End-tidal carbon dioxide tension reflects arterial carbon dioxide tension in the heat-stressed human with and without simulated hemorrhage

Abstract

End-tidal carbon dioxide tension (Pet_CO2) is reduced during an orthostatic challenge, during heat stress, and during a combination of these two conditions. The importance of these changes is dependent on Pet_CO2 being an accurate surrogate for arterial carbon dioxide tension (Pa_CO2), the latter being the physiologically relevant variable. This study tested the hypothesis that Pet_CO2 provides an accurate assessment of Pa_CO2 during the aforementioned conditions. Comparisons between these measures were made: 1) after two levels of heat stress (N = 11); 2) during combined heat stress and simulated hemorrhage [via lower-body negative pressure (LBNP), N = 8]; and 3) during an end-tidal clamping protocol to attenuate heat stress-induced reductions in Pet_CO2 (N = 7). Pet_CO2 and Pa_CO2 decreased during heat stress (P < 0.001); however, there was no group difference between Pa_CO2 and Pet_CO2 (P = 0.36) nor was there a significant interaction between thermal condition and measurement technique (P = 0.06). To verify that this nonsignificant trend for the interaction was not due to a type II error, Pet_CO2 and Pa_CO2 at three distinct thermal conditions were also compared using paired t-tests, revealing no difference between Pa_CO2 and Pet_CO2 while normothermic (P = 0.14) and following a 1.0 ± 0.2°C (P = 0.21) and 1.4 ± 0.2°C (P = 0.28) increase in internal temperature. During LBNP while heat stressed, measures of Pet_CO2 and Pa_CO2 were similar (P = 0.61). Likewise, during the end-tidal carbon dioxide clamping protocol, the increases in Pet_CO2 (7.5 ± 2.8 mmHg) and Pa_CO2 (6.6 ± 3.4 mmHg) were similar (P = 0.31). These data indicate that mean Pet_CO2 reflects mean Pa_CO2 during the evaluated conditions.

arterial carbon dioxide tension (Pa_CO2) is a primary regulator of cerebral perfusion, with hypercapnia and hypocapnia inducing increases and decreases in cerebral perfusion, respectively (7, 23). Studies using end-tidal carbon dioxide tension (Pet_CO2) to estimate Pa_CO2 during passive heat stress (i.e., increase in internal temperature ∼1.2–1.5°C) suggest that heat stress reduces Pa_CO2 ∼4–8 mmHg (3, 5, 15, 26). A potential limitation to these studies is the assumption that Pet_CO2 is an accurate surrogate for Pa_CO2, the latter being the physiologically relevant variable. In normothermic individuals, Pet_CO2 provides an accurate assessment of Pa_CO2 in supine, resting conditions (8, 9, 22). Upon assumption of the upright posture, however, there is a decrease in cardiac output that is accompanied by a ventilation/perfusion mismatch leading to much greater reductions in Pet_CO2 relative to Pa_CO2 (8, 9, 22). Thus, in this condition, there is a discrepancy between the values obtained from two measures of carbon dioxide tension such that reductions in Pet_CO2 overestimate reductions in Pa_CO2 (8, 9, 22). This issue has recently been minimized in normothermic individuals through the identification of a correction factor that can be used to predict Pa_CO2 from measures of Pet_CO2 (8). To our knowledge, however, there is no information regarding the relationship of Pet_CO2 to Pa_CO2 in individuals with elevated internal temperatures. This information is important because heat stress-induced reductions in Pa_CO2 likely contribute to reductions in cerebral blood flow (3, 5, 15, 26) and ultimately reduced orthostatic tolerance in this thermal condition (4, 12, 25). Because Pet_CO2 is commonly used as an index of Pa_CO2 in heat-stressed individuals (3, 5, 15, 26), it is important to identify the relationship between these variables in this thermal condition.

