Severity of hypoxic ischemic encephalopathy and heart rate variability in neonates: a systematic review

The specific search strategy for each database was developed in consultation with two medical librarians. Pubmed, Embase, Web of Science, Cochrane Database (CENTRAL), and Scopus were searched using subject headings and free text related to “hypoxic ischemic encephalopathy” and “heart rate variability”. The search was performed 6 April 2018. References of included studies were manually investigated. For PubMed and CENTRAL, the following search strategy was used: hypoxia ischaemia; hypoxic ischemia; hypoxic ischaemia; hypoxic ischemic; hypoxic ischaemic; HIE; asphyxia*; encephalopathy; asphyxia neonatorum[MeSH]; hypoxia-ischemia, brain[MeSH]. These were combined with: RR interval*; NN interval*; heart rate variation; HRV; time domain measure*; non-linear measure*; frequency domain measure*; heart rate variability; beat-to-beat variability; heart rate[MeSH]. The following Embase subject headings were: newborn hypoxia/exp; hypoxic ischemic encephalopathy/exp; heart rate variability/exp. For Scopus and Web of Science only free text was applied. The search results were then combined using Endnote X8.2® and Covidence systematic review software and duplicates were removed [33]. The complete search strategy is available in Additional file 1.

Two reviewers (MA and TCKA) independently screened titles and abstracts of the identified studies. Studies that met the eligibility criteria or provided insufficient information were subjected to full-text review. Assessment of potential eligibility and the subsequent data extraction was also performed by two reviewers (MA and TCKA). Any disagreements were resolved by discussion until consensus was reached or by consulting a third reviewer (TBH). The screening and selection process was documented in a PRISMA flowchart (Fig. 1) [28]. Data was extracted using a predefined data collection form based on the Cochrane Consumers and Communication Review Group Data Extraction Template [34]. To reduce errors and to identify missing data fields, piloting of the data collection form was performed by two reviewers (MA and TCKA). We did not contact authors of studies with missing or inadequate information.

The following information was extracted from the included studies [1]: title, author, country, year of publication, reference list, funding sources, and possible conflicts of interests [2]; type of methods including study design and unit of allocation [3]; characteristics of the neonates (gestational age, postnatal age, birth weight, and pathologies) including the eligibility criteria [4]; methods for assessing the HIE severity [5]; methods and postnatal age when assessing the HRV measures [6]; results and statistical significance [7]; information critical for providing an assessment of risk of bias and strength of evidence by the Newcastle-Ottawa Scale and the Grading of Recommendations Assessment, Development and Evaluation (GRADE).

The risk of bias assessment was independently performed by two reviewers (MA and TCKA). Any disagreements were resolved by discussion until consensus was reached or by consulting a third reviewer (TBH). To assess the risk of bias in the included studies, a modified version of the Newcastle-Ottawa Scale for cohort studies was used as proposed by the Cochrane Handbook [34, 35]. Using this scale, studies were awarded points with a maximum of nine. The points were awarded based on three domains [1]: selection, scored by the representativeness of the neonates with severe or moderate HIE, selection of the neonates with mild or no HIE, ascertainment of the HIE severity, and demonstration that outcome was not present at study start [2]; comparability, scored by the comparability of cohorts based on design or analysis in the studies with regards to malformations, infections, metabolic diseases, gender, birth weight, gestational age, postnatal age, therapeutic hypothermia, and medication [3]; outcome, scored by the assessment of the HRV measurements, sufficient length of follow-up, and adequacy of follow-up of the neonates. A maximum of four, two, and three points could be awarded in each domain, respectively. The quality of the studies was rated as ‘good’, ‘fair’, or ‘poor’ according to the points awarded. ‘Good quality’ was given 3–4 points in selection and 1–2 points in comparability and 2–3 points in outcome; ‘fair quality’ was given 2 points in selection and 1–2 points in comparability and 2–3 points in outcome; and ‘poor quality’ was given 0–1 points in selection or 0 points in comparability or 0–1 points in outcome.

