Tolsma KW, Allred EN, Chen ML, Duker J, Leviton A, Dammann O. Neonatal Bacteremia and Retinopathy of PrematurityThe ELGAN Study. Arch Ophthalmol. 2011;129(12):1555-1563. doi:10.1001/archophthalmol.2011.319
Author Affiliations: Division of Newborn Medicine, Floating Hospital for Children (Drs Washburn Tolsma, Chen, and Dammann), and Department of Ophthalmology (Dr Duker), Tufts Medical Center, Boston, Massachusetts; Neuroepidemiology Unit, Children's Hospital Boston (Ms Allred and Drs Leviton and Dammann); and Perinatal Neuroepidemiology Unit, Hannover Medical School, Hannover, Germany (Dr Dammann).
Objective To explore whether early or late and presumed or definite neonatal bacteremia are associated with an increased risk of severe retinopathy of prematurity (ROP).
Methods We evaluated 1059 infants born before week 28 of gestation for ROP. Infants were classified as having early (postnatal week 1) or late (weeks 2-4) definite (culture-proven) or presumed (antibiotics taken for >72 hours despite negative blood culture results) bacteremia. Severe ROP was defined as stage 3 to 5, zone 1, prethreshold/threshold, or plus disease. We used time-oriented risk models to adjust for confounders.
Results In univariable, but not multivariable, analysis, newborns with presumed early bacteremia were at increased risk for plus disease (odds ratio [OR], 1.7; 95% CI, 1.1-2.7), and those with definite early bacteremia were at increased risk for stage 3 to 5 disease (1.9; 1.1-3.2). Infants who had presumed or definite late bacteremia were at increased risk for all 4 indicators of severe ROP in univariable analysis. In multivariable analysis, newborns with presumed late bacteremia were at increased risk for prethreshold/threshold ROP (OR, 1.8; 95% CI, 1.02-3.2), and those with definite late bacteremia were at increased risk for prethreshold/threshold ROP (1.8; 1.1-2.9) and plus disease (1.8; 1.05-2.9).
Conclusions Definite late neonatal bacteremia seems to be an independent risk factor for prethreshold/threshold ROP and plus disease, and presumed late bacteremia seems to be related to prethreshold/threshold ROP.
Approximately 50 000 children worldwide are blind due to retinopathy of prematurity (ROP), and many more have significant visual disturbances.1 Although our knowledge about the pathogenesis of ROP has increased considerably, much remains unknown.
The classic risk factors for ROP include low gestational age, low birth weight, and exposure to an oxygen-rich environment.2- 13 The mainstay of current prevention is, therefore, stringent control of supplemental oxygen in the most immature newborns. Despite such efforts, the absolute number of infants diagnosed as having ROP has continued to increase, most likely due to improved survival rates of very preterm infants.3,14- 17
We recently began exploring potential roles for perinatal infection and inflammation in ROP etiology.12,18,19 In this article, we expand our search for modifiable risk factors of ROP by exploring the relationship between neonatal bacteremia in its various forms and ROP in infants born before 28 completed weeks' gestation.
The ELGAN (Extremely Low Gestational Age Newborn) study was designed to identify characteristics and exposures that increase the risk of structural and functional neurologic disorders in ELGANs.20 Between March 22, 2002, and August 31, 2004, women who gave birth before 28 weeks' gestation at 1 of 14 participating institutions in 11 cities in 5 states were asked to enroll in this study. The enrollment and consent processes were approved by the individual institutional review boards.
Mothers were approached for consent either on antenatal admission or shortly after delivery. A total of 1249 mothers of 1506 infants consented to enroll in this study (Table 1). Approximately 260 women were either missed or refused to participate. The sample for this study consists of 1059 ELGANs who survived until postnatal day 28 for whom we had information about early and late bacteremia, ROP status, and placental bacteriologic and histologic features.
A trained research nurse interviewed each mother after delivery in her native language using a structured data collection form and following procedures contained in a manual. The mother's report of her own characteristics and exposures and the sequence of events leading to preterm delivery were taken as truth, even when her medical record provided discrepant information.
