Effect of Sex, Menstrual Cycle Phase,and Oral Contraceptive Use on Circadian Temperature Rhythms

Summary: The circadian rhythm of rectal temperature was continuously recorded over
several consecutive days in young men and women on regular nocturnal sleep schedules.
There were 50 men, 21 women with natural menstrual cycles [i.e., not taking oral contraceptives
(OCs) (10 in the follicular phase and I 1 in the luteal phase)], and 14 women
using OCs (6 in the pseudofollicular phase and 8 in the pseudoluteal phase). Circadian
phase and amplitude were estimated using a curve-fitting procedure, and temperature
levels were determined from the raw data. A two-way analysis of variance (ANOVA) on
the data from the four groups of women, with factors menstrual cycle phase (follicular,
luteal) and OC use (yes, no), showed that temperature during sleep was lower during the
follicular phase than during the luteal phase. Since waking temperatures were similar in
the two phases, the circadian amplitude was also larger during the follicular phase. The
lower follicular phase sleep temperature also resulted in a lower 24-h temperature during
the follicular phase. The two-way ANOVA showed that temperature during sleep and
24-h temperature were lower in naturally cycling women than in women taking OCs. A
one-way ANOVA on the temperature rhythm parameters from the five groups of subjects
showed that the temperature rhythms of the men and of the naturally cycling women
in the follicular phase were not significantly different. Both of these groups had lower
temperatures during sleep, lower 24-h temperatures, and larger circadian amplitudes than
the other groups. There were no significant differences in circadian phase among the
five groups studied. In conclusion, menstrual cycle phase, OC use, and sex affect the
amplitude and level, but not the phase, of the overt circadian temperature rhythm. Key
Words: Circadian rhythms-Core body temperature-Menstrual cycle-Birth control
pills-Oral contraceptives.

There have been very few comparisons of circadian body temperature rhythms between
young men and women. Winget et al. (1) reported that men had larger circadian amplitudes
and lower mean levels than women, but they did not find a sex difference in circadian
phase. Wever (2) measured the temperature rhythms of free-running, internally
synchronized subjects and also found larger amplitudes and lower mean levels in men.
Rogacz et al. ( 3 ) studied women across the menstrual cycle and noted that they had
smaller amplitudes than men previously studied in their laboratory, regardless of menstrual
phase.
Hormonal changes across the female menstrual cycle affect the circadian temperature
rhythm, with progesterone having a thermogenic effect (4). The menstrual cycle
can be roughly divided into two phases: follicular (the time froin menses onset to ovulation)
and luteal (the time from ovulation to menses). Progesterone levels are low during
the follicular phase and high during most of the luteal phase, peaking in the middle
of the luteal phase. Estrogen levels are low at the beginning of the follicular phase, rise
during the last few days of the follicular phase, and are reduced but remain high during
the luteal phase (5,6). Women taking the commonly prescribed oral contraceptives
(OCs) receive an estrogen plus a progesterone for 2 I days of the 28-day cycle and no
active ingredients during the other 7 days (7).
There have been several studies of the circadian temperature rhythm in women with
natural menstrual cycles, i.e., in those not taking OCs. The amplitude of the rhythm IS
larger during the follicular phase than during the luteal phase, primarily due to a lower
nocturnal temperature, which also results in a lower daily mean temperature level
(3,8-13). Longitudinal graphs of temperature from single subjects across the entire
menstrual cycle show the interaction between the circadian and infradian (menstrual)
cycles (8,14). Most studies find no change in circadian phase across the menstrual cycle
(1 0,l I , 15,16), but a few preliminary studies suggest a phase delay during the luteal
phase (9,13). There is a paucity of data regarding the effects of OC on the circadian
temperature rhythm.
Given the differences in the temperature rhythm across the menstrual cycle, those
who make comparisons between women and men need to take female menstrual cycle
phase into account. The purpose of this report is to compare the circadian temperature
rhythms of men and women, taking menstrual cycle phase and OC use into account.
METHODS
The data were taken from the baselines of various studies performed in our laboratory
(1 7-20). During baseline, which was either 7 or 10 days long, participants adhered
to a regular sleep schedule, with fixed bed times and wake times and exactly 8 h in bed.
