We searched PubMed, Web Of Science and Embase on March 27, 2021 using relevant keywords and a Boolean search string (see Supplementary Materials S2 for the detailed search strategy): (((((“Exercise test”)) OR (“exercise test” OR “cardiopulmonary exercise test” OR “cycle exercise” OR “ergometer” OR “CPET” OR “CPX”)) AND (“position” OR “posture” OR “supine” OR “recumbent” OR “recline” OR “lean” OR “tilt” OR “clinostatism” OR “decubitus” OR “lie”)) AND (“erect” OR “upright” OR “orthostatic” OR “sit”)) AND (“VO₂” OR “oxygen uptake”). The search string consisted of MeSH and general search terms. Searches were restricted to English.

We included self-control trials, and selection criteria conformed to the PICOS approach, as described hereinafter.

After eliminating duplicate articles, XW and CL screened the titles and abstracts. Studies that did not mention VO_2peak or a synonym in the study title and/or abstracts, but were likely to have included them as a secondary measure, were also included. In the second step, XW and CL read the full texts of articles considered relevant based on title and abstract. There was full agreement on the inclusion of the full-text articles.

XW and CL assessed the quality of the included studies with a modified version of the Downs and Black checklist (see Supplementary Materials S3) (11). This checklist has been employed in several reviews in the field of sports science, which also uses cross-sectional studies for data retrieval. In our modified version, we considered four domains, including seven items, to evaluate the included studies' quality: (1) are the interventions of interest clarified? (2) are the test positions clarified? (3) is the time period the participants have between tests similar? (4) was the test order randomized? (5) how was VO_2peak defined? (6) was the VO_2peak test valid? and (7) are the characteristics of the participants included in the study clarified? All items were rated as “Yes,” “No,” or “Not sure.”

We extracted data on VO_2peak in the respective positions and the characteristics of the participants (number of participants, sex, age, body mass, types of disease) as well as the starting workload, rotation rate, duration and workload increase of the increments used during the test protocols.

We performed statistical analysis with RevMan version 5.3. The data are segmented or combined based on the angle between the upper body and the horizontal plane, or whether there is cardiopulmonary disease, and the calculation process is based on the formula in Cochrane 7.7. Continuous variables are expressed by means and standard deviation. We assessed heterogeneity across included studies using a Cochran chi-square test (with 0.1 as the cutoff for statistical significance) and an I² statistic test. We employed a random effects model where there was evidence of statistical heterogeneity (I² statistic > 50%). Otherwise, we used a fixed effects model. If there was heterogeneity, we first analyzed the source of heterogeneity, and then used subgroup analysis or other methods to deal with it. After excluding obvious clinical heterogeneity, we conducted the meta-analysis with a random response model. The meta-analysis test level was p = 0.05.

The detailed study search and selection process is outlined in Figure 1. In total, we retrieved 1,179 records from the database searches. After excluding duplicates, we screened 1,011 potentially relevant abstracts, and excluded 891 for failing to meet the inclusion criteria. We read the remaining 120 full texts, and deemed 32 self-control trials eligible for systematic review (9, 10, 12–41).

The quality of the included studies is shown in Table 1. Only nine studies (28%) clarified the participants' inclusion and exclusion criteria. Most of the studies (n = 21, 66%) clarified the exercise protocol, but only about half (n = 14, 44%) clarified the body position during exercise. Twenty studies (62.5%) mentioned the use of randomization methods to determine the sequence of positions used in the CPET. In addition, the interval between the two tests was similar in 18 (56%) of the studies. Although the VO_2peak was clarified in 14(44%) studies, VO_2peak was judged to be valid in 26 (81%) studies where the criteria for verification of maximal effort were stated explicitly.

Details of the participants' characteristics are listed in Table 2. 546 participants were included in the meta-analysis, of which about 62% were male. 21 studies included only healthy people, 9 studies included patients with cardiopulmonary disease, and 2 studies included both the healthy and the patients with cardiopulmonary disease. The ages of the included population ranged from 9 to 72 years old. There were four articles on children (13, 23, 25, 34), one on both adults and children (41), and the remaining 27 articles were on adults.

Details on the exercise protocols used in the included studies are summarized in Table 3. All of the included studies used the continuous incremental exercise program, except for Quinn's study (12). The angle between the upper body and the horizontal in the supine position ranged from −6 to 65°. In some studies, there was only one position and one test for supine exercise, while in other studies, supine exercise contained multiple positions and multiple tests. For example, Egaña's study showed that the angles between the upper body and the horizontal plane in the supine position included zero, 15 and 30° (21). The interval between the two tests was within 1 month in all studies, except for May's and Mizumi's (13, 14).

In terms of the relative VO_2peak, our pooled results showed that the upright VO_2peak was higher than the supine VO_2peak (relative VO_2peak: MD = 2.63 ml/kg/min, 95% CI: 1.66, 3.59, I² = 56%, p < 0.05; Figure 2). In the healthy subgroup, the upright VO_2peak remained higher than supine and the effect size was larger without heterogeneity (relative VO_2peak: MD = 4.04 ml/kg/min, 95% CI: 3.25, 4.83, I² = 0%; Figure 2). Similarly, in patients with cardiopulmonary disease, upright VO_2peak was also higher than supine, however, the effect size was lower compared to the healthy subgroup (relative VO_2peak: MD = 1.03 ml/kg/min, 95% CI: 0.29, 1.76, I² = 47%; Figure 2).

