Introduction

Hot weather and heat waves are associated with increased fatalities from adverse cardiovascular events1,2,3. The putative pathophysiology has been summarized in numerous narrative reviews4,5,6,7,8,9. Briefly, rising core and skin temperatures elicit reflex and locally mediated vasodilation of the cutaneous vasculature to facilitate transfer of blood-borne heat to the skin where it can be dissipated to the environment10. The resultant fall in vascular resistance necessitates an elevation in cardiac output to maintain arterial pressures and systemic blood flow11. These adjustments increase myocardial strain12,13,14,15 which, when combined with heat-induced dehydration and hypercoagulability16,17, could elevate the risk of acute coronary events, particularly in persons already vulnerable to such events (e.g., older adults)4,5,6,7,8,9.

Laboratory-based heat exposure studies have played a central role in delineating the cardiovascular adjustments to acute heat stress11. The limits of those adjustments were perhaps first shown in a seminal study by Rowell et al.18, in which cardiac output rose 5–7 L/min due to a 60–80 beats/min increase in heart rate in four young men heated to the limits of thermal tolerance using an encapsulated, tube-lined garment perfused with ~48–50 °C water. Water-perfused suits have since become a mainstay of laboratory research, primarily because they allow for rapid and precise control of body temperature11, which facilitates comparisons between different populations (e.g., young and older adults)15,19,20 and physiological states (e.g., central hypovolemia and normovolemia)12,13.

The extreme cardiac responses seen in studies employing water-perfused suits18,19 are often cited in translational reviews describing the impacts of hot weather and heat waves on the heart5,6,7,8,9. However, the encapsulated and conduction-mediated design of this heating modality impedes heat dissipation, meaning that extreme hyperthermia can develop rapidly and skin temperatures often reach supra-physiological levels (≥40 °C)18,19. Consequently, it has been suggested that perfusion suits (and other encapsulated modalities) are suboptimal for translational research aimed at quantifying the cardiac strain experienced during hot weather and heat waves6,11. To address this limitation, a growing number of studies have exposed participants to hot weather simulated in specialized climate-controlled chambers21,22,23,24,25,26,27,28, which are thought to provide a more ecologically relevant exposure model6,11. However, such studies are highly resource-intensive, which makes quantifying cardiac responses to a wide range of ambient conditions challenging.

With rising global temperatures, physiological research will play an increasingly important role in the development of robust climate-health models29,30,31 and evidence-based guidance for protecting heat-vulnerable persons5,7,32,33,34. Given the pervasive use of water-perfused suits, we were concerned that the substantial body of evidence on the physiological burden of heat has, in its current form, limited use in supporting these efforts. We therefore conducted this systematic review and meta-analysis to synthesize data from laboratory-based studies evaluating the cardiac responses to acute heat stress. Our primary objective was to compare the cardiac adjustments reported in studies employing water-perfused suits and other encapsulated and/or extreme modalities (hot water immersion and sauna) with those experienced during resting exposure to high ambient temperatures in climate-controlled chambers (supplemental Table S2 overviews common modalities). The time- and resource-intensive nature of thermophysiological research and the required controls (e.g., of skin temperature and hemodynamic responses) rendered approaching this problem experimentally intractable. Conversely, a large-scale meta-analysis permitted evaluations of between-modality differences, while adjusting for relevant confounders and non-heat-related mediators of the cardiac response to different heating modalities. Leveraging the large and unique dataset also allowed us to provide empirical estimates of body temperature and cardiovascular responses in a wide range of conditions experienced during hot weather and heat waves.

Results

Flow of studies through the review and report characteristics

After abstract and full text screening and data extraction (Fig. 1), 438 articles met the eligibility criteria. Of these, 28 reported secondary analyses of other included studies. These were omitted from primary analyses along with 10 studies employing only radiant or conductive heating modalities or heating participants in the standing posture or in high clothing insulation35,36,37,38,39,40,41,42,43,44. Effect estimates from an additional 2 studies meeting these criteria were removed from the primary analysis, but those studies also reported data for eligible conditions45,46. The records supplement summarizes studies that could not be classified in full-text screening, eligible studies for which data could not be sourced, reports of secondary analyses, and original study reports omitted from primary analyses.

Fig. 1: Flow of studies through the review.
Fig. 1: Flow of studies through the review.
Full size image

A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart showing the flow of studies through the review456. *Of 438 eligible reports, 28 reported secondary analyses of data already included in the review. These studies were not included in analyses. Data from 10 studies that only heated participants via radiant (e.g., infrared or radiofrequency) or conductive (e.g., heated blankets) heating modalities, in the standing posture, or in high clothing insulation (e.g., nuclear, biological, chemical suit) were also omitted from the primary analysis but included in sensitivity analyses.

Characteristics of the 400 reports included in primary analyses are summarized in Table 112,13,16,18,21,22,23,24,25,26,27,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176,177,178,179,180,181,182,183,184,185,186,187,188,189,190,191,192,193,194,195,196,197,198,199,200,201,202,203,204,205,206,207,208,209,210,211,212,213,214,215,216,217,218,219,220,221,222,223,224,225,226,227,228,229,230,231,232,233,234,235,236,237,238,239,240,241,242,243,244,245,246,247,248,249,250,251,252,253,254,255,256,257,258,259,260,261,262,263,264,265,266,267,268,269,270,271,272,273,274,275,276,277,278,279,280,281,282,283,284,285,286,287,288,289,290,291,292,293,294,295,296,297,298,299,300,301,302,303,304,305,306,307,308,309,310,311,312,313,314,315,316,317,318,319,320,321,322,323,324,325,326,327,328,329,330,331,332,333,334,335,336,337,338,339,340,341,342,343,344,345,346,347,348,349,350,351,352,353,354,355,356,357,358,359,360,361,362,363,364,365,366,367,368,369,370,371,372,373,374,375,376,377,378,379,380,381,382,383,384,385,386,387,388,389,390,391,392,393,394,395,396,397,398,399,400,401,402,403,404,405,406,407,408,409,410,411,412,413,414,415,416,417,418,419,420,421,422,423,424,425,426,427,428,429,430,431,432,433. These studies contributed a total of 586 effect estimates of the heat-induced increase in heart rate, representing 6858 participant-heat exposures, or approximately 8526 person-hours (~1 person-year). Most of the reports originated from laboratories in North America, Europe and Japan and were published with increasing frequency starting in 1961 through to 2024 (Fig. 2).

