We have four main findings. First, children who had documented early bacteremia had a subtle inflammation-related protein signal on day 1 (CRP and IL-8), a considerably stronger day-7 signal (CRP, SAA, TNF-alpha, TNF-R2, IL-8, ICAM-1, and VEGF-R2), and no evidence of systemic inflammation on day 14. Second, children who had documented late bacteremia were at increased risk of having a moderate inflammation-related protein signal on day 7 (CRP, SAA, TNF-R2, and IL-8), a stronger inflammatory signal on day 14 (CRP, SAA, IL-6, TNF-alpha, TNF-R2, IL-8, MIP-1B, I-TAC, ICAM-1, E-SEL, and VEGF-R2), and a moderate sustained/recurrent inflammatory signal (CRP, SAA, IL-6, TNF-R2, IL-8, and I-TAC). Third, children who had presumed early bacteremia showed a subtle inflammation signal on day 7 (MMP-9 and VEGF), and had a sustained/recurrent elevation of only one protein, IL-8. Fourth, children who had presumed late bacteremia were not at increased risk of having elevated concentrations of any protein on postnatal day 1, 7, or 14.
One of the definitions of content validity is that a test/assessment truly reflects what it is supposed to represent [17
]. Applied to our assessments, children exposed to bacteria in the blood should be much more likely than others to have elevated blood concentrations of inflammation-related proteins than their peers not exposed to inflammatory stimuli. That is what we found, providing additional documentation our protein measurements have content validity.
Differences between presumed and documented bacteremia
We have been faulted for classifying a disorder based on the therapy received, mainly because physicians vary widely in their propensity to use many therapies [18
]. Yet, a physicians’ decision to administer antibiotics to a very preterm newborn who does not have bacteria in his blood likely conveys information about the infant’s appearing to be septic, or to be at increased risk of developing sepsis.. Perhaps some of these culture-negative infants were exposed to a non-bacterial stimulus for systemic inflammation. Nevertheless, we did not find that these newborns had systemic inflammation.
Multiple proteins are associated with documented late bacteremia
Others have reported that CRP, SAA, IL-1beta, IL-6, IL-8, and TNF-alpha are good indicators of neonatal bacteremia [19
]. One of the especially attractive features of CRP and SAA, for epidemiologic studies at least, is their long half-life, which means they can be identified even when opportunities to sample blood are infrequent. We found that nine proteins with presumed shorter half-lives (IL-6, TNF-alpha, TNF-R2, IL-8, MIP-1B, I-TAC, ICAM-1, E-SEL, and VEGF-R2) are also associated with bacteremia. This is not at all surprising given that an inflammatory stimulus can increase the expression of more than a thousand genes [21
]. Although others encourage the quest for more specific biomarkers/tests [22
], our data support the concept that elevations of multiple proteins taken together may be a better indicator of a very preterm newborn’s systemic inflammation [23
Implications for postnatal inflammation leading to organ damage
Newborns who developed sepsis are at increased risk of bronchopulmonary dysplasia/chronic lung disease [24
] and of indicators of brain damage [26
]. In our group of children, prolonged/recurrent elevated concentrations of inflammation-related proteins are characteristic of those who developed indicators of lung [32
] and brain damage [33
]. Here we report that those who had late bacteremia are among the children most likely to have elevated concentrations of the proteins that have been shown to be associated with indicators of lung and brain damage. The resulting inference is that bacteria cultured from the blood sampled after the first postnatal week can be a stimulus for the inflammatory response we measured. We do not have the information to separate host and pathogen contributions to what we observed.
Strengths and limitations
Our study has several strengths. First, we included a large number of infants, making it unlikely that we have missed important associations due to lack of statistical power. Second, we selected infants based on gestational age, not birth weight, in order to minimize confounding due to factors related to fetal growth restriction [35
]. Third, we collected all of our data prospectively. Fourth, our protein data are of high quality [36
] and have high content validity [6
Our major limitation is that we did not record information about the organisms recovered from the blood of our subjects. Thus, our classification of documented bacteremia includes such highly virulent organisms as Staphylococcus aureus along with much less virulent organisms, such as Staphylococcus epidermidis.
The other major limitation of our study is that we are not certain that the bacteremia always occurred prior to the elevation of the protein concentration. We last measured proteins on day 14 and last recorded bacteremia for the week that included days 15–21, but did not know when during that week the blood that harbored the bacteria was collected.
A limitation for those who want to compare our study to others is our definition of early bacteremia as bacteremia during the first postnatal week, rather than bacteremia limited to the first three postnatal days as defined by others. To what extent we or others misclassified the last four days of the first week remains to be determined.