Does mild hypoxic ischemic encephalopathy adversely affect neurodevelopmental outcome

TITLE
Does mild hypoxic ischemic encephalopathy adversely affect neurodevelopmental outcome?

MANUSCRIPT CITATION:
Finder M, Boylan GB, Twomey D, Ahearne C, Murray DM, Hallberg B. Two-year neurodevelopmental outcomes after mild hypoxic ischemic encephalopathy in the era of therapeutic hypothermia. JAMA Pediatrics 2020;174: 48-55.

REVIEWED BY:

Megan W. Berube¹, Mihai Puia-Dumitrescu², Ryan M. McAdams¹

¹ Division of Neonatology and Newborn Medicine, Department of Pediatrics, University of Wisconsin School of Medicine and Public Health, Madison, WI, USA

² Division of Neonatology, Department of Pediatrics, University of Washington, Seattle, WA, USA

KEY WORDS: mild hypoxic ischemic encephalopathy; therapeutic hypothermia; cooling; neurodevelopmental outcomes

CORRESPONDENCE:

Megan W. Berube, MD

Assistant Professor of Pediatrics
Division of Neonatology and Newborn Medicine
University of Wisconsin School of Medicine and Public Health

H4/415 Clinical Science Center
600 Highland Avenue
Madison, WI 53792

mwberube@pediatrics.wisc.edu

(608) 265-8769

TYPE OF INVESTIGATION: Prognosis.

QUESTION: How do the neurodevelopmental outcomes of children with mild hypoxic-ischemic encephalopathy (HIE) at birth compare with the neurodevelopmental outcomes at age 18 to 42 months of 1) healthy children born at term or near-term, and 2) surviving children with moderate HIE who were treated with therapeutic hypothermia at birth? 
METHODS

Design: Multicenter cohort study.  This study is a secondary analysis of pooled data from four independent prospective cohorts and did not involve allocation. Infants were categorized into five groups: healthy controls, perinatal acidosis without HIE, mild HIE, moderate HIE, and severe HIE. HIE was graded using the highest modified Sarnat score recorded during the first 24 postnatal hours.

Blinding: Not blinded.

Setting: Two tertiary centers: Cork University Maternity Hospital, Cork, Ireland, and Karolinska University Hospital, Stockholm, Sweden. Participants were recruited prospectively to four independent cohorts: NeoCool, Biomarkers in Hypoxic-Ischemic Encephalopathy [BiHIVE1], Investigation and Validation of Predictive Biomarkers in Hypoxic-Ischemic Encephalopathy [BiHIVE2], and Algorithm for Neonatal Seizure Recognition Study [ANSeR1]. Recruitment occurred between January 2007 and August 2015.

Patients:  Infants were defined to have perinatal asphyxia if 1 or more of the following were observed: umbilical cord pH less than 7.1, Apgar score < 7 at 5 minutes, and use of assisted ventilation and/or cardiopulmonary resuscitation by 10 minutes of age.  Infants with neonatal encephalopathy of any cause other than HIE were excluded. Control cases were defined as infants with gestational age 36 and 0/7 weeks and greater who had uneventful deliveries, normal umbilical cord pH, normal Apgar scores, and normal newborn examination findings.

Follow-up period: Follow-up was performed at age 18 to 42 months.

Intervention: Encephalopathy among infants with perinatal asphyxia was graded according to the highest modified Sarnat score recorded during the first 24 hours of age by the treating physician.

Outcomes: Cognitive, language, and motor development at age 18 to 42 months were assessed using the Bayley Scales of Infant and Toddler Development, Third Edition (BSITD-III). A mean difference in any cognitive, language, or motor composite score of at least 5 points between groups was regarded as clinically meaningful. Rates of survival with normal composite score, defined as within 1 standard deviation of the standardized mean, were calculated for all three domains.