Table 1 Characteristics of reports and individual effect estimates included in meta-analysis
Full size table
Fig. 2: Geographical distribution and publication dates of included studies.
Fig. 2: Geographical distribution and publication dates of included studies.
Full size image

The size of the bubbles in the top panel indicates the number of studies from each location. The color shows the geographical density, which considers both the number of studies from each location and their proximity to other study locations. The world map was generated using the opensource R package rnaturalearth491.

Included studies received a median 5 (Interquartile range: 4–5) “yes” answers on the 12 questions of the National Heart, Lung and Blood Institute Quality Tool for Before-After Studies434, though no marked differences between heating modalities were observed (Figure S24). For studies including a control trial, we also evaluated risk of bias using the Cochrane Risk of Bias Tool for Randomized Trials435. A summary is provided in Figure S25.

Effect of heating modality on cardiac responses

Heart rate (primary outcome)

Comparisons of heart rate responses between heating modalities are shown in Fig. 3. A model including a heating modality by core temperature change interaction provided the most parsimonious fit. Relative to studies that heated participants via passive exposure to hot ambient conditions in a climate chamber, the overall pattern of response was statistically significantly different in studies employing water-perfused suits (P < 0.001), water immersion (P = 0.001), and saunas (P = 0.004). Heart rate increased, on average, 26 [95% CI: 22, 29] beats/min per 1 °C rise in core temperature in the climate chamber studies. Heat-induced increases in heart rate were exacerbated in perfusion suit studies at low levels of hyperthermia (intercept difference: 7 [4, 11] beats/min; P < 0.001), though the slope of the relation between core temperature and heart rate was not statistically different between these modalities (slope difference: −3 [−7, 1] beats/min/°C; P = 0.192). At low core temperatures, heart rate was also elevated to a greater extent in water immersion (intercept difference: 8 [4, 12] beats/min; P = 0.001) and sauna studies (24 [10, 37] beats/min; P < 0.001). For these modalities, however, elevations in heart rate were blunted by 8 [3, 13], and 10 [2, 18] beats/min/°C, respectively, compared to studies performed in a climate chamber (P ≤ 0.012).

Fig. 3: Association between core temperature and heart rate in each heating modality.
Fig. 3: Association between core temperature and heart rate in each heating modality.
Full size image

Figure shows the model-predicted relation between the change in heart rate and core temperature in studies heating participants via a climate chamber (grey, reference), water-perfused suit (red, top), water immersion (aqua, middle), and sauna (purple, bottom). Models are also adjusted for the core temperature measurement technique. Model predictions (left) are presented as means (lines) and 95% confidence intervals (shaded ribbons). The latter were produced via a sandwich estimator with small-sample correction. Analyses were also adjusted for the method used to measure core temperature. Model diagnostics are provided in supplemental Figures S27–S30 (pp 71–74). Included covariates explained a statistically significant portion of variance across included studies (omnibus test of covariates: two-sided P < 0.001). However, considerable residual heterogeneity was still observed (QE test: two-sided P < 0.001), with 93.5 [92.5, 94.4]% of total variance not explained by sampling variance (overall I2; 57.1 [50.1, 63.1]% between-study I2 and 36.4 [31.3, 41.9]% within-study I2). Model estimates and P-values (reported in the main text) were not adjusted for multiple comparisons. Individual effect estimates (right) are sized according to their weight in the meta-analytic model. The marginal distribution of individual effect estimates (unweighted) for each heating modality is depicted on the right side of the panel. The effect estimate for the seminal study by Rowell et al.18 is also highlighted, which was among the first to assess the physiological limits of cardiac responses to heat exposure and is also commonly cited in translational reviews describing heat’s effects on the heart.

Although differences in heating modality, baseline-to-end heating core temperature change, and the core temperature measurement technique explained a statistically significant portion of variance across included studies (omnibus test of covariates: P < 0.001), considerable residual heterogeneity was still observed (QE test: P < 0.001). Indeed, 93.5 [92.5, 94.4]% of the total variance was not explained by sampling variance (overall I2). Of total heterogeneity, 61% was attributable to differences between studies (57.1 [50.1, 63.1]% between-study I2), while the remaining 39% stemmed from within-study differences in extracted effect estimates (36.4 [31.3, 41.9]% within-study I2).

While adjustment for non-heat-related mediators (posture and clothing insulation), and sample demographics potentially confounding or altering cardiac responses to heating (age, sex, body mass index, climate zone) improved model fit (likelihood ratio test: P ≤ 0.041), modality model coefficients were comparable to the primary analysis (Figure S31). By contrast, when end-exposure mean skin temperature was included (likelihood ratio test: P ≤ 0.001) between-modality differences were no longer statistically significant (Figure S31). Among the additional covariates included, heart rate increased 2 [1, 3] beats/min per 1 °C increase in mean skin temperature (P < 0.001). Both posture (−4 [−6, −1] beats/min in supine vs. seated; P = 0.003) and sample mean age (−2 [−3, −2] beats/min per decade; P < 0.001) had small but statistically significant effects on heart rate (Table S18).

