Prognostic Value of Clinical Tests in Neonates With Hypoxic-Ischemic Encephalopathy Treated With Therapeutic Hypothermia: A Systematic Review and Meta-Analysis

Background and Objective: There remains an unmet clinical need for markers that predict outcomes in the hypothermia-treated (HT) infants with HIE. The aim of this meta-analysis was to investigate the prognostic accuracy of currently available clinical tests performed in the immediate post-natal period for predicting neurological outcomes between 18 months and 3 years of age in HT near-term and term infants with perinatal asphyxia and HIE. Methods: A comprehensive review of the Embase, Cochrane library, and PubMed databases was performed to identify studies that evaluated the prognostic value of clinical tests for neurological outcomes in HT near-term and term infants with perinatal asphyxia and hypoxic-ischemic encephalopathy. Pooled sensitivity and specificity with corresponding 95% confidence intervals and area under the receiver operating characteristic (ROC) curve (AUC) were calculated. Results: Of the 1,144 relevant studies, 26 studies describing four clinical tests conducted in 1458 HT near-term or term infants were included. For predicting an unfavorable neurological outcome, of the imaging techniques, MRI within 2 weeks of birth performed best on sensitivity 0.85 (95% CI 0.79–0.89), specificity 0.72 (95% CI 0.66–0.77), and AUC 0.88; among the neurophysiological tests, multichannel EEG (Electroencephalogram) demonstrated the sensitivity 0.63 (95% CI 0.49–0.76), specificity 0.82 (95% CI 0.70–0.91), and AUC 0.88, and for aEEG (amplitude-integrated electroencephalography) background pattern pooled sensitivity, specificity and AUC were 0.90 (95% CI 0.86–0.94), 0.46 (95% CI 0.42–0.51), and 0.78 whereas for SEPs (Somatosensory evoked potentials), pooled sensitivity and specificity were 0.52 (95% CI 0.34–0.69), 0.76 (95% CI 0.63–0.87), and AUC 0.84, respectively. Conclusions: In the wake of the era of TH, MRI and neurophysiological tests (aEEG or EEG) were promising predictors of adverse outcomes, while SEPs need high-quality studies to confirm the findings. Continued follow-up of the children and well-designed large prospective studies are essential to determine whether these benefits are maintained in later childhood.


INTRODUCTION
Hypoxic-ischemic encephalopathy (HIE) after perinatal asphyxia is the primary cause of death or long-term neurological impairment in infants born at term. Early predictive indicators of neurological outcomes in infants with HIE is essential for making rational clinical decisions. Before the era of therapeutic hypothermia, the prognostic value of MRI (first week) in neonates with HIE has been well-validated. aEEG or EEG recorded within the first 7 days of life in term infants may have potential as a neurophysiologic predicting tests. However, the prognostic value of SEP should be interpreted with caution due to small sample sizes (1).
Therapeutic hypothermia for 72 h provides neuroprotection that significantly improves survival and neurological outcomes in term infants with moderate to severe HIE (2). However, the prognostic capability of these parameters may vary under hypothermic conditions, and there remains an unmet clinical need for markers that predict outcomes in hypothermia-treated (HT) infants (3,4).
To the authors' knowledge, there are no published metaanalyses that investigate the prognostic capabilities of currently available clinical tests for predicting long-term neurological outcomes in HT term-infants with HIE. Therefore, a metaanalysis was conducted to evaluate the prognostic accuracy of currently available clinical tests performed in the immediate postnatal period for predicting neurological outcomes between 18 months and 3 years of age in HT near-term and term infants with perinatal asphyxia and HIE.

Search Strtegy
Three review authors (W.L, Q.Y and Z.H) independently searched the PubMed, Cochrane Library, and Embase databases from 2009 to February 2019 using the following keywords "hypoxic-ischemic encephalopathy" AND "newborn" AND "hypothermia"; "neonatal Encephalopathy" AND "newborn" AND "hypothermia" Searches were limited to literature published in the English language.

Inclusion and Exclusion Criteria
Inclusion criteria were: (1)observational prognostic studies that included infants ≥35 weeks of gestation with perinatal asphyxia and HIE diagnosed according to clearly described criteria; (2) that were treated with therapeutic hypothermia; and (3) that underwent neurological follow-up longitudinally until ≥18 months of age, with outcomes defined as good or adverse.

Study Selection
Three review authors (W.L, Q.Y, and Z.H) independently screened the titles and abstracts identified by the search for potentially relevant studies. Texts were retrieved and reviewed to determine which studies met the inclusion criteria. Disagreements about data extraction were resolved by discussion with a third review author (ZH) until consensus was reached.

