What is the “normal” fetal heart rate?

Introduction

Recording of fetal heart rate (FHR) via cardiotocography (CTG) monitoring is routinely performed as an important part of antepartum and intrapartum care. However, in several randomized trials it became evident that there is only limited efficacy in improving fetal outcome using CTG antenatally (Pattison & McCowan, 2004). A detailed meta-analysis of available studies on the use of intrapartum cardiotocogram showed reduction of perinatal mortality by 50%, but an increase of operative intervention by factor 2.5 (Vintzileos et al., 1995). One potential reason is the wide variability in clinical decision making associated with its use. Standardizing management of variant intrapartum FHR tracings was suggested to reduce this variability and to lead to improvement in fetal outcome (Downs & Zlomke, 2007). In a recent Cochrane review no difference in outcome could be found when looking at potential improvements through the use of CTG monitoring, but, remarkably, the conclusion was different when computerized interpretation of CTG traces was taken into account: “when computerized interpretation of the CTG trace was used, the findings looked promising” (Grivell et al., 2012). Therefore it seems natural to assume that further work on improving definitions and standardization by using computerized methods will further improve the monitoring systems. However, currently, there is not even agreement on the normal range of the baseline of the FHR, although, as Massaniev stated in 1996, “baseline rate provides valuable information on which we plan our further actions” (Manassiew, 1996).

The current international guidelines of the Fédération Internationale de Gynécologie et d’Obstétrique (FIGO) (Rooth, Huch & Huch, 1987), based on consensus during the 1985 conference, recommend a normal range of the FHR from 110 to 150 beats per minute (bpm). The FIGO guidelines, despite some well-known shortcomings, “remain the sole broad international consensus document in FHR monitoring” (Diogo & Joao, 2010). This consensus replaced the former range of 120 to 160 bpm, as there was evidence pointing to worse fetal outcome for baselines higher than 160 bpm (Saling, 1966). Up to now, ranges such as 110 to 150 bpm or 110 to 160 bpm (American Congress of Obstetricians and Gynecologists, 2009; Deutsche Gesellschaft für Gynäkologie und Geburtshilfe, 2010; Macones et al., 2008; Manassiev et al., 1998; National Institute for Health and Clinical Excellence (NICE), 2007; Perinatal Committee of the Japan Society of Obstetrics and Gynecology, 2009; Royal Australian and New Zealand College of Obstetricians and Gynaecologists, 2006; Society of Obstetrics and Gynaecologists of Canada, 2007) are also used, widely based on expert opinion rather than evidence.

This assessment of the situation and the existing “evidence base” is based on the following elements. We have published the plan to do the analysis and have publicly asked for feedback. We have done several literature searches mostly in Pubmed, Google Scholar, the Cochrane Library and have collected publications listed in various versions of published CTG guidelines and standard textbooks. In total we have collected more than 100 papers related to the topic. We have asked opinion leaders in Germany, the UK and the US about awareness of any recent and ancient work that would need to mentioned. In addition, stimulated by the reviewer’s comments, we have (March 2013) conducted a snowball search based on the original Manassiev paper, as well as a systematic search with the related topic of “electronic fetal monitoring”. We did not find any published work that would interfere with the findings in this manuscript.

Our aim was to first define what one should mean by “normal” fetal heart rate and then to give a data-driven answer to this question, as a basis for the more complicated question about the right choice of “alarm limits”.

Material and Methods

In order to reduce the probability of publishing false positive results, this study followed a strict analysis plan, published before onset of the analyses (Daumer et al., 2007). A similar methodology is now being recommended by ENCePP (www.encepp.org) of the European Medical Agency.

Results

Fetal heart rate analysis

The distribution of the FHR baseline measurements of the training data set over the whole range of possible frequencies is shown as a histogram in Fig. 1A, showing roughly the shape of a Gaussian distribution, but not the full symmetry. Distribution in steps of 5 bpm is summarized in Table 2 as a percentage of all measurements for the training data (Column 1).

Table 2:

Distribution of the fetal heart rate in the training and validation sets.

The number of singular fetal heart rate recordings under or above the given limits of fetal heart rate as a percentage of all measurements is displayed.

	Training	Validation I	Validation II	Validation III	Validation I - III
	TUM	TUM	WH	A
	2000–2004	2005–2006	06/2005–2007	09/2001–2005
Lower limit
<100 bpm	0.13%	0.15%	0.08%	0.17%	0.12%
<105 bpm	0.26%	0.26%	0.15%	0.37%	0.24%
<110 bpm	0.62%	0.64%	0.40%	0.78%	0.57%
<115 bpm	1.81%	1.79%	1.24%	1.68%	1.53%
<120 bpm	5.02%	4.90%	3.54%	4.45%	4.21%
Upper limit
>145 bpm	23.26%	23.81%	27.84%	22.33%	25.22%
>150 bpm	12.56%	13.13%	16.09%	12.04%	14.16%
>155 bpm	6.51%	6.96%	8.67%	6.23%	7.53%
>160 bpm	3.21%	3.55%	4.35%	3.11%	3.79%
>165 bpm	1.47%	1.76%	2.00%	1.51%	1.80%
>170 bpm	0.68%	0.78%	0.92%	0.70%	0.82%