Subjects were required to remain in bed, in the dark, for the entire 8 h, even if they could
not sleep. The time of the sleep period was similar to the subjects' habitual sleep schedule,
except that the sleep times were closer to the weekend sleep times to avoid masking
the minimum of the temperature rhythm by early wake times. The experimental
schedule was also fixed and regular, in contrast to the irregular schedules of most young
adults. To ensure compliance to the sleep schedule, subjects were required to call a timestamp
answering machine just before bed time, immediately after waking, and 30 min
after waking. They were also required to keep daily sleep logs of estimated sleep onset
and wake times. In general, the reported sleep times were similar to the scheduled sleep
times. Caffeine consumption was restricted to either no caffeine at all or else only in
the first 4 h after wake time and in the exact same amount every day. Alcohol and recreational
drugs were prohibited.
Before the studies began, female subjects were asked if they were using OC and when
their last menstrual period began. During the studies they checked a box on one of the
daily logs to indicate whether or not their period started on that day. For women not
using OC, conservative estimates of the two primary menstrual phases were made by
counting forward or backward from the day of menses onset. The average length of the
luteal phase is - 14 days and is rarely < 12 days (21). Therefore, we classified the 12
days prior to the onset of menses (reverse cycle days - 1 to - 12) as luteal days. The
average length of the follicular phase is - 13 days and is rarely < 10 days (22). Therefore,
we classified the day of menses onset and the following 9 days (forward cycle days
0 to +9) as follicular days. Since we used only data from the 22 days of the menstrual
cycle centered around the day of menses onset and did not include days close to the follicular
to luteal transition, we avoided classification errors. Given this method, if one
or both phases of a particular cycle were longer than usual, this would not cause a misclassification.
However, a misclassification could conceivably occur if the particular
Meal or follicular phase included was extremely short and happened to end during
baseline.
For women using OC, we defined the “pseudoluteal” phase as the 2 I days during
which exogenous estrogen plus progesterone are taken, to correspond to the natural
luteal phase with its increased levels of estrogen and progesterone. We defined the “pseudofollicular”
phase as the 7 days without exogenous hormones, to correspond roughly
to the natural follicular phase with low progesterone. For women taking OC, menstruation
occurs during the week with no exogenous hormones (7), beginning on about the
second day. Therefore, we made conservative estimates of pseudoluteal and pseudofollicular
days counting from the day of menses onset. Pseudofollicular days included
days - 1 to +2, and pseudoluteal days included all days except days -5 to +5.
For each subject, only the data from one phase of the cycle (luteal or follicular) were
used: the phase with more baseline data. In cases where participants had exactly equal
numbers of baseline days in their luteal and follicular phases, they were randomly
assigned to one of the groups and the data from the other phase were discarded.
Core body temperature was continuously measured using a rectal temperature probe
and either a Vitalog PMS-8 or Consumer Sensory Products AMS-1000 portable monitor
programmed to store measurements every minute. The probes were inserted and maintained
at a constant distance (1 0 cm). Subjects were told to take the probe out for exercise,
baths, showers, and sex so that extremely elevated temperatures were not recorded.
The temperature data of each subject were averaged into 15-min points and divided
into 24-h sections. The sections started at 2 1 :00 so that each section contained a complete
nocturnal sleep episode. If the data during sleep were missing or if half or more
data for any 24-h section were missing, the entire 24-h section was not used. We included
only subjects for which we had at least three 24-h sections.
The 24-h sections for each subject were averaged, and then two measures of temperature
level were calculated. The average temperature during sleep was calculated for
each subject using the scheduled bed time and wake time for each subject. Thus, this
measure was actually the average temperature during in-bed time. The 24-h temperature,
or mean temperature during the whole 24-h period, was also calculated.
A curve consisting of the fundamental cosine curve plus three additional harmonics
(the 12-, 8-, and 6-h harmonics) was fitted to the average curve of each subject. The
time of the minimum was used as a phase marker. The minimum was chosen, rather
than the maximum, because the minimum occurs during sleep, a time of relative inactivity,
and is thus less affected by masking from activity. Fitting a curve with three harmonics
smoothed the data and provided a single minimum. This method is similar to
visually picking the minimum, but more objective. More harmonics could have been
added to the fitted curve, but as more are added, the fitted curve approaches the actual
data curve and is not as smooth. Circadian amplitude was the difference between the
fitted curve’s maximum and minimum. Although some subjects were studied during
daylight savings time, all times are reported in central standard time.
Each variable was evaluated with a one-way analysis of variance (ANOVA) with the
factor group (women in the follicular phase, women in the luteal phase, women in the
pseudofollicular phase, women in the pseudoluteal phase, men) followed by Tukey’s
post hoc painvise comparisons of means. The same variables in only the four groups of
women were also evaluated with a two-way ANOVA with the factors OC use (yes, no)
and menstrual phase (follicular, luteal).