The purpose of this study was to test the hypothesis that Pet_CO2 provides an accurate assessment of Pa_CO2 during a moderate heat stress and during a more severe heat stress sufficient to decrease Pet_CO2 (aim 1) as well as during a heat stress combined with an orthostatic challenge (aim 2). Lastly, restoration of Pet_CO2 to preperturbation baseline is often used to assess the role of carbon dioxide on various physiological responses (3, 9). This procedure relies on the assumption that clamping-induced increases in Pet_CO2 are accompanied by comparable increases in Pa_CO2, but this assumption has not been assessed in heat-stressed individuals. Therefore, aim 3 tested the hypothesis that Pet_CO2 accurately tracks Pa_CO2 when the heat stress-induced reductions in Pet_CO2 are attenuated through Pet_CO2 clamping.

METHODS

Eleven healthy normotensive subjects participated in this study. Average (mean ± SD) subject characteristics were as follows: age, 34 ± 12 yr; height, 172 ± 7 cm; and weight, 70 ± 7 kg. Subjects were not taking medications and were free of any known cardiovascular, metabolic, or neurological diseases. Subjects were informed of the purpose and risks of the study before providing their informed written consent. The protocol and consent were approved by the Institutional Review Boards at the University of Texas Southwestern Medical Center at Dallas and Texas Health Presbyterian Hospital Dallas. Subjects refrained from alcohol, caffeine, and intense exercise for 24 h before the study.

Instrumentation and Measurements

Following arrival to the laboratory, each subject swallowed a telemetry pill for the measurement of intestinal temperature (HQ, Palmetto, FL). Mean skin temperature was measured from the weighted average of six thermocouples attached to the skin (24). Each subject was fitted with a water-perfused tube-lined suit (Med-Eng, Ottawa, Canada) and was placed in a lower body negative pressure (LBNP) chamber, sealed at the iliac crest, while in the supine position. The suit covered the entire body except for the head, face, hands, one forearm, and feet and permitted the control of skin and internal temperatures by adjusting the temperature of the water perfusing the suit. Heart rate was continuously obtained from an electrocardiogram (HP Patient Monitor; Agilent, Santa Clara, CA) interfaced with a cardiotachometer (CWE, Ardmore, PA). A 20-gauge catheter was inserted in the radial artery of the nondominant arm using sterile techniques under local anesthesia. The cannula was connected to a pressure transducer (Maxxim Medical, Athens, TX) that was positioned at the level of the heart. This catheter was used for direct continuous assessment of beat-by-beat arterial pressure as well as for blood samples for subsequent analysis of Pa_CO2.

Experimental Protocol

An outline of the entire experimental protocol is provided in Fig. 1. For all analyses, the alpha level was set at 0.05, and the results are reported as means ± SD.

Fig. 1.

Protocol schematic. The arrows indicate the timing of the blood draw for analysis of arterial carbon dioxide tension for each of the indicated aims. End-tidal carbon dioxide tension (Pet_CO2) was averaged over the 30-s period when arterial blood was drawn. T_c, core temperature.

Aim 1: Pet_CO2 vs. Pa_CO2 During Normothermia and Two Levels of Heat Stress

Following instrumentation, subjects rested quietly in the supine position while normothermic water (34°C) circulated through the suit. After a 10-min steady-state rest period, subjects were fitted with a nose clip and breathed room air through a mouth piece for 5 min while thermal, hemodynamic, and Pet_CO2 (VitalCap Capnograph Monitor; Oridion, Needham, MA) data were collected during spontaneous respiration. During the last minute of this period, blood was obtained from the arterial catheter and stored in a heparinized syringe on ice until subsequent analysis of Pa_CO2 was performed in triplicate (Gem Premier 3000; Instrumentation Laboratory, Lexington, MA). Following completion of normothermic data collection, heat stress began by circulating 49°C water through the suit. When internal temperature increased ∼1.0°C above baseline temperature, the Pet_CO2 and Pa_CO2 measures were repeated (heat stress 1). Subjects continued to be heat stressed until Pet_CO2 was reduced by at least 3 mmHg relative to normothermia. Once this decrease in Pet_CO2 was attained, the temperature of the water circulating the suit was slightly decreased in an effort to attenuate the rate of rise in internal temperature, as well as further decreases in Pet_CO2, at which point the Pet_CO2 and Pa_CO2 measures were repeated (heat stress 2).

