimation of the peak of V˙O2 by using Hansen’s equation that can overestimate this value in obese subjects [29]. In order to deal with this possible limitation, we excluded submaximal effort tests in addition to rigorous analysis of R, HR, perceived subjective exertion and exercise duration. Despite the fact that a convenience sample was an important limitation, our large sample was appropriate, including a sufficient sample in each group of non-obese and obese participants. Additionally, we adjusted our analysis for the main confounders. Finally, although our findings are clinically relevant, especially to improve CPET interpretation, we point out that these aforementioned parameters present a wide variability. Thus, despite our results, their implementation in clinical practice routine can still be challenging and needs to be addressed in future research.

Our results have practical implications for public health. Obese individuals, most often, have increased cardiovascular risk, and CPET is the gold standard for assessing the CRF related to numerous comorbidities. However, the maximal variables must be influenced by motivation besides being related to several biomechanical aspects, especially relevant regarding obesity. In contrast, submaximal data are more responsive to treatment and exercise training in comparison to maximal variables and, hence, may be useful in early diagnosis of limitations in exercise tolerance and development of strategies to health promotion and obesity-related prevention. Therefore, the measurement and interpretation of submaximal and dynamic responses in CPET must be encouraged in clinical practice routine for both non-obese and obese patients. Based on our findings, future investigations should analyze the variables presented here to reinforce their clinical value in obese individuals in different settings and age range based on our findings. It is also of our interesting that prospective studies assess and confirm the prognostic value of the studied variables, mainly V˙E/V˙CO2, ΔV˙E/ΔV˙CO2 and the y-intercepts of ΔV˙E/ΔV˙CO2, in obese individuals, in addition to its use in clinical practice routine.

Obesity expressively influences maximal and submaximal CPET physiological responses, mainly cardiovascular and metabolic. Only a few variables presented inconsistent findings, which can reduce its applicability in the test interpretation regarding overweight and obese subjects. In contrast, the prognostic values of the main ventilatory efficiency variables, e.g., Vt max, Vt/V˙E, Vt/IC, Vt/FVC, lowest V˙E/V˙CO2, ΔV˙E/ΔV˙CO2, and the y-intercepts of V˙E/V˙CO2 remained unchanged and, therefore, has the best prognostic value regardless of obesity. These dynamic responses are independent of maximum effort and may be useful in detecting incipient ventilatory disorder. Our results present great practical applicability in identifying exercise limitation, regardless of BMI.