Effects of Daylong Exposure to Indoor Overheating on Thermal and Cardiovascular Strain in Older Adults: A Randomized Crossover Trial

Abstract

Background:

Health agencies recommend that homes of heat-vulnerable occupants (e.g., older adults) be maintained below 24–28°C to prevent heat-related mortality and morbidity. However, there is limited experimental evidence to support these recommendations.

Objective:

To aid in the development of evidence-based guidance on safe indoor temperatures for temperate continental climates, we evaluated surrogate physiological outcomes linked with heat-related mortality and morbidity in older adults during simulated indoor overheating.

Methods:

Sixteen older adults [six women; median age: 72 y, interquartile range (IQR): 70–73 y; body mass index: $24.6(\text{IQR}:22.1–27.0){\mathrm{kg}/\mathrm{m}}^2$] from the Ottawa, Ontario, Canada, region (warm summer continental climate) completed four randomized, 8-h exposures to conditions experienced indoors during hot weather in continental climates (e.g., Ontario, Canada; 64 participant exposures). Ambient conditions simulated an air-conditioned environment (22°C; control), proposed indoor temperature upper limits (26°C), and temperatures experienced in homes without air-conditioning (31°C and 36°C). Core temperature (rectal) was monitored as the primary outcome; based on previous recommendations, between-condition differences $>0.3°\mathrm{C}$ were considered clinically meaningful.

Results:

Compared with 22°C, core temperature was elevated to a meaningful extent in 31°C [$+\mathrm{0}.\mathrm{7}°\mathrm{C}$; 95% confidence interval (CI): 0.5, 0.8] and 36°C ($+\mathrm{0}.\mathrm{9}°\mathrm{C}$; 95% CI: 0.8, 1.1), but not 26°C ($+\mathrm{0}.\mathrm{2}°\mathrm{C}$, 95% CI: 0.0, 0.3). Increasing ambient temperatures were also associated with elevated heart rate and reduced arterial blood pressure and heart rate variability at rest, as well as progressive impairments in cardiac and blood pressure responses to standing from supine.

Discussion:

Core temperature and cardiovascular strain were not appreciably altered following 8-h exposure to 26°C but increased progressively in conditions above this threshold. These data support proposals for the establishment of a 26°C indoor temperature upper limit for protecting vulnerable occupants residing in temperate continental climates from indoor overheating. https://doi.org/10.1289/EHP13159

Introduction

In the summer of 2021, an unprecedented heat wave in the Pacific Northwest of North America broke national temperature records in Canada by 5°C and resulted in the deaths of $>\mathrm{1,000}$ mostly older adults in Canada and the United States.^1–3 With climate change, coming years will see record-breaking temperatures occur with increasing regularity.⁴ Improving adaptive capacity and reducing susceptibility to heat stress in vulnerable sectors of the population (e.g., older adults) is critical for addressing the growing health burden of extreme heat.⁵

Estimates from North American and European cities indicate that 54%–98% of heat-related fatalities occur in the home.^3,6–8 The World Health Organization (WHO) therefore considers strategies to prevent indoor overheating a key component of heat-health action plans.^9,10 Although numerous health and building standards agencies, including the WHO, the American Society of Heating, Refrigeration and Air-Conditioning Engineers (ASHRAE), and the United Kingdom Department of Health and Social Care (DHSC), recommend maintaining indoor temperature below 24–28°C, experimental support for this guidance is lacking.^9–11 Indoor temperature guidelines are instead based on indirect evidence. In Canada, for example, associations between outdoor temperature and increases in mortality and emergency service calls^11–14 have been cited to support proposals by the Chief Public Health Officer¹⁵ as well as regional health and housing authorities^16,17 to establish a 26°C indoor upper limit for residential buildings. In fact, the City of Vancouver in British Columbia, Canada, has mandated that by 2025, all new multiunit residential buildings be equipped with mechanical cooling sufficient to maintain indoor temperature below 26°C.¹⁷ However, the relation between outdoor and indoor temperatures is modified by neighborhood-level factors (e.g., paved surfaces, vegetation), dwelling characteristics (e.g., design, insulation), and occupant behavior (e.g., air-conditioning use, closing or opening windows),¹¹ making outdoor temperature a poor proxy for indoor heat stress and associated health impacts.^18–20

Studies to elucidate the effect of indoor overheating on the health of vulnerable persons are needed to support evidence-based guidance and heat-health action plans.^10,11 To this end, recent work has evaluated associations between indoor temperature and heat-related symptoms (e.g., thirst, nausea) in community-living older adults.²¹ However, subjective self-assessments underestimate heat strain in this population because of age-related declines in body temperature and thirst perception.^11,22,23 To support the development of evidence-based policy and guidance on safe indoor temperatures in continental climates, like those characterizing large swaths of North America and Europe, we conducted a laboratory-based randomized trial to quantify the effect of daylong indoor overheating on surrogate physiological outcomes linked with heat-related mortality and morbidity in older adults, including indices of thermal and cardiovascular strain, cardiac autonomic function, and dehydration.²²

Methods

This randomized crossover trial (ClinicalTrials.gov identifier: NCT04348630) was approved by the University of Ottawa Health Sciences and Science Research Ethics Board (H-11-18-1186). Testing was conducted at the Human and Environmental Physiology Research Unit (HEPRU). The HEPRU is located in Ottawa, Ontario, Canada, which has a warm summer continental climate (Köppen Climate Classification subtype Dfb) characterized by cold winters and warm, humid summers (average summer high of 24–27°C).²⁴

Participants

Men and women recruited from the Ottawa region via community outreach activities (e.g., mailing lists, personnel visits to community centers) were screened for enrollment between August 2019 and October 2021. Eligible participants were 65–80 y of age, nonsmoking, and not diagnosed with clinical health conditions (e.g., type 2 diabetes, heart disease) or taking medications known to impair body temperature or cardiovascular regulation (e.g., anticholinergics, beta-blockers). Exclusion criteria were physical restriction (e.g., because of disease), use of or changes in medication judged by the patient or investigators to make participation inadvisable, and cardiac abnormalities or symptoms identified in screening. Written and informed consent was obtained from all volunteers prior to participation. Participant age, smoking status, physical restrictions, and medication use were evaluated through participant self-report. Cardiac abnormalities and symptoms were assessed in a preliminary screening session.

Preliminary screening.

During the preliminary session, prospective participants were familiarized with all procedures and measurements and completed the Canadian Society for Exercise Physiology (CSEP) Get Active Questionnaire (GAQ) and the American Heart Association Pre-Participation Screening Questionnaire to assess their eligibility to participate. Habitual physical activity levels and the types of activities performed (e.g., walking, biking, aerobics classes) were evaluated using the GAQ and the Kohl Physical Activity Questionnaire.^25,26

After completing the forms, arterial blood pressures were taken in triplicate ($\sim 30$ s between measures) via manual auscultation of the brachial artery according to American Heart Association recommendations.²⁷ Descriptive anthropomorphic data were then collected. Body height and mass were determined via a stadiometer (model 2391; Detecto) and a digital weighing terminal (model CBU150X; Mettler Toledo Inc.), respectively, and used to calculate body mass index (BMI) and surface area.²⁸ Thereafter, participants were screened for cardiac abnormalities via 12-lead echocardiogram during a semirecumbent cycling stress test to volitional fatigue. The echocardiogram was administered by an American College of Sports Medicine and Canadian Society for Exercise Physiology–certified exercise physiologist (see “Acknowledgments” section).

For participants entering the trial following resumption of testing after COVID-19–related lockdowns (November 2020; $n=10$), initial screening was conducted over the phone and the Rose Angina Questionnaire was used to screen for cardiac symptoms in place of the stress test.²⁹ For these participants, resting blood pressure and anthropomorphic data were collected in the morning of the indoor exposure simulations (described in the “Experimental Procedures” section). To ensure consistency, reported anthropomorphic data for all participants represent an average of those measured at baseline across the four indoor exposure simulations.

