Circadian rhythm of frequency-domain measures of heart rate variability in pregnancy


Correspondence: Dr E. M. K. Ekholm, Department of Obstetrics and Gynaecology, Turku University Hospital, SF-20520 Turku, Finland.


Objective To examine frequency domain measures of heart rate variability and their circadian rhythms in pregnancy.

Design A longitudinal study.

Setting University hospital in Turku, Finland.

Participants Sixteen healthy women between 11 and 27 weeks of pregnancy; 12 women before pregnancy; and four women postpartum.

Main outcome measures Heart rate variability as measured in frequency domain from 24-hour ambulatory electrocardiography.

Results Pregnancy was associated with a lower standard deviation of R-R intervals (P < 0.01), with reduced very low (P < 0.05), low (P < 0.01), and high frequency (P < 0.05) power spectral components of heart rate variability. The high frequency power was lower at night in pregnancy, but similar in the daytime in pregnant and nonpregnant women.

Conclusions Pregnancy is associated with an overall reduction in heart rate variability, most markedly reflected in the low frequency component. This suggests altered baroreflex or sympathetic modulation of heart rate, and decreased vagal activation at night.


Maternal haemodynamics undergo profound changes during pregnancy. This adaptation is deficient in pregnancies complicated with hypertension1 or maternal diabetes2. Pregnancy has been reported to exacerbate both supraventricular and ventricular arrhythmias3,4. The autonomic nervous system may be a principal system conforming the haemodynamics to various changes and facilitating arrhythmogenesis. However, there is little information about the effects of pregnancy on cardiac neural regulation. Previous results of autonomic modulation of heart rate in the laboratory environment show that both high frequency and low frequency components of heart rate variability are decreased in the second trimester of pregnancy5. High frequency variability tends to be smaller6, whereas low frequency and very low frequency has been reported to remain unchanged during the third trimester of pregnancy5,6. As the laboratory setting affects cardiovascular responsiveness7, we decided to study heart rate variability using ambulatory 24-hour electrocardiography. Twentyfour hour recordings permit a reproducible analysis of autonomic regulation and allow estimation of circadian changes in heart rate variability8. This study was specifically aimed at examining the effects of pregnancy on frequency domain measures of heart rate variability and their circadian rhythms without the interference of a laboratory setting or interindividual variability.


Sixteen healthy women (mean age 27 years) were examined during the 20th week of pregnancy (range 11 to 27 weeks). The nonpregnant state for mean 24-hour variabilities was assessed in 12 women 0 to 37 weeks prior to the last menstrual period before pregnancy and in four women 10 to 31 weeks postpartum. Postpartum measurements were included, since we have previously shown that autonomic cardiovascular control is similar before pregnancy and postpartum9. Thirteen of the women were nulliparous. The course of pregnancy was uncomplicated in all the women.

Oral consent was obtained from each woman and the study was approved by the Ethics Committee of the Turku University Hospital.

A 24-hour ambulatory electrocardiogram (ECG) was recorded from all the women during normal daily activities with their normal sleep-wake rhythm. The two-channel recordings were analysed with a Delmar Avionics scanner.

Analysis of heart rate variability

The ECG data was sampled digitally and transferred from the Delmar Avionics scanner to a microcomputer. Custom-made software (Heart Signal Co, Finland) was used for heart rate variability analysis. The intervals between consecutive pairs of R-waves in the ECG (R-R interval) were measured. A linear detrend was applied to the R-R interval data segments of 512 samples to make them more stationary. This was implemented by first fitting a straight line to each segment by a standard least-squares method and then subtracting it from the sample value. The R-R interval series was passed through a filter that eliminated unwanted premature beats and noise, and filled the resulting gaps with an average value computed in the immediate neighbourhood. An R-R interval is interpreted as a premature beat if it deviates from the previous qualified interval by more than a given tolerance level. This is a programmable parameter dependent on the prematurity index of ectopic beats for each woman. The details of this filtering technique have been described previously10,11. Only segments with > 90% qualified beats were included in the analysis.

Time domain analysis was done by calculating the standard deviation (SD) of successive R-R intervals. The average 1-hour and 24-hour R-R intervals, and the power spectrum components of heart rate variability

were calculated from 512 R-R interval segments. An autoregressive model was used to estimate the power spectrum densities of R-R interval variability. The computer program automatically calculated the autoregressive coefficients to define the power spectrum density. Power spectra were quantified by measurement in four frequency bands: 1. total power < 0.40 Hz, 2. high frequency power from 0.15 to 0.40 Hz, 3. low frequency power from 0.04 to 0.15 Hz and 4. very low frequency power from 0.0033 to 0.04 Hz. The ratios between low frequency and high frequency spectra in fractional units were also calculated (low frequency to high frequency ratio). The mean values for the sleeping hours and waking hours and total 24-hour period were calculated for each subject.


A Wilcoxon signed rank sum test was performed to compare the groups. P < 0.05 was considered significant.


Average 24-hour measures of heart rate variability

Mean heart rate during each 24-hour period was significantly higher in pregnancy than either before or after pregnancy (Table 1).