Analysis and Sample Size: Power or sample size calculations were not reported. The Kruskal-Wallis test was used for continuous variables and the Fisher exact test was used for categorical and binary variables. Stepwise linear regression was used to adjust for the effect of HIE grade, country of birth, maternal age, maternal work status, maternal tertiary educational level, gestational age at birth, sex, and birth weight on the outcome. An F-test was used for postestimation comparison of outcome in mild versus moderate HIE infants. Cohen d with adjustment for unequal variances was used to estimate effect sizes comparing the difference in the means in infants with mild HIE and controls. Odds ratios were estimated using logistic regression with and without adjustment for country of birth.  Logistic and predictive mean matching were used for missing covariates. Chained imputation with 20 imputed data sets was used, including neurodevelopmental outcome, HIE grade, country of birth, and an interaction term containing both exposure and outcome. A post hoc analysis was performed that excluded patients whose care deviated from current guidelines of therapeutic hypothermia for HIE.

Patient follow-up: 471/690 (68%) of the eligible infants received follow-up through at least 2 years of age.

MAIN RESULTS:

Of the 741 unique live-born infants recruited to the four cohort studies, neonatal and follow-up data were available for 471/690 eligible (68.2%) patients. The majority of infants in the study were recruited from the BiHiVE2 cohort (501/741). Infants were categorized into the following groups: healthy controls (152/471, 32.3%), perinatal asphyxia without HIE (185/471, 39.3%), mild HIE (55/471, 11.7%), moderate HIE (56/471, 11.9%), and severe HIE (23/471, 4.9%). Groups varied significantly with regard to country of birth, with more infants recruited from Cork, Ireland in the healthy controls group (66.4%), perinatal asphyxia without HIE group (62.2%), and mild HIE group (87.3%). Race/ethnicity was not reported as a baseline characteristic.

The crude mean difference in composite cognitive score was -6 points (95% CI, -9.9 to -2.1) for children with mild HIE in comparison to healthy controls. When adjusted for country of birth, maternal work status, sex, and birth weight, the mean difference in composite cognitive score was -5.2 (95% CI, -9.1 to -1.3). Mean differences in language and motor composite scores between the two groups did not exceed the a priori clinically meaningful difference threshold. Survival with normal cognitive composite score occurred in 141/152 (92.8%) of healthy controls, 47/55 (85.5%) of children with mild HIE, 43/56 (76.8%) of children with moderate HIE, and 9/23 (39.1%) of infants with severe HIE.

A significant proportion (219/690, 31.7%) of children were lost to follow-up. There were no significant differences in umbilical cord pH, Apgar scores, and maternal work status between patients with and without complete follow-up data. However, children in the healthy controls group who were lost to follow-up were significantly more likely to have been born to mothers without tertiary education (33/72, 45.8%).

All 3 BSITD-III mean composite scores were greater but not statistically different between the cooled and normothermic subgroups in the mild HIE cohort. After post hoc exclusion of 8 of the 55 children (14.5%) with mild HIE treated with therapeutic hypothermia, the crude mean difference in cognitive composite score between the remaining 47 children with mild HIE and the 153 healthy controls remained significant (-7 points, 95% CI, -11.0 to -3.0). There was no significant difference between BSITD-III mean composite scores among untreated children with mild HIE versus children with moderate HIE who were treated with therapeutic hypothermia. The crude mean difference in cognitive composite score between these groups was -2.2 (95% CI, -8.1 to 3.7). 

STUDY CONCLUSION:

Children with mild HIE at birth have lower cognitive composite scores measured with the BSITD-III at 18 to 42 months of age compared to healthy newborns, but similar language and motor composite scores. Cognitive outcomes were not significantly different between infants with mild HIE who were not treated with therapeutic hypothermia versus infants with moderate HIE who were treated with therapeutic hypothermia.