In terms of the absolute VO_2peak, our pooled results also showed that the upright VO_2peak was higher than the supine VO_2peak (absolute VO_2peak: MD = 0.18 L/min, 95% CI: 0.10, 0.26, I² = 63%, p < 0.05; Figure 3). Again, the effect size was higher in the healthy subgroup (absolute VO_2peak: MD = 0.22 L/min, 95% CI: 0.12, 0.32, I² = 69%; Figure 3), but lower in the cardiopulmonary disease subgroup (absolute VO_2peak: MD = 0.12 ml/kg/min, 95% CI: 0.02, 0.21 I² = 31%; Figure 3).

The purpose of this paper is to evaluate the effect of body position on VO_2peak in the CPET via meta-analysis. The results indicate that the VO_2peak measured by CPET in the upright position was higher than that of the supine position during an incremental cycling exercise test. Moreover, this difference exists in both healthy people and patients with cardiopulmonary disease.

Physiologic shifts in hemodynamic distribution occur with along with changes from supine to upright postures. This hemodynamic shift alters the preload to the right ventricle which subsequently affects preload to the left ventricle and the Frank-Starling mechanism associated with ventricular contraction and stroke volume. Thadani and Parker. have reported hemodynamic differences between the supine and upright positions in normal subjects. They found higher heart rates with lower left ventricular filling pressures and stroke volume index in the upright position, both at rest and during exercise. However, they noted no differences in cardiac index or peak work load between the two modes of exercise (42). Kramer's study of heart failure patients also showed significantly higher right atrial pressure in the supine position, but no differences in cardiac index or stroke index between the two positions. Further, these authors found significantly lower VO_2peak in the supine position (9). Moreover, in chronic thromboembolic pulmonary hypertension patients, Mizumi's study showed that CO at rest in the supine position was significantly higher than that in the upright. However, CO at peak exercise was comparable between the two positions, even though VO_2peak in the supine position tends to be lower than that of the upright (14). These studies suggest that increased ventricular preload caused by postural changes does not necessarily lead to an increase in stroke volume or CO as a key mechanism to facilitate increased oxygen uptake in the upright posture. In contrast, these data suggest that VO_2peak in the supine position may be limited by factors other than the effect of central hemodynamic shifts on CO.

According to the Fick principle, VO₂ = CO ^* C(a-v) O₂ (arteriovenous oxygen difference) (43). Therefore, the VO₂ is influenced by both central and peripheral components, including CO and C(a-v) O₂, during exercise. Studies have shown that VO_2peak in the upright position is higher than that in the supine position, and accompanied by higher C(a-v) O₂, while there is no difference in cardiac index between the two positions (9, 44). The change in C(a-v) O₂ is influenced by both muscle's ability to consume oxygen and the amount of muscle mass involved in the activity. The oxygen uptake in the muscles is determined by factors such as the amount of muscle work, perfusion pressure, blood flow and mitochondrial density and activity. With regards to muscle mass, the upright position requires greater muscle mass involvement, and thus more physiologic work. Research by Bouillon et al. has demonstrated that during upright cycling, the maximum voluntary isometric contraction of the gluteus maximus, gluteus medius, biceps femoris, lateral head of gastrocnemius, anterior tibialis, rectus femoris, lumbar erector spinalis and rectus abdominis exceed those measured in the semi-reclining position (45). Furthermore, when standing upright, the distribution of blood in the lower extremity vein increases due to increased orthostatic tension. Additionally, the arteries' diameter is wider than in the supine position, and the blood flow velocity is slower. This is conducive to the diffusion of more oxygen from hemoglobin to myoglobin (46). Furthermore, Eiken found that exposing working legs to sub-atmospheric pressure can increase perfusion pressure. This can improve motor ability in the supine position, and simulate upright exercise in normal gravity (47, 48). Further, several studies have used near-infrared spectroscopy illuminate the oxyhemoglobin relationship in the periphery during exercise. These studies suggest that muscle oxygen absorption capacity is higher in the upright position (49–51). Therefore, the C(a-v) O₂ during peak exercise is higher in the upright position, and this may contribute to the increased VO_2peak noted in the upright position.

In addition, studies have shown that the increased venous return in the supine position is conducive to increasing the fluid shift to the pulmonary circulation and interstitium. This may compress small airways and/or blood vessels, unbalancing the overall pulmonary V/Q ratio and reducing ventilatory efficiency (52). Bryan et al. (5) demonstrated that the change from an upright position to the supine position in healthy individuals results in a decrease in the V/Q of the entire lung from 0.83 to 0.76. Further, Sandoval et al. have shown that even in the resting state, the change in body position causes significant changes in alveolar-arterial oxygen partial pressure difference in patients with Eisenmenger syndrome. This decreases oxygen saturation in the supine position (53). In contrast, other studies have shown that the change from an upright position to a supine does not cause significant reduction in ventilatory efficiency in healthy people, or even in patients with stable heart failure (16, 18, 24, 37). As such, the extent to which these components affect peak oxygen uptake in the supine position, and contribute to the lower values when compared to those measured in the upright position, remains unclear.

It is important to acknowledge limitations associated with this study. First, the sample sizes of most studies in this meta-analysis were very small. Second, since this research is a self-control study, it is difficult to blind participants to the exercise condition, and therefore difficult to prevent unintentional bias. Third, the difference in VO_2peak between the upright position and the supine position may vary among patient groups, and the data for patients with cardiopulmonary disease was insufficient. Therefore, we could not conduct subgroup analysis according to specific disease types. In the future, with the increase in research data from patients with cardiopulmonary disease, subgroup analysis could be conducted according to specific disease types. Lastly, this analysis was unable to account for different supine position body angles. The body position of research participants should be clarified in future clinical studies to facilitate subgroup analysis.