Table 1.  Effect of pregnancy on mean heart rate variability. Values are given as mean (SD). HF = high frequency; LF = low frequency; VLF = very low frequency; NS = not significant
R-R interval (ms)   
 24-hour806 (69)736 (67)<0.01
 Daytime844 (84)675 (65)<0.0001
 Sleeping hours843 (84)674 (65)< 0.0001
 SD of R-R interval (ms)102 (22)79(26)<0.01
HF power (ms2)   
 24-hour1041 (802)690 (798)< 0.05
 Daytime530 (288)460 (460)NS
 Sleeping hours1760 (1621)1082(1581)<0.05
LF power (ms2)   
 24-hour1430 (551)757(720)<0.01
 Daytime712 (249)389(271)<0.01
 Sleeping hours1625 (892)861 (1057)<0.01
VLF power (ms2)   
 24-hour2567 (1279)1594(1058)<0.05
 Daytime2073 (U60)1347 (655)<0.05
 Sleeping hours3187 (1459)1930 (1694)<0.05
LF:HF ratio   
 24-hour1.72 (0.64)1.56 (0.70)NS
 Daytime1.62 (0.74)1.30 (0.61)NS
 Sleeping hours1.15 (0.48)1.20 (0.69)NS

The average 24-hour high frequency, low frequency and very low-frequency powers were all significantly lower in pregnancy compared with the nonpregnant state. Low frequency to high frequency ratio did not change in pregnancy (Table 1).

Circadian effects on heart rate variability

All power spectral components of heart rate variability showed a circadian rhythm with higher values during the sleeping hours than in the daytime in pregnancy as well as the nonpregnant state (Table 1). However, the circadian rhythm of the high frequency component was blunted: the high frequency power was lower at night in pregnancy than in the nonpregnant state, but similar in the daytime (Fig. 1).

Figure 1.

Circadian rhythm of mean (standard error of mean) values of high-frequency spectral area of heart rate variability was blunted in pregnancy (▾) compared with the nonpregnant state (○) (P < 0.05).


We found that the SD of successive R-R intervals and all 24-hour power spectral components of heart rate variability became reduced in pregnancy. Notably, the SD of successive R-R intervals and the low frequency component of heart rate variability were most markedly reduced in pregnancy, whereas the high frequency component of heart rate variability was reduced only during the sleeping hours.

High frequency oscillation of the heart rate reflects mainly vagal modulation. The low frequency component of heart rate variability is thought to be mediated by vagal activity at rest, and both by vagal and sympathetic activity in an upright posture due to baroreflex control. It is known to decrease during adrenergic stimulation12, perhaps due to saturation of the sympathetic tone resulting in a reduced capability of sympathetic modulation. The present findings of a reduced low frequency component of the heart rate may be explained by reduced baroreflex sensitivity and/or sympathetic modulation of the heart rate in pregnancy as compared with the nonpregnant state. However, the physiological origin of very low and low frequency components of heart rate variability without controlling for external conditions are not definitely known. Therefore, the physiological meaning of the present finding of a marked reduction of low frequency modulation of heart rate should be cautiously interpreted. Nonetheless, the findings may have clinical significance, because decrease of the very low frequency component of heart rate variability has been shown to predispose to arrhythmic events13.

There is growing interest in the analysis of circadian variation of heart rate variability with respect to abnormalities of cardiac neural regulation. The 24-hour high frequency component of heart rate variability was only marginally reduced in pregnancy, but the circadian fluctuation of this respiratory vagal component was blunted so that the high frequency component was reduced only during sleeping hours in pregnancy. These data suggest that pregnancy reduces the capability of the vagus to activate normally during sleep. Similar blunting of circadian rhythm of vagal activity has been described in patients with clinical heart disease14. This circadian alteration in autonomic control of the heart rate may play a role in the exacerbation of cardiovascular disorders in pregnancy.

A short heart rate recording during daytime did not reveal a pregnancy-induced change in very low frequency oscillation5. Furthermore, another study did not find any significant changes in heart rate variability in late pregnancy as compared with nonpregnant women6. Previously, heart rate variability was measured from short recordings performed in a laboratory setting during daytime. The different methodology of the earlier studies might explain the contradictory results, as the laboratory setting is known to affect cardiovascular responsiveness7. Further, the discrepancy between these and earlier results may be because of different study protocols: previous studies were cross-sectional, whereas this study evaluated the same women as pregnant and nonpregnant. A longitudinal study gives more reliable information, as there is a wide interindividual variation in cardiovascular responsiveness15.

The decrease in autonomic reserves may be essential for the mothers with cardiac problems or hypertension. The marked increase in blood volume with decreased autonomic reserves is dangerous especially for the mothers with impending heart failure4. Dysfunctional changes in haemodynamic and autonomic responsiveness might lead to the haemodynamic manifestations of pre-eclampsia. Indeed, pre-eclampsia is associated with signs of autonomic dysfunction, such as changes in short term haemodynamic control16, increased nocturnal motility17 and hypertensive crises at nightime18.


Marked impairment of low frequency modulation of the maternal heart rate, and blunted vagal activation during sleep, may reduce the adaptive capacity of the cardiovascular system during pregnancy.


This work was supported by grants from the Ida Montin Foundation, Finland, the Varsinais-Suomi Regional Fund of the Finnish Cultural Foundation, Finland, the Finnish Foundation for Cardiovascular Research, Helsinki, Finland, and the Medical Council of Academy of Finland, Helsinki, Finland