Does ventriculomegaly without hemorrhage impact neurologic and behavioral outcomes of premature neonates?

MANUSCRIPT CITATION:

Pappas A, et al. Neurodevelopmental and Behavioral Outcomes in Extremely Premature Neonates with Ventriculomegaly in the Absence of Periventricular-Intraventricular Hemorrhage. JAMA Pediatr. 2018;172(1):32-42. PMID: 29181530.

REVIEWED BY

Ashley M. Lucke, MD¹
Joseph L. Hagan, ScD²
Michael D. Weiss, MD³

¹Department of Pediatrics, Children's National Medical Center, George Washington University School of Medicine, Washington, District of Columbia
 ²Department of Pediatrics (Neonatology), Baylor College of Medicine, Houston, Texas
³Department of Pediatrics (Neonatology), University of Florida, Gainesville, Florida

TYPE OF INVESTIGATION: Retrospective Cohort Study

QUESTION

Among preterm neonates <27 weeks' gestation, is nonhemorrhagic ventriculomegaly detected prior to 36 weeks' postmenstrual age, compared to normal ventricle size, associated with a greater risk of neurodevelopmental and behavioral impairment at 18-22 months corrected age?

METHODS

Design: Retrospective cohort study 

Blinding: Certified Bayley examiners were blinded to clinical data

Follow-up period: 18-22 months corrected gestational age

Setting: 16 US academic centers of the Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network

Patients:

Inclusion criteria: <27 weeks Gestational age (GA), birth weight >401g, head ultrasound (HUS) completed prior to 36 weeks postmenstrual age (PMA)

Exclusion criteria: Death prior to 12 hours of age, periventricular-intraventricular hemorrhage (PIVH) within 28 days, cystic periventricular leukomalacia by cranial ultrasound

Exposure of interest: Nonhemorrhagic ventriculomegaly

Outcomes: The primary outcome was neurodevelopmental impairment defined as one or more of the following: BSID-III cognitive score <70, disabling cerebral palsy (Gross Motor Function Classification System score ≥2), blindness (acuity < 20/200 in both eyes) or hearing impairment. Secondary outcomes included BSID-III cognitive, language and motor subscores, behavioral scores according to the Brief Infant-Toddler Social and Emotional Assessment, risk of combined death or neurodevelopmental impairment and risk of death alone.

Analysis and Sample Size: Retrospective data from August 2013 to August 2017 was analyzed. No sample size analysis was performed given the retrospective nature of the study. Unadjusted comparisons were made using x² tests or Fisher exact tests (categorical variables) and Wilcoxon 2-sample tests (continuous variables). A combination of logistic (categorical variables) and linear (continuous variables) regression modeling was used to evaluate the association between nonhemorrhagic ventriculomegaly and neurodevelopmental impairment at 18-22 months. NRN center was controlled for as a random effect and gestational age, sex, antenatal steroids, maternal education, late-onset sepsis, BPD, NEC and ROP were controlled for as fixed effects.

MAIN RESULTS

Of the 4,193 infants who met inclusion criteria, 848 were excluded due to PIVH or cystic periventricular leukomalacia. Of the remaining 3,345 infants, 300 (7%) had nonhemorrhagic ventriculomegaly (281 (94%) with outcome data available) and 3,045 had a normal cranial ultrasound (2,727 (90%) with outcome data available). Characteristics of the 281 neonates in the nonhemorrhagic ventriculomegaly group were: 24.7 weeks GA, 58% male, 57% white, 62% delivered via Cesarean section, 88% received antenatal steroids, 74% born to mothers with high school graduation completion. Characteristics of the group with normal cranial ultrasounds were:  24.9 weeks GA, 50% male, 52% white, 64% delivered via Cesarean section, 90% received antenatal steroids, 77% born to mothers with high school graduation completion. 

            Neonates with nonhemorrhagic ventriculomegaly had a statistically but not clinically significant lower mean gestational age (24.7 vs 24.9, P<0.001), were more commonly male (58% vs 50%; P=0.01), had more RDS treated with surfactant (94% vs 90%; P=0.04), and more postnatal complications including: physiologic BPD, late onset sepsis, meningitis, NEC, stage 3 ROP, inflammation-initiating illnesses and histological chorioamnionitis. More neonates with nonhemorrhagic ventriculomegaly died prior to discharge (28% vs 21%; P=0.01). Maternal education, antenatal steroids, delivery mode and ethnicity were not significantly different.

            Nonhemorrhagic ventriculomegaly was detected within 28 days in 85 (30%), at 36 weeks' postmenstrual age (PMA) in 156 (56%) and at both time points in 59 (21%). Nonhemorrhagic ventriculomegaly was usually bilateral, identified after multiple ultrasounds and associated with a higher risk of neurodevelopmental impairment, moderate to severe cerebral palsy, and death or NDI, but not death alone. There was no statistically significant difference in parent-reported behavioral problems between the two groups.