In sensitivity analyses (Table S19–S28, Figure S32), interpretations of the primary analysis were not markedly altered when studies omitted from the primary analysis were included, a laboratory identifier was incorporated as a random effect, or we varied the within-study correlation modeled in robust variance estimation. Findings were also robust to removal of potentially influential effect estimates, modeling non-linear associations between core temperature and heart rate, restricting suit and immersion studies to those employing full body coverage, and variation in measurement timing for heart rate, study quality, and date of publication. By contrast, between-modality differences in heart-rate responses were blunted when core temperature was replaced with estimated mean body temperature. Finally, analyses of the 51 studies including a control arm were similar when data were analyzed for the heat stress arm only (as in the primary analysis) and as the difference in heart rate responses between the heat and control arms. That said, relations between heating modalities were altered considerably in these analyses compared to primary analysis of the 400 included studies.

Cardiac output, systolic blood pressure, rate pressure product (secondary outcomes)

Cardiac output, systolic pressure, and rate pressure product in the climate chamber and perfusion suit studies are shown in Fig. 4. At any given increase in core temperature, cardiac output was elevated an average of 1.0 [0.4, 1.6] L/min in the perfusion suit studies (P = 0.003). Systolic pressure changed by −8 [−15, 0] and 2 [−2, 7] mmHg per 1 °C rise in core temperature in the climate chamber and perfusion suit studies, respectively. While those within-modality associations between systolic pressure and core temperature were not statistically significant (P ≥ 0.054), the between-modality difference in slopes was (slope difference: 10 [1, 18] mmHg/°C, P = 0.025). Heat-induced increases in rate pressure product were exacerbated 1841 [32, 3650] mmHg·beats/min/°C in the suit studies (P = 0.046).

Fig. 4: Association between core temperature and secondary cardiac outcomes in the climate chamber and perfusion suit studies.
Fig. 4: Association between core temperature and secondary cardiac outcomes in the climate chamber and perfusion suit studies.
Full size image

Figure shows the model-predicted relations of the change in cardiac output (top; 79 studies, 114 effect estimates), systolic blood pressure (middle; 212 studies, 337 effect estimates), and rate pressure product (bottom; 205 studies, 303 effect estimates) as a function of the change in core temperature from baseline to end-heating in studies employing a climate chamber (gray) or perfusion suit (red) to induce heat stress. Model predictions (left) are presented as means (lines) and robust 95% confidence intervals. Analyses were also adjusted for the method used to measure core temperature. Model estimates and P-values (reported in the main text) were not adjusted for multiple comparisons. Data for hot water immersion and sauna studies are shown in supplemental Figures S33-S35 (pp 89–90), respectively. Model diagnostics are provided in supplemental Figures S35-S46 (pp 92−103). Model estimates and P-values (reported in the main text) were not adjusted for multiple comparisons. Individual effect estimates (right) are sized according to their weight in the meta-analytic model. The marginal distribution of individual effect estimates (unweighted) for each outcome modality is depicted on the right side of the panel. The effect estimate for the seminal study by Rowell et al.18 is also highlighted, which was among the first to assess the physiological limits of cardiac responses to heat exposure and is also commonly cited in translational reviews describing heat’s effects on the heart.

Secondary outcome model comparisons for hot water immersion and sauna studies are summarized in Figures S33-S34. Cardiac output was elevated, on average, 0.9 [0.2, 1.6] L/min in the immersion studies (P = 0.020), whereas reductions in systolic pressure were blunted (slope difference vs. chamber: −11 [−20, −2] mmHg/°C, P = 0.016). Despite a statistically significant overall response pattern for rate pressure product (P = 0.046), model intercept and slope coefficients were not different between modalities (P ≥ 0.145). For studies performed in a sauna, no differences in cardiac output were seen relative to the chamber studies (P = 0.562), though only 2 sauna studies reported this secondary outcome. By contrast, associations between core temperature and systolic pressure were significantly different between climate chamber and sauna studies (P = 0.001) such that systolic pressure fell 8 [0, 15] mmHg/°C in the chamber (P = 0.054) but rose 16 [8, 24] mmHg/°C in the sauna (P = 0.001; slope difference: 16 [8, 24] mmHg/°C, P = 0.001). While the overall rate pressure product response was significantly different between climate chamber and sauna studies (P = 0.001), the individual model intercept and slope coefficients were not (P ≥ 0.083).

Statistically significant residual heterogeneity was detected for all secondary cardiac outcomes (P < 0.001). The overall I2 was 92.7 [89.6, 95.0]% for cardiac output (66 [46.9, 77.0]% and 26.7 [16.7, 41.2]% between- and within-study, respectively), 92.9 [91.4, 94.3]% for systolic pressure (76.2 [70.6, 80.8]% and 16.8 [12.1, 22.8]%), and 94.0 [92.7, 95.1]% for rate pressure product (62.1 [53.4, 69.3]% and 31.9 [25.5, 39.2]%).

In sensitivity analyses, between-modality differences in cardiac output were not markedly altered when analysis was adjusted for the measurement methods employed (Table S30). That said, the method used to measure cardiac output was found to have strong impacts on model outputs (P < 0.001). Compared with more direct and invasive methods (e.g., direct Fick, thermodilution), the reported increase in cardiac output for a given rise in core temperature was reduced 1.2 [0.2, 2.2] L/min with imaging techniques (e.g., echocardiogram, magnetic resonance imaging; P = 0.026), 2.1 [1.3, 3.0] L/min with rebreathing methods (e.g., Innocor®; P < 0.001), and 1.8 [0.9, 2.6] L/min with indirect estimations (e.g., Modelflow®; P = 0.001). Like cardiac output, between-modality differences in the heat-induced change in systolic pressure were not altered when the measurement method was considered in analysis (Table S31). Again, however, the measurement method had strong influences on observed responses (P < 0.001). Relative to measurements made at the brachial artery (e.g., auscultation, oscillometry), systolic pressure was 16 [11, 22] and 7 [4, 10] mmHg/°C lower in studies employing an intra-arterial pressure transducer (P = 0.002) and indirect methods (P < 0.001), respectively. Systolic pressure (Table S32) and rate pressure product responses (Table S33) were no longer statistically significantly different between climate chamber and hot water immersion studies when core temperature was modeled using a spline (P ≥ 0.186); comparisons for the other two modalities remained statistically significant when a non-linear response function was modelled (P ≤ 0.022).

