Journal of Applied Physiology

Thermoregulatory responses to cold transients: effects of menstrual cycle in resting women

Richard R. Gonzalez, Laurie A. Blanchard


Effects of the menstrual cycle on heat loss and heat production (M) and core and skin temperature responses to cold were studied in six unacclimatized female nonsmokers (18–29 yr of age). Each woman, resting supine, was exposed to a cold transient (ambient temperature = mean radiant temperature = 20 to −5°C at −0.32°C/min, relative humidity = 50 ± 2%, wind speed = 1 m/s) in the follicular (F) phase (days 2–6) and midluteal (L) phase (days 19–23) of her menstrual cycle. Clothed in each of two ensembles with different thermal resistances, women performed multiple experiments in the F and L phases. Thermal resistance was 0.2 and 0.4 m2 ⋅ K ⋅ W−1for ensembles A andB, respectively. Esophageal temperature (Tes), mean weighted skin temperature (T̅ sk), finger temperature (Tfing), and area-weighted heat flux were recorded continuously. Rate of heat debt (−S) and integrated mean body temperature (T̅ b,i) were calculated by partitional calorimetry throughout the cold ramp. Extensive peripheral vasoconstriction in the F phase during early periods of the ramp elevated Tesabove thermoneutral levels. Shivering thermogenesis (ΔM =MM basal, W /m2) was highly correlated with declines inT̅ sk and Tfing(P <0.0001). There was a reduced slope in M as a function ofT̅ b,i in the L phase with ensembles A(P < 0.02) andB (P< 0.01). Heat flux was higher and −S was less in the L phases withensemble A(P < 0.05). An analytic model revealed thatT̅ sk and Tes contribute as additive inputs and Tfing has a multiplicative effect on the total control of ΔMduring cold transients (R 2 = 0.9). Endogenous hormonal levels at each menstrual cycle phase, core temperature andT̅ skinputs, vascular responses, and variations in body heat balance must be considered in quantifying thermoregulatory responses in women during cold stress.

  • clothing
  • regional heat flux
  • thermoregulatory model

the reproductive system has a clear and important role in altering thermoregulation in women, particularly when internal body temperature becomes elevated during the luteal phase compared with the follicular phase of the menstrual cycle. In the luteal phase, thermoregulatory responses of women are characterized by alterations in core temperature (Tc) thresholds affecting the onset of specific physiological effector responses during exercise, heat exposure, and cold exposure (17, 20, 24, 27). Elevated Tc thresholds controlling the onset of sweating and skin blood responses are consistent with a higher internal body temperature reference point evident in the luteal phase, which may compromise thermoregulation during prolonged exercise or warm exposures in the luteal phase (20, 24, 27). Several studies (17, 19,20, 27) carried out in warm conditions during exercise show that when Tc and skin temperature (Tsk) are elevated, heat loss mechanisms become activated, and sudomotor drive and blood flow to the skin surface increase in the mid- to late luteal phase of some eumenorrheic women.

Few data exist quantifying female responses to cold stress. During cold stress, shivering by gross muscular contraction may or may not be sufficient to maintain Tc. Tc and Tsk interact in a unique fashion as constant temperature multipliers or in a summative fashion to increase metabolism (5, 15). Other than limited studies (1, 17, 28), information on the role of reproductive hormones during various stages of a woman’s menstrual cycle is scarce. Also there are few studies on how various stages of the menstrual cycle influence thermoregulatory responses to cold. This thermosensitivity may be described by a change in slope to a decreasing internal temperature or Tsk affecting a given heat loss response or an M response. The integrated mean body temperature (T̅ b,i) or core thermosensitivity of a thermoregulatory response is defined here as the amount of change in the specific dependent response for each unit change inT̅ b,iduring cooling. The change in slopes of the respective dependent response curves above a given referenceT̅ b,idescribes the various thermosensitivities. The shivering response is generally determined by a plot of excess shivering vs. Tsk or Tc (2, 4, 15, 17). Peripheral thermosensitivity is described as the shivering thermogenesis influenced by skin surface temperature decreases solely (4, 5, 15).

Various studies (2, 4, 15) have revealed an interaction of hypothalamic temperature with skin and deep-body temperatures in the initiation and control of various thermoregulatory responses. Skin cooling has been shown to alter the hypothalamic thermosensitivity driving increases in metabolic heat production and cutaneous vasoconstriction (15, 17, 28). Fluctuations of estradiol and progesterone levels and changes in their relative ratios during a woman’s menstrual cycle also participate in mediating cutaneous vascular responses during cold challenges (1, 17).

When Tc is offset to a higher temperature level in a woman’s midluteal phase, there may exist competitive inhibition in the processing of thermal information in the thermoregulatory system (4, 8). This competition of thermal afferent information from central and peripheral receptors may blunt the shivering response to cold stress at a given mean Tsk(T̅ sk) (17). Additionally, there is strong evidence that hypothalamic neurons show changes in their local hypothalamic thermosensitivity during displacements in cutaneous temperature that closely parallel whole body thermoregulatory responses (5). That is, the highest slope (e.g., “gain”) in a shivering response (2, 15) occurs when cold Tc is combined with cold Tsk and their combined rate of change is rapid.

Ability to tolerate cold stress is also affected by use of extra clothing (e.g., thermal resistance added to a fixed tissue resistance). Here, effect of added clothing is considered an extrinsic factor that only functions as part of the passive means in the regulation of energy exchange between the body and the environment (12).