Data analysis for aim 1.

Data were sampled at 50 Hz via a data-acquisition system (Biopac System, Santa Barbara, CA). For normothermia and both levels of heating, data during the 60-s period before the blood draw, for each respective thermal condition, were averaged, the exception being that Pet_CO2 was averaged over the 30-s period when arterial blood was drawn. Hemodynamic data among the three conditions (i.e., baseline normothermia, heat stress 1, and heat stress 2) were analyzed via one-way repeated-measures ANOVA, followed by a Tukey post hoc analysis when a main effect was identified. Values of Pet_CO2 and Pa_CO2 were evaluated via a two-way repeated-measures ANOVA, with main factors of carbon dioxide measurement method and thermal condition. Comparisons between Pa_CO2 and Pet_CO2 measures, during the three thermal conditions, were also evaluated using the methods of differences as described by Bland-Altman (2). The Bland-Altman plot was subsequently analyzed by linear regression analysis to determine if the slope of the relationship was greater than zero, which would indicate that a proportional bias was present [i.e., difference between two methods changes as the average values from the two methods becomes smaller or larger (6, 14, 16)]. Furthermore, a fixed bias is present if the 95% confidence limit of the Bland-Altman plot does not include zero (14, 16). Lastly, a linear regression analysis was performed to further characterize the relationship between Pa_CO2 and Pet_CO2 measures during the three thermal conditions.

Aim 2: Pet_CO2 vs. Pa_CO2 During a Simulated Hemorrhage Challenge While Heat Stressed

Immediately after the Pet_CO2 and Pa_CO2 measures during moderate heating (i.e., heat stress 1), eight subjects were exposed to 4 min of ∼30 mmHg LBNP. Pet_CO2 and Pa_CO2 were measured during the final minute of LBNP.

Data analysis for aim 2.

Absolute values of Pet_CO2 and Pa_CO2 during the LBNP challenge were compared using a paired t-test as well as by linear regression analysis (N = 8). In a subset of subjects (N = 5), this level of LBNP was sufficient to further reduce Pet_CO2 at least 2 mmHg below the heat stress value (i.e., pre-LBNP). In these individuals, the magnitude of the reduction in Pet_CO2 and Pa_CO2 during LBNP relative to the pre-LBNP values was compared using a paired t-test and linear regression analysis.

Aim 3: Does Pet_CO2 Track Pa_CO2 When the Heat Stress-Induced Reductions in Pet_CO2 are Attenuated?

Immediately after the Pet_CO2 and Pa_CO2 measures during pronounced heating (i.e., heat stress 2), a subset of subjects (N = 7) was exposed to a Pet_CO2 clamping protocol. This was accomplished using a computer-controlled gas blender, sequential gas delivery, and rebreathing circuit (RespirAct; Thornhill Research, Toronto, Canada), which has been described in detail elsewhere (3, 10, 17–19). The Respiract device was programmed to return Pet_CO2 to tensions measured during normothermia while simultaneously maintaining normoxia via administration of a mixture of nitrogen, oxygen, and carbon dioxide gases in a closed-loop sequential rebreathing circuit. Once this was achieved, the aforementioned Pet_CO2 and Pa_CO2 measures were repeated.

Data analysis for aim 3.

The magnitude of the increase in Pet_CO2 and Pa_CO2 during the clamping protocol relative to preclamped heat stress values was compared using a paired t-test and linear regression analysis.

RESULTS

Thermal and hemodynamic data. Before any thermal perturbations, internal and mean skin temperatures were 37.0 ± 0.4 and 34.5 ± 0.5°C, respectively. Heat stress 1 and heat stress 2 increased mean skin temperature to 38.5 ± 0.6 and 38.7 ± 0.9°C, respectively (both variables P < 0.001 relative to normothermia), and the magnitude of this increase was similar between the two heat stress conditions (P = 0.66). Internal temperature was increased to 38.0 ± 0.5°C during heat stress 1 (P < 0.001 relative to normothermia) and was further elevated to 38.4 ± 0.4°C during heat stress 2 (P < 0.001 relative to normothermia and heat stress 1). Heart rate was increased from 63 ± 8 beats/min during normothermia to 102 ± 18 and 106 ± 14 beats/min during heat stress 1 and heat stress 2, respectively (both variables P < 0.001 relative to normothermia). Mean arterial pressure was reduced from a normothermic value of 89 ± 7 mmHg to 77 ± 7 and 76 ± 6 mmHg during heat stress 1 and heat stress 2, respectively (both variables P < 0.001 relative to normothermia).