RESULTS
The data from 50 male and 35 female subjects were analyzed (see Table I ) . For illustrative
purposes only, the average temperature curves of the individual subjects were
averaged together to produce average curves for the five groups of subjects (Figs. 1 and
2). These graphs were made from raw (as opposed to fitted) data. The graphs show that
the differences among the groups were most pronounced at night, during sleep.
Temperature Level
The two-way ANOVA produced a significant main effect of menstrual phase for temperature
during sleep (F = 24.182, df= 1,3 I , p < 0.001) and for 24-h temperature ( F =
13.959, df= 1,3 I , p < 0.0 1) and a significant main effect of OC use for temperature
during sleep (F= 14.476, df= 1,3 1, p < 0.01) and for 24-h temperature (F= 10.372, df’
= 1,3 I, p < 0.01). There were no significant two-way interactions. These analyses showed
that temperature was significantly lower during the follicular phase than during the
luteal phase. The analyses also showed that temperature was significantly lower in non-
OC users than in OC users.
The one-way ANOVA was significant for temperature during sleep (F = 47.08 I , df
= 4,80, p < 0.001) and 24-h temperature (F= 23.971, df= 4,80, p < 0.001). There were
two groups that had similar temperature levels and differed from the other three: women
non-OC users in the follicular phase and men. They had the lowest temperatures during
sleep and the lowest 24-h temperatures. The post hoc tests showed that the differences
in temperature level between each of these two groups and the other three groups
were statistically significant (see Table 2). It is also interesting to note that although
there were no significant differences between OC users and nonusers during the luteal
phase, there was a significant difference between them during the follicular phase, with
the nonusers having lower temperatures.
Temperature Amplitude
The two-way ANOVA produced a significant main effect of menstrual phase (F =
1 1.322, dj= 1,3 1, p < O.Ol), but the main effect for OC use and the two-way interaction
were not significant. This analysis showed that amplitude was larger during the follicular
than during the luteal phase.
The one-way ANOVA was significant (F = 13.326, df= 4,80, p < 0.001). There were
two groups that had similar and larger temperature amplitudes than the other three
groups: women non-OC users in the follicular phase and men. The differences between
the men and each of the other three groups of women were statistically significant. The
differences between the women non-OC users in the follicular phase and the other three
groups of women reached significance for two of the three groups.
Temperature Phase
Temperature minima occurred around 5 a.m., -3 h before wake time .
There were no significant differences in time of temperature minima, in time of temperature
minima in relation to wake time, or in wake times.
Amplitude and Level: Comparisons Among Women
This report confirms previous studies of women that found luteal phase temperature
levels to be elevated, particularly during sleep, which in turn produced a smaller circadian
amplitude during the luteal phase (3,8-13). The increase in temperature levels during
the luteal phase is attributed to the thermogenic effect of progesterone (4-6). However,
it is not clear why the progesterone-induced temperature elevation would be greater
during sleep. Rogacz et al. (3) discussed the idea that temperature levels might be limited
by a physiologic ceiling that prevents the already high day time temperatures from
increasing past a certain point, so that only the lower nocturnal temperatures can rise.
They also suggested that progesterone might act directly on the endogenous pacemaker
to reduce amplitude. Another possibility could be that the effects of progesterone,
like those of drugs in general, depend on circadian phase. In other words, progesterone
might produce a larger rise in body temperature during one part of the circadian cycle
than another.
The increase in temperature level at the transition from the follicular to the luteal
phase is the basis for using basal body temperature as a marker of ovulation (23). However,
Fig. l reminds us that the term “basal” can be misleading, since temperature does
not remain constant during inactivity or sleep, but instead has a circadian rhythm. This
figure also shows how much the temperature recorded on waking in the morning could
change if the woman woke up earlier or later than usual, i.e., at an earlier or later point
on the temperature curve.
We found that women OC users had higher temperatures during the pseudofollicular
phase than women with natural menstrual cycles during their follicular phase. Progesterone
cannot be responsible for this difference because both groups have low levels
during their follicular phases. However, the naturally cycling women produce estrogens,
which can be high near the end of the follicular phase. Therefore, estrogen is probably
exerting a temperature-lowering effect (24). Lee (10) also reported relatively high temperatures
during the follicular phase in three OC users. Evidently, there is a need for
more research to determine how exogenously administered hormones affect body temperature
rhythms.