Body temperature and cardiac responses to high ambient temperatures

Associations between ambient heat index and 103 effect estimates for the heat-induced rise in core temperature and heart rate from 68 climate chamber studies are shown in Fig. 5 (1433 participant-exposures, approximately 3519 person-hours). Elevations in thermal and cardiac strain were exacerbated with increasing heat index, particularly at high levels of ambient heat stress (P ≤ 0.001). Heat-induced rises in core temperature also increased with sample age (0.05 [0.02, 0.07]°C/decade, P = 0.001) and the duration of heat exposure (0.05 [0.01, 0.10]°C/h, P = 0.027).

Fig. 5: Changes in core temperature and heart rate with increasing ambient heat stress in the climate chamber studies.
Fig. 5: Changes in core temperature and heart rate with increasing ambient heat stress in the climate chamber studies.
Full size image

Figure shows the model-predicted relation between the change in core temperature (top, red) and heart rate (bottom, purple) with increasing heat index in studies exposing participants to ambient heat stress in a climate chamber (k = 68 studies, n = 103 effect estimates). Model predictions are presented as means (lines) and robust 95% confidence intervals (shaded ribbons). Heat index was modeled via a restricted cubic (natural) spline. Predictions for a model in which heat index was included as a linear term are also shown (dark grey dashed lines). Models were adjusted for sample mean age and the duration of heat stress. Model outputs are shown for a hypothetical study with a sample mean age of 27 years and 100 min of heat exposure (based on the sample size-weighted medians of the analyzed studies). Model diagnostics are provided in supplemental Figures S51–S66 (pp 112–127). Model estimates and P-values (reported in the main text) were not adjusted for multiple comparisons. Individual effect estimates (grey bubbles) are sized according to their weight in the meta-analytic model. The marginal distribution of the individual effect estimates (unweighted) is depicted on the right side of the panel.

For the secondary outcomes (Table S34), linear associations between heat index and cardiac output and systolic blood pressure were not statistically significant (P ≥ 0.193). For the latter, however, we observed significant effects of age and duration. Systolic pressure was reduced 1 [1, 2] mmHg per decade increase in age (P < 0.001) and 1 [0, 1] mmHg per hour of exposure (P = 0.021). Rate pressure product was elevated 257 [45, 469] mmHg·beats/min/5 °C heat index (P = 0.021) but not influenced by age or heating duration (P ≥ 0.236).

Discussion

We synthesized the findings of over 400 laboratory-based studies to compare heat-induced changes in heart rate, cardiac output, systolic pressure, and rate pressure product between studies using encapsulated and extreme heating modalities and those exposing participants to hot ambient conditions in a climate-controlled chamber. The employed modality had marked effects on cardiac responses. Of key importance, water-perfused suits, the most common passive heating modality in thermophysiological research, were associated with exacerbated cardiac responses compared to resting exposure in climate chambers. While perfusion suits play an integral role in mechanistic investigations, our findings suggest that they are less useful for quantifying the cardiac effects of more natural exposures. Ecologically-minded studies simulating or conducted in real-world settings are needed to improve scientific understanding of the cardiac impacts of hot weather and support the development of robust climate-health models. As a step toward this goal, we provide empirical quantifications of the acute thermal and cardiac strain experienced in conditions reflective of deadly heat waves.

Following foundational work by Rowell et al.18,357,436, water-perfused suits quickly became the most common heating modality in research evaluating the cardiovascular adjustments supporting thermoregulation and homeostasis during acute heat stress11. In addition to their accessibility, their popularity is owed to their encapsulated design, which, when combined with the high thermal conductivity of water perfused in near-direct contact with the skin, allows for precise control of body temperature. While these features have clear benefits for mechanistic research (Table S2), we questioned whether perfusion suits provide a suitable model for evaluating the cardiovascular consequences of more natural heat exposures like those occurring during hot weather and heat waves6. To address this knowledge gap, we leveraged data from over 400 studies conducted over 60 years to evaluate how cardiac responses were modified by the employed heating modality. As anticipated, we found exacerbated elevations in heart rate, cardiac output, and rate pressure product along with blunted reductions in systolic pressure in studies using perfusion suits compared to those exposing participants to hot conditions in climate-controlled chambers (Figs. 2–4).

These findings have important implications for translational efforts directed at linking cardiovascular adjustments to acute environmental heat stress11 with the elevated risk of cardiovascular events seen during heat waves1,2. Discussions in this area have focused on the profound cardiovascular strain observed in studies using water-perfused suits4,5,6,7,8,9, most commonly the early work by Rowell et al.18  which showed increases in cardiac output of 5-7 L/min. Our analyses, however, suggest that rises in cardiac output do not surpass 3 L/min when participants are heated with a climate chamber. In fact, only 1 in 14 effect estimates of the heat-induced elevation in cardiac output in the chamber studies exceeded 2.0 L/min (Fig. 4)26,27,70,140,157,164,226,229,244,284,346,392,405 and those data were collected under extreme conditions (42 °C, 70% relative humidity)70. Further, despite comparable heart rate responses (Fig. 3), systolic pressure and rate pressure product were elevated >10 mmHg and >1800 mmHg·beats/min in perfusion suits relative to chamber studies at core temperature elevations >1.0 °C (Fig. 4), differences that, on average, exceeded previously suggested thresholds for clinical significance21. These findings highlight that while perfusion suits have played a central role in developing mechanistic understanding of the physiological adjustments to extreme heat11, they have limited utility for quantifying the cardiovascular burden of hot weather and heat waves.