Data Extraction
Three review authors (W.D, H.W., and Y.F) independently extracted data from the included studies using a data extraction sheet. The following information was recorded: authors, year of publication, number of study subjects, gestational age, birth weight, Apgar score, blood pH, clinical tests (e.g., imaging, neurophysiological, other), neurological tests, and length of follow-up. If both hypothermic and normothermic infants were included in a study, data about the hypothermic infants were extracted separately. According to the review of the studies, we established the MRI subgroup and the aEEG subgroup.

Statistical Analysis
Statistical analyses were conducted with RevMan v5.3 and Meta-Disc 1.4. Three review authors (W.L, Q.Y, and Z.H) examined the quality of the included studies using Quadas 2, which evaluates four key domains, including patient selection, index test, reference standard, and flow and timing and independently extracted individual patient data from each of the studies into a predefined database. Personal patient data from all studies were pooled to create 2 × 2 tables and pooled sensitivity and specificity with 95% confidence intervals (CIs) and area under the receiver operating characteristic curve (AUC) were calculated. Heterogeneity between studies was tested with the inconsistency (I 2 ) index and χ 2 -test. A fixed-effect model was used if there was no evidence of heterogeneity between studies (I 2 < 50%, P > 0.05); otherwise, a random-effects model was used.

Study Identification
The searches identified 1,144 articles. Following the removal of duplicates, the titles and abstracts of 925 studies were screened, and 82 studies were considered potentially eligible for inclusion in this meta-analysis. Full-text articles were retrieved and reviewed; among these, 11 studies were excluded because they were not published in the English language, 27 studies were excluded due to missing data. Fifteen studies were excluded due to uncertain follow-up time and follow-up time is <18 months, and three studies were not analyzed due to small sample size. Finally, 26 reviews were included in the pooled analyses (Figure 1 and Supplementary Table 1).

Characteristics of Included Studies
The characteristics of the included studies are shown in Tables 1, 2. The 26 eligible studies included 1458 HT near-term or term infants with perinatal asphyxia and HIE and follow up available after 18 months of age. Clinical tests performed in the immediate postnatal period included brain magnetic resonance imaging (MRI), EEG, aEEG, SEPs. Neurological examinations were performed between 18 months and 3 years of age.

Methodological Quality of Included Studies
The risks of bias in the index test and patient selection were low (Figures 2, 3). Sixteen studies (

Prognostic Value of Clinical Tests
The results of the meta-analysis are shown in Table 3 (pooled sensitivities and specificities with confidence intervals) and Figure 4 (forest plots of sensitivity and specificity as calculated from the original reports).

