Preliminary screening.

During the preliminary session, participants were familiarized with all procedures and measurements and completed the Canadian Society for Exercise Physiology Get Active Questionnaire (GAQ) (29) and the American Heart Association Pre-Participation Screening Questionnaire. The general types of physical activity performed were assessed using the Kohl Physical Activity Questionnaire (30). Thereafter, participants’ body height and mass were measured with a stadiometer (model 2391, Detecto) and digital weighing terminal (model CBU150X, Mettler Toledo), respectively. Body mass index and surface area were subsequently calculated from these measures (28). Thereafter, participants performed an exercise stress test to volitional fatigue (semirecumbent cycling) while monitored via 12-lead electrocardiogram by an American College of Sports Medicine and Canadian Society for Exercise Physiology-certified exercise physiologist (P.B.). For trials completed after November 2020 (resumption of testing following COVID-19-related lockdowns), initial screening was conducted over the phone, the Rose Angina Questionnaire was used to screen for cardiac symptoms (34) in place of the exercise stress test, and anthropomorphic data were collected in the morning of the experimental heat exposure trial. To ensure consistency, the anthropomorphic data reported in Table 1 are those measured at the start of the experimental trial.

Experimental heat exposure trial.

After arriving at the laboratory (0630–0900), participants inserted a rectal temperature probe and were instrumented with wireless skin temperature sensors and a 5-lead electrocardiogram (see Outcome Measurement and Analysis for more details). Baseline hemodynamic responses were then evaluated via a brief test battery (∼45 min), which was performed as follows. First, cardiac output was measured via inert gas rebreathing (Innocor and Innovision). Next, participants rested quietly for a 10-min electrocardiogram recording, after which arterial systolic and diastolic pressures were measured in triplicate via manual auscultation (each measurement separated by ∼30 s). A second measurement of cardiac output was then taken (allowing for ≥10 min between the first and second measurements to ensure full gas washout). Thereafter, forearm blood flow was measured on the right side of the body with venous occlusion plethysmography (AI6, D.E. Hokanson). Throughout the tests, participants remained quietly seated (slightly reclined) with both feet flat on the floor and their hands resting comfortably on their laps. The only exception was during the measurements of forearm blood flow, where the instrumented limb was raised slightly above the heart level to facilitate venous drainage.

Following baseline hemodynamic measurements, a venous blood sample was obtained, and the participant entered a climate-controlled chamber to begin the 9-h heat exposure. The chamber was regulated to 40°C dry-bulb temperature (∼9% RH; 25.5°C wet-bulb globe temperature, WBGT). These conditions are consistent with maximum temperatures recorded in many North American cities, although they are on the low end of peak conditions during the 2021 North American heat dome event in the Pacific Northwest (daily peak: 38.2°C–49.6°C and 9%–20% RH) (25, 26). Airflow within the climate chamber was low (<0.3 m/s), and there were no substantial sources of radiant heating (35).

For the first 3 h of exposure (hours 1–3), participants were seated within the Snellen air calorimeter, a unique device able to directly measure whole body dry and evaporative heat loss (36). At the 3 h mark, participants exited the calorimeter (but remained in the climate chamber), and the brief hemodynamic test battery was repeated. Hours 4–6 were spent seated in the climate chamber (but outside the calorimeter). During this time, tap water (∼16°C) was available to the participants ad libitum from a self-service insulated water cooler, and they could eat a light, self-provided lunch with low water content (e.g., peanut butter sandwich). The hemodynamic test battery was conducted again just before the 6 h mark, at which point participants were transferred back to the calorimeter where they spent the final 3 h of exposure (hours 7–9). At the end of this period, participants completed a fourth and final hemodynamic test battery, and a venous blood sample and final nude body mass measurement were procured before leaving the climate chamber.

Whole body heat exchange and storage.

Whole body dry and evaporative heat loss were measured during the first and final 3 h of heat exposure (hours 1–3 and 7–9) using the Snellen direct calorimeter (35–38). Calorimeter inflow and outflow air temperature and absolute humidity were measured every 8 s with resistance temperature detectors (Black Stack model 1560, Hart Electronics) and high-precision dew-point hygrometry (Model 373H, RH Systems). Air mass flow was measured via differential thermometry over a known heat source in the effluent air stream. Under its current operation, airflow in the climate chamber and in the calorimeter where the participant is seated is <0.3 m/s (35). Data were recorded with LabVIEW software (Version 7.0, National Instruments).