COMMENTARY:

Multi-center international randomized-controlled trials of therapeutic hypothermia for moderate-to-severe HIE have reported a clinically significant reduction in death or survival with disability. Infants with mild encephalopathy, defined using the modified Sarnat examination, were excluded from treatment arms of early clinical trials due to potential risks associated with hypothermia and the assumption that outcomes among this group of infants were likely to be favorable¹.  However, more than half of all infants with HIE are categorized as mild², and in recent years, reports have recognized that survivors of mild HIE manifest both acute and chronic neurological impairment. Recent analysis of data from the pre-hypothermia era suggests an increased risk for behavioral dysregulation and attention deficits among older school-aged children with a history of mild HIE at birth^3,4. A large proportion of infants (28/54, 52%) enrolled in the Prospective Research on Infants with Mild Encephalopathy (PRIME) study had abnormalities on aEEG, MRI, or neurological evaluation at time of discharge from neonatal intensive care⁵.  A meta-analysis conducted by Conway et al found an abnormal neurodevelopmental outcome among 25% of infants pooled from 4 randomized controlled trials and 16 prospective observational cohort studies of infants with mild HIE⁶. The study by Finder et al demonstrates a clinically meaningful deficit in cognitive outcomes at two years of age among children with mild HIE at birth in comparison to healthy controls. While BSITD-III composite scores fell within the standardized normal range, the divergence and leftward shift in score distribution supports accumulating evidence that the developmental trajectory of children with mild HIE diverges from that of their healthy peers.

Despite more than a decade of experience with therapeutic hypothermia for HIE, evidence-based guidance is lacking on the best management approach to infants who present with mild HIE after birth^1,7. Concerns for abnormal development and the existing “evidence vacuum” have likely contributed to the progressive therapeutic drift of hypothermia treatment for infants with mild HIE⁸. Contrary to existing guidelines, hundreds of infants with perinatal acidemia and mild HIE have been treated with therapeutic hypothermia despite incomplete safety and efficacy data for its use in this population^8,9. In fact, 15% of infants with mild HIE in the Finder et al study were treated with therapeutic hypothermia. It is impossible to generalize on comparative outcomes of infants with mild HIE treated with hypothermia given that these studies are diverse and report small numbers of infants.

This study describes outcomes of a cohort created by pooling data from 4 separate trials. Limitations of this study that pertain, but are not limited to the use of a composite cohort, warrant discussion. First, that 68% of infants came from a single trial limits the generalizability of study results. Second, 32% of participants were lost to follow-up, which may jeopardize the validity of the study through differential retention of subjects¹⁰. The authors offered little information with regard to two major delivery models of outpatient special services: early intervention programming and individualized rehabilitative therapies. As neonatologists, we are aware that access to developmental-behavioral care, frequency of in-home rehabilitative therapies, and social determinants of health influence neurodevelopment during childhood. Third, the exposure of primary interest, mild HIE, was noted in only 12% of infants with available follow-up data. Fourth, inference regarding the effect of therapeutic hypothermia on the outcomes of infants with mild HIE was limited by an insufficient sample size of 8 infants.

A clinically meaningful deficit in cognitive function at two years of age was observed in infants with mild HIE, suggesting that mild HIE does adversely affect neurodevelopmental outcomes. Compositive cognitive scores were similar between untreated infants with mild HIE and infants with moderate HIE who were treated with therapeutic hypothermia. It remains unknown whether treatment with therapeutic hypothermia or other emerging neuroprotective therapies may improve cognitive outcomes for infants with mild HIE. The authors of this paper join several other investigators in recognizing the limitations of retrospective cohort and prospective observational studies with regard to solving the “conundrum of mild HIE⁷.” We agree that well-designed randomized controlled trials of therapeutic hypothermia and other neuroprotective therapies for infants with mild HIE at birth that can recruit an adequately powered representative population and achieve a high rate of follow-up are urgently needed.

EBM LESSON: Multiple Imputation for missing data

Missing data is problematic for multiple reasons; the loss of data decreases statistical power, may lead to biased inference, and jeopardizes the generalizability of results¹¹. In this study, follow-up data was available for 471/690 (48%) participants with neonatal data, and only 392/471 (83%) participants with follow-up data had complete data. Finder et al addressed missing data within their combined dataset via the statistical method of multiple imputation with chained equations (MICE).

