Elevated endogenous erythropoietin concentrations are associated with increased risk of brain damage in extremely preterm neonates

Abstract

Background: We sought to determine, in very preterm infants, whether elevated perinatal erythropoietin (EPO) concentrations are associated with increased risks of indicators of brain damage, and whether this risk differs by the co-occurrence or absence of intermittent or sustained systemic inflammation (ISSI).

Methods: Protein concentrations were measured in blood collected from 786 infants born before the 28th week of gestation. EPO was measured on postnatal day 14, and 25 inflammation-related proteins were measured weekly during the first 2 postnatal weeks. We defined ISSI as a concentration in the top quartile of each of 25 inflammation-related proteins on two separate days a week apart. Hypererythropoietinemia (hyperEPO) was defined as the highest quartile for gestational age on postnatal day 14. Using logistic regression and multinomial logistic regression models, we compared risks of brain damage among neonates with hyperEPO only, ISSI only, and hyperEPO+ISSI, to those who had neither hyperEPO nor ISSI, adjusting for gestational age.

Results: Newborns with hyperEPO, regardless of ISSI, were more than twice as likely as those without to have very low (< 55) Mental (OR 2.3; 95% CI 1.5-3.5) and/or Psychomotor (OR 2.4; 95% CI 1.6-3.7) Development Indices (MDI, PDI), and microcephaly at age two years (OR 2.4; 95%CI 1.5-3.8). Newborns with both hyperEPO and ISSI had significantly increased risks of ventriculomegaly, hemiparetic cerebral palsy, microcephaly, and MDI and PDI < 55 (ORs ranged from 2.2-6.3), but not hypoechoic lesions or other forms of cerebral palsy, relative to newborns with neither hyperEPO nor ISSI.

Conclusion: hyperEPO, regardless of ISSI, is associated with elevated risks of very low MDI and PDI, and microcephaly, but not with any form of cerebral palsy. Children with both hyperEPO and ISSI are at higher risk than others of very low MDI and PDI, ventriculomegaly, hemiparetic cerebral palsy, and microcephaly.

Fig 1. Odds ratios (and 95% confidence intervals) for ventriculomegaly (A.) and hypoechoic lesion (B.) calculated with logistic regression models.

The three risk groups: ISSI only: (an inflammation-related protein concentration in the highest quartile on two days); hyperEPO only (an EPO concentration in the highest quartile on day 14); and ISSI+hyperEPO are each compared to the referent group that consists of newborns who had neither ISSI nor hyperEPO. All models are adjusted for gestational age.

Fig 2. Odds ratios (and 95% confidence intervals) of quadriparesis (A.), diparesis (B.) and hemiparesis (C.) calculated with multinomial logistic regression models with risk groups and adjustment for gestational age as described in Fig. 1.

Missing values indicate an inability to estimate odds due to complete separation of outcomes among exposed and unexposed in light of the small sample size.

Fig 3. Odds ratios (and 95% confidence intervals) of Very Low Mental Development Index (A.) and Very Low Psychomotor Development Index (B.) calculated with multinomial logistic regression models as described in Fig. 2.

Note: Only children with a GMFCS < 1 are included in the MDI analysis.

Fig 4. Odds ratios (and 95% confidence intervals) of 24 month head circumference Z-score < -2 calculated with logistic regression models as described in Fig. 1.

Note: Children with a birth head circumference Z-score < -2 are excluded from this analysis.

Fig 5. Summary of Associations.

We identify three patterns of increased risk of indicators of brain damage associated with hyperEPO and ISSI. hyperEPO only is identified with yellow, ISSI only is identified with orange, and the combination of ISSI+hyperEPO is identified with red. Reduced risk of an echolucent lesion associated with hyperEPO only is identified with green. Boxes with 2 or 3 separate colors indicate that 2 or 3 patterns were identified. Note: Cells in this table identify patterns of results of unique multivariable regression models fitted to answer whether or not children in each of three mutually exclusive study groups (hyperEPO only, ISSI only, or hyperEPO+ISSI) were at higher or lower risk of the brain damage indicator identified at the top of each column, relative to those in a referent group who did not have hyperEPO or ISSI, adjusting for gestational age category.