Experimental Procedures

Enrolled participants completed four randomized indoor exposure simulations. No testing was performed between March and November 2020 because of the COVID-19 pandemic. Most participants completed the exposure simulations in the spring, summer, and early autumn ($n=10$). This was done for the sake of ecological relevance since individuals regularly exposed to high summer temperatures (indoors and outdoors) gain some degree of natural acclimatization.^30–32 However, six participants completed their trials during late fall or winter because of disturbances stemming from COVID-19–related testing restrictions. Exposure simulations for each participant were separated by a minimum of 48 h. Most participants completed their exposures on the same weekday over four consecutive weeks ($n=11$), resulting in a median 21 d between the first and final laboratory visit (min: 11 d, max: 35 d).

Preexposure instructions and standardizations.

Participants were instructed to avoid strenuous physical activity and alcohol for 24 h and to eat a light meal 2 h before each laboratory visit. Participants were also asked to consume a minimum of $500\mathrm{mL}$ of water the night before and morning of each session to ensure adequate hydration, defined as a urine specific gravity $\le 1.025$ (Reichert TS 400 total solids refractometer; Reichert).³³ Finally, during the baseline measurement period for each indoor exposure simulation, participants were asked a series of questions regarding heat exposure, physical activity and sleep habits, meals, and general well-being in the days leading up to the session. This information was recorded using a custom form and emailed to the participant $\sim 48$ h prior to the subsequent trial. The participant was asked to follow the activity pattern for the previous trials as closely as possible before each laboratory visit.

Indoor exposure simulations.

Each indoor exposure simulation commenced between 0700 h and 0900 h. On arrival at the laboratory, hydration status was confirmed via measurement of urine specific gravity, and the participant inserted a rectal thermocouple temperature probe. Following a measurement of nude body mass, participants dressed in light summer clothing (sandals, shorts, and a light shirt) and were transferred to a climate-controlled chamber regulated to 22°C and 45% relative humidity where they performed a specialized battery of cardiac autonomic response tests ($\sim 60–75\text{\hspace{0.17em}min}$; see the “Cardiac autonomic response tests” section).

Following the baseline cardiac response test battery, the participant rested in a seated posture (reclined slightly) in a $\sim 22°\mathrm{C}$ room adjacent to the climate chamber for $\sim 30\text{\hspace{0.17em}min}$. During this time, they were instrumented with digital skin temperature sensors and answered the series of questions on activity patterns in the days leading up to the trial (described in the “Preexposure instructions and standardizations” section). Baseline measurements were then recorded (see the “Outcome Assessment and Processing” section).

During the 30-min baseline period, the climate chamber was heated to the conditions for that day’s indoor exposure simulation. Each participant completed four 8-h exposures, the order of which was determined via simple randomization according to a key generated using commercially available spreadsheet software. The random allocation sequence was generated by R.D.M., who was also responsible for enrolling participants. Participants were blinded to the experimental conditions until instrumentation and baseline measurements were completed.

The ambient conditions for each exposure simulation were:

• 22°C, 45% relative humidity (heat index: 21°C) to simulate an air-conditioned room.²

• 26°C, 45% relative humidity (heat index: 26°C), corresponding to the upper temperature limits for indoor environments recommended by international, national, and local health agencies^9,10 including the new limits proposed for residential buildings in major Canadian cities (e.g., Toronto, Vancouver).^11,16,17 These conditions are also similar to those measured in air-conditioned residences in Detroit, Michigan, United States during hot summer days (hot summer continental climate, Köppen subtype Dfa; $\sim 750\mathrm{km}$ from Ottawa).²⁰

• 31°C, 45% relative humidity (heat index: 32°C) to simulate indoor temperatures during hot summer days. This temperature was chosen based on indoor temperatures measured in the homes of older adults during hot summer days in Detroit.²⁰ It is also comparable to conditions measured in non-air-conditioned homes in Montérégie, Quebec, Canada (warm summer continental climate, Köppen Subtype Dfb, $\sim 200\mathrm{km}$ from Ottawa)²¹ and in low-rent share housing (doss-houses) occupied by older adults in Seoul, South Korea (hot summer continental climate, Köppen subtype Dfw, $\sim \mathrm{10,500}\mathrm{km}$ from Ottawa).³⁴

• 36°C, 45% relative humidity (heat index: 41°C) to simulate peak indoor temperatures during extreme heat waves when ambient cooling (e.g., air conditioning) is unavailable or otherwise not used. This condition is consistent with indoor temperatures measured in non-air-conditioned houses and communal living spaces in Detroit²⁰ and Seoul³⁴ and during the 2021 heat dome event in the North American Pacific Northwest.² Although the Pacific Northwest is primarily characterized by a warm-summer Mediterranean climate (Köppen subtype Csb), the hottest areas during the heat dome event (e.g. Lytton, British Columbia, Canada, $\sim \mathrm{4,150}\mathrm{km}$ from Ottawa; maximum temperature: 49.6°C³⁵) have a warm-summer continental climate (Köppen Subtype Dfb).

Relative humidity was set to 45% for all trials based on outdoor conditions in recent heat waves in Ontario³⁶ and indoor humidity standards by the American Society of Heating and Air-Conditioning Engineers.³⁷ Airflow in the chamber was $\le 0.3$ m/s, the chamber had no external windows, and there were no substantial sources of radiation. The employed conditions were therefore reflective of summer heat exposure in a poorly ventilated room without windows (or shaded windows to prevent solar heating).

Each participant remained in the climate chamber for 8 h. Most of this time was spent resting in the seated position (reclined slightly), except during the seventh hour when the participant completed the cardiac autonomic response test battery a second time (see the “Cardiac autonomic response tests” section). Drinking water (tap, $\sim 15–20°\mathrm{C}$) was available ad libitum. Participants could eat a self-provided lunch ($\sim 300\mathrm{g}$, with low water content, e.g., peanut butter sandwich) between hours 1.5 and 3.5 of exposure. The mass of consumed water and food was measured every hour using a digital kitchen scale. Participants were able to briefly perform up to 5 min of light stretching at the end of each hour (after a measurement of body mass) and were provided with a portable urinal for washroom breaks. Those unwilling or unable to use the portable urinal were allowed to briefly leave the chamber and use the laboratory washroom (maximum 5 min).

Cardiac autonomic response tests.

Heat-related mortality and morbidity in heat-vulnerable populations are primarily attributable to adverse cardiovascular events secondary to increased cardiac strain and impaired blood pressure regulation.^22,38 Older adults exhibit impaired cardiovascular responses to heat stress,²² including attenuated autonomic blood pressure regulatory responses.^39–41 Autonomic function is also an important determinant of functional capacity to perform activities of daily living.⁴² To evaluate autonomic cardiovascular function, each participant performed a specialized battery of autonomic response tests prior to and during (hour 7) each exposure simulation. The tests included: a) analysis of heart rate variability,⁴³ b) Ewing’s battery (modified),^44,45 and c) analysis of integrated baroreflex sensitivity during squat-stand maneuvers.⁴⁶ The selected tests reasonably emulate autonomic challenges experienced during activities of daily living (e.g., standing from a lying position, squatting down to reach a lower shelf).

Prior to each test battery, participants were instrumented with an integrated 3-lead electrocardiogram (ECG) and noninvasive blood pressure monitor for the simultaneous measurement of beat-to-beat heart rate and arterial blood pressures.^47,48 The left arm was affixed to the chest using a medical sling, with the finger held at the approximate level of the left ventricle (fourth intercostal space). Participants were instructed to breathe normally and, without speaking, follow the instructions given by the test administrator (R.D.M.). A short practice period was allotted to ensure the participant understood the instructions and could perform each test correctly.

The cardiac response tests were performed as follows:

1. Resting heart rate variability—Heart rate variability was evaluated during supine rest on a medical bed located in the climate chamber. Participants were instructed to breathe for 8 min in time with a metronome set to 30 beats/min (15 breaths/min), to remove the influence of respiratory sinus rhythm.⁴³ The rate of breathing was monitored throughout (Go Direct Respiration Belt; Vernier). One minute of recovery was given prior to the Ewing’s battery.