The present experiments were designed to perturb the thermoregulatory system by employing a controlled decrease in ambient temperature (Ta) (14). The technique allows quantification of dynamic Tskresponses and provides information characterizing cold reactions in women. The purposes of this study were1) to determine the heat exchange responses during cold transients at two stages of the menstrual cycle in women clothed in two different clothing systems,2) to ascertain whether heat loss responses are influenced by dynamic changes in Tsk and internal body temperature during a cold transient at two stages of a woman’s menstrual cycle, and 3) to establish effective model parameters in the control of shivering thermogenesis during cold transients. First, it was hypothesized that a constant level of thermal resistance affects heat loss but should not alter thermoregulatory mechanisms within each menstrual cycle phase. Second, it was hypothesized that the cold stress would principally stimulate skin thermoreceptors rather than central Tc driving the shivering responses. Finally, it was hypothesized that control of thermoregulatory responses (shivering and heat loss) activated by peak levels of estradiol and progesterone in the luteal phase should have significant effects on the Tcreference point (20, 27) but not the slope of shivering thermogenesis.


Six highly motivated women, recruited from the military test subject pool, volunteered to participate in the study after being informed of the risks and purpose of the study and after giving their written, informed consent. The protocol was approval by the US Army Research Institute of Environmental Medicine Human Use Review Committee (HURC). The women completed a questionnaire that incorporated various signs and symptoms related to their menstrual cycle. A complete medical examination was done on each woman before any experimentation. Each woman had no history of cardiovascular or respiratory disease or complications from irregular menstruation. Before any testing began, each subject had a blood scan to ensure the absence of anemia and a pregnancy test, which was verified as negative.

Each woman completed all experiments. The characteristics of the subjects were as follows (means ± SD): age = 21.2 ± 3.9 yr, height = 1.65 ± 0.10 m, weight = 60.9 ± 7.9 kg, body surface area = 1.66 ± 0.12 m2, and body fat = 23.9 ± 2.5%. Experiments on each subject were done in the late fall, winter, and early spring between 0700 and 0900 to control for circadian variation in temperature regulation. The women were normal early risers. Each of the six women displayed a normal menstrual cycle as defined by regular periodicity (∼28- to 30-day cycles) kept in a daily log book, and no subject was taking oral contraceptives. To verify that ovulation had taken place in a given month, each subject recorded her daily basal body temperature on awakening for the month throughout the experimental study. Oral temperature was measured twice while the subject rested at the same time each morning (with mouth completely shut) with a calibrated, fast-responding, automated oral thermometer. Data from an entire menstrual cycle were collected and graphed before the study to determine whether oral temperature increased after ovulation (18). Although basal body temperature monitoring is not a wholly sufficient method to predict ovulation time, higher basal internal temperature is closely correlated with the higher plasma progesterone concentration evident in the luteal phase of the menstrual cycle (9) and directly applicable to our resting study.

Testing in the luteal phase occurred only on days when the resting Tc was elevated (approximatelydays 19–23). Testing in the early follicular phase occurred on days 2–6 (day 1 = 1st day of menstrual flow). The calendar dates initially chosen to correspond with a given stage of the menstrual cycle of the woman were verified by post hoc blood analyses of estradiol and progesterone levels. Hormonal data, basal body temperature records, and subject logs were matched for proper cycle phase at the end of the total study protocol. Experimental data authenticated with the respective hormonal data from each woman were used in the statistical analyses. Hormonal data were culled according to appropriate phase and test day to determine mean differences in precision of measurement by ANOVA (e.g., interassay variability) for the two separate test samples from each woman. Percent differences between test days are shown in Table1, along with the mean concentrations of the reproductive hormones at each cycle phase.

View this table:
Table 1.

Concentration levels of estradiol and progesterone during basal resting periods

Clothing Ensembles

The women dressed in each of two clothing ensembles devised to add two constant, fixed resistances to their skin surface layer.Ensemble A consisted of the US Army-issue physical training shorts and underwear plus the US Army T-shirt worn under a temperate battle dress uniform.Ensemble B consisted of a US Army battle dress overgarment worn over ensemble A. The clothing insulation values were evaluated separately at three different wind speeds to establish effective clothing heat and water vapor transfer coefficients employed in partitional calorimetric analysis. These heat transfer coefficients were estimated using a regional copper manikin resting supine on a wood-framed cot supported by parachute nylon webbing, which simulated the conditions of the human experiments. The total thermal resistance of ensemble A measured at chamber wind speed of 1 m/s (paralleling the human experiments) was 0.21 m2 ⋅ K ⋅ W−1(1.33 clo). The clothing resistance of ensemble A is equivalent to normal civilian trousers and shirt. The thermal resistance of ensemble B, also tested at a wind speed of 1 m/s, was 0.4 m2 ⋅ K ⋅ W−1(2.58 clo).

During all experiments, subjects wore a standard US Army light-duty work glove with a five-finger woolen insert, which added a constant thermal resistance to the hands. The glove was separately evaluated on the US Army Research Institute for Environmental Medicine copper hand model and had a thermal resistance of 0.13 m2 ⋅ K ⋅ W−1(0.86 clo). During all experiments a standard-issue US Army black boot and US Army cushion-sole socks were worn on the feet with each clothing system. The thermal insulation of the boot with a sock was analyzed (1.8 clo, thermal resistance = 0.28 m2 ⋅ K ⋅ W−1) using a regionally heated copper foot in our laboratory. During all experiments, subjects were not allowed to open or ventilate their garments by opening closures or zippers or by excessive movement, thereby serving to control against disparate changes in skin surface temperatures and heat flow.