Aim 1: Pet_CO2 Provides an Accurate Assessment of Pa_CO2 During Normothermia and Two Levels of Heat Stress

The two-way repeated-measures ANOVA revealed a significant main effect of thermal condition in reducing carbon dioxide tension (P < 0.001); however, this reduction was similar between measurement techniques [main effect of carbon dioxide measurement technique (Pet_CO2 and Pa_CO2; P = 0.36)]. While not significant, there was a trend toward an interaction between thermal condition and measurement technique (P = 0.06). To verify that this nonsignificant trend was not due to a type II error, Pet_CO2 and Pa_CO2 values at each of the three thermal conditions were also compared using paired t-tests. The results were consistent with the two-way repeated-measures ANOVA in that there was no difference between Pet_CO2 and Pa_CO2 while normothermic (P = 0.14), during heat stress 1 (P = 0.21), and during heat stress 2 (P = 0.28; Fig. 2A). Linear regression analysis demonstrated a significant correlation between measures of Pet_CO2 and Pa_CO2 when the three thermal conditions were analyzed together (r = 0.94, P < 0.001; Fig. 2B). The Bland-Altman plot revealed a small bias (bias, −0.4 mmHg) between measures of Pet_CO2 and Pa_CO2 when the three thermal conditions were analyzed together (Fig. 2B). Neither a proportional (r = 0.01, P = 0.94, slope −0.004) nor a fixed (95% confidence interval = −4.2 mmHg, 3.4 mmHg) bias was detected from the Bland-Altman plot (Fig. 2C).

Fig. 2.

Relationship between Pet_CO2 and arterial carbon dioxide tension (Pa_CO2) during normothermia, as well as elevations in core body temperature of 1.0 ± 0.2°C (heat stress 1) and 1.4 ± 0.2°C (heat stress 2). Measures of Pet_CO2 (white bars) and Pa_CO2 (gray bars) were similar for each thermal condition (P values are from paired t-tests; A). The linear regression analysis (B) and the Bland-Altman plot (C) also revealed a close relationship between Pa_CO2 and Pet_CO2 within the three thermal conditions, which was particularly evident in the hypocapnic range (B and C). For the Bland-Altman plot, there was a small bias between measurement devices (bias, −0.4 mmHg) and 95% confidence limits of −4.2 and 3.4 (B).

Aim 2: Pet_CO2 Provides an Accurate Assessment of Pa_CO2 During a Simulated Hemorrhage Challenge While Heat Stressed

A paired t-test revealed no difference between absolute values of Pet_CO2 (31.4 ± 7.2 mmHg) and Pa_CO2 (31.8 ± 7.0 mmHg) during LBNP while subjects were heat stressed (P = 0.61; Fig. 3A). This finding was supported by linear regression analysis which revealed a significant correlation between absolute values of Pet_CO2 and Pa_CO2 during LBNP (r = 0.96, P < 0.001; Fig. 3B). Likewise, a paired t-test revealed that the magnitude of the reduction in Pet_CO2 (5.3 ± 2.6 mmHg) and Pa_CO2 (5.4 ± 4.2 mmHg) in the subset of subjects (N = 5) who exhibited decreases in Pet_CO2 during LBNP was also similar (P = 0.89; Fig. 3C), which was further confirmed by linear regression analysis (r = 0.93, P = 0.02; Fig. 3D).

Fig. 3.

