Short-term variation in abnormal antenatal fetal heart rate records P. Street, MRCS, G.S. Dawes, DM, M. Moulden, and C.W.G. Redman, MB Oxford, England In a retrospective study the relation of reduced fetal heart rate variation to fetal acidemia was analyzed with a computerized system for numeric analysis. Between 1983 and 1987, 78 pregnancies were identified in which at least one record of the fetal heart rate had very low long-term variation. The outcome was analyzed to determine the numeric criteria of fetal heart rate variation that most efficiently detect a fetus that will die (preterminal) or is dying (terminal). Because fetal compromise was found on occasion to be associated with a slow sinusoidal fetal heart rate rhythm that increased measures of long-term variation. It was necessary to define a new index of short-term fetal heart rate variation (the V,S minute epoch-epoch variation). This was closely related to long-term variation (r = 0.9) but provided better detection of preterminal records as judged by metabolic acidemia at delivery or intrauterine death. (AM J OSSTET GVNECOL 1991 ;165:515-23.)

Key words:Fetal heart rate variation, terminal fetal heart rate records, computer analysis, intrauterine death

Evaluation of the fetal heart rate (FHR) is a clinical test of well-being, that is based on empiric observations, of which many aspects are poorly understood. Grossly abnormal patterns have been identified by visual analysis but with large interobserver and intraobserver variation.'" The degree of such variation is so great as to cast serious doubts on the accuracy of current guidelines . To overcome the problems of subjective analysis, we have developed a system for numeric analysis of the human FHR, with a small on-line computer. The analysis depends on accurate fitting of the baseline and on simple techniques for measuring heart rate variation, accelerations, and decelerations. By quantifying signal loss, the quality of different recording systems can be assessed 5 ; the introduction of a wide-range ultrasonographic transducer, combined with autocorrelation, improved record quality." Originally we started from the observations of Visser and Huisjes 7 that identified decelerative FHR traces as a cause for concern. However, the feature that was outstandingly associated with acidemia, on delivery by cesarean section in the absence of labor, proved to be a decline in FHR variation; hence the focus of this article. Its shortest component (beat-to-beat variation) is too small to be measured accurately by the present From the Nuffield Department of Obstetrics and Gynaecology, University of Oxford. Supported by the Medical Research Council and the Department of Health . Received for publication October 16, 1990; accepted March 22,

1991 . Reprint requests: G. S. Dawes, Nuffield, Dept. ofOb/Gyn, University ofOxf01'd,john Radcliffe HospitaL, Oxford, England OX3 9DU. 611 /29775

generation of FHR monitors and cannot be reliably estimated by external ultrasonographic monitoring." 9 Measurement of longer-term FHR variation provided accurate information on the rest and activity cycles of the normal fetus 5 . 10. 11 and evidence of compromise associated with chronic hypoxemia. An index of longterm variation, the mean minute range, was devised , in which the effects of decelerations are excluded to identify records that are otherwise flat. As gestational age advances, the mean minute range increases a little, but much less than the incidence of accelerations ; at term its normal value is about 50 milliseconds corresponding to about 17 beats/min." Abrupt changes in the magnitude of the mean minute range identify episodes of low or high FHR variation, coincident with changes in fetal rest and activity from 28 weeks' gestation onward. We have shown that a 5-minute episode of active fetal sleep, identified in this way, is a better index of health than reactivity defined in terms of FHR accelerations. 12 FHR variation was much reduced in association with intrauterine growth retardation without acidemia. 13 This resulted from attenuation oft he amplitude ofvariation rather than from prolongation of episodes of low variation. 14 The mean minute range was reduced from a normal value of 44 ± 1.5 milliseconds at 32 weeks to 20.6 ± 1.2 milliseconds, and this was associated with a decrease in fetal Pao 2 measured at cesarean section before labor and with hypoglycemia, hyperalaninemia, and a rise in amniotic fluid erythropoietin as evidence of chronic hypoxemia .15 Hypoxemia in these circumstances has been confirmed by cordocentesis. I" In these studies the attributes of the FHR of the

515

516

Street et al.