In analysis, cubic splines were used to smooth calorimeter inflow and outflow temperature and humidity data (“smooth.spline” function in R). Minute averages for evaporative heat loss (insensible heat loss via sweat evaporation) were calculated by multiplying the smoothed outflow-inflow difference in absolute humidity by the latent heat of vaporization of sweat (2,426 J/g) and air mass flow. Dry heat loss (sensible heat loss via convection; radiative and conductive heat loss are negligible in the calorimeter) was determined similarly from the smoothed outflow-inflow air temperature difference and specific heat capacity of air (1,005 J/kg/°C). Dry heat loss was measured as a negative value for all participants since ambient temperature was greater than skin temperature during exposure. Negative dry heat loss will be referred to as dry heat gain hereafter.

During each 3-h calorimetry measurement period, oxygen and carbon dioxide concentrations in expelled air were measured with electrochemical gas analyzers (AMETEK model S-3A/1 and CD 3A, Applied Electrochemistry) from air drawn from a 6-L fluted mixing box located within the calorimeter. Expired air was recycled back into the calorimeter to account for respiratory heat exchange. Minute ventilation was measured with an integrated turbine ventilometer. The gas analyzers and ventilometer were calibrated ∼30 min before each of the two 3-h calorimetry measurement periods. Oxygen and carbon dioxide exchange data were smoothed with cubic splines (as described above), converted to 1-min averages, and used to calculate metabolic rate (37). Metabolic heat production was assumed to equal metabolic rate since no external work was performed.

The rate of body heat storage was calculated as total heat gain (heat production + dry heat gain) minus evaporative heat loss (37). For analysis, total heat gain, evaporative heat loss, and the rate of heat storage were taken as 15-min averages at hours 0 (1–15 min), 3 (166–180 min), 6 (361–375 min), and 9 (526–540 min) and presented relative to body surface area (i.e., in W/m²). Cumulative heat storage (in kiloJoules, kJ) during each calorimetry measurement period (hours 1–3 and 7–9) was calculated as the temporal summation of the rate of body heat storage (37).

Body core and skin temperatures.

Rectal temperature was measured as an index of body core temperature using a general-purpose thermocouple temperature probe (Mon-a-therm General Purpose Temperature Probe, Mallinckrodt Medical Inc.) inserted ∼12 cm past the anal sphincter. Data were collected in 15 s intervals using LabVIEW software (Version 7.0, National Instruments). Rectal temperature was not measured in four older participants due to technical difficulties or participant refusal to insert the rectal thermocouple. For these individuals, core temperature was estimated via a temperature capsule (VitalSense ingestible capsule thermometer; Mini Mitter Company) ingested ∼2 h before heat exposure (upon arrival to the laboratory). A hip-worn recording device recorded temperature data from the pill at 1 min intervals (VitalSense Monitor, Mini Mitter Company). These capsules demonstrate low systematic bias when assessed against rectal temperature (39). Skin temperature was measured at eight body regions using surface temperature sensors (DS1922L Thermochron, OnSolution Pty Ltd) and used to calculate mean skin temperature based on the weightings described in ISO 9886:2004: 7% forehead, 17.5% right scapula, 17.5% upper left chest, 7% upper right arm, 7% right forearm, 5% left hand, 19% right anterior thigh, and 20% left calf (40).

Core and skin temperature data were converted to 15-min averages at the end of each hour of heat exposure after manual cleaning and interpolation [linear interpolation using the “na.approx” function for R (41)]. To ensure consistency with our previous report, data for hours 4 and 6 were taken as the average of mins 31–45 (instead of mins 46–60) (14).

Hemodynamic responses.