*NDI=neurodevelopmental impairment

STUDY CONCLUSION

In this retrospective cohort study, nonhemorrhagic ventriculomegaly was associated with a 3-fold increased odds (OR 3.07) of neurodevelopmental impairment at 18-22 months but not abnormal behavioral outcomes.

COMMENTARY

Ventriculomegaly and PIVH affect 20% of infants born ≤1500g¹. Fifteen percent of infants born at 22-28 weeks gestation have severe PIVH (grade 3 or 4)². In 2002 the American Academy of Neurology and Child Neurology Society released a practice parameter on neuroimaging to screen for PIVH in premature neonates. Consequently, the identification of post-hemorrhagic ventriculomegaly and understanding of it's neurodevelopmental impact has improved. However, the incidence and implications of ventriculomegaly alone, in the absence of any hemorrhage, remains poorly described. Previous studies of this specific population have been limited to a few small cohort studies with conflicting results^4,5. Fox et al showed ventricular dilatation diagnosed by cranial ultrasound at one month of age in neonates less than 30 weeks gestation correlated with poorer neurodevelopmental outcomes at 2 years of age⁴. On the contrary, in a Finnish cohort of VLBW and VLGA neonates Maunu et al did not find an association between isolated ventricular dilatation at term and developmental impairment at 2 years⁵. This study aimed to characterize the incidence and neurodevelopmental outcomes of neonates with nonhemorrhagic ventriculomegaly compared to those with normal cranial ultrasounds.

Cranial ultrasound timing was not specified by the study protocol. However, neonatal research network centers follow the American Academy of Neurology and Child Neurology Society practice guidelines. These guidelines recommend obtaining the first screening ultrasound between 7-14 days of life and repeating screening at 36-40 weeks PMA in all neonates born ≤30 weeks GA³. Ultrasounds were analyzed in two groups: early at <28 days of age and late at ≥36 weeks PMA. If multiple ultrasounds were obtained the one with the most severe result in the first 28 days was used for the early screen and the ultrasound closest within 7 days of 36 weeks PMA was used for the late screen.

 Nonhemorrhagic ventriculomegaly occurred in 7% of participants and was associated with a greater risk of neurodevelopmental impairment but not parent-reported behavioral problems. Specifically, the risks for cerebral palsy and/or impairment in cognitive, language, motor and visual function were higher in the ventriculomegaly group. Strengths of the study include a large sample size, multi-center, prospective data collection and minimal loss to follow up with a comprehensive neurodevelopmental assessment. There are some important limitations of the study to consider when interpreting the results. First, although the definition of ventriculomegaly is generally accepted as larger than one centimeter, there can be variability in center reporting and specific measurements of ventricle size. Second, the progression of ventricle size over time and it's rapidity of development were not recorded. The precise timing of onset of ventriculomegaly was also not available and some neonates may have had early PIVH that resolved prior to obtaining the first ultrasound or was not detected. Lastly, the participant demographics of RDS treated with surfactant, meningitis, inflammation-initiating illnesses and chorioamnionitis were significantly higher in the ventriculomegaly group and were not accounted for in the regression analysis. These potential mediating factors are counterbalanced though by more infants in the ventriculomegaly group having died prior to discharge and a disproportionate number of participants with chorioamnionitis being lost to follow up. Both would be expected to worsen neurodevelopmental outcome in the ventriculomegaly group^6,7, possibly underestimating the effect of ventriculomegaly on neurodevelopmental impairment. 

Although less common than post-hemorrhagic ventriculomegaly, prognostic counseling of nonhemorrhagic ventriculomegaly is particularly challenging due to the lack of available data. This study provides useful epidemiologic information and insights into the range of neurodevelopmental outcomes of nonhemorrhagic ventriculomegaly. The pathophysiology of ventriculomegaly in this specific population remains unclear. Potential etiologies include unrecognized systemic inflammatory processes,⁸ undetected post-hemorrhagic ventriculomegaly or compensatory ventricular enlargement secondary to white matter injury⁹. Further research is needed to understand the mechanisms of injury to alter management in hopes of minimizing these newly recognized risks of neurodevelopmental impairment. The application of magnetic resonance imaging in this population remains controversial. Although there is evidence that MRI imaging in cases of neonatal ventriculomegaly detects undiagnosed brain pathology¹⁰, it often does not change clinical management. All parents of premature neonates should be counseled on the risks for neurodevelopmental impairment and need for careful follow up by the pediatrician and early childhood intervention services, regardless of the presence of ventriculomegaly or attainment of an MRI.