2. Ewing’s battery—The modified Ewing’s battery included two clinically validated tests designed to assess autonomic control of heart rate and blood pressure.^44,45 Consistent with the optimal sequence identified by Stranieri et al.,⁴⁹ the battery was performed as follows:

   • First, participants performed 1 min of deep diaphragmatic breathing at a rate of 6 breaths/min. Participants were instructed to breathe to through a maximal inspiration and expiration (breathe to vital capacity), and the rate of breathing was monitored (Go Direct Respiration Belt; Vernier). The test was followed by 5 min of recovery.
   • Next was the lying-to-standing test (orthostatic response test). The test began with 1 min of supine rest. Thereafter, the participant was instructed to assume a standing posture beside the medical bed. The participant stood quietly for 3 min while blood pressure and heart rate were monitored. The participant was instructed before each test to sit down on the bed if they felt they could not maintain a standing position and was closely monitored for signs of orthostatic intolerance (e.g., pale face, exaggerated postural sway, delayed or no response to investigator instruction). Reported symptoms were recorded by the study investigators.
   • Heart rate variability measurements and the Ewing’s battery were performed again before the squat-stand procedures (after 10 min of supine recovery). After the second Ewing’s battery, the participant transitioned directly into the squat-stand procedure. Including the stand test and set up (e.g., explanation of the tests, adjustment of the sling and arm position, etc.), the participant was standing for at least 5 min before the start of the squat-stands.

3. Squat-stand maneuvers—Five minutes of squat-stands were performed at a rate of 6 cycles/min. Each cycle consisted of a squat, held for 5 s, followed by standing for 5 s. Participants were instructed to squat as low as was comfortable, breathe normally to avoid expiratory straining, and limit excessive forward flexion at the hip. Participants were verbally aided in timing for the first $\sim 2–3$ squat-stand cycles and self-timed thereafter by an electronic stopwatch located within their direct line of sight.

Outcome Measurements and Data Processing

Data were processed with R (version 4.2.0; R Development Core Team) and, for the cardiac autonomic response tests, LabChart physiological data analysis software (version 8.1.16.2019; ADInstruments). The analyst was blinded to participant ID, experimental condition, and, for the cardiac response tests, timepoint (pre- and end-exposure).

Body core and skin temperatures.

The primary purpose of guidance on maintaining a cool ambient environment is to help individuals maintain core temperature within safe limits.⁵⁰ Rectal temperature was monitored as an index of core temperature (primary outcome) using a general-purpose thermocouple temperature probe inserted $\sim 12\mathrm{cm}$ past the anal sphincter (Mon-a-therm General Purpose Temperature Probe; Mallinckrodt Medical Inc.). Data were collected in 15-s intervals using LabVIEW software (version 7.0; National Instruments). A temperature capsule (VitalSense ingestible capsule thermometer; Mini Mitter Company) was used to monitor gastrointestinal temperature in place of the rectal thermocouple in one participant who refused the rectal thermocouple.⁵¹

Skin temperature was assessed every minute using digital temperature sensors (DS1922L Thermochron; OnSolution Pty Ltd.) affixed to eight body regions, using double-sided adhesives and medical tape. Mean skin temperature was calculated according to the weightings recommended 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.⁵²

Blinded core and skin temperature data were manually cleaned, interpolated,⁵³ and converted to 15-min averages at the end of each hour of exposure. Plots of the raw, smoothed, and averaged data were exported and inspected. Baseline data were taken as the lowest 15-min average during the last 30-min of the baseline cardiac autonomic response test battery and the subsequent 30-min seated instrumentation and measurement period prior to the indoor exposure simulation. Mean body temperature was estimated as the weighted average of rectal (80%) and mean skin temperature (20%).⁵⁴ Data used for statistical analyses correspond to the 15-min average at baseline (hour 0) and end-exposure (hour 8).

Resting cardiovascular responses.

Heart rate and arterial systolic and diastolic blood pressures were taken as the average of triplicate measurements ($\sim 30$ s between measures) at baseline and at the end of each hour of the indoor exposure simulations via an independently validated (ISO 81060-2:2013 and the European Society of Hypertension)⁵⁵ brachial artery oscillometer (UM-211; A&D Medical). Consistent with American Heart Association guidance, participants were seated $>15$ min before the measurement, rested quietly with both feet flat on the floor, and the cuff was at the approximate level of the left ventricle.²⁷ Rate–pressure product, an index of myocardial strain,^56,57 was derived as heart rate multiplied by systolic pressure.

Cardiac autonomic response tests.

During the cardiac autonomic response tests, blood pressure was estimated from beat-to-beat recordings of the arterial pressure waveform measured at the left middle or ring finger (noninvasive blood pressure monitor; ADInstruments),⁴⁸ and heart rate was derived from RR interval measured via an integrated 3-lead ECG (FE231 Bio Amp; ADInstruments). Blood pressure and RR interval data were sampled at $200\text{\hspace{0.17em}Hz}$ using an analog-to-digital converter (PowerLab; ADInstruments) and stored offline for subsequent analysis (LabChart Pro, version 8; ADInstruments). Analysis was performed by a single analyst (N.V.K.) blinded to the participant ID, condition, and timepoint.

Time–domain heart rate variability estimates were calculated using dedicated software (LabChart HRV; ADInstruments). These included the standard deviation (SD) of normal-to-normal RR intervals (SDNN) and the root mean squared of successive differences (RMSSD). SDNN is an index of overall variability, whereas RMSSD is more reflective of short-term high-frequency fluctuations in heart rate, mediated primarily by the parasympathetic nervous system.⁴³

Ewing’s battery tests were analyzed according to the guidance by Spallone et al.⁴⁵:

• Deep breathing heart rate difference: Data collected during deep breathing were analyzed as the difference between the average of the three highest heart rates recorded during inspiration and three lowest heart rates recorded during expiration. A reduction in the deep breathing heart rate difference reflects attenuated parasympathetic input to the respiratory sinus arrhythmia.⁴⁵

• Cardiac response to standing: The 30:15 ratio was calculated from data collected during the lying-to-standing test as the longest RR interval (lowest heart rate) measured between the 25th and 35th heartbeat after standing divided by the shortest RR interval (highest heart rate) between the 10th and 20th beat after standing. A reduction in the 30:15 ratio primarily reflects attenuated parasympathetic control of heart rate in response to perturbations in arterial blood pressure (via the cardiac baroreflex).⁴⁵

• Systolic response to standing: Systolic arterial blood pressure was measured $\sim 1$ min and $\sim 2$ mins after participants assumed a standing position (average of data recorded from 45 to 75 s and 115 to 135 s post stand, respectively). The lying–standing change in systolic pressure was taken as the lowest of these two values minus systolic pressure prior to standing. The absolute value of systolic blood pressure in the standing position was also evaluated, as was the lying–standing change in heart rate. Exacerbated reductions in systolic blood pressure during standing reflect attenuated effectiveness of sympathetic blood pressure regulation.⁴⁵

Cardiac baroreflex sensitivity (baroreflex gain, the change in RR interval in milliseconds for each mm Hg change in systolic pressure) was evaluated via transfer function analysis of data collected during the 5-min squat-stand procedure as described by Zhang et al.⁴⁶ Time series of RR interval and systolic blood pressure were linearly interpolated and resampled at $4\text{\hspace{0.17em}Hz}$. Data were detrended and subdivided into 256-point segments with 50% overlap (Welch procedure). Fast Fourier transformation was implemented over each Hanning-windowed segment (30 s overlap), and transfer function gain was derived from the cross spectrum.^58,59 Transfer function gain provides an estimate of baroreflex sensitivity around the operating point of the baroreflex curve^59–61; a reduction in this variable reflects attenuated vagal modulation of heart rate in response to perturbations in arterial pressure.^62,63

Body fluid status.

Nude body mass was measured to an accuracy of $10\mathrm{g}$ using a high-performance digital weighing terminal before and after each exposure simulation (model CBU150X; Mettler Toledo Inc.). Fluid losses through sweat and urine were quantified as the change in nude body mass. Net fluid loss was presented as a percentage change from baseline values, corrected for consumed food measured to $1\mathrm{g}$ using a digital kitchen scale (MK200; Mafiti). Fluid loss was not corrected for fluid consumption or urination to provide an index of total dehydration. Average whole-body sweat rate was estimated as the average hourly change in body mass corrected for urination and food and fluid consumption and presented relative to body surface area. Average hourly fluid consumption was also calculated.