Before the beginning of the study, volunteers were thoroughly familiarized with all experimental techniques, and their standing height was measured. During these training sessions, body weights were measured so that on the four test days (twice inensemble A and twice inensemble B), preexperiment body weights could be easily traced back within 1% of the mean body weight measured during preliminary testing to control for possible effects of hypohydration.

Experimental Testing

On arriving at the laboratory each morning, the subject rested on a chair for 30 min, and a 10-ml blood sample was taken by venipuncture for the measurement of serum 17β-estradiol and progesterone concentrations (by RIA; Coat-a-Count, Diagnostic Products, Los Angeles, CA) to accurately define menstrual cycle phases in each woman (Table1). Blood samples were quickly processed when taken and frozen. Samples from all women were analyzed in the same batch assay (in duplicate) to obviate potential interassay variability, as pointed out previously. After blood was drawn, the previously trained subject inserted through the nostril a polyethylene-encased thermocouple channeled through the pharynx into the esophagus to a depth ∼25% of her height (in cm). Exact placement of the thermocouple at the midheart level was verified by following a real-time thermal recording on a computer screen as she slightly inserted and retracted the thermocouple into the esophagus past the initial “hot” spot demarcating a correct heart-level point (27). The women were asked to avoid swallowing by spitting saliva into a cup during the experiments to obviate spurious recordings. Any inadvertent swallows, shown by immediate dips on the computer monitor scan, were later smoothed in the data file by using an exponential smoothing of esophageal temperature (Tes) to predict a value based on the forecast for the immediately preceding 15-s period.

Surface thermistors with calibrated heat flow disks (Concept Engineering, Old Saybrook, CT) were placed at six skin surface sites (midchest, midthigh, lateral calf, upper hand, upper arm, and midforearm) and area weighted to estimateT̅ sk(23). Separate 28-gauge copper-constantan thermocouples were also placed in the middle fingernail bed and big toenail bed. The calibrated heat flow disk surrounding each embedded thermistor element determined heat flux from each of the skin sites. Weighted heat flux (W /m2) was calculated from each respective skin site area weighting (10, 23).

Environmental temperatures, Tesand Tsk, and heat flow data were recorded every 15 s with a personal computer. Wind speed in the chamber was controlled at 1 m/s. O2 uptake (V˙o 2, l/min) was measured by 2-min collection of all expired gases into a Douglas bag sampled every 20th min of the transient and twice at 20°C, once before the decrease of Ta and once after ∼10–15 min at ∼10°C. Heat production (W /m2) was calculated from the expired respiratory parameters obtained, the caloric equivalent of 1 liter of O2, and the DuBois surface area equation (12). Subjects were exposed to the lowest air temperature (−10°C) for an additional ∼10–15 min. Experiments ceased if a woman voluntarily withdrew or if she was withdrawn because fingertip Tsk(Tfing) reached ≤5°C or Tes plummeted toward 35°C. These lower limits were advisory guidelines set by our HURC for removal of a given subject from the test for that day.

Heart rates were obtained and recorded every 5 min from the electrocardiogram measured continuously using chest electrodes (CM5 placement) interfaced to a telemetry system (models 78510A and B, Hewlett-Packard).

The subjects were to refrain from active exercise and were not to consume food, caffeine, and medication (including aspirin or any analgesic-antipyretic compounds) ∼10 h before the experimental testing. If the subject had unintentionally taken any medication, an experiment was rescheduled for the next appropriate calendar day. Body weight and composition were determined by repeated nude weighings and skin-fold measurements (27).

Steady-State and Transient Exposure

All subjects rested supine on a specially designed wooden cot for 15 min of baseline data collection at 20°C air temperature. After complete instrumentation, an additional resting period began; it lasted ∼20–30 min, until equilibration occurred, i.e., the woman’s Tsk and Tc remained constant within ±0.1°C for 10–15 min. After the initial equilibration period, each subject was exposed to the thermal transient by decreasing the environmental chamber Ta(Ta =T̅ r = To, whereT̅ r is mean radiant temperature and To is operative temperature) (12) in a controlled downward ramp [for the 24 runs: To = 17.5 − 0.316 ⋅ (min) + 1.088e −3 ⋅ (min)2,R 2 = 0.98, SE of estimate ±1.2°C]. Tois the critical variable describing the ambient environment when clothing is worn (12). Dew point temperature was allowed to fall passively during the room temperature decreases. The cooling phase at the lowest target To typically continued for 80–120 min (the latter time point when the subject was dressed in ensemble B) at a constant decreased exponential ramping rate of −0.32°C/min. A final exposure time of 10–20 min at To of −5 to −10°C was completed before the experiments were ended by a subject’s request or in conformity with our HURC recommendations. All data were truncated to an 80-min time point to facilitate statistical analyses.