Unless otherwise stated, the summary measure was expressed as mean difference with 95% confidence interval (CI). Statistical significance refers to a two-sided p-value of less than 0.05. Calculations were performed in Review Manager 5.3 [36]. The heterogeneity between studies did not allow for meta-analysis. Hence, a narrative synthesis of the results was performed. The synthesis was made in accordance with Popay et al. [37]. We arranged our results based on the different combinations of comparisons between the neonates with severe or moderate HIE and the neonates with mild or no HIE. For each of these comparisons we present the results both tabulated and narratively summarised. Furthermore, we sought relations between each of these comparisons such as dose-response. We used the GRADE approach for assessing the strength of evidence [38]. The number of included studies was insufficient to allow construction for a funnel plot and formal testing of asymmetry. Selective reporting bias was assessed by comparing the outcomes reported in the method section and the result section of the included studies.

A total of 1440 studies were identified across the different databases. After removal of duplicates, 1048 studies were screened by title and abstract. A total of 43 full-text studies were assessed for eligibility. Of these, four studies met the inclusion criteria and were submitted to data extraction and analysis [19‐22]. Furthermore, 139 studies from the reference lists of the included studies were screened but none of these were included (Fig. 1). The included studies were published between 2012 and 2017. An overview of excluded articles with reasons for exclusion is available in Additional file 2.

The points awarded to each study by the Newcastle-Ottawa Scale are summarized in Table 2. The representativeness of the neonates with severe or moderate HIE was appropriate and the comparators was drawn from the same hospitals. The assessment of both HIE severity and HRV measures were done by examining the appropriate recordings. Only one study stated a blinded assessment of the HRV measures [19]. This study was also the only study reporting adjustment for potential confounders such as gestational age, birth weight, gender, and different pathological conditions. Vergales et al. [20] reported adjustment for administration of phenobarbital, while Goulding et al. 2015 [21] reported adjustments for both phenobarbital and morphine. These adjustments failed to change the estimated associations apart from the association between TINN and LF and the HIE severity which became statistically insignificant. Within all four studies, the neonates were comparable with respect to postnatal age and therapeutic hypothermia.

The studies were heterogeneous which made direct comparisons between the studies problematic. Number of neonates or epochs in each statistical analysis and mean differences with 95% CI are reported in Table 3. Five different comparisons and ten different HRV measures were presented in the studies. Non-linear measurements, usually presented by Poincaré plots, were not reported.

We found no indication of selective reporting bias as no discrepancies between outcomes reported in methods and results were found in the included studies. The results included in this review are from relatively small studies and all found an association between the severity of HIE and HRV. Therefore, we were unable to rule out publication bias. This may lead to an overestimation of the association between HIE severity and HRV.

One of the main findings in this systematic review was the lack of evidence in this specific area of interest especially due to the variety of HRV measures and methodology for measuring. For every outcome, the strength of the evidence was very low according to the GRADE assessment mainly due to the risk of bias and imprecision of estimates. Still, our review provides three key findings that may be of importance [1]: in three out of four studies, neonates with severe or moderate HIE had reduced time domain measures compared with neonates with mild or no HIE [2]; in three out of four studies, neonates with severe or moderate HIE had reduced frequency domain measures compared with neonates with mild or no HIE with exception of an increased HF in relative power [3]; in three out of four studies, a dose-response like pattern was seen as a tendency towards HRV being more reduced with increasing HIE severity.

Currently, no standardized methodology for measuring HRV in neonates exist. This is needed as the methods used may affect HRV measures. Time domain measures are affected by the sampling length of the ECG [31, 32]. For example, this issue hampered the comparison of HRV measures from Vergales et al. [20] and Goulding et al. [21] [22]. Frequency domain measures may be affected by differences in methodology (i.e., parametric or non-parametric) and the width of the frequency bands used [41]. Doyle et al. [42] has defined the frequency bands in healthy neonates less than 12 h of age as: VLF: 0.01–0.04 Hz; LF: 0.04–0.2 Hz; HF: > 0.2 Hz. These frequency bands were used by Goulding et al. [21, 22] but not by Aliefendioglu et al. [19], which again limit comparisons and meta-analyses. Therefore, we emphasize the need for standard guidelines for future studies on HRV in neonates. In the future, such guidelines would allow for collaborations and centre-to-centre data comparison.