Venous blood was collected for determination of plasma volume changes. Samples drawn from the antecubital vein at baseline and hour 8 were transferred directly into plasma Vacutainer tubes ($5.4\mathrm{mg}$ K2EDTA; BD), mixed via inversion, and analyzed for hematological parameters in duplicate (Ac·T diff2; Beckman Coulter). The change in plasma volume from baseline to the end of exposure was estimated from the average values for hemoglobin and hematocrit.⁶⁴ Participants rested in the seated posture (slightly reclined) for at least 20 min prior to each blood sample to ensure that plasma volume was relatively stable.⁶⁵ However, because the baseline blood draw was procured at the end of the 30-min baseline measurement period, participants were typically seated for 30 min before sample collection. Likewise, participants were generally seated for 45–60 min before the end-exposure blood draw.

Statistical Analyses

Statistical analyses were conducted using R (version 4.2.0), including packages tidyverse,¹⁰² emmeans (R package, version 1.8.9), lme4,¹⁰³ marginaleffects (R package, version 0.16.0) performance,¹⁰⁴ ggplot2 and patchwork.⁶⁶

Sample size estimations.

An a priori power analysis indicated that a sample size of 16 was required to detect a $>0.3°\mathrm{C}$ difference in body core temperature between the 22°C simulation and any one of the other three conditions with 80% statistical power, after adjusting for multiple comparisons (Bonferroni). The effect size of interest, 0.3°C, was determined from the typical day-to-day variation in core temperature⁶⁷ and the threshold for clinical significance proposed in recent analyses assessing common personal cooling strategies.^68,69 The SD (0.3°C) was determined from prior studies exposing older adults to extreme heat for 2–3 h.⁷⁰

Primary and secondary outcomes.

The effect of indoor heat stress on core temperature at the end of the indoor exposure simulations was evaluated using a linear mixed effects model. Experimental condition was modeled as a categorical predictor (22°C, 26°C, 31°C, and 36°C) with basal core temperature included as a continuous covariate. A random intercept was modeled for each participant (participant ID) to account for repeated measurements because of the crossover design. Homoskedasticity and normality of residuals (fixed and random effects) were evaluated via inspection of diagnostic plots.

Preplanned contrasts were conducted to evaluate the difference in core temperature between the 22°C indoor exposure simulation and the other three conditions (26°C, 31°C, and 36°C). We also evaluated whether the between-condition differences exceeded the 0.3°C threshold for clinical significance. In addition, we performed a post hoc linear trend analysis to estimate the slope of the relation between core temperature and ambient temperature.^71–73 A two-sided $p<0.050$ was considered statistically significant. No adjustment for multiplicity was made because core temperature was the primary outcome.⁷⁴ Data for each condition were presented as means and SDs of raw data. Between-condition comparisons and slope of the ambient temperature core temperature relation were presented as means and 95% confidence intervals (CIs; lower limit, upper limit).

Secondary outcomes.

The analyses of skin temperature, mean body temperature, heart rate, systolic and diastolic pressures, rate-pressure product, and baroreflex sensitivity were identical to that of core temperature. Similar models were used to evaluate the percent change in body mass and plasma volume from baseline values as well as average hourly sweat rate, except pretrial body mass was modeled as a covariate for the former variable and no baseline covariate was included for the latter three.

Resting heart rate variability, deep breathing heart rate difference, and lying-to-standing cardiac (30:15 ratio) and systolic blood pressure (lying–standing change and standing systolic pressure) responses were also analyzed with linear mixed effects models. The fixed effects were experimental condition and baseline values, as described above. However, because these tests were performed twice during each cardiac autonomic response test battery, a nested random intercept was modeled for each participant, with test number nested within participant ID.

For all secondary outcomes, we performed the same post hoc contrasts as those performed in the primary analysis but employed the Holm-Bonferroni correction to control the family-wise error rate. All secondary outcomes corresponding to each grouping of variables were considered a family of comparisons (body temperatures, resting cardiovascular responses, cardiac autonomic function, fluid regulation variables). This approach meant, for example, that in the evaluation of the effect of indoor heat stress on resting cardiovascular strain, we corrected $p$-values for heart rate, systolic and diastolic blood pressures, and rate–pressure product simultaneously (4 contrasts $\times$ 4 variables for 16 comparisons total).

Deviations from trial registration.

For participants tested prior to the COVID-19 pandemic, a Valsalva maneuver was performed during the cardiac autonomic test battery. This test was removed after resumption of testing because of concerns over aerosol generation. We also did not originally plan to conduct the linear trend analysis but decided this analysis was an important addition to the study because it allowed us to define the study outcomes in relation to ambient temperature while avoiding many between-condition comparisons. Further, we did not originally specify multiplicity corrections. However, given the large number of secondary outcomes, we decided to adjust $p$-values using the Holm-Bonferroni procedure at the hypothesis level, as described above. Although this approach is conservative, we felt it was appropriate to improve the robustness of our inferences.

Missing data.

Two participants displayed normal rhythm but interspersed with 2–4 heart beats in more rapid succession, causing inflated heart rate variability (e.g., RMSSD $\sim 8$ SD from the mean in the 22°C condition). Data from these participants were excluded, and heart rate variability, cardiac autonomic responses to standing and the inspiration/expiration heart rate difference (tests requiring analysis of single RR intervals) were analyzed with a reduced sample size (all $n=14$). We did not impute missing data.

Sensitivity analyses.

Heart rate and indices of cardiac autonomic modulation are naturally inversely related.⁷⁵ To evaluate the potential that our findings of reduced indices of cardiac autonomic function with increasing simulated indoor temperature were due to an increase in resting heart rate, we reperformed our analysis of heart rate variability indices (SDNN and RMSSD) and cardiac autonomic responses to standing (30:15 ratio) with an additional covariate for average resting RR interval measured just before the cardiac autonomic response tests.

As described above, in post hoc analyses we evaluated the linear trend between ambient conditions and physiological responses. We opted to test for a linear trend rather than a polynomial trend because of the number of conditions and for the sake of simplicity. In sensitivity analyses, we evaluated the robustness of this assumption by modeling quadratic and cubic terms for primary and secondary variables [using R package emmeans (version 1.8.9)].

Finally, body temperature exhibits a circadian rhythm such that core temperature is elevated in the afternoon/evening in comparison with temperature in the early morning,⁷⁶ though the magnitude of this variation is likely blunted with aging.⁷⁷ Although we standardized participant arrival to the laboratory, in sensitivity analysis we evaluated whether small variations in the time exposure commenced (due, for example, to delays during instrumentation or baseline measurements) influenced our analysis of core temperature. This evaluation was accomplished by repeating the primary analysis but with the time participants entered the climate chamber included as a covariate.

Results

Participants

The flow of participants through the study is presented in Figure 1. Thirty-five individuals were screened for eligibility. Of these, 19 were excluded because they were not eligible for the study ($n=1$), they did not wish to participate after explanation of the protocol ($n=6$), or they ceased contact with the study investigators ($n=12$). Consequently, 16 adults 66–78 y of age participated in each of the four 8-h indoor exposure simulations, for a total of 64 participant exposures. Participant baseline characteristics are reported in Table 1. The median age was 72 y (interquartile range: 70–73 y), and BMI was $24.6{\mathrm{kg}/\mathrm{m}}^2$ ($22.1–27.0{\mathrm{kg}/\mathrm{m}}^2$). Six participants were women (37.5%), all of whom were postmenopausal; none had had a hysterectomy or were undergoing hormone replacement therapy. Reported medications included statins ($n=2$), topical creams/ointments ($n=1$), and medications treating gastrointestinal reflux ($n=2$), osteoporosis ($n=2$), essential tremor ($n=1$), benign prostatic hyperplasia ($n=1$), and glaucoma ($n=1$). None have been shown or suggested to affect physiological responses to heat.²²

Figure 1.