Heat Exchange Variables

Partitional calorimetric analyses (12, 30) were conducted at 20-min intervals with each avenue of heat exchange from the heat balance equation (e.g., respiratory and convective heat loss responses combining all respective clothing and heat transfer coefficients) taken into account, in which±S=MEsk(R±C)K(W/m2) Equation 1 where ±S is rate of body heat storage (in this study −S is heat debt);M is metabolic heat production calculated from each 20-minV˙o 2 interval;E sk is evaporation or insensible (wet) heat exchange, which is set by the clothing moisture properties (im/clo, where im is an evaporative constant) (12) and evaporative heat transfer coefficients determined from parallel copper manikin evaluations of the garments, the Tsk-Tagradient (T̅ sk− To), and the change in body weight loss and respiratory heat loss;R is radiation;C is convection; andK is conduction.R andC combine as sensible (dry) heat exchange, which was determined by the environment, thermal conductance, and insulative properties of the ensembles and their respective heat transfer coefficients.

Details of the analysis to ascertain integrated mean body temperature from partitional calorimetry are addressed in the .

Statistical Methods

Values are means ± SD. For the baseline equilibration periods, mean values of Tes and Tsk, Tc and Tsk changes from neutral (ΔTes, ΔT¯ sk, and ΔTfing), and metabolic heat production were analyzed by univariate ANOVA techniques with repeated measures. Whenever a significantF-ratio was found, Tukey’s critical difference was employed for post hoc analysis (α = 0.05) (25).

Mean data of Tes,T̅ sk, heat content (kJ), and heat flux as a function of time during the thermal transient were analyzed by two-way (ensemble × menstrual phase and time × menstrual phase) or three-way (time × ensemble × menstrual phase) ANOVA for repeated measures. When the ANOVA indicated significant main or interactive effects, Tukey’s studentized range, honestly significant difference (HSD) was used to compare means and locate minimum significant differences (α = 0.05) between factors and among repeated measurements (25).

Regression methods.

Regression analyses of metabolic heat production vs.T̅ sk and Tfing were determined using a linear regression for repeated measures, with between-subjects differences taken into account (13, 25). Dummy variables were used to encode the different subjects, and all the data were pooled to estimate a single regression equation (13, 25).

A two-parameter, piecewise linear regression analysis was used to fit the data from each subject’s shivering response (ΔM =MM 0, where M 0 is the basal metabolic rate) as a function of integrated mean body temperature. Each regression coefficient (slope) and threshold point were examined by ANOVA for repeated measures and Tukey’s post hoc test, as described above (6,25).

Parameter model estimation.

Maximum-likelihood parameter estimation (MLE) was used to determine the respective control coefficients (parameters) derived by independent effects of Tc and Tsk likely to be considered influencing shivering thermogenesis (ΔM =L). This technique was chosen as an ideal method of describing ΔM =L on the basis of the estimates of three principal control variables: Tes,T̅ sk, and Tfing (15, 28) driving the shivering response throughout the cold transient. It was assumed that the product of a parameter value times a given variable (ΔTes, ΔT̅ sk, and ΔTfing) in the data set are independent terms that equally weight the final likelihood function (ΔM =L). Also assumed was an initial model statement that shivering thermogenesis is based on proportional control (13) by summed linear effects or multiplicative (28) effects from the three variables. In the MLE analysis, starting values of control parameters (P1, P2, and P3) and their respective constants ( ΔTsk,x1 , ΔTes,x2 , and ΔTfing,x3 ) are first set to default values on the basis of a given model statement. Each subsequent iteration seeks to find the minimum sum of squares residual (SSR) by a quasi-Newtonian derivative procedure. This is done by differentiating with respect to a given parameter and equating each derivative to zero (e.g., ∂ logL/∂P1= ∂ logL/∂P2= ...∂ logL/∂Pn= 0). Iterations terminate when the values of the parameter estimates in the iteration procedure fail to change. This yields the maximum-likelihood estimator of all the parameters. The method also generates R 2 = (1 − residual SSR/total SSR) of each model equation.

MLE is a useful method to optimize unknown parameters in a probability model where the response variables are dichotomous and the predicted variable is a probability (6).


Tsk and TesResponses

Extensive peripheral vasoconstriction and rate of fall in Tsk(T˙skt) of about −0.1°C/min occurred during all thermal transient runs. During the initial thermoneutral equilibration periods,T̅ sk was not significantly different between phases for a given ensemble. FinalT̅ sk in both phases was lower in ensemble Athan in ensemble B experiments (23.2 ± 0.6 and 27.4 ± 0.3°C,P < 0.05).

In ensemble A experiments a higher Tes was evident in the luteal phase than in the follicular phase (36.9 ± 0.05 vs. 36.6 ± 0.16°C, P < 0.01) during the basal resting period (at −30 min). During the initial phases of the transient the drops inT̅ sk were associated with elevations in Tes. Figure 1 shows that ΔTes initially rose in the early time periods of the transients with ensemble A. ΔTes was elevated higher at 50, 70, and 80 min of the transient in the follicular phase than in the luteal phase (P < 0.05).

Fig. 1.

Esophageal temperature (Tes) during cold transient in 6 women. A:ensemble A in follicular (Fol, ○) and luteal phase (Lut, □). B:ensemble B in follicular (•) and luteal phase (•). Values are means ± SD; SD bars are plotted at 10-min intervals for clarity.a Significant difference between phases at each time point (P< 0.05).

In experiments with ensemble B, ΔTes was higher (P < 0.05) during the follicular phase than during the luteal phase at 70 and 80 min of the transient. Toward the final time points (70–80 min) ΔTes began to decline markedly with the cold stress.