For each of the four cardiovascular test batteries (occurring at ∼3 h intervals), cardiac output was taken as the average of the two measurements made using inert gas rebreathing (Innocor, Innovision). In four older participants, cardiac output was estimated from beat-to-beat measurements of the arterial pressure waveform of the right middle or ring finger during the 10-min resting electrocardiogram measurements (described below) via the volume-clamp technique (Finometer Pro, Fina-press Medical Systems) (42, 43) due to technical issues with the rebreathing unit. Forearm blood flow was taken as the average of four consecutive measurements made on the right side of the body using venous occlusion plethysmography (8-s cuff inflation cycle; AI6, D.E. Hokanson). The instrumented limb was supported slightly above heart level to facilitate venous drainage. Electrocardiogram during 10 min of seated rest was recorded using a Holter monitor (DigiTrak XT Holter Monitor, Philips) and downloaded and analyzed using Philips Zymed Software (Version 3.0, Philips). Heart rate was extracted from the electrocardiogram tracing, smoothed with cubic splines (see Whole body heat exchange and storage), and presented at each of the four resting measurement periods occurring at ∼3 h intervals (before and after each calorimetry period). Arterial systolic and diastolic blood pressures were taken as the average of three values measured at the brachial artery (∼30 s between measurements) via manual auscultation. Consistent with American Heart Association guidance, participants were seated for at least 15 min before blood pressure measurement, rested quietly with both feet flat on the floor, and the arm was supported at the approximate level of the fourth intercostal space (44). Stroke volume was calculated as cardiac output ÷ heart rate × 1,000 (conversion from L to mL). Rate pressure product, an index of myocardial oxygen demand, was derived as heart rate × systolic pressure (45, 46).

Body fluid status.

Participant nude body mass was measured to an accuracy of 10 g using a high-performance digital weighing terminal before and after each heat exposure trial (CBU150X, Mettler Toledo). Net fluid loss through sweat and urine over the 9-h heat exposure were quantified as the percentage change in body mass from baseline values (corrected for the mass of consumed food).

Venous blood collected before and at the end of heat exposure was used to estimate plasma volume changes. Additional blood was collected to assess age-related changes in the cellular stress response in peripheral blood mononuclear cells, as reported in the companion article (27). Each ∼40 mL sample was drawn from the antecubital vein and transferred directly into plasma Vacutainer tubes (5.4 mg K2 EDTA, BD), mixed by inversion, and used to measure hematological parameters in duplicate (Ac·T diff2, Beckman Coulter). Intravenous fluid was not provided to replace drawn blood. The change in plasma volume was estimated from the average values for hemoglobin and hematocrit using the method by Dill and Costill (47). Since venous blood draws were taken following the first and final hemodynamic test batteries, participants were seated for ∼45 min before sample collection (48).

Sample size determination.

An a priori power analysis determined that a sample of 19 young and 19 older adults was required to detect a 6 W/m² between-group difference in the rate of whole body heat storage, equivalent to a 0.1°C–0.2°C difference in mean body temperature if sustained over an hour, at the end of each calorimeter session (i.e., hours 3 and 9) with 80% statistical power (without correcting for multiple comparisons, α = 0.05). The expected mean difference and standard deviation in each group were estimated from the rate of body heat storage in young and middle-aged-to-older adults over the final 30 min of the 3-h heat exposure in our previous work [young: 0 (4) W/m², older: 6 (8) W/m²] (9). With the analyzed sample size (20 young and 37 older, see Missing data), we had an estimated 96% power to detect a 6 W/m² difference in the rate of body heat storage between the young and older adults using a two-sided test without multiplicity adjustment.

Missing data.

Technical difficulties with the recording equipment meant that calorimetry data were missing for two older participants. In addition, we were not able to procure venous blood samples for three participants (n = 1 young and n = 2 older). Missing data were not imputed. Since data were missing primarily due to technical difficulties, we were comfortable assuming that they were missing at random. Complete case analysis gives unbiased effect estimates under this assumption (49). Furthermore, techniques for estimating missing data (e.g., multiple imputation) can improve statistical power relative to the complete case approach when covariates are imputed (50), but provide minimal benefit to statistical power when outcome data are imputed (51).

Primary analysis: comparing physiological responses between young and older adults.