Time domain measures are more accessible and easier to interpret compared with most other HRV measures [31]. SDRR was only reported in a single study comparing neonates with moderate and severe HIE with neonates with mild or no HIE determined by EEG after an initial clinical diagnosis of moderate or severe HIE [20]. However, only discrimination of the various grades of HIE will provide detailed information on HRV measures in HIE. This study also reported HRC monitor data. They found that the HRC index had an inverse correlation to SDRR and could predict adverse outcomes in neonates with HIE, indicating a possible clinical use of HRV in the assessment of HIE. SDNN describes the total variability and is believed to be controlled by both the sympathetic- and parasympathetic nervous system [31, 43]. Thus, SDNN may be the most clinically relevant HRV biomarker for HIE severity. Overall, neonates with severe or moderate HIE had a reduced SDNN compared with neonates with mild or no HIE. One study [22] found no statistically significant reduction in SDNN when neonates with moderate HIE were compared with neonates with mild HIE. The main difference between this study and the previous study from the same group [21] was treatment with therapeutic hypothermia in the later. However, other studies have found no effect of temperature on SDNN in neonates with HIE but have reported an increase in HF in absolute power during rewarming and a negative association between LF in relative power and temperature [26, 44].

None of the studies reported HRV measures from the first 6 h after birth. Assessment of the grade of HIE in this period is essential when considering therapeutic hypothermia since this treatment must be initiated as early as possible and preferably before 6 h after birth to be neuroprotective [45, 46]. Yamaguchi et al. [24] reported an increased SDNN within 0 to 6 h after delivery in preterm fetal sheep with severe hypoxia-ischemia compared with healthy controls or mild hypoxia-ischemia. In the secondary phase of brain injury, considered to be 6 to 72 h after delivery, SDNN decreased and was significantly reduced compared with the controls. These findings are in keeping with the findings of the clinical studies included in this systematic review.

Frequency domain measures also appeared to be potential markers of HIE severity. LF and HF may be used to describe the autonomic regulation of HRV but the interpretation of the individual frequency measures is still uncertain. LF may reflect a combination of parasympathetic activity, sympathetic activity, and baroreceptor regulation, or baroreceptor regulation only, while it is believed that HF mainly reflects the parasympathetic activity [31, 47‐53]. Neonates with severe or moderate HIE had reduced frequency domain measurements in absolute power. The absolute power may reflect the overall autonomic activity, which therefore appears to be reduced [47]. Furthermore, neonates with severe or moderate HIE had a reduced LF/HF ratio in relative power. LF and HF in relative power reflects the proportion of the total power minus VLF, which may accentuate the balance between the different neuronal regulations of the HRV [31]. As HF is believed only to reflect the parasympathetic activity and LF may reflect different neural components, these findings may indicate a preponderance of parasympathetic activity in these neonates [19].

In frequency domain measures in relative power, a dose-response-like relationship was indicated by a larger difference when comparing neonates with severe HIE with no HIE than when comparing neonates with moderate HIE with no HIE. The same pattern was seen in time domain measures and most frequency domain measures in absolute power, although not statistically significant. Still, these findings indicate that HRV may be a potential biomarker for HIE severity. This is supported by other studies who found that the HRV could distinguish neurodevelopment outcome in neonates with HIE. Goulding et al. 2015 [21] also investigated two-year neurodevelopmental outcome and found a correlation between adverse outcomes and reduced HRV measurements from 12 to 48 h after birth. Massaro et al. [25] compared neonates with HIE with either favorable or adverse outcomes at 15 months. Frequency domain measures were measured in relative power and assessed between 5 to 110 h after birth. In neonates with an adverse outcome, LF was reduced while HF was increased, which corroborated with the findings in Aliefendioglu et al. [19] HRV seemed to provide the best discrimination between those with adverse outcome and those without if measured at 24 to 80 h of life. A systematic review [27] has evaluated HRV as a long-term prognostic marker. They concluded that HRV may be used as a long-term prognostic marker for neurodevelopmental outcome in HIE. However, further research was warranted.