Metabolic Heat Production

Steady-state metabolic heat production was not significantly different during equilibration time periods (−30 min) when the women woreensemble A orB in the follicular and luteal phases: 47.16 ± 5.11 and 47.28 ± 4.69 (SD) W /m2 in the follicular and luteal phases, respectively, with ensemble Aand 44.2 ± 5.13 and 45.9 ± 6.2 W /m2 in the follicular and luteal phases, respectively, withensemble B. These values, which ranged from 5 to 13%, are not significantly different from the basal heat production values (41.85 W /m2) reported for 25-yr-old women (17).

Figures 2 and 3 show that total heat production was closely correlated (P < 0.0001) with the reducedT̅ sk and Tfing in all the cold transient experiments.

Fig. 2.

Metabolic heat production (M) plotted as a function of mean skin temperature (T̅ sk; linear regression for repeated measures, Ref. 14).A: ensemble A in follicular phase [M = −5.93 ± 0.59(T̅ sk) + 227, R 2 = 0.83,P < 0.0001].B: ensemble A in luteal phase [M = −3.93 ± 0.51(T̅ sk) + 169, R 2 = 0.76,P < 0.0001].C: ensemble B in follicular phase [M = −6.56 ± 0.95(T̅ sk) + 2,248, R 2 = 0.66, P < 0.0001].D: ensemble B in luteal phase [M = −3.75 ± 0.48(T̅ sk) + 166, R 2 = 0.71,P < 0.0001].

Fig. 3.

M plotted as a function of finger temperature (Tfing; linear regression for repeated measures, Ref. 14).A: ensemble A in follicular phase [M = −1.97 ± 0.21(Tfing) + 99.2,R 2 = 0.70,P < 0.0001].B: ensemble A in luteal phase [M = −1.21 ± 0.12(Tfing) + 80.3,R 2 = 0.81,P < 0.0001].C: ensemble B in follicular phase [M = −1.33 ± 0.0.19(Tfing) + 83,R 2 = 0.60,P < 0.0001].D: ensemble B in luteal phase [M = −0.75 ± 0.11(Tfing) + 68,R 2 = 0.60,P < 0.0001].

Shivering thermogenesis (i.e., MM basal, W /m2) was determined from each woman’s response to the cold ramp and plotted as a function of integrated mean body temperature (T̅ b,i), as described in the , to derive regression coefficients (6). Table 2shows the mean results.

View this table:
Table 2.

Comparison of slopes of shivering response to Embedded Image and Embedded Image during the cold ramp

The slope of shivering thermogenesis response toT̅ b,iM/ ΔT¯ b,i) in the luteal phase of the menstrual cycle in each woman was attenuated during experiments with ensemble A(P < 0.02) andensemble B(P < 0.01). TheT̅ b,ithreshold (T̅ b,i,0) was not significantly different between phases or ensembles.

Heat Flux

Figure 4 shows the weighted heat flux data plotted at 20-min intervals from all the experiments. During the 12–13 min of the cold ramp with ensemble A in the luteal phase, the heat flux through the garment was not significantly different from values observed in the follicular phase. Into the 20th min (To ≈ 12 ± 1°C) and throughout the cold transient, heat flux was displaced toward a higher level in the luteal phase than in the follicular phase (P < 0.05). During experiments withensemble B, heat flux values were increased by 60–70% (P < 0.05) above basal values, but mean heat flux at each time point of the transient was not significantly different between phases.

Fig. 4.

Mean weighted heat flux plotted at 20-min intervals fromtime 0 of cold transient at respective experiments.

Three-way ANOVA (time × phase × ensemble) followed by Tukey’s HSD indicated that mean weighted heat flow values were higher at each time point and cycle phase with ensemble A (P < 0.05).

Rate of Heat Debt

Figure 5 shows the results of whole body rate of heat debt calculated for each 20-min period of the transient. Heat debt was determined by partitional calorimetric analyses, with each woman’s metabolic heat production, body weight, percent body fat and surface area, Tes, andT̅ sktaken into account, as explained in the .

Fig. 5.

Rate of heat debt determined by partitional calorimetry as a function of cold transients. Values are means ± SD.

ANOVA revealed significant differences in rate of heat debt from basal time periods. The women in the follicular phase withensemble A exhibited a higher rate of heat debt at 40 and 60 min and final time points (P < 0.05) than at similar time points in the luteal phase. In experiments withensemble B the rate of heat debt was greater than values at basal time points (P < 0.05), but there were no phase differences throughout the transient.

Prediction Equations for Shivering Thermogenesis

The variables Tes, ΔTes,T̅ sk, and Tfing were shown to be independent factors influencing the shivering response (Figs. 1-3). For this reason, an attempt was made to develop a unifying algorithm to describe shivering thermogenesis as a function of the respective Tsk and Tc thresholds and control constants. Because shivering thermogenesis was highest in the experiments with ensemble A, this data set was examined for distinct differences possibly related to cycle phase. All data were analyzed by an iteration procedure employing MLE, as explained in methods. An acceptable criterion for a model equation based on summative or multiplicative construct of the independent variables was a derivedR 2 ≥ 0.9 from the lowest SSR (3).