Data for the rates of whole body heat gain, loss, and storage, core and skin temperatures, hemodynamic variables (cardiac output, forearm blood flow, systolic and diastolic blood pressures, and rate pressure product), and indices of body fluid status (changes in body mass and plasma volume) were analyzed using linear mixed-effects models. The fixed effects for analysis of heat exchange were age group (young and older) and time (rates: hours 0, 3, 6, 9; cumulative: hours 1–3 and 7–9). Similarly, the models for body temperature and hemodynamic responses included fixed effects for age group and time (3 levels: hours 3, 6, and 9), with baseline values modeled as a continuous covariate. Fluid status variables were reported as changes from baseline values, so time was excluded from the model. Akaike’s information criterion was used to determine the fixed-effect structure (additive or multiplicative) for the final model (52). All models included a random intercept for each participant. Homoscedasticity and normality of residuals (fixed and random effects) were evaluated via visual inspection of diagnostic plots.

Model estimated marginal means were compared between groups and over time. Reported P values were adjusted for multiplicity using the Holm–Bonferroni procedure (all comparisons for each variable were treated as a family of comparisons). Raw data are presented as means (standard deviations) and estimated between-group differences as means [95% confidence limits]. A P < 0.050 (two-sided) was considered statistically significant. Data were analyzed using R (Version 4.2.0, R Core Team) (53–58).

Secondary and sensitivity analyses.

In secondary analyses, we explored the effects of reported chronic health conditions and sex on heat exchange and body temperature responses during the 9-h heat exposure. For chronic conditions, the primary analyses described earlier were performed but with the age group variable replaced with an indicator variable for either the presence of type 2 diabetes (i.e., participants with vs. without diabetes), hypertension (i.e., participants with vs. without hypertension), or either condition (i.e., participants with diabetes and/or hypertension vs. those without either condition). This analysis was conducted only with data from the older participants, as no young individuals with diabetes or hypertension were recruited.

To evaluate the effect of sex on physiological responses to heat exposure, we used likelihood ratio tests to compare the best fitting model for each outcome variable from the primary analysis to a new model where female sex was included as an indicator variable (59). Post hoc comparisons were then explored in models where consideration of sex led to a statistically significant improvement in model fit. Note that we only evaluated an additive effect of sex (i.e., we did not model an interaction between sex and time and/or age group) due to concerns over statistical power (60).

As another method to explore age-associated changes in thermoregulatory function, we evaluated associations between participant age and the change in rectal temperature over the 9 h exposure within the entire study sample and among the older adults using linear regression. In additional models, we assessed whether the association between age and the increase in core temperature was modified by the presence of chronic health conditions and participant sex (via inclusion of an age by condition or age by sex interaction). Finally, in sensitivity analyses, we performed the primary analyses again but limited the sample to the 20 young and 19 older participants tested in our original trial (NCT04353076).

Whole body heat exchange and body temperatures.

Whole body heat exchange and storage in the young and older adults over the 9-h heat exposure are shown in Fig. 1. Total heat gain (metabolic + dry; Fig. 1A) was, on average, 6 W/m² [95% confidence interval: 1, 10] lower in the older compared with young adults (P = 0.029; Fig. 1A). This was driven by an age-related difference in metabolic heat production, which was 6 W/m² [3, 9] lower in the older adults (P = 0.001); dry heat gain was not different between groups (older − young difference: 0 W/m² [−3, 3]; P = 0.926). Heat loss via sweat evaporation increased over the first 3 h in the young and older adults (P < 0.001; age by time interaction: P < 0.001; Fig. 1B) but was not different between subsequent time points (P ≥ 0.264). Evaporative heat loss was 18 W/m² [11, 25] lower in the older adults at the start of heat exposure (hour 0; P < 0.001) but not statistically significantly different between groups at hours 3, 6, or 9 (P ≥ 0.138).

