All neonates with a GA at birth <35 weeks hospitalized at the Leiden University Medical Center neonatal intensive care unit between January 2010 and December 2022 were included in the study. Part of this cohort has previously been included in a study investigating the occurrence of thrombocytopenia in SGA neonates [20]. From this cohort, all neonates with a complete blood count sampled at a single timepoint within the first 72 hours after birth were selected. Neonates without a complete blood count within 72 hours after birth were excluded. Neonates who presented with clinical signs of infection for whom a C-reactive protein (CRP) >10 mg/L was measured were excluded to remove infection as a cause for altered complete blood count. Monochorionic twins diagnosed with twin-twin transfusion syndrome or twin anemia polycythemia sequence were excluded from the study, as were neonates diagnosed with polycythemia, hemolytic disease of the fetus and newborn, fetal and neonatal alloimmune thrombocytopenia and congenital anomalies to reduce confounding. Neonates were grouped based on birthweight, with neonates with a birthweight <3rd percentile for GA and sex classified as extreme SGA, neonates with a birthweight >3rd percentile and <10th percentile for GA and sex classified as moderate SGA, and those with higher birthweights as controls. Birthweight curves from Perined, the Dutch national organization for perinatal care registration, were used as reference [21]. For this retrospective study, the requirement of review and approval by the Medical Ethical Review Committee was waived by non-WMO Review Committee of the LUMC, as the Medical Research Involving Human Subjects Act did not apply (reference number 23-3064).

Clinical data was extracted from electronic patient files and included pre- and perinatal records and care, mortality, neonatal morbidities and interventions. Peripheral full blood counts and CRP measurements were performed as part of routine care within the first 3 days after birth. A neonate was diagnosed with asphyxia after meeting one of the following criteria: Apgar score <5 after 5 minutes, umbilical cord or peripheral blood pH<7.0, base excess < -16 mmol/L in umbilical cord or peripheral blood, umbilical cord or peripheral blood lactate > 10.0 mmol/L, resuscitation or ventilation for at least 10 minutes after birth.

A Kruskal-Wallis test was used for analysis of continuous variables between the 3 groups. P values of <0.05 were considered statistically significant. For cell types with significant Kruskal-Wallis test results, a Dunn test pairwise analysis with Bonferroni multiple comparisons adjustment was performed. After this, multiple linear regression analyses were performed to determine the effect of clinical variable on blood parameters. Eight clinical variables possibly influencing the blood parameters were used in univariable linear regression, namely sex, GA, birthweight, SGA group, multiple pregnancy, birth method, asphyxia and time between birth and blood sampling. The variables with a P value of <0.1 were included in the multivariable linear regression, to ensure no potentially relevant clinical variables were missed. There, P values of <0.005 were considered to be statistically significant, as an adjustment for multiple testing was applied. For the regression model of the neutrophils, monocytes and lymphocytes, the cell counts were transformed using a natural logarithmic transformation. For the interpretation of the effect sizes, the estimate is exponentiated to calculate the percent change in cell count per unit change of the clinical variable. The leukocyte, basophil and eosinophil counts were not included in the multiple linear regression. The leukocytes were excluded as its subsets were analyzed individually, whereas the basophil and eosinophil counts contained natural zeros, making transformation difficult. Since we hypothesized basophils and eosinophils not essential to explain the clinical differences between SGA neonates and controls, these were excluded. Z scores were calculated for birthweight and used in the regression models to reduce collinearity with both the SGA grouping and GA at birth. The birthweight z score was calculated using the following formula: (birthweight–mean birthweight)/birthweight standard deviation. The mean birthweight and standard deviation per sex and GA at birth from the Perined birthweight curves were used [21]. All statistics and visualizations were performed using R ver. 4.4.0 (R Foundation for Statistical Computing, Austria). Results from the Kruskal-Wallis test were represented in violin plots. These are an extension of the summary statistics displayed box-and-whisker plot to also include local density estimates [22]. These plots were chosen to better display the distribution of the cell count data per SGA group.

Perinatal characteristics of the included 1,940 neonates are presented in Table 1. The neonates in the extreme and moderate SGA groups differ from the control group by birthweight, as expected, as well as the GA at birth. The number of births in current pregnancy and the method of delivery also differed between the SGA groups and controls. The extreme and moderate SGA groups had fewer neonates with a GA<27 weeks than the control group, which may explain the significant difference between these groups (Supplementary Fig. 1). It is likely that neonates who are both extreme SGA and extremely preterm have decreased chances of survival, explaining this phenomenon in our cohort.

We analyzed the leukocyte and leukocyte differentiation cell counts in the 3 SGA groups. As expected from previous literature, the leukocyte numbers were lowest in the extreme SGA group, increasing through the moderate SGA and control groups (Fig. 1A; Supplementary Table 1) [18,19]. Similarly, the lowest number of neutrophils occurred in the extreme SGA group, with cell numbers increasing slightly although non-significantly in the moderate SGA group and highest in the controls, as expected based on literature (Fig. 1B; Supplementary Table 1) [18,19,23]. The neutrophil count in both the moderate and extreme SGA groups was found to be significantly decreased compared to controls. When looking at the monocyte and lymphocyte counts, their cell number increased from extreme to moderate SGA, with highest numbers in the control group (Fig. 1C and D; Supplementary Table 1). With regard to both basophils and eosinophils, while no significant differences were found between the extreme and moderate SGA groups, cell numbers were lower in both SGA groups when compared to controls (Fig. 1E and F; Supplementary Table 1).

We did not find a significant difference between any of the groups in the number of erythrocytes in the Kruskal-Wallis test (Fig. 2A; Supplementary Table 2). In contrast, the concentration of hemoglobin (Hb) was significantly higher in both SGA groups compared to controls, but not between the SGA groups (Fig. 2B; Supplementary Table 2). Because of this disparity between erythrocyte count and Hb concentration in SGA groups compared to controls in our cohort, we also analyzed the mean corpuscular Hb (MCH) and mean corpuscular volume (MCV). Both the MCH and MCV were significantly increased in neonates with either extreme or moderate SGA compared to controls (Fig. 2C and D; Supplementary Table 2). The increased MCH of SGA neonates may explain why the Hb concentration was increased while no differences were seen in erythrocyte counts between SGA groups. Finally, when investigating thrombocytes, it was determined that the neonates with extreme SGA had the lowest count, increasing in neonates with moderate SGA and highest in controls (Fig. 2E; Supplementary Table 2).