Predicting mental and psychomotor delay in very pre-term infants using machine learning.

BACKGROUND: Very preterm infants are at elevated risk for neurodevelopmental delays. Earlier prediction of delays allows timelier intervention and improved outcomes. Machine learning (ML) was used to predict mental and psychomotor delay at 25 months. METHODS: We applied RandomForest classifier to data from 1109 very preterm infants recruited over 20 years. ML selected key predictors from 52 perinatal and 16 longitudinal variables (1-22 mo assessments). SHapley Additive exPlanations provided model interpretability. RESULTS: Balanced accuracy with perinatal variables was 62%/61% (mental/psychomotor). Top predictors of mental and psychomotor delay overlapped and included: birth year, days in hospital, antenatal MgSO4, days intubated, birth weight, abnormal cranial ultrasound, gestational age, mom's age and education, and intrauterine growth restriction. Highest balanced accuracy was achieved with 19-month follow-up scores and perinatal variables (72%/73%). CONCLUSIONS: Combining perinatal and longitudinal data, ML modeling predicted 24 month mental/psychomotor delay in very preterm infants ½ year early, allowing intervention to start that much sooner. Modeling using only perinatal features fell short of clinical application. Birth year's importance reflected a linear decline in predicting delay as birth year became more recent. IMPACT: Combining perinatal and longitudinal data, ML modeling was able to predict 24 month mental/psychomotor delay in very preterm infants ½ year early (25% of their lives) potentially advancing implementation of intervention services. Although cognitive/verbal and fine/gross motor delays require separate interventions, in very preterm infants there is substantial overlap in the risk factors that can be used to predict these delays. Birth year has an important effect on ML prediction of delay in very preterm infants, with those born more recently (1989-2009) being increasing less likely to be delayed, perhaps reflecting advances in medical practice.

Should clinical suspicion shift to ampicillin resistance organisms in certain preterm premature rupture of membranes scenarios?: Case of early-onset sepsis caused by E coli.

This case highlights the importance antibiotic stewardship, including reexamining current regimens use for intrapartum antibiotic prophylaxis, which may be accountable for the shift in organism responsible for some cases of early-onset sepsis.

Neonatal brainstem function and 4-month arousal-modulated attention are jointly associated with autism.

The authors evaluated the contribution of initially abnormal neonatal auditory brainstem responses (ABRs) and 4-month arousal-modulated attention visual preference to later autism spectrum disorder (ASD) behaviors in neonatal intensive care unit (NICU) graduates. A longitudinal study design was used to compare NICU graduates with normal ABRs (n = 28) to those with initially abnormal ABRs (n = 46) that later resolved. At 4 months postterm age, visual preference (measured after feeding) for a random check pattern flashing at 1, 3, or 8 Hz and gestational age (GA) served as additional predictors. Outcome measures were PDD Behavior Inventory (PDDBI) scores at 3.4 years (standard deviation = 1.2), and developmental quotients (DQ) obtained around the same age with the Griffiths Mental Development Scales (GMDS). Preferences for higher rates of stimulation at 4 months were highly correlated with PDDBI scores (all P-values < 0.01) and the GMDS Hearing and Speech DQ, but only in those with initially abnormal ABRs. Effects were strongest for a PDDBI social competence measure most associated with a diagnosis of autism. For those with abnormal ABRs, increases in preference for higher rates of stimulation as infants were linked to nonlinear increases in severity of ASD at 3 years and to an ASD diagnosis. Abnormal ABRs were associated with later reports of repetitive and ritualistic behaviors irrespective of 4-month preference for stimulation. The joint occurrence of initially abnormal neonatal ABRs and preference for more stimulation at 4 months, both indices of early brainstem dysfunction, may be a marker for the development of autism in this cohort.

Early medical and behavioral characteristics of NICU infants later classified with ASD.

OBJECTIVES: Recent evidence suggests higher prevalence of autism spectrum disorder (ASD) in NICU graduates. This aim of this study was to identify retrospectively early behaviors found more frequently in NICU infants who went on to develop ASD. METHODS: Twenty-eight NICU graduates who later received a diagnosis of ASD were compared with 2169 other NICU graduates recruited from 1994 to 2005. They differed in gender, gestational age, and birth cohort. These characteristics were used to draw a matched control sample (n=112) to determine which, if any, early behaviors discriminated subsequent ASD diagnosis. Behavioral testing at targeted ages (adjusted for gestation) included the Rapid Neonatal Neurobehavioral Assessment (hospital discharge, 1 month), Arousal-Modulated Attention (hospital discharge, 1 and 4 months), and Bayley Scales of Infant Development (multiple times, 4-25 months). RESULTS: At 1 month, children with ASD but not control children had persistent neurobehavioral abnormalities and higher incidences of asymmetric visual tracking and arm tone deficits. At 4 months, children with ASD had continued visual preference for higher amounts of stimulation than did control children, behaving more like newborns. Unlike control children, children with ASD had declining mental and motor performance by 7 to 10 months, resembling infants with severe central nervous system involvement. CONCLUSIONS: Differences in specific behavior domains between NICU graduates who later receive a diagnosis of ASD and matched NICU control children may be identified in early infancy. Studies with this cohort may provide insights to help understand and detect early disabilities, including ASD.