Extreme increases in skin temperature are the most likely explanation for the exacerbated cardiac strain seen with perfusion suits. In our primary analyses, we normalized cardiac responses to the rise in core temperature, the most relevant body temperature when discussing the clinical implications of heat4,5,6,7,8,9. However, cardiovascular adjustments are also influenced by skin temperatures through autonomic- and locally-mediated mechanisms acting to facilitate blood redistribution from the central to peripheral circulation to aid in heat dissipation378,437,438. Since perfusion suits heat participants by passing hot water over the skin surface, cutaneous heat gain is substantial6,11. As a result, skin temperatures often reach levels considerably higher than those seen during exposure to hot conditions where the skin can readily participate in heat exchange (Figure S83). Our secondary analyses clearly demonstrate the importance of skin temperature. While correcting for posture, worn clothing, and participant demographics had little impact on heart rate differences between heating modalities, adjusting for skin temperature nullified between-modality differences (Figure S31).

Some translational reports have acknowledged skin temperature’s role in cardiovascular adjustments to heat. For example, in their review on heat waves and cardiovascular health, Kenney et al.9 cite the >40 °C skin temperatures achieved in the work by Rowell et al.18, as the key mechanism underpinning the observed doubling of cardiac output. However, skin temperatures of this magnitude are unlikely to develop outside of the laboratory. In conditions experienced in most hot weather and heat waves, evaporation of sweat ensures continual skin-environment heat transfer in unencapsulated persons (e.g., wearing summer clothing), limiting potential rises in skin temperature. This can be seen in the modality-specific distributions of average end-exposure skin temperatures. Over half of the analyzed perfusion suit studies reported skin temperatures exceeding 38.0 °C compared to less than 15% of the climate chamber studies (Figure S84). While there were some examples of skin temperatures in a climate chamber approaching those seen in water-perfused suits, they were measured in wet-bulb temperatures of 36–39 °C (heat index 80–96 °C)67,166,377. In these conditions, thermoregulation is impossible30. Thus, even if modality-related differences in skin temperature are accounted for, the marked whole-body hyperthermia induced by perfusion suits severely limit their generalizability to naturally occurring hot weather.

Our broad inclusion criteria also allowed us to explore differences in cardiac responses between climate chambers and other extreme heating modalities (Table S2). Compared to resting exposure in a climate chamber, heat-induced elevations in heart rate were blunted with water immersion (Fig. 3), whereas cardiac output was elevated, particularly at lower increases in core temperature (Figure S34; differences in systolic pressure and rate pressure product were not statistically significant when non-linear responses were considered, Tables S32–S33). Given the apparently similar body temperature distributions between the two modalities (Figure S83), elevated cardiac output in the immersion studies is likely related to the hydrostatic effects of water immersion, which promote blood displacement to the central circulation, supporting cardiac filling pressure439, increasing stroke volume for a given heart rate440,441.

We also found that at small increases in core temperature, average heart rates were elevated considerably in studies heating participants in a sauna relative to those employing climate chambers (Fig. 3). The high ambient temperatures and/or humidities generated by saunas are a likely explanation for these findings (Figure S5), as even very short exposures (~15 min) can cause highly elevated skin temperatures (Figure S83). Interestingly, however, changes in systolic pressure followed opposing response patterns, increasing with core temperature in the sauna studies but falling in climate chamber studies (Figure S34). It could be that heat-induced reductions in arterial pressure are established over durations longer than those typical of sauna studies. Another possibility is that studies performed in climate chambers are more likely to enroll older adults (Figure S22), who, on average, experience greater reductions in systolic pressure during heat exposure (Table S34). While both explanations are consistent with our data (Figure S85), further investigation is required to establish their validity. Underpinning mechanism(s) notwithstanding, the disparate findings between relatively similar heating modalities, both involving resting exposure to hot conditions, further underscores that cardiac responses to heat stress are complex and are often not readily generalizable across populations and contexts.

With the planet warming rapidly, there is growing demand for climate-health models predicting how changes in climate systems affect the health of individuals29,30,31 as well as for evidence-based heat-mitigation strategies for protecting at-risk groups5,7,32,33,34. Central to these efforts is a need for robust physiological data describing how environmental conditions (e.g., air temperature, humidity, air flow, solar radiation) and other factors (e.g., health conditions, socioeconomic status, physical activity) influence body temperature regulation, cardiovascular function, and susceptibility to adverse health events. However, physiological inputs for recently proposed models have generally been based on select studies, often with few participants exposed to conditions unreflective of hot weather. This severely limits their utility, as model inferences can be highly sensitive to the methods used to derive key inputs33. It is therefore our view that a more systematic approach synthesizing large amounts of physiological data is needed to better quantify relations between environmental factors, personal characteristics, and physiological outcomes. Such an approach would also benefit rational physiological models (e.g., the human energy/heat balance models described in references31,33) in deriving key inputs and defining their uncertainty for use in sensitivity analyses.

As a step toward this goal, we quantified core temperature and cardiac responses across a wide range of ambient conditions based on data collected from over 1400 participant-heat exposures in climate-controlled chambers. These analyses highlight that considerable hyperthermia and cardiovascular strain can be experienced during exposure to heat wave conditions. As an illustrative example, Fig. 6 presents model estimates for hypothetical groups of young and older adults resting for 2 h in peak conditions measured during recent extreme heat waves. During even a short and sedentary exposure, our model predicts that resting core temperature can exceed 38.0 °C, commonly recommended as the safe upper limit in occupational contexts442, while heart rates can surpass 100 beats/min and blood pressure can be reduced by as much as 10 mmHg, on average, in older adults (though the 95% confidence intervals are also consistent with group mean systolic blood pressures as low as ~95 mmHg). The contribution of such reductions in blood pressure to heat-related cardiovascular mortality and morbidity is an important area for future exploration. Translational discussions on the mechanisms underpinning increased cardiac risk during heat exposure tend to focus on tachycardia-induced elevations in cardiac strain4,5,6,7,8,9. However, acute hypotension may precipitate end-organ malperfusion, leading to myocardial infarction and ischemic strokes443,444,445, the risks for both of which are elevated during hot weather1,446.