Flow of participants through the study. A CONSORT diagram showing participant enrollment, randomization, and analysis. Of the 35 individuals evaluated for inclusion, 16 adults aged 66–78 y were enrolled. Each completed four, randomized 8-h indoor exposure simulations covering a range of ambient conditions experienced indoors during hot weather and heat waves in continental climates (total of 64 participant-exposures).

Table 1.

Baseline characteristics of participants who completed four randomized, 8-h exposures to simulated indoor overheating.

	All participants ($n=16$)
Participant characteristics
Age [median (IQR) (y)]	72 (70–73)
Sex [$n$ (%)]
Women	6 (37.5%)
Men	10 (62.5%)
Height [median (IQR) (cm)]	171 (166–174)
Body mass [median (IQR) (kg)]	71.9 (60.7–81.5)
Body mass index [median (IQR) (${\mathrm{kg}/\mathrm{m}}^2$)]a	24.6 (22.1–27.0)
Body surface area [median (IQR) (${\mathrm{m}}^2$)]b	1.8 (1.7–2.0)
Self-reported physical activity [median (IQR) (min/week)]c	155 (105–300)
Taking prescription medications [$n$ (%)]d
No	7 (44%)
Yes	9 (56%)
Smoking statuse
Never	6 (36%)
Past	10 (63%)
Baseline physiological outcomesf
Core temperature [mean (SD) (°C)]	36.7 (0.2)
Heart rate [mean (SD) (beats/min)]	60 (7)
Systolic arterial blood pressure [mean (SD) (mm Hg)]	135 (19)
Diastolic arterial blood pressure [mean (SD) (mm Hg)]	78 (9)
Rate–pressure product [mean (SD) (mm Hg·beats/min)]	8,036 (1,517)
Heart rate variabilityg
SDNN [mean (SD) (ms)]	21.8 (4.9)
RMSSD [mean (SD) (ms)]	14.4 (4.1)
Lying-to-standing testh
Cardiac response (30:15 ratio) [mean (SD) (unitless)]	1.20 (0.10)
Blood pressure response ($\mathrm{\Delta }\text{SBP}$) [mean (SD) (mm Hg)]	4 (14)
Standing systolic pressure [mean (SD) (mm Hg)]	123 (18)
Baroreflex sensitivity [mean (SD) (ms/mm Hg)]i	10.1 (0.6)

Note: Values are median and IQR or No. participants (%) for physical characteristics and mean and SD for baseline physiological outcomes. IQR, interquartile range; RMSSD, root mean square of successive differences in normal-to-normal RR intervals; SBP, systolic blood pressure; SD, standard deviation; SDNN, standard deviation of successive normal-to-normal RR intervals.

^a Body mass index calculated as weight in kilograms divided by the square of height in meters.

^b Body surface area calculated according to the Du Bois equation.²⁸

^c Participant self-reported physical activity level determined via the Canadian Society for Exercise Physiology Get Active Questionnaire (GAQ).²⁵ The most commonly reported physical activities were walking ($n=11$) or light biking, jogging, or swimming ($n=8$).²⁶

^d Medications included statins ($n=2$), topical creams/ointments ($n=1$), and medications treating gastrointestinal reflux ($n=2$), osteoporosis ($n=2$), essential tremor ($n=1$), prostatic hyperplasia ($n=1$), and glaucoma ($n=1$).

^e Smoking status determined via participant self-report. Prospective participants were excluded if they were currently smoking. All past smokers quit $\ge 15$ y prior to participation.

^f Baseline outcome data for each participant was calculated as an average of data collected during the baseline measurement period of each of the four indoor exposure simulations (22°C).

^g Resting heart rate variability was measured during 5 min of paced breathing (15 breaths/min). Data for heart rate variability indices reported for $n=14$.

^h Cardiac autonomic function was evaluated during an orthostatic test consisting of standing from a supine position.⁴⁴ The ratio of RR intervals at the 30th and 15th heartbeats ($\pm 5$ beats) after standing (30:15 ratio) is an index of parasympathetically mediated modulation of heart rate to a fall in blood pressure. The lying–standing difference in SBP ($\mathrm{\Delta }\text{SBP}$) is an index of sympathetically mediated augmentation of cardiac output to restore arterial blood pressure. Data for cardiac responses to standing reported for $n=14$.

ⁱ Baroreflex sensitivity reflects the parasympathetically mediated modulation of heart rate determined during oscillations in arterial blood pressure elicited by 5 min of repeated squatting (6 squats per min).⁴⁶

Core and Skin Temperatures

Core temperature during the 8-h indoor exposure simulations is shown in Figure 2. Core temperature increased 0.07°C (95% CI: 0.06, 0.09; $p<0.001$) per 1°C rise in ambient temperature and was statistically significantly elevated in 26°C ($+0.2°\mathrm{C}$, 95% CI: 0.0, 0.3; $p=0.011$), 31°C ($+0.7°\mathrm{C}$, 95% CI: 0.5, 0.8; $p<0.001$), and 36°C ($+0.9°\mathrm{C}$, 95% CI: 0.8, 1.1; $p<0.001$) in comparison with the 22°C control exposure. The between-condition difference in core temperature exceeded the threshold for clinical significance (0.3°C) in 31°C ($p<0.001$) and 36°C ($p<0.001$) but not 26°C ($p=0.170$).

Figure 2.

Mean changes in core temperature (rectal) in adults aged 66–78 y during 8-h exposures to conditions experienced indoors during hot weather and heat waves in continental climates (A) and statistical analysis of between-condition differences (B). Graphical data are presented as mean and SD (error bars) and individual points (women are shown in open circles, $n=6$; men in filled circles, $n=10$). Differences between the 26°C–36°C exposure simulations and the 22°C control condition are presented as a mean and 95% CI (lower limit to upper limit), adjusted for baseline values. Post hoc linear trend analysis indicates the mean (95% CI) increase in core temperature for each 1°C increase in ambient temperature (${\mathrm{T}}_{\text{amb}}$). The data used to produce this figure are summarized in Tables S1 and S2. Note: CI, confidence interval; RH, relative humidity; SD, standard deviation.

Mean skin and body temperatures are shown in Table 2. Both variables were elevated during exposure to 26°C, 31°C, and 36°C in comparison with the 22°C control condition. For every 1°C rise in air temperature, mean skin temperature rose 0.31°C (95% CI: 0.28, 0.34; $p<0.001$), whereas mean body temperature increased 0.12°C (95% CI: 0.10, 0.13; $p<0.001$).

Table 2.

Mean skin and body temperature in 16 older adults at the end of randomized 8-h exposures to a range of conditions simulating those experienced indoors during hot weather and heat waves (26–36°C).

Variable	Ambient temperature ($\text{mean}\pm \text{SD}$)				Difference from 22°C [mean (95% CI)]a			Linear trend ($\mathrm{\Delta }/°\mathrm{C}$ ${\mathrm{T}}_{\text{ambient}}$)a
	22°C	26°C	31°C	36°C	26°C	31°C	36°C
Mean skin temperature (°C)	31.7 (0.9)	32.9 (0.4)	34.9 (0.3)	35.9 (0.5)	1.2 (0.8, 1.6) $p<0.001$	3.2 (2.8, 3.5) $p<0.001$	4.2 (3.8, 4.6) $p<0.001$	0.31 (0.28, 0.34) $p<0.001$
Mean body temperature (°C)	35.9 (0.4)	36.3 (0.3)	37.0 (0.2)	37.4 (0.3)	0.4 (0.2, 0.6) $p<0.001$	1.2 (1.0, 1.3) $p<0.001$	1.6 (1.4, 1.7) $p<0.001$	0.12 (0.10, 0.13) $p<0.001$

Note: Values are mean and SD or mean and 95% CI of data measured at the end of the 8-hour indoor exposure simulations or during a specialized battery of cardiac autonomic response tests performed during the seventh hour of exposure. Data reported for $n=16$ unless otherwise indicated. CI, confidence interval; SD standard deviation.

^a Estimated marginal mean difference and linear trend derived from a linear mixed-effects model adjusted for baseline values. $p$-Values are adjusted for multiplicity using the Holm-Bonferroni technique. All variables were considered a family of comparisons and adjusted simultaneously.