Threshold values initiating excess heat production due to shivering thermogenesis were generated from the data set. In the follicular phase, thresholds were 26.5 ± 0.3°C for Tfing(Tfing,0) and 32.5 ± 0.2°C forT̅ sk(T̅ sk,0). In the luteal phase, thresholds were 21.5 ± 0.3°C for Tfing,0 and 31 ± 0.9°C forT̅ sk,0. These values were found to be offset to lower temperatures (P < 0.002). Tes,0 thresholds were not significantly different between phases (36.9 ± 0.1°C). Prediction equations established by MLE areΔMF=[P1,F·(T¯sk32.5)+P2,F·(Tes36.9)] ·P3,F·(Tfing26.5)(W/m2) (P1,F=0.35,P2,F=0.85,P3,F=0.9) R2=0.91,SSR=1,776 Equation 2for the follicular phase (days 1–6), with ΔMM F) ≥ 0, andΔML=[P1,L·(T¯sk31.0)+P2,L·(Tes36.9)] ·P3,L·(Tfing21.5)(W/m2) (P1,L=0.65,P2,L=2.28,P3,L=0.59) R2=0.86,SSR=2,200 Equation 3for the luteal phase (days 19–25), with ΔMM L) ≥ 0.

The analyses indicate that Tes/ T¯ skthermal sensitivity (P2,F> P1,F) governing the shivering responses was 2.42 W ⋅ m−2 ⋅ °C−1in the follicular phase and 3.5 W ⋅ m−2 ⋅ °C−1in the luteal phase (increase of ∼44.6%). That is, sensitivity to a lowering of Tc is still a stronger factor in the control of the shivering response than Tsk alone (Fig. 1). Tfing was shown to be a significant peripheral multiplier to the summed effects of core and skin thermal inputs affecting shivering thermogenesis in both phases. P3 was higher in the follicular than in the luteal phase (by ∼53%).


The general results of this study suggest that a woman’s metabolic response to a lowered Tsk and a loweredT̅ b,ibecomes significantly attenuated during the midluteal phase of her menstrual cycle when endogenous levels of estradiol and progesterone become elevated. The mechanism that determines the level of shivering per change inT̅ b,iduring a woman’s luteal phase may be associated with a reduced requirement of heat production but enhanced skin blood flow to the cutaneous sites (16, 17, 19). It is possible to infer from our study’s heat balance results that elevated reproductive hormones induce a warmer central core (Tc), resulting in a lower rate of heat debt, which would extend tolerance times during cold transients.

The general effect of the added clothing insulation in combination with enhanced central warming in the luteal phase is that, for a given decrease in Tsk, less heat production is required to maintain deep body temperature and rate of heat debt is diminished as a function of time of the cold stress. The practical application is that women during their midluteal phases are able to tolerate lower Tc during cold transient exposures.

Clothing Resistance

We hypothesized that intrinsic thermoregulatory mechanisms activated by the hormonal changes appearing during the two menstrual cycle phases would probably not become modified during cold stress by addition of clothing. This hypothesis was confirmed by the heat flux and rate of heat debt alterations evident during the thermal transients (Figs. 4and 5). The higher thermal resistance of ensemble B significantly decreased sensible heat loss from the skin surface to the ambient. Additionally, the added clothing resistance diminished heat debt by adjusting the lower critical temperature (e.g., Ta where net metabolic heat production rises) necessary for shivering thermogenesis. Adding a finite thermal resistance to the skin layer by clothing blunts somewhat the heat exchange properties but not the consequences of thermoregulatory effector responses (shivering, vasoconstrictor activity). Changes in heat production and heat loss responses remained controlled by alterations in Tcand Tsk during the two stages of the menstrual cycle.

Shivering Responses

Because the Tes of the women in each phase was elevated in the cold transients, more so in the follicular phase with ensemble A, it would appear that the stimulus for increased heat production was, to a large extent, peripheral in origin. The hypothesis was confirmed that a strong peripheral drive becomes active in controlling shivering thermogenesis during supine rest in the women. The results suggest that as long as the Tes remains near thermoneutral levels (or ΔTesfrom basal is approximately ±0.15°C), the summed effect of shivering thermogenesis is highly correlated with a combined effect ofT̅ sk(weighted without fingers and toes) and temperature of cold acral sites (Tfing). As evident in Figs. 2and 3, absolute heat production was strongly correlated with the decreases inT̅ sk and Tfing during the cold transient.

The reciprocal response of a rise in Tc to skin cooling has been frequently displayed in small animals and humans, in which the heat capacitance is limited due to small surface area-to-volume ratio (2,15, 17, 30). Hessemer and Brück (17) found that a reciprocal effect of Tc to Tsk decreases in women exposed to cold temperature step changes. Our study showed that the rise in Tes was most clearly apparent during the follicular phase when the women were clothed inensemble A, which has the lower thermal resistance (Fig. 1). Rises in ΔTes accompanying Tsk cooling were less (P < 0.05) during experiments in the luteal phase with ensemble A than during experiments in the follicular phase at 40–80 min of the transient.

This study also revealed important new information regarding the effectiveness of a change in Tcrelative to a change inT̅ sk in producing alterations in shivering response during the two stages of the menstrual cycle. Previously, Hessemer and Brück (17) found, in unclothed women exposed to a step change of room temperature from 32 to 12°C, that the shivering response was strongly altered by Tc and Tsk, and the threshold for shivering shifted to a higher level in the luteal phase. They attributed the shift in threshold for the initiation of shivering to an increase in basal metabolic heat production. They revealed that a decreased electromyogram response to cold stress also occurs in the luteal phase. Unresolved from their study was whether the slope of the shivering response to Tsk or Tc becomes modified as a consequence of the hormonal changes during the luteal phase.