Multiple imputation (MI) is a stepwise analytical strategy to account for missing data that uses a statistical model with a random error component to generate predicted values for missing data within a dataset. Unlike single imputation, MI allows for uncertainties in the imputations process.  With MI, the missing data is replaced by predicted values determined by an imputation model to form a complete dataset. This process is replicated multiple times, resulting in multiple complete datasets. The imputation model typically includes each variable used in the planned analysis as well as other variables that may be potential predictors of missingness and/or have missing data. The analysis is performed on each complete dataset separately and a single multiple imputation estimate (and its standard error) is calculated by combining the estimate (and its standard error) from each completed dataset. The number of imputations performed is generally proportional to the missing value rate in the dataset¹². MICE specifies separate, conditional imputation models for each variable with missing data and may be used for data of different distribution types. An important assumption of MI is that the missing data are missing at random. Sensitivity analyses should be conducted to measure any departures from the ‘missing at random assumption’ and to assess the impact of such departures on the MI results¹¹.

Acknowledgement: The Journal Club is a collaboration between the American Academy of Pediatrics - Section of Neonatal Perinatal Medicine and the International Society for Evidence-Based Neonatology (EBNEO.org)

REFERENCES:

1. El-Dib M, Inder TE, Chalak LF, Massaro AN, Thoresen M, Gunn AJ. Should therapeutic hypothermia be offered to babies with mild neonatal encephalopathy in the first 6 h after birth? Pediatr Res 2019;85:442-448.
   
   
2. Murray DM, O’Connor CM, Ryan CA, Korotchikova I, Boylan GB. Early EEG Grade and Outcome at 5 years After Mild Neonatal Hypoxic Ischemic Encephalopathy . Pediatrics 2016;138:e20160659.
   
   
3. van Handel M, Swaab H, de Vries LS, Jongmans MJ. Behavioral outcome in children with a history of neonatal encephalopathy following perinatal asphyxia. J Pediatr Psychol. 2010;35(3):286-295.
   
   
4. van Handel M, deSonneville L, de Vries LS, Jongmans MJ, Swaab H. Specific memory impairment following neonatal encephalopathy in term-born children. Dev Neuropsychol. 2012;37(1): 30-50.
   
   
5. Prempunpong C, Chalak LF, Garfinkle J, Shah B, Kalra V, Rollins N et al. Prospective research on infants with mild encephalopathy: the PRIME study. J Perinatol 2018;38:80-85.
   
   
6. Conway JM, Walsh BH, Boylan BG, Murray DM. Mild hypoxic ischaemic encephalopathy and long term neurodevelopmental outcome – A systematic review. Early Hum Dev 2018;120:80-87.
   
   
7. Chalak L, Latremouille S, Mir I, Sanchez PJ, Guilherme S. A review of the conundrum of mild hypoxic-ischemic encephalopathy: Current challenges and moving forward. Early Hum Dev 2018;120:88-94.
8. Saw CL, Rakshasbhuvankar A, Rao S, Bulsara M, Patole S. Current Practice of Therapeutic Hypothermia for Mild Hypoxic Ischemic Encephalopathy. J Child Neurol. 2019;34(7):402-409.
   
   
9. Goswami IR, Whyte H, Wintermark P, et al. Characteristics and short-term outcomes of neonates with mild hypoxic-ischemic encephalopathy treated with hypothermia. J Perinatol 2020;40:275-823.
10. Howe CJ, Cole SR, Lau B, Napravnik S, Eron Jr. JJ. Selection bias due to loss to follow up in cohort studies. Epidemiology 2016;27:91-97.
    
    
11. Rezvan RH, Lee KJ, Simpson JA. The rise of multiple imputation: a review of the reporting and implementation of the method in medical research. BMC Medical Research Methodology 2015;15:30.
    
    
12. Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York: John Wiley & Sons, 1987.