Relationship between Pet_CO2 and Pa_CO2 during combined lower body negative pressure (LBNP) and heat stress 1. Absolute Pet_CO2 and Pa_CO2 during LBNP while subjects were heat stressed were similar, as identified by a paired t-test (N = 8; A) and linear regression analysis (N = 8; B). Likewise, the magnitude of the reduction in Pet_CO2 (5.3 ± 2.6 mmHg) and Pa_CO2 (5.4 ± 4.2 mmHg) tensions to LBNP was similar, as identified by a paired t-test (N = 5; C) and linear regression analysis (N = 5; D), indicating that Pet_CO2 tension accurately reflects Pa_CO2 tension during these perturbations.

Aim 3: The Magnitude of the Increase in Pet_CO2 and Pa_CO2 During a Pet_CO2 Clamping Procedure is Similar in Heat-Stressed Subjects

The Pet_CO2 clamping procedure was successful at returning Pet_CO2, which decreased during heat stress 2, to normothermic levels (normothermia: 38 ± 4 mmHg; heat stress preclamp: 31 ± 5 mmHg; heat stress clamp: 38 ± 5 mmHg; P < 0.001 between heat stress values, whereas there was no difference between normothermia and heat stress + clamp values; P = 0.97). Importantly, the magnitude of the increase in Pa_CO2 that occurred during this clamping procedure was similar to the increase in Pet_CO2 (P = 0.31; Fig. 4A). This finding was confirmed by linear regression analysis (r = 0.86, P = 0.01; Fig. 4B).

Fig. 4.

Relationship between Pet_CO2 and Pa_CO2 tensions during the clamping procedure directed toward restoring Pet_CO2 tension to preheat stress levels. The clamping procedure similarly increased Pet_CO2 (7.5 ± 2.8 mmHg) and Pa_CO2 (6.6 ± 3.4 mmHg) tensions relative to their respective preclamp heat stress values, indicated numerically in each column (i.e., heat stress 2; A). This finding is further confirmed by linear regression analysis, which identified a significant correlation between Pet_CO2 and Pa_CO2 (B). These finding indicate that Pet_CO2 tension accurately reflects Pa_CO2 tension during this clamping procedure.

DISCUSSION

Pet_CO2 is reduced in individuals with an elevated internal temperature, and the reduction in Pet_CO2 is exacerbated when this thermal stress is combined with an orthostatic challenge (3, 15, 26). These physiological occurrences likely contribute to decreases in cerebral perfusion and thus the reduction in orthostatic tolerance that occur during heat stress (1, 4, 11–13, 25). The precise understanding of the relationship between carbon dioxide tension and the cerebral vasculature is dependent on the assumption that Pet_CO2 accurately reflects Pa_CO2, the latter of which is the physiologically relevant variable. The current results indicate that Pet_CO2 provides an accurate noninvasive index of Pa_CO2 during normothermic conditions, differing degrees of heat stress, and during a simulated hemorrhage challenge combined with heat stress. Furthermore, the magnitude of the increase in Pa_CO2 upon returning Pet_CO2 to normothermic levels (i.e., the Pet_CO2 clamping protocol) is similar to that of Pet_CO2. These data suggest that Pet_CO2 provides an accurate assessment of Pa_CO2 under the testing conditions without the added cost and risk of arterial catheterization.

In normothermic supine resting individuals, Pet_CO2 provides an accurate reflection of Pa_CO2 (8, 9, 22), which is consistent with the present data (Fig. 2, A, B, and C). That said, a discrepancy in the relationship between these variables has been reported following the assumption of the upright posture in normothermic individuals, such that reductions in Pet_CO2 are greater than Pa_CO2 (8, 9). This response is due to changes in distribution of blood flow throughout the lungs secondary to postural reductions in cardiac output, which subsequently alters the ventilation-perfusion ratio (8). In the current study, subjects were exposed to a simulated orthostatic challenge during heat stress conditions by LBNP. Unlike what is observed during normothermia (8, 9), the present results indicate that Pet_CO2 is similar to Pa_CO2 during the LBNP challenge (Fig. 3, A and B), and, furthermore, the magnitude of the reductions in Pet_CO2 and Pa_CO2 during this challenge are similar (Fig. 3, C and D). While speculative, it is possible that differing normothermic and heat stress responses to an orthostatic stress are related to the differences in cardiac output between thermal conditions. Cardiac output was not measured in the current study; however, values of 10–13 l/min have been reported during similar degrees of heat stress (20, 21, 26). The imposed LBNP during heat stress reduces cardiac output by ∼3 l/min (26), resulting in final cardiac outputs of ∼7–10 l/min. In contrast, during an orthostatic challenge in normothermic conditions, cardiac output is reduced by ∼1.5 l/min (26), resulting in a final cardiac output of ∼4 l/min. Thus it is possible that, despite greater reductions in cardiac output during LBNP while heat stressed, the prevailing cardiac output (i.e., ∼7–10 l/min) and thus blood flow distribution throughout the lungs remains sufficient to maintain similarities between measures of Pet_CO2 and Pa_CO2, relative to when cardiac output has been reduced to ∼4 l/min during LBNP while normothermic.

