515 - Tests and Models for Early Prediction of Cerebral Palsy In Children Born Preterm: A Systematic Review

Poster Number: 515
Publication Number: 515.1584

Background: Cerebral palsy (CP), a neuromotor condition, is the most common childhood motor disability with preterm infants at 50 to 100-fold higher risk. Currently, CP cannot accurately be diagnosed until 18 to 24 months of age, yet these early years are critical for neuroplasticity and impact a child’s disease progression. Thus, there is a need for a model that can accurately predict CP within the first few months of age so that interventions can be initiated early in high-risk patients to reduce CP severity.

Objective: Identify and critically appraise literature analyzing all the available tests/models for prognostication of CP within the first six months of age for preterm infants.

Design/Methods: We conducted a systematic literature review of five databases (Fig.1). Study eligibility included a focus on cohort or nested case-control study of preterm infants ( < 37 weeks gestational age), reported minimum of 10 CP cases, prognostic test/model studied prior to 6 months corrected age, and cerebral palsy diagnosed at 18 months of age or later. Non-English studies and ones only reporting General Movement Assessments were excluded. We used 29 medical subject headings (MeSH) and non-MeSH terms to identify relevant articles (Fig. 1). We used the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) and recorded our methods and results using the PRISMA 2020 Checklist guidelines.

Results: Out of 1,331 studies reviewed, 32 studies fit the inclusion criteria for the systematic review (Fig. 1). We found large discrepancies in the reported predictive accuracy of commonly used tests such as cranial ultrasound and MRI. Cranial ultrasound had sensitivities varying from 18% to 82% and specificities from 50% to 96% amongst the studies included in our review (Fig. 2). MRI was reported to have sensitivities ranging 63% to 100% and specificities from 80% to 93% (Fig. 3). Critical appraisal of studies indicated that tests with high reported prognostic properties suffered from selection bias and/or were not true cohort studies, the preferred study design for prognostic studies. None of the tests/models were sufficiently accurate for early CP prediction.

Conclusion(s): Available tests and models are insufficient for accurate prediction of cerebral palsy in preterm infants early in life. There is an urgent need for a validated model with higher prognostic capabilities that can enable individual level early diagnosis. Machine learning models that combine promising tests (e.g., diffusion and functional MRI) with clinical risk factors have the potential to achieve this goal.