Fig. 6: Estimated core temperatures, heart rates, and systolic pressures during recent heat waves.
Fig. 6: Estimated core temperatures, heart rates, and systolic pressures during recent heat waves.
Full size image

Figure shows model predictions (points) and robust 95% confidence intervals (error bars) for hypothetical groups (studies) of young (mean age: 25 years; sample weighted median below 65 years) and older adults (mean age: 70 years; sample weighted median above 65 years) resting for 2 h in peak conditions measured in recent heat waves in British Columbia, Canada (2021, gray)493, Bangkok, Thailand (2023, aqua)494, Asaluyeh, Iran (2022, purple)495, and Sikkim, India (2022, blue)496,497. Predictions are presented as means (points) and robust 95% confidence intervals (error bars) produced by adding the estimated change in each variable from the models presented in Fig. 5 (core temperature and heart rate, k = 68 studies, n = 103 effect estimates) and supplemental Table S34 (systolic pressure; k = 34 studies, n = 55 effect estimates) to the sample-weighted model median baseline values (grey horizontal lines).

Though somewhat crude, ours are among the first empirical estimates of heat strain to systematically consider the breadth of physiological data collected using highly specialized equipment and, perhaps more importantly, quantify uncertainty in model outputs. That said, there are important knowledge gaps that must be addressed to improve the utility and generalizability of such empirical models. Although accounting for design characteristics and demographic factors and between-modality differences in skin temperature reduced unexplained variance, most notably on the between-study level, considerable heterogeneity was still detected (Figure S86). These findings highlight the well-established variability in physiological responses to heat stress447,448,449. Delineating sources of this variation, particularly under ecologically relevant exposure conditions and durations (e.g., over several hours to days)449,450, is needed to not only improve our understanding of heats effects on the body, but also to identify factors for improving the accuracy and representativeness of physiological models.

It is important that future studies and analyses collect and consider data from individuals and locations most vulnerable to heat stress. For example, while age-related increases in heat-vulnerability and associated cardiovascular mortality are well-established2,3,6, 90% of our dataset, including studies aimed at defining responses to simulated heat waves or evaluating cooling strategies for protecting vulnerable groups32,33, was comprised of studies in adults, predominantly males, with mean ages under 50 years (Figure S22). Relatedly, few studies have assessed physiological responses in participants with heat-sensitive chronic health conditions447 and fewer still under environmental conditions relevant to the general population (e.g., those experienced during hot weather)6. Because of this, we excluded studies with participants reporting diagnosis of, or taking medications for, chronic health conditions linked with heat vulnerability. Future research and analyses should also evaluate the impacts of chronic diseases and associated medications on cardiac responses under ecologically relevant conditions, especially in light of recent research showing that age-associated functional decline and multimorbidity are associated with increased risk of heat-related hospitalization and deaths451,452. Finally, the geographical areas most strongly represented in our dataset (Fig. 1) are not those facing the most immediate threats of extreme heat and climate change (e.g., countries highlighted in Fig. 6). Research is needed to expand understanding of cardiovascular responses to heat exposure to encompass a greater range of climate regions (and associated differences in natural acclimatization among inhabitants), particularly those in the Global South. More studies performed in ecologically relevant environmental conditions would also allow for more thorough consideration of environmental factors like solar radiation and airflow33,453 and facilitate the expansion of models from resting conditions to incorporate physical activity (e.g., activities of daily living)454.

Another consideration is that not all studies reported data for secondary cardiac outcomes. Cardiac output, for instance, was reported in only ~20% of studies, and almost two-thirds of effect estimates were derived from participants heated using perfusion suits. This resulted in wide confidence intervals for estimates of the average heat-induced change in cardiac output and between-modality differences (Fig. 4). A likely contributor to this uncertainty is variation in employed measurement techniques. While Rowell et al.18 measured cardiac output using the highly accurate dye dilution technique (described in reference455). Common estimation techniques like Modelflow® can underestimate cardiac output in hyperthermic individuals relative to more direct methods like thermodilution (which is similar to dye dilution)313. Compared to direct methods, cardiac output was ~1–2 L/min lower when measured via imaging techniques (e.g., echocardiogram), rebreathing, or other indirect methods like Modelflow® (Table S30). However, direct measurements were almost entirely restricted to studies employing perfusion suits (Figure S87). While it is unlikely that measurement technique explains the large differences in cardiac output between Rowell et al.18 and the reviewed climate chamber studies, the lack of direct measurements in the latter further highlights the challenges in using extant laboratory data to inform our understanding of the real world.

Laboratory-based heat exposure research is becoming increasingly leveraged in the development of climate-health models and evidence-based heat-health guidance. However, our findings suggest that the substantial body of evidence on the cardiovascular burden of heat has, in its current form, limited use in supporting these efforts. Ecologically minded physiological research, conducted in both laboratories and natural settings, is urgently needed to develop and refine empirical physiological models and facilitate linkages to larger health databases and heat-health action plans and policies.

Methods

This review was prospectively registered (PROSPERO: CRD42020192887) and conducted and reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines456. The full review protocol, statistical analysis plan, and supporting analyses, are provided in the supplemental materials.

Literature search and study eligibility criteria

Cochrane search strategies were employed to identify eligible studies457. PubMed, Embase, Web of Science, and the Cochrane Central Register of Controlled Trials were searched from database inception until January 9th, 2023, for terms related to different laboratory-based passive (non-exercising) heating modalities, indices of body core temperature, and cardiac outcomes. Search criteria for each database are reported in Table S1 of the supplemental materials (pp 3–5). Systematic searches were supplemented with hand-searches of reference lists from topical narrative reviews and PubMed performed until February 8th, 2024.