EBM LESSON

Logistic Regression Analysis

Logistic regression is used to model a binary outcome as a function of one or more predictor variables which can be either categorical or continuous. Logistic regression can be used to estimate the probability of the event of interest occurring for given values of the predictor variables, and can also be used to examine the association of a predictor variable with the outcome after controlling for the effects of the other covariates in the regression model.   The latter application is particularly useful in observational studies that compare outcomes between groups.  The groups being compared might not be balanced in terms of covariates that are associated with the outcome, so it is important to compare the groups after controlling for the effects of covariates that could confound the relationship between the groups and the outcome.  For example, when examining the association of nonhemorrhagic ventriculomegaly with death or NDI, it was important for the authors to control for gestational age since infants with nonhemorrhagic ventriculomegaly were born at a significantly younger gestational age, which in itself is a risk factor for death or NDI. Selection of variables to be included in the logistic regression model can be done based on clinical rationale (e.g., including variables known to be related to the outcome), statistical rationale (e.g.,, the analyst might include  variables significantly associated with the outcome or use automated variable selection methods such as forward, backward or stepwise selection when prediction is the primary objective,¹¹ or the analyst could adjust for varaibles that are significantly different between groups in an effort to achieve covariate balance to reduce potential confounding when assessing group associations with the outcome is the primary objective),¹² or a combination of clinical and statistical rationale can be employed.  The authors state "Logistic regression modeling was used to evaluate the association between nonhemorrhagic ventriculomegaly and adverse outcomes while controlling for center differences… as a random effect and for other potentially confounding variables… as fixed effects."  Since the model included both fixed and random effects, it is called a mixed effects logistic regression model.  Fixed effects are used when factor levels are not treated as random parameters (e.g., the measured gender and gestational age of infants) whereas observed factor levels are considered to be randomly sampled from a larger population for random effects (e.g., the 16 centers in the study represent a random sample obtained from a larger population of all centers and inference regarding the effects of specific centers is not of interest).  

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. Stoll BJ, Hansen NI, Bell EF, Shankaran S, Laptook AR, Walsh MC, et al. Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network.  Neonatal outcomes of extremely preterm infants from the NICHD Neonatal Research Network. Pediatrics 2010; 126(3): 443-456.

2. Stoll BJ, Hansen NI, Bell EF, et al. Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network. Trendsin Care Practices, Morbidity, and Mortality of Extremely Preterm Neonates, 1993-2012. JAMA 2015; 314(10): 1039-1051.

3. Ment LR, Bada HS, Barnes P, et al. Practice parameter: neuroimaging of the neonate: report of the Quality Standards Subcommittee of the American Academy of Neurology and the Practice Committee of the Child Neurology Society. Neurology 2002; 58(12): 1726-38.

4. Fox LM, Choo P, Rogerson SR, Spittle AJ, Anderson PJ, Doyle L, et al. The relationship between ventricular size at 1 month and outcome at 2 years in infants less than 20 weeks' gestation. Arch Dis Child Fetal Neonatal Ed 2014; 99(3): F209-F214.

5. O'Shea TM, Allred EN, Dammann O, Hirtz D, Kuban KC, Paneth N, et al; ELGAN study investigators. The ELGAN study of the brain and related disorders in extremely low gestational age newborns. Early Hum Dev 2009; 85(11): 719-725.

6. Natarajan G, Shankaran S, Saha S, Laptook A, Das A, Higgins R, et al. Antecedents and Outcomes of Abnormal Cranial Imaging in Moderately Preterm Infants. J Pediatr 2018; 195: 66-72.

7. Shevell A, Wintermark P, Benini R, Shevell M, Oskoui M. Chorioamnionitis and cerebral palsy: Lessons from a patient registry. Eur J Paediatr Neurol 2014; 18(3): 301-7.

8. Kuban KCK, O'Shea TM, Allred EN, Fichorova RN, Heeran W, Paneth N, et al. The breadth and type of systemic inflammation and the risk of adverse neurological outcomes in extremely low gestational age newborns. Pediatr Neurol 2015; 52(1): 42-8.

9. Kidokoro H, Neil J, Inder TE. A New MRI Assessment Tool to Define Brain Abnormalities in Very Preterm Infants at Term. Am J Neuroradiol 2013; 34(11): 2208–2214.

10. Hintz SR, Barnes PD, Bulas D, et al. Neuroimaging and Neurodevelopmental Outcome in Extremely Preterm Infants. Pediatrics 2015; 135(1): e32-42.

11. Kleinbaum DG, Kupper LL, Muller KE. (2008) Applied Regression Analysis and Other Multivariable Methods, 3^rd edition. Belmont, California: Thomson Learning, Inc.

12. Pourhoseingholi MA, Baghestani AR, Vahedi M. How to control confounding effects by statistical analysis. Gastroenterol Hepatol Bed Bench 2012; 5(2): 79-83.