INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

PHYSIOLOGISTS often use daily measurements of morning core temperature (T_c) as a marker to predict different phases of the menstrual cycle. Despite its variability and relative inaccuracy, this method has been used to differentiate the follicular or luteal phases of the menstrual cycle, enabling significant work on temperature regulation (2, 16), fluid volume regulation (9, 15), metabolism (14), and exercise responses (10) to be conducted. T_c monitoring has also been used in fertility assessment (3, 14) to detect impending ovulation. However, the key marker of impending ovulation, a drop in T_c, was only observed in ~50% (7) to 80% (19) of patients studied. In those studies, a single measurement of oral or rectal temperature was taken at fixed times in the morning. However, the combined influence of circadian and menstrual cycles on T_c made it difficult to identify ovulation, because a single temperature measurement was located on the sliding scale of the circadian rhythm.

The purpose of the current study was to determine whether T_c telemetry can be used in ambulatory women to identify menstrual cycle phases. We hypothesized that the sensitivity of T_c telemetry would enable detection of impending ovulation. We present a method whereby we record T_c at 1-min intervals over several days during different phases of the menstrual cycle. This methodology allows us to separate the menstrual cycle and circadian effects on T_c. We also present a methodology and procedure for quick computer analysis of the data. These methods enable us to track the changes in the circadian T_c rhythm and to identify particular phases of the menstrual cycle, such as preovulation. This technology is noninvasive, and the methods appear to be robust enough to withstand the vagaries of the normal lifestyle of ambulatory subjects. Using this methodology, we have been able to confirm both the elevation in mean T_c and the reported (5) decrease in amplitude of the circadian T_c rhythm in the luteal compared with the follicular phase. In addition, we have documented an overall decrease in T_c and the predicted minimum T_c (T_c-min) just before the urinary LH peak.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Eumenorrheic female subjects between the ages of 22 and 43 yr were recruited from a population of college students and staff from Wellesley College and the US Army Research Institute of Environmental Medicine, without bias for race or ethnicity. The study was approved by the Institutional Review Boards of both institutions, because recruitment and work occurred at the two locations. Subjects were nonsmokers; were not taking oral contraception or chronic non-steroidal anti-inflammatory drugs such as aspirin or ibuprofen, or melatonin or herbal supplements; and maintained a regular sleep-wake cycle. The presence of ovulatory cycles was documented by obtaining oral baseline temperatures with a digital thermometer at normal arousal time (usually between 0600 and 0730) for a complete cycle before testing and also during testing. T_c was monitored in each subject under her normal living conditions at four phases of the menstrual cycle (Table 1) by use of a calibrated temperature telemetry pill (Human Technologies, St. Petersburg, FL). The silicone-coated pill was easily swallowed with water and passed through the gastrointestinal tract in an average time of 3 days. A small amount of carbohydrate was eaten to move the pill into the intestines. The pill transmitted the local internal temperature to a receiver/datalogger (FitSense Technologies, Wellesley, MA), which was worn in a padded waist pack. Data continued to be collected while the subjects showered and dressed, if the receiver was placed within 4 feet of the subject. From our knowledge of the oral temperatures in the previous cycle and from continued monitoring of oral temperature, we attempted to begin data collection on the days noted in Table 1. Each subject received 3-4 precalibrated pills per session. She monitored the receiver for a loss of signal and then ingested another pill while in contact with one of the research team. Collection continued for 3-4 days or until the last pill was lost (7 days). The receivers stored 8 days of data, which were collected each minute. Data were downloaded into an IBM-compatible computer at the end of each session and transferred into Excel files. Subjects tested their urine for luteinizing hormone (LH) each morning and evening with a commercially available test stick (courtesy of Carter-Wallace, Cranbury, NJ) to assess the periovulatory period. Such tests are a semiquantitative indicator of rising and falling urinary LH levels during ovulation.

The T_c data were subjected to cosinor analysis with custom software (8) developed in the MatLab (Mathworks, Natick, MA) programming environment by use of an IBM-compatible computer (>300 MHz). The program works on Excel data files and automatically removes extraneous data (temperatures <35°C and >40°C), codes the sleep periods, calculates the days relative to the LH peak, and compiles the results in Excel worksheets. Filtering the data (1, 2, or 4 h) with fast-Fourier transform had no effect on circadian parameters. Several 24- and 36-h data segments (noon to noon, midnight to midnight, noon ± 18 h, and midnight ± 18 h) were tested. We selected the 36-h segments of the data between noon and ±18 h, because these had the best fit as judged by correlation (r²) values. The program returns 24-h cosine segments for plotting midnight to midnight, as well as the mesor, amplitude, period, and acrophase for each segment. The program also provides the maximum (T_c-peak) and the T_c-min for the cosine curve fit. All parameters were tested for significant differences using a one-way ANOVA for a repeated-measures experimental design. A Student-Newman-Keuls post hoc test was done to determine critical differences between phases.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Data were collected from four subjects over 11 menstrual cycles (Table 2). In several cases we missed the periovulatory changes on the first attempt and had to collect during a subsequent cycle. The circadian rhythms in T_c for each subject are graphed in Fig. 1, A-D. The mean T_c for each graph is indicated by the dotted line; however, this value could be biased by the number of observations in each phase of the cycle. Solid horizontal lines at the top of each graph represent the sleep periods. The women in this study maintained a relatively consistent sleep/wake schedule, when one considers the vagaries of normal family life. The shifts in mean T_c and change in amplitude of the circadian rhythm throughout the cycle can be clearly recognized in each of the graphs. On visual inspection, several consistent patterns are evident in the graphs. First, the daily mean temperature (mesor) was higher during the luteal phase than during the follicular phase. Second, the daily amplitudes were damped in the luteal compared with the follicular phase. Third, before the LH peak, the daily T_c-min reached its lowest value (nadir). Although T_c maximum values were also suppressed during this phase, these differences were more obvious when we looked at the nadir for each day rather than the peak. Statistically, the mesor (P < 0.001) and T_c-min (P < 0.0001) were significantly higher, and the amplitude was significantly lower (P < 0.005) in the luteal phase of the cycle compared with the three other phases (Table 3). The preovulatory phase was notable in that both the mesor (P < 0.001) and the T_c-min (P < 0.0001) were the lowest temperatures seen throughout the cycle (Table 3). There were no differences in the period of the circadian T_c rhythm among the four phases of the cycle. The r² values in Table 3 are averaged over all of the daily r² values for 36 h and over all subjects for the given phase. The overall mean (±SD) for the r² values was 0.83 ± 0.1, n = 72. 