chronically compromised fetus were defined, but the absence of acidemia at birth confirmed that the fetal state was not terminal. The concept of preterminal and terminal FHR records 7 has helped to identify the fetus that requires immediate delivery. In modern practice the qualities of the terminal trace cannot be studied formall y because, once recognized visually, immediate intervention is required. However, to examine the limits of intrauterine survival in relation to FHR variation, we reviewed all cases encountered over 4 years (1983 to 1987) in which at least one record had an abnormally low mean minute range «20 milliseconds), much less than the lowest with normal outcome. I7 We also identified a small number of fetu ses with a terminal heart rate pattern who could not be delivered because of immaturity and who died in utero. In one a slow sinusoidal rhythm was superimposed on an otherwise flat FHR, in such a way as to yield a high mean minute range. We therefore explored the relation between short-term and long-term measures of FHR variation in preterminal and other records and derived a new index of short-term variation to improve diagnostic discrimination. Methods

FHR records were taken with Hewlett-Packard model 8040 fetal monitors. The data were captured and analyzed by a microcomputer on line. From 1983 to 1985 the records were made with Nascom microcomputers. 5 . 18 From 1986 these were progressively replaced with Apricot XEN microcomputers, which accessed the RS232 port of the HP8040. The archive of Nascom disks was transferred to the new system and reanalyzed. The date and time (to the nearest minute) of the start of the record was entered manually (Nascom) or automatically (Apricot XEN). The analysis proceeded as follows . The autocorrelation function, generated by the fetal monitor as a measure to discriminate between background noise and valid signal, was sampled. Accuracy was further checked by an error algorithm . 19 Valid pulse intervals were averaged over epochs of 3.75 seconds Wl6 minute) and a baseline was fitted. 20 Signal loss was calculated as the proportion of epochs for which no valid pulse interval was available. The fetal monitor measures pulse intervals, not heart rate; therefore calculations of FHR variation were reported in milliseconds. Large changes in heart r ate from the baseline were identified as accelerations (> 10 beats/min and 15 seconds) or decelerations (> 10 beats/min and::::: 1 minute, > 20 beats / minute and :::::0.5 minute, or >25 beats / min and :::::0.25 minute). The area of decelerations was calculated as the number of lost beats,21 and large decelerations were defined as having an area >20 lost beats. The temporal association of decelerations with uterine contractions (where present) and the lag time between

September 1991

Am

J Obstet Gynecol

peak contraction and FHR trough were recorded. In each minute the computer acquired 16 measurements of fetal pulse interval; each was averaged over a period of 3.75 seconds. The difference between minimum and maximum was calculated as the minute range. For a record of25 minutes' duration, for example, there were 25 such values, whose average (the mean minute range) was calculated as an index of long-term FHR variation. Several adjustments were made. First, minutes in which a deceleration (defined above) occurred were excluded, so that decelerative variation did not amplify this index. In the same way minutes were excluded in which the FHR was wholly below baseline. Second, if the minute under consideration formed part of a prolonged acceleration, then the range was c alculated from the baseline. These adjustments ensured that long-term variation (the mean minute range) was biased toward recording accelerative FHR variation, the main characteristic of the normal record .22 There was a close correlation (r = 0.97) with the root mean square value of FHR variation. 6 Short-term FHR variation was calculated as the average of successive epochal (I/16-minute) differences, decelerations being excluded as for the calculation of long-term variation . Certain decelerations and accelerations, with abrupt changes in FHR (>35 beats/min followed by an equally abrupt change in the opposite direction) were identified and excluded as unreliable evidence.23 Decelerations and accelerations with high signal loss (>50%) also were eliminated as unreliable. Episodes of high or low FHR variation were identified when in 5 of 6 consecutive minutes the mean minute range of pulse intervals was >32 or 5 beats / min and 0.5 minute) and their temporal relation to uterine contractions to be established. The choice of these criteria has been explained elsewhere,s. 9, 10 and the application to fetuses of different ages has been checked by us 6, 11 and by joint studies with colleagues in other centers in Great Britain, The Netherlands, Canada, Italy, and the United States (unpublished observations). Basal heart rate was measured as the mean value during episodes of low FHR variation or, if no such episode was available, it was measured as the modal or peak value of pulse intervals (as previously described), calculated from the frequency distribution!O In this data set all records contained episodes of low FHR variation. During the record the mother signaled all fetal movements with a hand-held button. Each epoch (3.75 seconds) containing one or more perceived movements was identified. The duration of the FHR trace was determined by the on-line analysis. If an episode of high