After hospital discharge, the research nurse reviewed the maternal medical record using a second structured data collection form. The medical record was relied on for events after hospital admission. The clinical circumstances that led to each maternal hospital admission and ultimately to each preterm delivery were operationally defined using data from the maternal interview and data abstracted from the medical record.21
Gestational age estimates were based on a hierarchy of the quality of available information. Most desirable were estimates based on the dates of embryo retrieval or intrauterine insemination or fetal ultrasonography before week 14 (62%). When these were unavailable, reliance was placed sequentially on a fetal ultrasonogram at 14 weeks or later (29%), last menstrual period without a fetal ultrasonogram (7%), and gestational age recorded in the log of the neonatal intensive care unit (1%).
The birth weight z score is the number of standard deviations the infant's birth weight is above or below the median weight of infants at the same gestational age in a standard data set.22 Mode of ventilation was defined as the highest level of support on each day. After the first week, this information was collected on days 7, 14, 21, and 28 and 36 weeks after menstruation. The number of days each infant received supplemental oxygen was recorded.
Clinicians selected times for blood gas measurements based on their own criteria. ELGANs were classified by their extreme blood gas measurements on postnatal days 1 through 3. In the present sample, the blood gas measurement that defined the extreme quartile varied by gestational age and postnatal day. We consequently classified infants by whether their extreme value was in the extreme quartile for their gestational age on each day, and we require that an infant be in the extreme quartile on at least 2 of the 3 days to be considered “exposed” to such extremes.
Medications were recorded if given on any day during the first 28 days and included surfactant, analgesics (ie, morphine sulfate, fentanyl citrate, and methadone hydrochloride), sedatives (ie, lorazepam, midazolam, and chloral hydrate), corticosteroids (ie, hydrocortisone and dexamethasone), and antibiotics.
The newborn's medical record was reviewed for receipt of blood products on postnatal days 7, 14, 21, and 28. These assessments were for those days only and not the intervening days. Of the 1003 infants for whom we had information about transfusions, only 7% did not receive a transfusion on the 4 days sampled.
Definite early bacteremia was defined as recovery of an organism from blood collected during the first week, and late bacteremia as recovery of an organism from blood collected during week 2, 3, or 4. Specific organisms were not identified. Presumed infections were culture negative, but the infant received antibiotics for more than 72 hours.
Eighty-two percent of the samples were obtained within 1 hour of delivery. Placentas were placed into a sterile basin and were transported to a sampling room, where a biopsy specimen was obtained under sterile conditions. The microbiologic and histologic procedures are described in detail elsewhere.23- 26
Participating ophthalmologists helped prepare a manual and a standardized data collection form and then participated in efforts to minimize observer variability. Definitions of terms were those accepted by the International Committee for the Classification of Retinopathy of Prematurity.27 In keeping with guidelines,28 the first ophthalmologic examination was within postmenstrual weeks 31 and 33. Follow-up examinations were as clinically indicated until normal vascularization began in zone 3.
We evaluated the generalized null hypothesis that the risk of severe ROP, defined as stage 3 to 5, zone 1, prethreshold/threshold, or plus disease, is not associated with bacteremia. In the entire ELGAN study sample, bacteremia29 and severe ROP13 varied with gestational age at delivery. In early sets of analyses, gestational age was adjusted for in 2 ways, by week of gestation (23, 24, 25, 26, and 27) and by groups of weeks (23-24, 25-26, and 27). Each method provided almost identical results. In this article, we present data adjusted for gestational age in groups of weeks.
Multivariable models were created to identify the contribution of relevant characteristics and exposures to the outcome of interest. To account for the possibility that infants born at a particular hospital are more like each other than like infants born at other hospitals, a hospital cluster term was included in all models.30
Because postnatal phenomena, such as the need for ventilatory assistance, can be affected by antepartum phenomena, we created logistic regression models in which risk factors are ordered temporally so that the earliest occurring predictors/covariates of an outcome are entered first and are not displaced by later occurring covariates.31- 36 For these time-oriented risk models, we categorized sets of antecedents/covariates by the time they occurred or were identified. We used a step-down procedure seeking a parsimonious solution without interaction terms. The contributions of the forms of bacteremia are presented as odds ratios (ORs) with 95% CIs.