Cardiovascular Strain and Autonomic Function

Indices of cardiovascular strain and cardiac autonomic function are reported in Table 3. There were no statistically significant differences between the 22°C and 26°C exposures. Heart rate was significantly higher, whereas seated arterial blood pressures, heart rate variability, cardiac responsiveness to standing (30:15 ratio), and standing systolic blood pressure were reduced in both 31°C and 36°C relative to 22°C. The lying–standing reduction in systolic blood pressure increased with ambient temperature, though no between-condition differences were detected. Three participants experienced symptoms of orthostatic intolerance during the lying-to-standing tests in the 36°C exposure simulation (Table 4).

Table 3.

Indices of cardiovascular strain and autonomic function in 16 older adults at the end of randomized 8-hour exposures to a range of conditions simulating those experienced indoors during hot weather and heat waves (26–36°C).

Variable	Ambient temperature ($\text{mean}\pm \text{SD}$)				Difference from 22°C [mean (95% CI)]a			Linear trend ($\mathrm{\Delta }/°\mathrm{C}$ ${\mathrm{T}}_{\text{ambient}}$)a
	22°C	26°C	31°C	36°C	26°C	31°C	36°C
Resting cardiovascular responses
Heart rate (beats/min)	60 (8)	62 (7)	69 (9)	74 (10)	2 ($-1$, 5) $p=0.676$	9 (6, 12) $p<0.001$	13 (10, 16) $p<0.001$	1.0 (0.8, 1.2) $p<0.001$
Systolic pressure (mm Hg)	142 (26)	132 (20)	117 (14)	114 (15)	$-12$ ($-21$, $-3$) $p=0.086$	$-26$ ($-35$, $-17$) $p<0.001$	$-29$ ($-38$, $-20$) $p<0.001$	$-2.1$ ($-2.7$, $-1.5$) $p<0.001$
Diastolic pressure (mm Hg)	81 (9)	78 (9)	73 (8)	70 (9)	$-3$ ($-6$, 0) $p=0.393$	$-9$ ($-12$, $-6$) $p<0.001$	$-12$ ($-15$, $-8$) $p<0.001$	$-0.9$ ($-1.1$, $-0.6$) $p<0.001$
Rate–pressure product (mm Hg × beats/min)	8,488 (1,642)	8,158 (1,554)	8,115 (1,478)	8,424 (1,509)	$-377$ ($-873$, 119) $p=0.546$	$-466$ ($-963$, 30) $p=0.393$	$-298$ ($-797$, 200) $p=0.676$	$-20$ ($-53$, 14) $p=0.676$
Cardiac autonomic function
SDNN (ms)b	30.6 (11.8)	27.3 (9.1)	21.2 (9.5)	16.5 (5.8)	$-3.8$ ($-7.7$, 0.10) $p=0.592$	$-9.4$ ($-13.3$, $-5.6$) $p<0.001$	$-14.1$ ($-18.0$, $-10.3$) $p<0.001$	$-1.0$ ($-1.3$, $-0.8$) $p<0.001$
RMSSD (ms)b	20.4 (8.5)	17.3 (6.9)	13.5 (5.6)	9.3 (5.9)	$-3.2$ ($-6.4$, 0.0) $p=0.526$	$-6.9$ ($-10.0$, $-3.7$) $p<0.001$	$-11.1$ ($-14.2$, $-7.9$) $p<0.001$	$-0.8$ ($-1.0$, $-0.6$) $p<0.001$
Exp/Insp heart rate difference (beats/min)c	8 (5)	7 (6)	7 (4)	6 (5)	$-1$ ($-2$, 1) $p>0.999$	$-1$ ($-2$, 1) $p>0.999$	$-2$ ($-3$, 0) $p=0.389$	$-0.1$ ($-0.2$, 0.0) $p=0.509$
Lying-to-standing 30:15 ratio (unitless)d	1.22 (0.10)	1.17 (0.10)	1.09 (0.08)	1.07 (0.10)	$-0.04$ ($-0.08$, 0.01) $p>0.999$	$-0.12$ ($-0.16$, $-0.07$) $p<0.001$	$-0.15$ ($-0.19$, $-0.10$) $p<0.001$	$-0.01$ ($-0.01$, $-0.01$) $p<0.001$
Lying-to-standing $\mathrm{\Delta }\text{SBP}$ (mm Hg)e	1 (13)	6 (14)	0 (16)	$-4$ (14)	5 ($-1$, 10) $p=0.326$	0 ($-5$, 4) $p>0.999$	$-5$ ($-10$, $-1$) $p=0.341$	$-0.5$ ($-0.8$, $-0.2$) $p=0.032$f
Standing systolic blood pressure (mm Hg)d	128 (29)	136 (19)	106 (20)	96 (18)	6 ($-3$, 15) $p>0.999$	$-22$ ($-31$, $-13$) $p<0.001$	$-36$ ($-45$, $-26$) $p<0.001$	$-2.9$ ($-3.5$, $-2.3$) $p<0.001$f
Baroreflex sensitivity (ms/mm Hg)e	9.8 (0.9)	9.6 (1.3)	10.0 (1.3)	9.8 (1.0)	$-0.2$ ($-1.1$, 0.6) $p>0.999$	0.1 ($-0.7$, 0.9) $p>0.999$	0.0 ($-0.8$, 0.8) $p>0.999$	0.01 ($-0.05$, 0.06) $p>0.999$

Note: Values are mean and SD or mean and 95% CI of data measured at the end of the 8-h indoor exposure simulations or during a specialized battery of cardiac autonomic response tests performed during the seventh hour of exposure. Data reported for $n=16$ unless otherwise indicated. CI, confidence interval; Exp, expiration; Insp, inspiration; min, minute; RMSSD, root mean square of successive normal-to-normal R–R intervals; SBP, systolic blood pressure; SD, standard deviation; SDNN, standard deviation of successive normal-to-normal R-R intervals.

^aEstimated marginal mean difference and linear trend derived from a linear mixed-effects model adjusted for baseline values. $p$-Values are adjusted for multiplicity using the Holm-Bonferroni technique. All variables under each subheading were considered a family of comparisons and adjusted simultaneously.

^bResting heart rate variability was measured during 5 min of paced breathing (15 breaths/min). Data for heart rate variability indices reported for $n=14$. Data were not imputed.

^cExpiration–inspiration heart rate difference was evaluated as the difference in the average of the 3 maximum heart rates and 3 minimum heart rates measured during 1 min of breathing to vital capacity at 6 breaths per min. Reported for $n=14$. Data were not imputed.

^dCardiac autonomic function was evaluated during an orthostatic test consisting of standing from a supine position.⁴⁴ The ratio of RR intervals at the 30th and 15th heartbeats ($\pm 5$ beats) after standing (30:15 ratio) is an index of parasympathetically mediated modulation of heart rate to a fall in blood pressure. The lying–standing difference in systolic blood pressure ($\mathrm{\Delta }\text{SBP}$) is an index of sympathetically mediated augmentation of cardiac output to restore arterial blood pressure. Data for cardiac responses to standing reported for $n=14$. Data were not imputed.

^eBaroreflex sensitivity reflects the parasympathetically mediated modulation of heart rate determined during oscillations in arterial blood pressure elicited by 5 min of repeated squatting (6 squats per min).⁴⁶

^fIn sensitivity analyses, we ran a post hoc polynomial contrast for variables for which a statistically linear trend was detected in primary analysis. Although the linear term remained significant for all variables ($p<0.001$), inclusion of a quadratic and/or cubic term improved fit for the lying-to-standing $\mathrm{\Delta }\text{SBP}$ (quadratic: $p=0.003$, cubic: $p=0.001$) and standing SBP (quadratic: $p=0.002$, cubic: $p=0.111$).

Table 4.

Incidence of self-reported symptoms of orthostatic intolerance during cardiac autonomic response test battery performed in 16 older adults at the end of randomized 8-h exposures to a range of conditions simulating those experienced indoors during hot weather and heat waves (22–36°C).