DOI: 10.7717/peerj.82/table-2

The criterion for definition of the best interval is ${\displaystyle arg\underset{i=1,\dots ,5}{min}{\left(\widehat{F}\right(Z_{lower}^{\left(i\right)})-(1-\widehat{F}\left(Z_{upper}^{\left(i\right)}\right)\left)\right)}^2.}$ (for further details see our analysis plan (Daumer et al., 2007)).

Analyzing the training set, the selected interval of 40 to 45 bpm width was 115 to 160 bpm (criterion: (0.0181−0.0321)² = 0.20⋅10⁻³). The criterion for the interval with 120 to 160 bpm was only marginally bigger (criterion: (0.0502−0.0321)² = 0.33⋅10⁻³) (Table 4, Column 1), such that the lower bound, in contrast to the upper bound, is not stable.

Table 3:

Calculation of the criterion for definition of the best interval in the training and validation data sets.

Square of difference between upper and lower tail of the distribution ([i]), as shown in Table 3. All values have to be multiplied with 10^-3. The best criterion for each data set is marked in bold letters.

	Training	Validation I	Validation II	Validation III	Validation I - III
	TUM	TUM	WH	A
	2000–2004	2005–2006	06/2005–2007	09/2001–2005
110–150	14.24	15.60	24.62	12.69	18.48
110–155	3.46	3.99	6.83	2.97	4.85
115–155	2.21	2.68	5.51	2.07	3.61
115–160	0.20	0.31	0.97	0.20	0.51
120–160	0.33	0.18	0.07	0.18	0.02
120–165	1.26	0.98	0.24	0.86	0.58

DOI: 10.7717/peerj.82/table-3

Table 4:

Distribution of FHR baseline during gestation.

(A) 95% confidence intervals for mean FHR baseline are displayed for intervals of several gestational weeks. All pairwise comparisons are significant (p < 0.01) with both t-test and Mann-Whitney tests. The comparisons between gestational age of > = 37 and other groups are the most significant. (B) 95% confidence intervals for mean FHR baseline within the group of gestational age of 37 weeks or more.

Gestational age	n	95% confidence interval
A
<28	1230	140.7538	–	141.9422
28 – <32	1059	139.1587	–	140.3843
32 – <37	2248	138.1575	–	138.9322
>=37	8478	136.0104	–	136.4295
B
37	1090	136.7176	–	137.8588
38	1793	135.5575	–	136.4720
39	1962	135.9786	–	136.8404
40	2325	135.2181	–	136.0158
41	1199	135.9135	–	137.0438
42	109	133.2492	–	137.8009

DOI: 10.7717/peerj.82/table-4

Hence the following hypotheses were formulated and tested during validation:

1. The upper limit of the FHR should be 160 bpm.

2. The lower limit should be either 115 or 120 bpm.

Results of each of the validation data sets and of a combination of all three of them revealed the range of 120 to 160 bpm as the best interval (Fig. 1B, Tables 2 and 3, Columns 2, 3, 4, and 5). Hence, both hypotheses were validated.

The mean FHR baseline plotted against gestational age is shown in Fig. 2. Table 4 shows 95% confidence intervals for mean FHR baseline in different gestational weeks. Regression analysis with the median FHR baseline as dependent variable and the gestational age (in weeks) as independent variable yielded a slope estimate of −0.378 (p < 0.001), meaning that the median FHR decreases on average by 0.4 bpm per week of pregnancy. The assumptions underlying the linear regression model were approximately fulfilled.

Discussion

Analyzing about 1.5 billion individual single baseline fetal heart rate measurements from 78,852 CTG tracings in three German medical centers, we found that “normal” ranges – normality in a statistical sense - are 120 to160 bpm. By this data-driven definition of the normal FHR we aimed to generate a solid basis for the clinically important attempt to eventually further reduce the rate of false alarms in CTG monitoring in general and electronic decision support systems in particular. This might help to avoid unnecessary interventions such as Cesarean sections. The FHR baseline in our analysis decreases slightly during gestation, in line with results of other groups (Nijhuis et al., 1998; Serra et al., 2009). There are well-known physiological changes in fetal development that are consistent with this empirical finding (Karolina & Edwin, 2011), essentially due to the increasing opposed effect of the sympathetic nervous system as gestational age increases.