Despite differences in Pet_CO2 and Pa_CO2 during normothermic head up tilt (8, 9), Immink et al. (9) reported that the magnitude of increase in both of these variables was similar when Pet_CO2 was clamped at the pretilt level. Additionally, using the same clamping device as in the current study, Ito et al. (10) recently reported that the accuracy of Pet_CO2 as an estimate of Pa_CO2 was sustained in seated normothermic individuals across a wide range of hypocapnic and hypercapnic stimuli. The present findings are in agreement with the cited findings given that the magnitude of the increase in Pet_CO2 and Pa_CO2 was similar during the clamping protocol (Fig. 4, A and B).

Considerations/Limitations

While not significant, there was a trend toward an interaction between the thermal condition and measurement technique in aim 1 (P = 0.06). This is most likely the result of a slightly greater decrease in Pa_CO2 (8.8 ± 4.6 mmHg) during heat stress 2 relative to the decrease in Pet_CO2 (7.2 ± 4.2 mmHg). Nonetheless, when paired comparisons were made at each thermal condition, the measures of CO₂ were similar regardless of the measurement technique (Fig. 2).

The Bland-Altman plot in Fig. 2C depicts the relationship of the difference between the two measurement techniques (y-axis) and the average between the two techniques (x-axis) during resting normothermia and two levels of heat stress. This plot further reveals a close relationship between measures of Pet_CO2 and Pa_CO2, as demonstrated by a small measurement bias of −0.4 mmHg. A proportional bias exists if the slope of the relationship between the difference of two measurements and the mean of the two measurements is significantly different from zero [i.e., the difference between the two methods changes as the average values from the two methods becomes smaller or larger (6, 14, 16)]. In the current study, a proportional bias was not detected (r = 0.01, P = 0.94, slope −0.004). That being said, it appears that the relationship between the Pet_CO2 and Pa_CO2 weakens at higher CO₂ tensions as indicated by a wider dispersion of data points with respect to the measurement bias (in both the positive and negative direction). Therefore, it is possible that greater differences would be identified if comparisons between carbon dioxide measurement methods were evaluated in the hypercapnic range.

Perspectives and Significance

In conclusion, absolute Pet_CO2 and Pa_CO2 were similar during normothermic conditions, differing degrees of heat stress, as well as during LBNP combined with heat stress. Furthermore, the heat stress-induced reductions in Pet_CO2 and Pa_CO2 were of similar magnitude, and subsequent further decreases in Pet_CO2 and Pa_CO2 imposed by LBNP were also of a similar magnitude. Based on these findings, Pet_CO2 measures can be used to estimate Pa_CO2 during heat stress studies that use the imposed perturbations. These findings are important to the clinical community who are interested in the impact of changes in carbon dioxide concentration on physiological responses, and to researchers who do not have access to skilled personnel necessary for arterial catheter placement and to the subjects by reducing the risks associated with experimentation where estimates of Pa_CO2 are needed.

GRANTS

This research was supported by National Heart, Lung, and Blood Institute Grants HL-61388, HL-84072, and HL-092761 and by the Research and Education Institute of Texas Health Resources.

DISCLOSURES

No conflicts of interest are declared by the authors.