Variable	Ambient temperature [no (% of tests)]
	22°C	26°C	31°C	36°C
Number of participants reporting symptoms [No (% of participants)]	0 (0%)	0 (0%)	0 (0%)	3 (19%)
Participant felt lightheaded [No (% of tests)]	0 (0%)	0 (0%)	0 (0%)	2 (6%)
Test terminated early due to inability to remain standing [No (% of tests)]	0 (0%)	0 (0%)	0 (0%)	2 (6%)
Total instances of self-reported symptoms of orthostatic intolerance [No (% of tests)]	0 (0%)	0 (0%)	0 (0%)	4 (13%)

Note: Data reported as No. participants or tests (%). The four instances of self-reported symptoms of orthostatic intolerance occurred among $n=3$ participants. One participant reported feeling lightheaded on their first lying-to-standing test, and then the second test was terminated early because of participant inability to remain standing.

Body Fluid Regulation

Indices of body fluid regulation are shown in Table 5. No statistically significant differences in cumulative fluid consumption, fluid loss, or plasma volume responses were observed between 22°C and 26°C. Fluid consumption increased with ambient temperature and was significantly greater in 36°C in comparison with 22°C. No between-condition differences in fluid loss were observed. Plasma volume was significantly elevated in 31°C and 36°C in comparison with 22°C.

Table 5.

Indices of fluid consumption and hydration status in 16 older adults at the end of randomized 8-h exposures to a range of conditions simulating those experienced indoors during hot weather and heat waves (26–36°C).

Variable	Ambient temperature ($\text{mean}\pm \text{SD}$)				Difference from 22°C [mean (95% CI)]a			Linear trend ($\mathrm{\Delta }/°\mathrm{C}$ ${\mathrm{T}}_{\text{ambient}}$)a
	22°C	26°C	31°C	36°C	26°C	31°C	36°C
Fluid consumption (L)	0.48 (0.34)	0.54 (0.44)	0.76 (0.48)	1.22 (0.75)	0.06 ($-0.17$, 0.29) $p>0.999$	0.28 (0.05, 0.50) $p=0.148$	0.74 (0.51, 0.96) $p<0.001$	0.05 (0.04, 0.07) $p<0.001$
Net fluid loss (% body weight)b	$-0.6$ (0.8)	$-0.7$ (0.5)	$-0.5$ (0.7)	$-0.5$ (1.2)	$-0.1$ ($-0.4$, 0.3) $p>0.999$	0.1 ($-0.2$, 0.5) $p>0.999$	0.1 ($-0.2$, 0.5) $p>0.999$	0.01 ($-0.01$, 0.04) $p>0.999$
Average sweat rate (${\mathrm{g}/\mathrm{h}/\mathrm{m}}^2$)c	20 (30)	26 (15)	31 (15)	80 (24)	7 ($-6$, 20) $p>0.999$	22 (8, 35) $p=0.011$	60 (47, 73) $p<0.001$	4 (3, 5) $p<0.001$
Change in plasma volume (% baseline)	$-2.4$ (2.5)	$-0.5$ (3.2)	2.5 (2.5)	1.1 (4.1)	1.8 (0.2, 3.5) $p=0.214$	4.8 (3.2, 6.5) $p<0.001$	3.5 (1.8, 5.1) $p=0.001$	0.28 (0.17, 0.39) $p<0.001$

Note: Values are mean and SD or mean and 95% CI of data measured during the 8-h indoor exposure simulations ($n=16$). —, no data; CI, confidence interval; SD, standard deviation.

^a Estimated marginal mean difference and linear trend derived from a linear mixed-effects model adjusted for baseline values. $p$-Values are adjusted for multiplicity using the Holm-Bonferroni technique. All indices of body fluid status were considered a family of comparisons and adjusted simultaneously.

^b Net fluid loss was calculated as the percentage change in nude body weight from baseline to the end of exposure and was not corrected for fluid consumption or urination.

^c Average hourly sweat rate calculated from the change in body mass corrected for urination and food and fluid consumption and presented relative to body surface area.

Sensitivity Analyses

Although adjustment for heart rate slightly reduced the effect of indoor heat stress on heart rate variability indices, interpretation of the data was unchanged from the original analysis (Figure S1). Neither SDNN or RMSSD were statistically significantly different between the 22°C and 26°C exposures ($p\ge 0.093$) but were reduced in 31°C ($p\le 0.026$) and 36°C ($p\le 0.003$) in comparison with 22°C. Effect estimates for the cardiac response to standing were slightly increased after correction for resting heart rate. Consistent with the primary analysis, 30:15 ratio was reduced in 31°C ($p<0.001$) and 36°C ($p<0.001$) but not 26°C ($p=0.097$) in comparison with the 22°C control condition.

When we reperformed our post hoc trend analyses for core temperature allowing for a nonlinear association with ambient temperature, we found that the linear term was statistically significant ($p<0.001$) but quadratic and cubic terms were not ($p\ge 0.058$). Likewise, we did not observe statistically significant quadratic or cubic associations between ambient temperature and resting heart rate, systolic, pressure, diastolic pressure, SDNN, RMSSD, or the cardiac autonomic responses to standing ($p\ge 0.139$), whereas the linear terms were all statistically significant ($p<0.001$). We did, however, observe significant linear ($p<0.001$), quadratic ($p=0.003$) and cubic terms ($p<0.001$) for standing systolic blood pressure and significant linear ($p=0.004$) and quadratic terms ($p=0.002$) for the lying–standing difference in systolic pressure (cubic: $p=0.111$). This difference was primarily driven by an elevated systolic pressure responses to standing in the 26°C in comparison with the 22°C condition; however, data still appear generally linear (Figure S2).

Finally, including the time participants entered the climate chamber in the analysis of core temperature did not influence the primary findings (effect of exposure start time: $p=0.978$; see Table S1 for more details).

Discussion

We evaluated surrogate physiological outcomes linked with heat-related mortality and morbidity in older adults exposed to simulated indoor overheating. Core temperature rose in proportion to ambient temperature, but differences between 22°C and 26°C were small and unlikely to be clinically meaningful. Exposure to temperatures exceeding 26°C increased heart rate, reduced arterial blood pressure, and blunted cardiac autonomic responses to postural changes consistent with activities of daily living.

Our findings have important implications for the development of evidence-based guidance on safe indoor temperatures. Canada’s Chief Public Health Officer¹⁵ and regional health and housing authorities, including Toronto Public Health (Ontario, Canada)¹⁶ and the City of Vancouver (British Columbia, Canada)¹⁷ have recently endorsed a 26°C upper temperature limit for residential buildings. These recommendations are based on population-level studies showing elevated heat-related mortality and emergency service calls in Canadian cities when daily mean outdoor temperature exceeded 26°C.^11–14 Here we demonstrate that maintaining indoor temperature below this threshold is effective for limiting thermal and cardiovascular strain and dehydration in community-living older adults. Consistent with these findings, Teyton et al.²¹ reported progressive increases in the incidence of heat-related symptoms (e.g., thirst, reduced urine output, headache, nausea) with elevations in indoor ambient temperature in older persons living in Montérégie, Quebec, Canada (warm summer continental climate, Köppen Subtype Dfb, $\sim 200\mathrm{km}$ from Ottawa). These reports, when considered alongside the current data, support a 26°C indoor temperature limit for reducing physiological strain and discomfort and protecting health and well-being for individuals living in continental climates,¹⁵ like those characterizing large areas of North America and Europe.