Validation of the results in an independent data set is a crucial step to avoid the publication of false positive research findings (Daumer et al., 2008; Ioannidis, 2005). Both temporal validation (based on data collected later than the training data) and external validation (based on data collected in another medical center), used in our study, are known to be essential (König et al., 2007). Furthermore, the strict blind validation procedure was adopted and described in a detailed analysis plan in the pre-publication platform Nature Precedings (Daumer et al., 2007) before starting the analyses. The results about the normal range are very robust, indicating that neither the type of hospital which is potentially linked to special selection criteria for the pregnant women nor the time as measured roughly in 5–10 year intervals seems to play a role – an argument for the external validity of the findings in the exploratory part.

For user acceptance we used steps of 5 bpm as possible borders of the normal FHR as recommended in the consensus meeting of the National Institute of Child Health and Human Development (Macones et al., 2008; National Institute of Child Health and Human Development Research Planning Workshop, 1997). The width of the interval of 40 to 45 bpm was traditionally used in many international guidelines. As we planned the study, we chose no other intervals, as narrowing of the interval would increase the false alarm rate and wider intervals could miss pathologic conditions of the fetus.

The upper limit of 160 bpm raised concerns in the FIGO meeting in 1985, as Saling described abnormal findings in 24% of scalp blood analyses if the baseline was higher than 160 bpm (Saling, 1966). It could be shown that the current FIGO guidelines based on computerized analyses of the CTG show a high sensitivity to detect fetal acidosis in case of a suspect or pathological classification of the baseline level. It may turn out that a modification of the normal ranges further improves sensitivity and specificity of fetal acidosis during labor (Schiermeier et al., 2008). Also, multivariate modeling involving fetal and maternal outcome data may improve evidence-based online decision support tools.

Data from a recently published study in a different context (Serra et al., 2009) is compatible with the findings of our exploratory analysis with a lower limit of 115 or 120 bpm for the gestational ages. Data for the 97th and 99th percentiles are not shown in this study. But shifting the lower limit to 120 will increase the number of false alarms whereas a lower limit of 115 will inevitably increase the risk to misinterpret maternal heart rates as fetal heart rate. This last problem has raised many concerns and discussions about technical solutions for differentiation of maternal and fetal heart rate, as fatal consequences for the fetus could occur (Murray, 2004). The new German guideline (Deutsche Gesellschaft für Gynäkologie und Geburtshilfe, 2012) recommends therefore simultaneous recording of fetal and maternal heart rate, technically possible either by maternal pulse oxymetry integrated in a CTG device or simultaneous ECG recording of mother and fetus.

As FHR tracings of prenatal care patients were included, our study population consists of a fraction of pregnancies remote from term, eventually resulting in higher baselines as suggested before. As our analysis according to gestational ages shows, the upper limit of 160 bpm is valid for younger and for later gestational ages. A lower limit of 120 bpm leads only near term to more false alarms since normal FHR decreases further, and is more appropriate, as discussed above, to avoid misinterpretation of maternal heart beat as FHR. There are no different guidelines for scoring cardiotocograms of early gestational ages as this would be too difficult in daily practice. Only computerized algorithms could use boundaries without rounding based on multivariate modeling and correlate these results to fetal outcome.

FIGO guidelines defined boundaries from 110 to 150 bpm, representing the approximately 0.6th to 86th percentile from our study. Current guidelines released by the American College of Obstetricians and Gynecologists (American Congress of Obstetricians and Gynecologists, 2009), the National Institute of Child Health and Human Development (National Institute of Child Health and Human Development Research Planning Workshop, 1997), the Society of Obstetricians and Gynaecologists of Canada (Society of Obstetrics and Gynaecologists of Canada, 2007), the United Kingdom’s National Institute for Health and Clinical Excellence (National Institute for Health and Clinical Excellence (NICE), 2007), the Royal Australian and New Zealand College of Obstetricians and Gynaecologists (Royal Australian and New Zealand College of Obstetricians and Gynaecologists, 2006) and the Japan Society of Obstetrics and Gynecology (Perinatal Committee of the Japan Society of Obstetrics and Gynecology, 2009) define a very wide range of normal FHR with 110 to 160 bpm, representing the approximately 0.6th to 96th percentile. We raised concerns about the broad width of the range of 50 bpm and the lower limit of 110 bpm. As these guidelines are in use for some years in many countries at the moment, we assume that this range is still safe for detection of fetal compromise. In contrast, specificity of the CTG for fetal acidosis becomes better. But safety-analyses should confirm this assumption.

Our results have stimulated discussions within the corresponding German society “Deutsche Gesellschaft für Gynäkologie und Geburtshilfe” (Deutsche Gesellschaft für Gynäkologie und Geburtshilfe, 2010) having led to a recent update of the previous guidelines (Deutsche Gesellschaft für Gynäkologie und Geburtshilfe, 2012), based on data from the exploratory analysis. We hope that our study will trigger a process of continuous improvement of evidence based clinical decision making in fetal monitoring – perhaps a task to be triggered by the HTA working group of ENCePP (http://www.encepp.eu/structure/documents/ENCePPWGHTA_Mandate.pdf).