The present study confirmed the results of the study of Hessemer and Brück (17), in which shivering responses were strongly correlated with Tsk. However, our results indicate that, in the luteal phase, decreases in the shivering thermogenesis are strongly affected by decreasingT̅ b,idetermined from summed heat balance. Unlike Hessemer and Brück, we found that, during the luteal phase, basal metabolic heat production was unaltered and there was no statistically significant shift in theT̅ b,ireference point controlling the initiation of shivering. This latter result was contrary to our initial hypothesis.

In past studies using animal models to study shivering, it was found that hypothalamic warming reduces or even suppresses shivering in a cold environment (4, 5, 15). Boulant and Gonzalez (5) showed in rabbits that heating the preoptic/anterior hypothalamic area decreased the hypothalamic thermosensitivity, inducing the shivering thermogenesis when colonic temperature and Tskwere warm. Preoptic cooling, however, increased the hypothalamic thermosensitivity when Tc and Tsk were very cold. Our study showed that shivering response to a givenT̅ b,isignal was reduced in the luteal phase of the women when the warming activity on the hypothalamus would be most likely prevalent. In response to cold skin calling for increased metabolic heat production, the greater weighted influence of deep body thermoreceptors, possibly acting within a warmed hypothalamic area during the midluteal phase, could have a dominant role in inhibition of the shivering response. The mechanisms for the diminished ΔM/ΔTb,iand lack of anyT̅ b,ireference shift suggest a readjustment of overall heat balance during the luteal phase possibly influenced by the thermogenic activity of peak hormonal levels of estradiol and progesterone or other mediators (8, 24, 26).

Skin Heat Flux Responses

The weighted heat flux obtained in these experiments represents an independent assessment of cutaneous blood flow stemming from extremity (arms, thighs, and calf) and chest skin surface sites purported to be responsive to neural efferent drive controlling skin blood flow (10, 12). If cutaneous heat flux reflects a concomitant effect of augmented peripheral blood flow throughout the body’s system, which we believe it certainly does, then the mean weighted heat flux responses found in the present study (Fig. 4) definitely concur with studies showing increased arm blood flow by direct vasodilatory action or attenuation of vasoconstrictor activity prevalent in the luteal phase of a woman’s menstrual cycle (1, 16, 17, 20).

During experiments in the luteal phase in which the women woreensemble A, increases in mean weighted heat flux were initiated earlier and rose to a much higher level at 20–80 min of the cold transient (Fig. 4) than during the follicular phase (P < 0.05). Decreases in heat flux were evident with ensemble B during the cold transient. This lowered heat flux would be a requirement, owing to the summed effect of increased tissue resistance during vasoconstriction coupled with trapped heat within the extra layers of clothing. This trapped heat likely caused a reduction in the Tsk-Tagradient, thereby constraining skin heat flow to the ambient.

The increased heat flux responses observed in the luteal phase with ensemble A are at variance with steady-state studies of unclothed women exposed to 90 min of a 28°C Ta showing decreased thermal conductance in the midluteal phase (11) and other resting studies showing reductions in thermal conductances (3). These latter studies proposed that the offset in Tc apparent in the luteal phases of women is the outcome of an increase in tissue thermal insulation. Our data suggest, on the other hand, that such observations of reliance on thermal insulation per se are probably not in effect when peripheral cold receptor activity (e.g., dynamic skin receptors) dominates in resting clothed women.

The elevated heat flux response displayed in the luteal phase withensemble A suggests that the response results from augmented endogenous levels of estradiol and/or progesterone acting centrally (4, 5) or in peripheral vascular sites (1, 16). In our study it was not possible with our experimental techniques to deduce whether the heat flux response derives from a direct central nervous system response activating a peripheral vasodilatory response or occurs primarily via release of vasoconstrictor response at various peripheral vascular sites (1, 16).

Body Heat Debt

Analysis of rate of heat storage in the heat balance equation (in this case, heat debt) (12, 30) (Fig. 5) provided a highly valuable indicator of the summed effects of cold stress. Decreases in rate of heat debt were readily apparent after 40 min of the cold stress, when a marked rise in sensible heat loss accompanied the decreased shivering thermogenesis (Figs. 2 and 5). In the follicular phase with prolonged inactivity, sufficient heat cannot be generated to prevent excessive body cooling when an ensemble with thermal resistance equivalent to street clothing (ensemble A) is worn. With ensemble A, rate of heat debt was significantly less in the women in the luteal phase than in the follicular phase at 40–80 min of exposure. Body heat debt in the luteal phase was also less at 40, 60, and 80 min (P < 0.05, 3-way ANOVA with Tukey’s HSD post hoc test) than in the follicular and luteal phases withensemble B. However,ensemble B, with a higher thermal resistance than ensemble A, also has outer semipermeable layers that impede insensible heat loss. Vapor accumulation potentially causes condensation within the layers, lowering the intrinsic insulation and increasing skin wettedness (12). This vapor accumulation may have contributed to elevated heat debt evident with ensemble B during the final time points of the cold transient. During the luteal phase when the women were dressed in ensemble A, sensible heat loss was also impeded by the intrinsic insulation of the ensemble higher than bare skin. However, ensemble A is vapor permeable, and without moisture condensation, any trapped heat apparently was sufficient to prevent excessive body cooling during the transient.