There is considerable global heterogeneity in the association between ambient temperature and the risk of heat-related mortality.⁷⁸ Areas with higher mean summer temperatures experience smaller elevations in heat-related mortality during heat waves (defined based on local temperature thresholds)⁷⁹ due, in part, to physiological acclimatization.^30–32 Our study is therefore most applicable to public health agencies in Ontario and surrounding areas (e.g., Quebec, Canada, and the Midwest and Northeast regions of the United States). There is, however, indirect evidence that our findings may be generalizable to other geographical locations with continental and/or temperate climates. For example, our data are consistent with those of Kim et al.,³⁴ who observed that aural temperature rose 0.2°C and systolic pressure fell $3\mathrm{mm}$ Hg per 1°C increase in home temperature in older adults ($\sim 70$ y of age) living in communal housing in Seoul, South Korea (hot summer continental climate, Köppen Subtype Dfw). The mean indoor temperature (31.5°C)³⁴ exceeded the yearly 99th percentile outdoor temperature in Seoul, which has been associated with a 4% greater risk of mortality in comparison with 25°C,⁸⁰ nearly identical to the increase in heat-related mortality between 26°C and 31°C in Ontario.^12,16

Exposure to temperatures exceeding 26°C caused progressive elevations in core temperature and heart rate and marked reductions in arterial blood pressure and cardiac autonomic function. The latter findings are consistent with Ferreira et al.,⁸¹ who observed attenuated vagal modulation of heart rate in young adults ($\sim 27$ y of age) during a lying-to-standing test after 30 min in a hot (36°C) environment in comparison with a thermoneutral environment (24°C). Unlike our findings, however, those changes did not translate to reduced systolic pressure.⁸¹ This discrepancy is likely explained by the fact that we employed longer exposures and enrolled older adults, who are at elevated risk of heat-related mortality because of age-related declines in physiological function.²² To our knowledge, we are the first to show that indoor overheating blunts cardiac autonomic function and blood pressure regulation (Table 3), manifesting in symptoms of orthostatic intolerance (Table 4) during postural changes consistent with activities of daily living.

Remaining Knowledge Gaps and Future Directions

Broadly speaking, our study adds to a growing body of research employing laboratory-based hot weather exposure simulations.⁸² By enabling evaluations of surrogate physiological outcomes linked with heat-related mortality and morbidity under ecologically relevant conditions, laboratory-simulated exposures are a promising and increasingly important tool for complementing more traditional forms of public health research to inform guidance and policy for protecting vulnerable persons from extreme heat, particularly where more direct assessment has proven difficult or unfeasible.^68,69,82 However, these studies are time- and resource-intensive and require specialized equipment (e.g., climate chambers, beat-to-beat blood pressure monitoring), limiting the number of participants and range of conditions that can be tested in a single study. Thus, although our findings have important implications for the development of evidence-based indoor temperature upper limits for continental climates, there remain important knowledge gaps that must be addressed to refine and generalize these recommendations. The following section highlights some of these future directions in context of the current findings.

Factors affecting body–environment heat exchange.

We evaluated physiological responses in a large range of ambient temperatures but fixed other variables known to modulate thermal strain. For example, humidity and airflow are well-established modifiers of whole-body heat loss. Increased humidity reduces the skin–environment water-vapor pressure gradient, the driving force for sweat evaporation, whereas elevations in airflow (e.g., via use of an electric fan) can have variable effects on heat exchange and body temperature regulation, depending on complex interactions between ambient conditions and factors affecting the physiological control of sweating (e.g., age, medication use).^70,83–85 That said, the impacts of these factors on body–environment heat exchange are likely reduced in older adults and other vulnerable groups^84,86 because of reductions in the capacity for sweat secretion.²² Nonetheless, they warrant consideration in future trials evaluating the physiological and health impacts of indoor overheating.

In addition to elevations in ambient humidity and airflow, radiant heat load is a particularly important consideration for future research and guidance development. The human body absorbs almost all (95%–100%) of thermal radiation reaching the skin.⁸³ This means that a person exposed to radiant heat through direct or reflected solar radiation or other hot objects in the immediate environment (e.g., ovens, water heaters) experiences considerably greater heat stress for a given air (dry-bulb) temperature than an unexposed person.⁸⁷ As such, the upper limit safe indoor temperature is likely appreciably lower than that identified in the current report for homes without the ability to reduce internal solar load during the day (e.g., without external shutters, blinds).¹¹ Unfortunately, however, few studies have considered the impact of sun exposure and other sources of thermal radiation on physiological responses or its implications to the development of recommendations on indoor temperature upper limits.

Other important areas for future research include evaluating the physiological effects of elevated metabolic heat production from physical activity through, for example, simulated activities of daily living,⁸⁸ as well as impairments in body–environment heat exchange due to clothing insulation. The latter factor is especially important for older adults, who are more likely to overdress in hot weather.⁸⁹

Individual modifiers of physiological responses to heat.

Numerous individual factors modify physiological responses to heat stress. Arguably the most important in the context of indoor temperature upper limits is the history of heat exposure. There is considerable global heterogeneity in heat-related mortality,⁷⁸ with individuals living in cooler climates more susceptible to the adverse impacts of heat waves.⁷⁹ Reduced heat vulnerability in persons residing in hotter climates stems, at least in part, from the well-established improvements in thermoregulatory function that occur with repeated heat exposure (i.e., acclimatization).^30,31 Indeed, we have previously shown that middle-age-to-older adults from the current study’s catchment area (Ottawa, Ontario, Canada) exhibit improvements in whole-body heat dissipation during resting heat exposure following the summer relative to before it.³¹ We accounted for this factor by testing most participants in the summer months (though some participants were tested in cooler months because of COVID-19–related restrictions; see the “Experimental Procedures” section). However, future research is needed to determine whether a 26°C upper temperature limit is equally protective early in the heat season, when the health risks of heat waves are at their greatest.⁹⁰ Consideration of the physiological impacts of acclimatization in future studies is also critical for determining the extent to which the 26°C upper limit generalizes to nontemperate climates. It is likely that 26°C is overprotective for individuals living in warmer areas (e.g., tropical or hot, dry climates), and enforcing this limit in those regions could have negative consequences because of current economic and environmental costs of mechanical cooling.⁸²

Although habitual heat exposure achieved through, for example, regular physical activity in hot conditions or sauna use can confer beneficial physiological effects, extreme heat stress for prolonged periods can impair body temperature regulation during subsequent exposures. Our group and others have observed reduced whole-body heat loss and greater increases in core temperature on a second in comparison with the first day of arduous work in hot conditions.^91–95 These effects appear to be exacerbated in older adults and might be related to inadequate rehydration,^92–94 consistent with age-associated alterations in body fluid regulation that reduce fluid consumption during heat exposure.²² Whatever the mechanisms, evaluating whether a 26°C upper limit indoor temperature remains protective over multiple days is another key area of future inquiry, especially given that heat waves, by definition, occur over extended periods.

Further, we enrolled habitually active participants without chronic health conditions known to increase heat vulnerability. Although some health conditions, including type 2 diabetes and heart disease, are associated with increased physiological strain during exercise in hot conditions,²² it is unlikely that core temperature responses would be markedly altered under resting conditions. Supporting this postulate, we recently observed that although type 2 diabetes is associated with reduced heat loss during moderate exercise in the heat,⁹⁶ physiological responses did not differ between older adults with or without well-controlled type 2 diabetes during a resting 3-h exposure to 44°C and 35% relative humidity⁹⁷ conditions exceeding those measured indoors during heat waves.^20,21,34 Likewise, we recently showed no differences in body temperature⁹⁸ or cellular strain⁹⁹ between older adults with and without well-controlled type 2 diabetes and/or hypertension during 9-h of exposure to 40°C, 9% relative humidity. Furthermore, although maintaining high levels of aerobic fitness has been shown to prevent the age-related decline in thermoregulatory function during exercise heat-stress, these effects are considerably reduced at lower levels of heat stress like those that might be experienced by vulnerable groups indoors during hot weather.¹⁰⁰ Whether physiological responses to indoor heat stress are exacerbated in sedentary individuals or those with less well-controlled chronic health conditions (e.g., diabetic individuals with poor blood sugar control¹⁰¹) is an important area for future inquiry.

Conclusions and Future Directions

Maintaining ambient temperature at or below 26°C limited thermal and cardiovascular strain in adults 66–78 y of age during laboratory-based indoor exposures simulating conditions experienced during hot weather and heat waves in temperate continental climates. These findings support recommendations that a 26°C indoor temperature upper limit is effective for protecting vulnerable occupants living in continental climates from indoor overheating. Larger trials are needed to corroborate our findings and improve generalizability to more at-risk populations residing in different climate types. Future work should also consider differing physical activity patterns and evaluating physiological responses to indoor overheating over multiple days and in other heat-vulnerable groups (e.g., children).