Consistent with the data for evaporative heat loss, the rate of heat storage fell from hours 0 to 3 in both groups (P < 0.001; age by time interaction: P = 0.016; Fig. 1C) but was not different between subsequent time points (P > 0.230), except heat storage fell 7 W/m² [2, 11] from hours 6 to 9 in the older adults (P = 0.041). Heat storage was 11 W/m² [5, 16] greater in the older compared with young adults at the beginning of exposure (P = 0.002) but was not significantly different between groups at hours 3–9 (P > 0.999). Consequently, the older adults stored 88 kJ [29, 147] more heat over the first 3 h of exposure [P = 0.006; young: 275 (123) kJ, older: 363 (122) kJ; age by time interaction: P = 0.026], whereas no between-group differences in cumulative heat storage were observed during the final 3 h [young: 60 (80) kJ, older: 63 (96) kJ; older − young difference: 3 kJ [−56, 61], P = 0.926].

Body core and mean skin temperatures are depicted in Fig. 2. After the initial rise over the first 3 h of exposure, core temperature in the young adults increased 0.1°C [0.0, 0.1] and 0.1°C [0.0, 0.2] from hours 3 to 6 and hours 6 to 9, respectively, although the changes over each 3 h period were not statistically significant (P ≥ 0.216; age by time interaction: P = 0.022; Fig. 2A). In contrast, there was a statistically significant increase in core temperature of 0.2°C [0.1, 0.3] in the older adults between hours 3 and 6 (P < 0.001). Core temperature in the older adults did not change over the final 3 h of exposure (0.0°C [−0.1, 0.0]; P = 0.455). At the end of the first 3 h of exposure, core temperature was not statistically different between the young and older adults (older − young difference: 0.1°C [0.0, 0.3]; P = 0.216). By contrast, core temperature was elevated in the older compared with young adults at 6 (0.3°C [0.1, 0.4]; P < 0.001) and 9 h (0.2°C [0.0, 0.3]), although the latter comparison was not statistically significant after correcting for multiple comparisons (P = 0.061). Mean skin temperature was not different between the young and older adults (P = 0.176; Fig. 2B).

Hemodynamic responses and body fluid status.

Hemodynamic responses in the young and older adults are shown in Table 2. For all variables, an additive model best fit the data (i.e., without age-by-time interaction). Cardiac output and forearm blood flow were, on average, 1.0 L/min [0.3, 1.6] and 1.3 mL/100 mL_tissue/min [0.6, 2.0] lower in the older compared with young adults (after adjusting for baseline values; P ≤ 0.004). Similarly, the reduction in systolic blood pressure was 5 mmHg [0, 10] greater in the older adults, although the between-group difference was not statistically significant after correction for multiple comparisons (P = 0.173). No other between-group differences were observed.

We also found small but statistically significant changes in hemodynamic variables after the initial change from baseline to the 3 h time point. Cardiac output increased an average of 0.6 L/min [0.4, 0.8] from hours 3 to 6 (P < 0.001) but fell the same amount (0.6 L/min [0.4, 0.8]) over the final 3 h of exposure (P < 0.001). The change in forearm blood flow from hours 3 to 6 (0.3 mL/100 mL_tissue/min [0.0, 0.6]) was not statistically significant after adjustment for multiple comparisons (P = 0.078). However, forearm blood flow fell 0.6 mL/100 mL_tissue/min [0.2, 0.9] over the final 3 h of exposure (P = 0.002). Heart rate increased 2 beats/min [1, 3] from hours 3 to 6 (P = 0.010) but did not change over the final 3 h of exposure (P = 0.872). Stroke volume increased 6 mL [3, 9] from hours 3 to 6 (P < 0.001) and fell by 8 mL [5, 11] from hours 6 to 9 (P < 0.001). Diastolic blood pressure fell by 4 mmHg [2, 6] over the middle 3 h and increased by 6 mmHg [4, 7] over the final 3 h of exposure (both P < 0.001). Systolic blood pressure and rate pressure product did not change from 3 to 6 h (P ≥ 0.607) or from 6 to 9 h (P ≥ 0.768).

The reduction in plasma volume was 4.0% [1.6, 6.4] greater in the older [−3.9% (SD 3.9%)] compared with younger adults [−0.1% (5.1%); P = 0.002]. By contrast, no between-group differences in the change in body weight from baseline values were observed [young: 1.9% (0.9%), older: 1.9% (0.9%); difference: 0.0% [−0.5, 0.5], P = 0.971].