Predefined criteria for population, exposure (intervention), comparator, and outcomes, (PECOs) were used to identify eligible studies457,458. Included studies needed to be published in a peer-reviewed journal. Conference papers were ineligible. There were no limits on date of publication. Given the size of the review, we decided not include study reports without an already available English translation.

Population

The population of interest was adults aged ≥18 years without reported acute or chronic disease linked with heat vulnerability (e.g., heart disease). Relatively few studies have evaluated thermal and cardiovascular responses to heat exposure in persons with chronic health conditions. We therefore opted to exclude such reports. Studies including individuals with risk factors for cardiovascular disease (e.g., hypertension, hypercholesteremia, cigarette smoking, obesity, etc.) were included, and we did not restrict inclusion based on sex/gender, anthropometric characteristics, habitual physical activity participation, or aerobic fitness.

Exposure

Studies were eligible if they included a passive (resting) heat stress intervention (no physical activity) using any heating modality (supplemental Table S2 provides an overview of common modalities) without concurrent continuous secondary intervention (e.g., intravenous saline infusion). All studies needed to report an index of core temperature (e.g., rectal temperature, oral temperature) and only those in which core temperature increased ≥0.1 °C were included. This ensured that only heat stress studies, and not studies using water-perfused suits or water immersion for other reasons (e.g., cold-stress, central volume manipulation) were included. Studies with acute secondary interventions (e.g., isometric hand grip, cognitive function tests) were included provided that resting data were reported and the intervention was not expected to influence resting responses. Secondary exposure variables included mean skin temperature and mean body temperature (weighted average of core and skin temperatures459).

Comparator

Included studies were required to report cardiovascular outcomes prior to and at the end of the heat stress intervention. Data from studies that included a non-heat stress control arm (e.g., rest in a thermoneutral environment, temperate water immersion) were also included for use in sensitivity analyses (described below).

Outcomes

All studies needed to report heart rate (primary outcome). Secondary outcomes included cardiac output, stroke volume, systolic blood pressure, and rate pressure product (an index of cardiac work calculated as heart rate × systolic pressure).

Study screening, data extraction, and risk of bias assessment

Study screening, data extraction, and risk of bias assessments were independently performed by at least two review authors. Disagreements were resolved via group consensus. Screening was conducted using commercially available software460. Relevant data on experimental design, participant characteristics, and study outcomes were extracted using a custom spreadsheet based on a previously tested form461. A full list of extracted variables is included in supplemental Tables S5–S11. A similar form was used to record study quality and risk of bias assessments (Tables S12–S14).

The method to induce heat stress (i.e. the heating modality) was extracted and classified as either ambient exposure in a climate chamber, water-perfused suit, water immersion, or sauna (Table S2). Body temperature (exposure) and cardiovascular (outcome) data were extracted as mean and dispersion (e.g., standard deviation or error, 95% confidence interval) before and at the end of the heat stress intervention (and control intervention, if included) and/or the change score and its dispersion (i.e., the baseline-end difference). Data were extracted for all eligible experimental conditions. If studies employed a multi-stage protocol wherein body core temperature was stabilized (clamped) at multiple levels of heat stress, data for each level were extracted. Otherwise, only the pre- and end-heating data were recorded. For studies with multiple experimental trials in which drinking water was provided during heating, only the condition eliciting the smallest fluid loss (change in body weight) was included. Open-source software was used to extract numerical data presented in graphical format462. For studies that met the inclusion criteria but did not present core temperature and heart rate data in a usable format, the corresponding author was contacted (supplemental Box S1).

Information on numerous additional study-level variables was also extracted, including participant characteristics and posture, duration of heat stress, additional indices of body temperature (mean skin temperature and mean body temperature), indices of body fluid status (change in body weight and/or plasma volume), and methods of exposure/outcome ascertainment. Regarding the latter, the body location in which core temperature was measured was recorded and grouped into three categories: i) direct, slow-responding (rectum, gastrointestinal tract); ii) direct, fast-responding (pulmonary artery, esophagus, or tympanic membrane); or iii) indirect (oral cavity, aural canal). This was done to account for differences in the absolute and temporal responses to body heating among common core temperature indices (see “Meta-analyses” below)459. Ascertainment methods for all exposure and outcome variables are described in supplemental Table S9 (pp 17–18).

The quality of studies including only a heat stress arm were assessed using the National Heart, Lung and Blood Institute Quality Tool for Before-After Studies434. Risk of bias in the controlled studies was evaluated using the revised Cochrane Risk of Bias Tool for Randomized Trials (RoB 2)435. Information on the use of these scales along with adaptations made for the current review are described in the supplement (Tables S12–S14).

Meta-analyses

Individual study effect estimates (unit of analysis)

Analyzed effect estimates were the change in heart rate (primary outcome) and secondary cardiac outcomes from pre-heating (baseline) values to the end of heating. If the baseline-to-end-heating change and its standard deviation were not directly reported, they were derived from data presented for the baseline and end-heating timepoints463. Data presented in figures were extracted using open-source software (WebPlotDigitizer)462. Secondary variables reflecting a composite of other exposures (e.g., mean body temperature) or outcomes (e.g., cardiac output, rate pressure product) were imputed from the composite variables if sufficient data were available (Tables S15–S16). Secondary outcomes reported in the main article included cardiac output, systolic pressure, and rate pressure product, with analyses for stroke volume confined to the supplement.

Modifying effect of heating modality on cardiac responses

We evaluated how the employed heating modality modified heat-induced cardiac responses in a series of multilevel mixed-effects meta-analyses. Robust variance estimation was used to account for statistical dependencies among effect estimates derived from the same study and to correct for small sample bias464. This approach allows for inclusion of multiple estimates from individual studies in a single analysis, increasing statistical power to evaluate the contribution of study-level covariates (e.g., heating modality) to the observed variation in study findings (heterogeneity)464,465.