View larger version (55K): [in this window] [in a new window]   Fig. 1.   Core temperatures collected every minute during selected phases of the menstrual cycle. The fine gray line is a plot of the individual data points; the darker line is a sequence of 24-h segments of individual 36-h cosinor fits. The light dashed horizontal line is mean temperature for the combined data in each graph. The dashed heavy line (top) indicates periods of sleep. Arrows indicate either the beginning of menses or day of the urinary luteinizing hormone (LH) peak and are so marked. Phases of the menstrual cycle are indicated below graphs. A: subject S01. Data are from cycles L (luteal/menses and follicular), J (periovulatory), and A (luteal). B: subject S03. Data are from phases A (follicular, luteal/menses) and D (periovulatory). C: subject S05. Data are from phases B (follicular, luteal/menses) and A (periovulatory). D: subject S07. Data are from phases A (menses, follicular, luteal) and C (periovulatory).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have described and tested a methodology for recording the circadian core temperature rhythm in ambulatory subjects for extended periods of time. This has been coupled with a fast, reliable method for applying cosinor analysis to the resulting data files. Using these methodologies, we have confirmed the effects of the female reproductive hormones on the mesor and amplitude of the circadian T_c rhythm during the follicular and luteal phases. More importantly, this method clearly demonstrates the fall in both the mesor and T_c-min 1-2 days before the surge of urinary LH.

Clinically, in fertility studies, a rise in oral morning temperature has been used to confirm ovulation and can also be used to grossly predict a time frame for conception. The present data confirm the rise in mesor and the reported decrease in amplitude of the circadian temperature rhythm in the luteal phase of the cycle, which were based on single 24-h segments of measurement (5, 11, 13). However, the current data indicate that both the rise in mesor and decrease in amplitude are not immediate responses but adjust over a 4- to 6-day time frame, which makes it difficult to pinpoint ovulation.

On the basis of observation of oral temperature, prior evidence suggested that ovulation might be preceded by a drop in T_c (7, 14, 19). More recently a decreased regulated body temperature has been associated with the time of the preovulatory estrogen rise (17) and with estrogen replacement in postmenopausal women (4, 18). The data reported here unequivocally demonstrate a significant decrease in both the mesor and the T_c-min within 1 and sometimes 2 days before the urinary LH surge. This temperature drop occurs about 24-48 h before ovulation, and if measured accurately, could provide an important noninvasive tool for predicting ovulation for fertilization studies.

Although we did not measure the plasma levels of estrogen directly, we predict that the fall in T_c-min just before the urinary LH surge is due to the preovulatory estrogen surge. A good correlation has already been established between increased estrogen levels and lowered core temperature in postmenopausal women (4, 12, 18), and the addition of progesterone to these subjects reversed the hypothermic estrogen effect (12).

Although the transition in circadian T_c rhythm from the follicular to luteal phase may take up to 6 days, the transition from luteal phase to menses is relatively rapid and consistent across subjects. For example, the mesor decreased significantly within 24 h and continued to fall during the next 24 h. In three of the four subjects, the amplitude had increased within 24-48 h as well. This rapid and clear transition that takes place when both estrogen and progesterone levels are dropping is in contrast to the complex interactions and changes seen during the transition at ovulation.

In studies in which ambient light, eating schedules, and activity have been controlled, there was a demonstrable delay in the acrophase in the luteal portion of the cycle (5). Although our data suggest this trend, none of the differences in acrophase were statistically significant. This is not surprising, because changes in acrophase are masked in ambulatory subjects (1) and modified by changes in ambient lighting (6).

Ambulatory monitoring of T_c with the temperature pill can be used post hoc to identify characteristics of changing menstrual cycle phases in individual women. Although measurements of morning oral temperature have been useful for gathering general patterns, there is a great deal of variability, because the core temperature is rising rapidly from its nightly nadir. Consequently, differences in measurement time of 1-2 h can substantially alter the measured temperature and mask important changes in T_c, such as those associated with ovulation.

Regarding the analysis method, although it is based on the cosinor analysis, which has been used extensively for biological rhythm studies, we were able to test the effects of using different data segments and different levels of data filtration. From a qualitative analysis of these results, we found that the best analysis of circadian variability is to use data from 1800 of one day through 1800 on the next day to 0600 of the following morning (a total of 36 h). This provides enough data to fit both the peak and the nadir of the 24-h period adequately. We also found that filtering the data before running the cosinor analysis had little or no effect.

In summary, we propose that the use of an ingestible temperature-sensing pill in conjunction with automated cosinor analysis is an effective tool in evaluating circadian T_c rhythms in human subjects. It is particularly effective in identifying the preovulatory decline in T_c and T_c-min. In addition, we are impressed that the effects of the female reproductive hormones on the circadian pattern of T_c rhythms are so robust that the pattern can be easily quantitated in ambulatory women who are not subjected to controlled lighting, sleep/wake patterns, or activity.

Perspectives