DISCUSSION
HIE after perinatal asphyxia is a significant cause of mortality and morbidity in newborns, accounting for approximately 20% of cerebral palsy cases (31,32). Evaluating long-term neurological outcomes based on clinical evaluations in the immediate postnatal period can be challenging. However, a reliable, evidencebased prognosis is essential for parental counseling regarding possible long-term sequelae. Hypothermia is generally accepted as the safest method for reducing neurological morbidity and mortality in infants with perinatal asphyxia and HIE (33,34). Although the literature is replete with studies evaluating novel but relatively unknown test modalities, to ensure clinical relevance. The present review focused on tests that are widely used in clinical practice and did not attempt to provide an overview of all available procedures (1).
Current MRI injury scoring systems published in the literature include the Barkovich, the National Institutes for Child Health and Development (NICHD) and Rutherford systems (5,9,35). Each scoring system has some limitations. The previous does not explicitly address posterior limb of the internal capsule injury, whereas the latter two systems do not include diffusionweighted imaging (DWI) (25), and Clinicians often evaluate the neonatal brain MRI as a whole test rather than specific brain MRI components. So, we considered parenchymal (gray or white matter) abnormalities on T1, T2, and diffuse weighted images to define an abnormal MRI. In the systematic review, brain MRI predicted an unfavorable neurological outcome in HT infants with perinatal asphyxia and HIE with a sensitivity of 0.85. Accordingly, in the TOBY trial, the accuracy of prediction of death or disability to 18 months of age by MRI was 0.84 (0.74-0.94) in HT infants and 0.81 (0.71-0.91) in a normothermic group (5). Evidence suggests that therapeutic hypothermia without affecting the overall predictive value of MRI as a marker of neurological impairment (5,13).
In the era of predates hypothermia treatment, late MRI (8-30 days) had higher sensitivity but lower specificity than early MRI (1-7 days) (36). However, the current literature does not provide detailed individual data on the time of neonatal brain MRI in HT (37). In the present review, late MRI predicted an unfavorable neurological outcome with a sensitivity and specificity of 0.88 and 0.88, respectively, whereas early MRI showed less specificity of 0.73. However, Charon et al. reported that the specificity of MRI for predicting death or disability to 18 months of age in HT infants with HIE was 96.3% in the first week and 89.3% in the second week (18). The discrepancy between our findings and those reported by Charon et al. might be explained by the various abnormal findings and the thresholds for the index tests. Results from the present meta-analysis indicate that within 2 weeks of birth correctly predict neurological outcomes at 18 months of age in HT infants with HIE (pooled sensitivity, 0.85 [95% CI, 0.79-0.89]. Deep gray matter lactate/N-acetyl aspartate (Lac/NAA) peak/area ratio is the most quantitative biomarker for prediction of neurodevelopmental outcomes in magnetic resonance spectroscopy (38). However, the equipment is not widely used in clinical practice, and based on limited available studies, the validity of the results has not been quantified in previous reviews of the literature.
aEEG is the commonly used neurophysiological tests for assessment of HIE severity, for monitoring improvement over time, and for predicting neurological outcomes in infants (39). aEEG can be performed at the bedside, and background patterns and voltage classification have been considered an early predictor of neurological outcomes in HT infants with HIE. In normothermic infants, a persistently abnormal aEEG between 6 and 24 h of age is considered predictive of adverse outcomes (6,40). However, the predictive value of aEEG for subsequent neurological impairment is altered by hypothermia. In one study, the positive prognostic value of an abnormal aEEG increased from 6 to 48 h of age in HT term infants with HIE. This shift in prognostic accuracy may be explained, at least in part, by the neuroprotective effects of therapeutic hypothermia (41). In the present review, aEEG background pattern predicted an unfavorable neurological outcome with a sensitivity and specificity of 0.90 and 0. 46 (12). In accordance with our findings, Shany et al. reported that background pattern was more sensitive than voltage classification for predicting neurological outcomes in infants with  HIE, although assessment of background pattern may be more subjective (42). Multichannel EEG is the gold standard for assessment of the severity of HIE and for monitoring improvement over time (39). In our review, EEG predicted subsequent neurological impairment in HT infants with HIE with a sensitivity of 0.63 (95% CI, 0.49-0.76), specificity of 0.82 (95% CI, 0.70-0.91), and an AUC of 0.88, yielding a higher specificity than aEEG. However, EEG is a relatively complex technique. Technicians are required to site EEG leads and specialists are needed to interpret neurophysiology, but these resources may not be readily available (39). High seizure burden and sleep-wake cycling have been independently associated with poor outcomes in HT infants with HIE (41). Unfortunately, lack of data meant that these parameters could not be examined in the present metaanalysis.
SEPs assess the deep brain structures such as the thalami and brainstem, which are known to be selectively vulnerable to hypoxia and ischemia (43). Several studies had shown that normal SEPs were strongly predictive of a favorable outcome, and absent SEPs were strongly predictive of an unfavorable outcome (44,45). In our review, SEPs predicted an unfavorable neurological outcome in HT infants with HIE with a sensitivity of 0.52 (95% CI, 0.34-0.69) and a specificity of 0.76 (95% CI, 0.63-0.87). Although SEPs may provide complementary prognostic resources to HIE after perinatal asphyxia, the predictive value of SEPs investigated in this review should be interpreted with caution due to small sample sizes.
Although not explored in this review, we propose that a combination of the above prognostic tests performed would provide the greater prognostic accuracy for predicting long-term neurological outcomes in the HIE infants undergoing HT.

Strengths and Weaknesses
To the authors' knowledge, this is the first meta-analysis to investigate the prognostic value of clinical tests performed in the immediate post-natal period as predictors of adverse neurological outcomes in HT near-term and term infants with perinatal asphyxia and HIE. Importantly, to investigate which HIE infants would have neurological sequelae, this review only included studies with long-term follow-up, whereby infants were followed until ≥18 months of age. In our opinion, outcomes couldn't be accurately assessed in infants younger than 18 months as neurological sequelae usually manifest at ≥12 months of age, and mental and behavioral disabilities might appear even later (46). This meta-analysis has several limitations. First, the sample size was small. Second, this review was restricted to articles published in the English language, which may have led to an overestimation or underestimation of prognostic reliability. Finally, heterogeneity was low to moderate, possibly due to differences in thresholds for the index test, the severity of HIE and the design of the studies. Subgroup analyses were not performed to investigate the source of this heterogeneity. Welldesigned more extensive studies are required to determine an accurate estimate of the value of clinical tests performed in

CONCLUSIONS
This systematic review and meta-analysis provided insight into the prognostic value of clinical tests performed in the immediate post-natal period as predictors of adverse neurological outcomes in HT near-term and term infants with perinatal asphyxia. MRI and neurophysiological tests (aEEG or EEG) were promising predictors of the adverse outcomes, whereas SEPs need highquality studies to confirm the findings. Given the heterogeneity in the tests' performance, continued follow-up of the children and well-designed large prospective studies are essential to determine whether these benefits are maintained in later childhood.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.