Differential regulation of the tyrosine hydroxylase and enkephalin neuropeptide transmitter genes in rat PC12 cells by short chain fatty acids: concentration-dependent effects on transcription and RNA stability.

At physiologic concentrations, butyrate regulates the expression of individual genes involving at least three mechanisms: (i) through induction of cis- and trans-acting butyrate-dependent transcription factors for selected genes, (ii) by inhibition of histone deacetylation and attendant chromatin remodeling and (iii) by affecting turnover of mRNAs. Our previous work illustrated gradual accumulation of mRNA for tyrosine hydroxylase (TH), the rate-limiting enzyme in catecholamine biosynthesis and the neuropeptide transmitter proenkephalin (ppEnk) in butyrate-differentiated PC12 cells (Nankova, B.B., Chua, J., Mishra, R., Kobasiuk, C.D., La Gamma, E.F. 2003. Nicotinic induction of preproenkephalin and tyrosine hydroxylase gene expression in butyrate-differentiated rat PC12 cells: a model for adaptation to gut-derived environmental signals. Pediatr. Res. 53, 113-118.). However, at higher physiological concentrations (6 mM), TH mRNA levels are significantly reduced while ppEnk mRNA transcripts remained elevated. These differential effects suggest suppression of endogenous TH gene transcription, targeted degradation of TH mRNA or both. By using nuclear run-on assays, we found that transcription increased for both endogenous TH and ppEnk genes, even at time points and concentrations when reduced steady-state levels of TH mRNA were observed. The reduction in TH mRNA was blocked by cycloheximide consistent with a protein-dependent mechanism. We also observed a dose-dependent accumulation of luciferase reporter molecules driven by TH promoter in transient transfection experiments, data that provide additional support for separate regulatory pathways. Significantly, butyrate-dependent decreases in TH mRNA were also reflected in a reduction in TH protein. Our results suggest a novel mode of regulation for TH by butyrate operating via both transcriptional and post-transcriptional mechanisms. We speculate that, depending on plasma concentrations of butyrate, this naturally occurring signaling molecule can function as an in vivo molecular switch to alter levels of TH mRNA, its protein and thus the biosynthesis of endogenous catecholamines.

Short chain fatty acids induce TH gene expression via ERK-dependent phosphorylation of CREB protein.

Butyrate modulates specific gene expression through various second-messenger signal transduction systems including activation of the PKA/cAMP pathway (Decastro, M., Nankova, B.B., Shah, P., Patel, P., Mally, P.V., Mishra, R., La Gamma, E.F., 2005. Short chain fatty acids regulate tyrosine hydroxylase gene expression through a cAMP-dependent signaling pathway, Brain Res. Mol. Brain Res. 142 28-38; Mally, P., Mishra, R., Gandhi, S., Decastro, M.H., Nankova, B.B., Lagamma, E.F., 2004. Stereospecific regulation of tyrosine hydroxylase and proenkephalin genes by short-chain fatty acids in rat PC12 cells, Pediatr. Res. 55 847-854). In the current report, we provide additional evidence that exposure to butyrate causes a rapid activation of the MAP kinase pathway, associated with increased phosphorylation of CREB. Under these conditions, no changes in relative amounts of CREB protein were observed by Western blot. Pre-treatment with the MAPK specific inhibitor (U0126) or the adenylate cyclase inhibitor dideoxyadenosine (ddA) abolished the butyrate-induced: (i) accumulation of TH mRNA, (ii) the phosphorylation of ERK1/2 as well as (iii) CREB phosphorylation. PC12 cells transfected with a TH promoter-luciferase reporter gene showed a robust induction in response to butyrate that was significantly reduced after co-transfection of either of two dominant-negative CREB expression vectors. Nuclear run-on assays demonstrated that butyrate increases endogenous TH gene transcription. We conclude that the initial steps of butyrate-induced gene activation are mediated through the CREB/CREB family of transcription factors which are coupled to both the MAP kinase and cAMP-dependent second messenger systems. Our data delineate a molecular mechanism through which short chain fatty acid's, their related drug-congeners (e.g., valproate) or even diet-derived butyrate (from fermentation of carbohydrates in the gut) can in principle, modulate brain catecholaminergic systems by modifying TH gene expression, dopaminergic levels and the corresponding animal behavior. These molecular relationships also offer a plausible explanation of how the well-recognized clinical effects of ketogenic diets can alter human behavior via the same central mechanisms.