Control Mechanisms During Cold Transients

Another interesting finding from the parameter estimation analyses of data during both menstrual cycle phases is that there appears to be a separate contribution from Tsk, Tes, and Tfing (representing vasoconstriction in acral sites) (1) controlling the shivering response. Shivering responses have been shown to be predicted by linear dependency betweenT̅ sk and Tes or tympanic, spinal, or rectal temperatures (2, 15, 28, 29). In various model schemes, output of the thermocontroller (shivering thermogenesis) is described as being proportional to the difference between the sensed value (skin and core thermal signals), the controller variable (e.g., hypothalamic temperature), and various thermal reference or “set” points (15,28). Hammel (15) proposed that shivering thermogenesis could be regulated by such a proportional control scheme with different thermal sensitivities emanating from different thermal inputs. Several thermal models predict metabolic response to cold with use of such a proportional control scheme from measurements ofT̅ sk and Tc in which there is a linear dependence of these two variables (12, 15). Others (28, 29) have shown that shivering thermogenesis is closely associated by a multiplication of thermal signals from the body core and mean weighted skin sites and heavily influenced by percent body fat.

The general results of a decrease in shivering thermogenesis but increase in heat flux responses shown in the luteal phase are consonant with Boulant’s (4) model and Hammel’s concept (15) that activation of warm-sensitive neurons is accompanied by reciprocal inhibition of cold-sensitive neurons. Elevated levels of estradiol and/or progesterone functioning during a woman’s luteal phase may possibly trigger heightened activity in warm-sensitive neurons in the preoptic/anterior hypothalamus during cold stress (4, 21). Activation of warm-sensitive neurons stimulating heat loss responses has been shown to cause a dampening of cold-sensitive neurons that stimulates shivering (4, 26). However, extensive interaction (and possible competition) between thermal inputs from skin and deep body thermoreceptors probably exists, tempering the final summed shivering responses in women as a function of the menstrual cycle (4, 5, 15).

In summary, during cold transient stress in resting women at two stages of the menstrual cycle, 1) a decreased slope in the luteal phase was observed when shivering thermogenesis was plotted as a function of integrated mean body temperature, 2) in the luteal phase, women dressed in an ensemble equivalent to street clothing displayed an increase in cutaneous heat flux, but rate of heat debt, assessed by partitional calorimetry, was reduced, which extended cold tolerance, and 3) the control of shivering thermogenesis in both phases was found to be influenced by a linear combination of Tc(Tes) andT̅ sk. Tfing (e.g., thermal inputs from acral areas) is multiplicative with other thermal signals influencing total control of the shivering response.


We thank the subjects for participation in the study; W. F. Allison, Nisha Charkoudian, Janet E. Staab, Michelle Mayo, and Jane P. Deluca for help with data collection, data reduction, graphical presentation, and general biochemical analyses; and especially Robert F. Wallace for statistical analyses and consultation.


  • Address for reprint requests: R. R. Gonzalez, Biophysics and Biomedical Modeling Div., USARIEM, 5 Kansas St., Natick, MA 1760-5007.

  • This study was funded, in part, by a Congressional Special Interest Medical Research Program Grant focused on Defense Women’s Health.

  • Human subjects participated in these studies after giving their free and informed voluntary consent. US investigators adhered to US Army Medical Research and Materiel Command (USAMRMC) Regulation 70-25 on Use of Volunteers in Research. Citations of commercial organizations and trade names in this report do not constitute an official Department of the Army endorsement or approval of the products or services of these organizations. The views, opinions, and/or findings contained in this report are those of the authors and should not be construed as an official Department of the Army position, policy, or decision unless so designated by other official documentation.


Rate of heat storage and integrated mean body temperature.

In the cold, initial mean body temperature (T̅ b,0 ) is often calculated from a steady-state weighting ratio ofT̅ sk to rectal temperature (T̅ sk/Tre) of 1:2 (7). In the heat or during exercise the probable ratio varies from 1:4 to 1:9 when Tes or Tre is used as a measure of Tc (12, 28). During exercise or Ta transients when body temperatures are in a non-steady state, coefficients for mean body temperature change appreciably (12, 21). The calculation of the classic Burton (7) mean body temperature by a simple weighting of Tc and Tsk is inaccurate during thermal transients and probably not applicable in women, since time and Ta, as well as Tsk and Tc, vary.

Rate of storage of body heat (S), evaluated by partitional calorimetry (3, 12, 30), is directly associated with the rate of change of integrated (i.e., weighted signals from peripheral and central thermoreceptors) mean body temperature ( ΔT¯ bt). This form was used in the present study to quantify responses during the cold ramps (12, 30), in which±S=(λ·mb/AD)·ΔT¯b/Δt(W/m2) Equation A1 where λ is the specific heat of the body in 0.965W ⋅ h ⋅ °C−1 ⋅ kg−1(or 3.49 kJ ⋅ °C−1 ⋅ kg−1),m b is the body weight in kg, and Δt is time in hours.

For this study, during the resting period at Ta of 20°C before Ta drops,T̅ b,0 was first calculated by a 1:4 weighted average ofT̅ sk and Tes.T̅ b,i(°C) was then determined asT¯b,i=T¯b,0+0t(ΔT¯b/Δt)dt Equation A2a or=T¯b,0+[(S·AD)/(λ·mb)]·Δt Equation A2b where Δt is the time interval (tx t 0) of a run taken at eachT̅ b(min/60) in hours.

b,i at each time point is evaluated byΔTb,i/Δt=[M(W)(R+C)Esk]·ADλ·mb(°C) Equation A3 where λ is the specific heat constant,A D is the DuBois surface area (in m2), and the energy exchange terms in brackets (all in W /m2) are evaluated by partitional calorimetry (12, 30).


View Abstract