Higher temperatures during sleep could conceivably cause or be a result of poorer
sleep quality (25). Our data are not ideal for detecting differences in sleep duration,
because all subjects were required to remain in bed in the dark for exactly 8 h. Furthermore,
we did not have a measure that could reveal fine distinctions in sleep quality.
There is some evidence of sleep disturbance during the luteal phase (see 26 for
review), but more research is necessary to determine whether endogenous or exogenously
administered hormones affect sleep.
Amplitude and Level: Comparisons Between Men and Women
In this study, the women with natural menstrual cycles (non-OC users) who were in
the follicular phase had circadian temperature rhythms that resembled the men’s more
than they resembled those of the other groups of women. This group of women and the
men had the lowest temperatures during sleep, and consequently the lowest 24-h temperatures
and largest circadian amplitudes. The low temperatures in these two groups
are consistent with both having little or no circulating progesterone. In contrast, Rogacz
et al. (3) reported that women in both the follicular and the luteal phases had smaller
amplitudes than men. The other studies that compared temperature rhythms in men and
women did not account for menstrual phase ( I ,2). Thus, further research is necessary
to clarify the differences in temperature rhythm amplitude and temperature level between
men and women. Future studies should also take possible sex differences in physical
fitness into account, since larger temperature rhythm amplitudes have been found in
very physically fit than in inactive or average men, even when the temperature rhythm
was measured during standardized conditions (27,28).
Temperature Phase
We did not find any differences in the phase of the circadian rhythm of temperature
among the various groups of subjects. The only other study we know of that compared
temperature rhythms in young men and women also did not find a phase differcnce (1).
Studies of temperature phase in women across the menstrual cycle are more common.
Most found no phase differences ( 10,11,15,16), but two suggest a phase delay during
the luteal phase (9,13). In both of these studies, the data were collected during a constant
routine using a within-subject design. These methodological refinements may be
necessary to reveal phase differences. However, Wagner et al. ( 16) used both a constant
routine and a within-subject design, but did not find phase differences. More research
is necessary to resolve the issues of whether temperature phase changes with the menstrual
cycle and whether there is a sex difference. The use of other phase markers of the
circadian pacemaker besides the temperature rhythm would also help to clarify the issue.
However, there have been two studies using the circadian rhythm of melatonin, and these
did not reveal phase changes across the menstrual cycle (29,30).
The time of the minimum of the circadian temperature rhythm is currently of great
interest, because it is used as a marker of the crossover point between delays and advances
in the human phase response curve to light (e.g., 3 1,32). Thus, the clock time of the
minimum is important for practical applications of light treatment (e.g., 33). Constant
routine methods (3 1,34) may provide the most accurate measure of the temperatiire minimum
and are good for measuring the phase shift in situations in which the endogenous
temperature minimum does not occur during sleep and is thus subject to masking, e.g.,
during shift work. However, it is not clear whether the constant routine method offers
any benefit for determining the temperature minimum of normal subjects sleeping at
normal times. A few studies compared temperature phase measured during a constant
routine with those obtained during days with normal sleep. Two found no difference in
phase ( 1 3,35), one found a slightly later phase (36), and one found a slightly earlier
phase (37) during the constant routine. In our study, the clock times of the temperature
minima were similar to those from constant routine studies: For men: our study-5: 18;
constant routine studies4:31 to 5:02 (38), 5:19 (39), and 6:48 (40). For women not
taking OC: our study-5:06 in the follicular stage and 4:36 in the luteal stage; constant
routine s t u d y 4 : 18 during the follicular phase and 4:27 during the luteal phase ( 16).
Some of the variations in the time of the temperature minimum within and between
studies are undoubtedly due to factors such as individual differences in niorningnesseveningness
(4 1 ), previous sleep schedules, and daylight exposure schedules. For example,
in a previous study of ours in which wake times were -2 h earlier than in the present
study, the temperature minima were also -2 h earlier (42). Variations in the time
of the temperature minimum are also produced by the different methods for determining
the minimum, such as cosine curve fitting, smoothing procedures, visually picking
the minimum, etc. Another complication is daylight savings time. Most authors do not
indicate whether times are reported in daylight savings or standard time. It seems likely
that these factors are more important determinants of circadian phase than sex or
menstrual cycle phase.
In conclusion, more research is needed to compare the circadian temperature rhythms
of women in their various hormonal states and men. Meanwhile, our study of temper
ature rhythms shows that although sex, menstrual cycle phase, and contraceptive use
influence temperature rhythm amplitude and level, these factors may not have a significant
influence on temperature phase.