The analysis of each cardiac outcome followed the general workflow for meta-analyses with complex dependency structures466. First, an approximate variance-covariance matrix of the sampling variance of dependent effect estimates was constructed. A unique identifier for study (study id) was used as the clustering variable and the within-study correlation was assumed constant at 0.6 (different values were considered in sensitivity analyses, described below)464. The resultant variance-covariance matrix was then used in a series of multilevel mixed-effects meta-analyses, which differed with respect to prespecified covariates (fixed effects). Model 1 included heating modality as a categorical predictor (4 levels: ambient exposure in a climate chamber [reference], water-perfused suit, water immersion, sauna). To account for differences in the level of heat strain, model 2 included, in addition to heating modality, the baseline-to-end heating change in core temperature (in °C), and the core temperature measurement technique as a categorical predictor (three levels: direct-slow [reference], direct-fast, indirect). Model 3 was identical to model 2 except an interaction term between heating modality and core temperature change was specified. All models included a multilevel random effects structure with an effect identifier nested within the study identifier467,468. Variance components (random effects) for all models were initially derived using the maximum likelihood (ML) estimator. Models were then compared using Akaike’s Information Criterion (AIC)469 and the most parsimonious (lowest AIC) was refit using the restricted maximum likelihood (REML) variance estimator470.

Approximate normality and homogeneity of model residuals were assessed via visual inspection of residual diagnostic plots. Identifiability of the variance components (random effects) was evaluated using profile likelihood plots466. After confirming the models were well specified, robust confidence intervals and significance tests were obtained using a cluster-robust (sandwich) estimator with small-sample correction471,472. Modification of cardiac responses by heating modality was interpreted with respect to significance tests considering both the overall pattern of response between the climate chamber studies and other heating modalities (i.e., omnibus test of both model intercept and slope coefficients) and the individual model coefficients for the between-modality differences in intercept and slope (per °C change in the pre-end heating rise in core temperature). Statistical significance was taken as a two-sided robust P < 0.050. No adjustments were made for multiple comparisons.

Residual heterogeneity was assessed via the QE test and the I2 statistic473,474,475. The latter was evaluated for the overall model as well as individually at the between- and within-study levels. Influence statistics were generated to identify effect estimates with potentially undue influence on model outputs (Cook’s distances and differences in regression coefficient values (DFBETAS))476. Relations between model residuals and their associated standard errors were visualized using funnel plots477.

Potential mediators and confounders of heat-induced changes in heart rate

In secondary analyses, we evaluated the association between heating modality and heart rate (primary outcome), controlling for potential non-heat stress-related mediators and confounders of this relation. A directed acyclic graph (DAG) describing assumed relations between these factors is provided in supplemental Figure S21 (pp 63). Three models were produced. Model 4 included the same covariates from the best-fitting model amongst models 1–3 along with participant posture (2 levels: seated [reference], supine) and the level of clothing insulation (2 levels: low [reference], medium). Model 5 was comprised of the same covariates as model 4, as well as participant characteristics known to influence thermoregulatory function. These included age (in years), sex (proportion of females in sample), body mass index (in kg/m2), and climate zone (Köppen climate classifications478), which was taken as a proxy for natural heat acclimatization (4 levels: temperate [reference], continental, tropical, polar). Model 6 included the same covariates as model 5 along with end-heating mean skin temperature (in °C) also included. These analyses allowed us to isolate the extent to which modality-related differences in heat strain explained the primary findings.

Body temperature and cardiac responses to high ambient temperatures

A secondary goal of this project was to leverage the large number of captured studies to provide empirical estimates of the thermal and cardiac strain that may be experienced in hot weather and heat waves. To achieve this, we modeled associations between ambient heat index (apparent temperature index considering both air temperature and humidity) and core temperature, heart rate, and secondary cardiac outcomes in studies exposing participants to ambient heat stress in a climate chamber (see supplement pp 66). For core temperature and heart rate, separate models were produced with heat index included as a linear term and as a restricted cubic (natural) spline with knots placed at the 10th, 50th, and 90th quantiles479. Analyses for all other outcomes included only the linear term. Models for core temperature, heart rate, systolic pressure, and rate pressure product were adjusted for sample mean age (in decades) and heating duration (in hours). Analysis was restricted to experimental conditions with heat index ≤80 °C, the highest value ever recorded480.

Sensitivity analyses

A series of sensitivity analyses were conducted to evaluate the robustness of our findings to key assumptions made in data processing and analysis (see supplement pp 66–69). For heart rate, we evaluated whether findings were robust to: a) inclusion of the studies omitted from the primary analysis, b) consideration of laboratory in the multilevel random effect structure, c) variation in the within-study correlation input in robust variance estimation, d) removal of potentially influential effect estimates flagged in influence analyses, e) differences in the bodily coverage of perfusion suits and water immersion in studies that employed those modalities, f) inclusion of covariates describing when the baseline and end-heating measurements were procured, g) modeling non-linear associations between core temperature and heart rate, h) replacing core temperature with estimated mean body temperature, i) including a covariate for study quality, j) including a covariate for publication date, and k) whether the outcome was analyzed as the pre-end heating change or the between-arm difference in studies including a control group. We also evaluated whether findings of secondary analyses were altered by the technique used to measure l) cardiac output and m) systolic blood pressure, and when modeling non-linear associations between core temperature and n) systolic pressure and o) rate pressure product.

All data processing and statistical analyses were performed using R (Version 4.2.0, R Core team)466,481,482,483,484. The full analysis plan is provided supplemental materials (pp 30–69). R was also used for all data visualizations484,485,486,487,488,489,490,491,492.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.