Volume 165 Number 3

variation was identified, with no large deceleration and at least one fetal movement or three accelerations, a message was displayed advising that recording may be stopped. Otherwise it lasted 60 minutes . The average record length in ordinary use was about 15 minutes. At the end of each record all FHR variables were printed, and if there were five or more records available on that patient, a synoptic display was presented in which the significance of the trend in short-term variation with time was calculated (e.g. , Fig. 3). The computer program used was the same as that previously described:· 15 with the addition of algorithms to exclude unreliable data"' and to measure short-term FHR variation. Clinical selection. We analyzed 7396 FHR records obtained from 2582 women from June 1983 through 1987, during which there were 23,010 births to 22,742 women in the John Radcliffe Hospital. The patients were monitored because they were identified as at high risk from medical disorders such as hypertension, diabetes, or renal disease; because of a poor obstetric history; or because of current obstetric problems, mainly protein uric preeclampsia, or because the fetus was small for gestational age. These were cases of preterm placental vascular insufficiency, often associated with preeclampsia, in which closely judged decisions were needed to avoid premature delivery. In 78 women the mean minute range was

521

FETAL MOVES per hour

50

A

,'1,\r" J

~

B

,/

1\

50

\



C

;'

,

I

o

o

," I

I'

'I"

J

.'

/\

I

I

~"

,/

SHORT TERM FHR VARIATION msec

8

8

6

6

4

~-~

2

o 200

. :"'\, ".

4

" '>I'

2

o

BASAL FHR bpm

200 + ___ • __ .. ___ • ___ .... .-If'-

100

100 28

OCT

30

3

5 NOV

12

14

16

18 20 22

JULY

24

8

10

12

14

16

MAY

Fig. 3. Three synoptic displays of number of fetal movements per hour (toP), of short-term FHR variation (middle), and of basal heart rate (bottom). Horizontal lines are drawn at 10 movements per hour, short-term variation 2.5 milliseconds, and basal heart rate 160 beats/min. A, From pregnancy at 32 weeks with significant downward trend in variation (- 0.58 msecl day; p < 0.0 I). B, Slower decline (P < 0.0 I) between 32 and 34 weeks. C, Significant upward trends in variation (+ 0.20 mseclday; tJ < 0.02) at 27 to 28 weeks' gestation; delivery was undertaken for maternal reasons. In no instance was there metabolic acidemia on delivery.

can be measured accurately from a scalp clip electrocardiogram in labor but not by ultrasonographic Doppler methods. Even from a scalp clip the normal value is low, only 2.1 :±: 0.2 milliseconds' (mean pulse interval variation:±: SE) or 1.98 :±: 0.32 milliseconds" in two separate studies of normal fetuses during labor. This is at the limit of accuracy of measurement. The FHR cannot be detected reliably, between 26 and 34 weeks' gestation, by abdominal electrocardiogram. This precludes the accurate measurement of beat-to-beat variation. The problem is aggravated by the use of autocorrelation, a process that uses two successive pulse intervals to achieve an estimate of heart rate and so reduces (literally) the value of beat-to-beat FHR variation." We therefore defined a new measure of short-term variation as that between successive 3.7 5-second (1/16minute) epochs during which valid FHR values are av-

eraged (epoch-to-epoch variation). There are on average nine fetal pulses during each epoch at 144 beats/min. Epochs lasting '116 minute were first used in 1979 to reduce the data as much as possible (by a factor of 9) without compromising detection of brief changes in FHR from the baseline (small accelerations or decelerations) and to minimize the effect of signal loss. The epoch-to-epoch variation uses the least interval compatible with our data collection system, yet the mean value is sufficiently large to provide a good measure of short-term variation as it declines before death in utero. The mean value of epoch-to-epoch variation for normal fetuses near term (corresponding to a long-term variation of 50 milliseconds, Fig. 2) is about 9 milliseconds. We have previously demonstrated that a mean minute range >30 milliseconds is associated with normal outcomes!7 and that it must fall to

Short-term variation in abnormal antenatal fetal heart rate records.

In a retrospective study the relation of reduced fetal heart rate variation to fetal acidemia was analyzed with a computerized system for numeric anal...
822KB Sizes 0 Downloads 0 Views