Vaginitis was the only predelivery maternal characteristic associated with zone 1 disease, prethreshold/threshold ROP, and plus disease, and aspirin was the only drug taken during pregnancy that was associated with a near doubling of prethreshold/threshold ROP and plus disease (Table 2).
Maternal fever during labor was associated with an increased the risk of early bacteremia but not with severe ROP. Antenatal corticosteroid use, duration of labor, rupture of membranes, mode of delivery, magnesium exposure, cervical insufficiency, and fetal indication for delivery were not associated with significant differences in bacteremia or ROP status (data not shown).
By and large, the lower the gestational age, birth weight, and birth weight z score the higher the risk of all categories of bacteremia and ROP. Similar but less prominent trends were seen with head circumference z scores (Table 3).
Recovery of 2 or more organisms, or of anaerobes, from placental parenchyma was the only placental microbiologic characteristic associated with a minimally increased risk of plus disease but not bacteremia. Thrombosis of fetal stem vessels was associated with a doubling of the risk of zone 1 disease (data not shown).
Exposure to any medication (surfactant, corticosteroids, analgesics, and others) and therapeutic intervention (blood transfusion, patent ductus arteriosus, and others) was associated with increased risks of all categories of ROP. Blood transfusion in weeks 3 and 4 was associated with a near tripling of definite late bacteremia and with severe forms of ROP (Table 4).
Hyperoxemia was more common in infants with bacteremia and severe ROP, however defined. Insertion of a central venous catheter after the first week of life was associated with an increased risk of all categories of ROP (Table 5). Increasing the number of days of mechanical ventilatory assistance was associated with greatly increased risk of late bacteremia, zone 1 disease, prethreshold/threshold ROP, plus disease, and stage 3 to 5 ROP. Culture-proven tracheal colonization was associated with a nearly doubled risk of definite late bacteremia and less prominent increases in the risk of stage 3 to 5 ROP. Diagnoses of chronic lung disease/bronchopulmonary dysplasia, pulmonary interstitial emphysema, pulmonary hemorrhage, and early and persistent pulmonary dysfunction were associated with increased risks for all categories of ROP.
The incidence of stage 3 to 5 ROP was 22% to 25% in the absence of neonatal bacteremia and 33% to 42% in its presence. Late bacteremia was associated with a 2-fold increase in the incidence of zone 1 disease from 5% to 11% to 12%, plus disease from 8% to 16% to 18%, and prethreshold/threshold ROP from 11% to 21% to 24% (Table 6).
In multivariable time-oriented risk model analyses that combined presumed and definite sepsis (data not shown), late bacteremia was associated with a significantly increased risk of prethreshold/threshold ROP (OR, 1.8; 95% CI, 1.2-2.7) and plus disease (1.7; 1.1-2.7). The same was true for stage 3 to 5 ROP. However, once conventional or high-frequency ventilation on postnatal day 28 was added, the 95% CI for the multivariable point estimate included 1 (OR, 1.4; 95% CI, 0.98-1.9). No significant association was found between early bacteremia and the various severe forms of ROP.
To determine whether presumed and definite late bacteremia provided similar or different risk information, we evaluated each separately (Table 7). In separate multivariable time-oriented risk model analyses of presumed and definite late bacteremia, both were associated with a significantly increased risk of prethreshold/threshold disease. Definite late bacteremia was also associated with plus disease.
Perhaps the most important finding was that late bacteremia, whether presumed or definite, was associated with an increased risk of prethreshold/threshold ROP. This relationship remained statistically significant in the multivariable time-oriented risk model analyses adjusting for confounders. A second important finding was that presumed and definite bacteremia do not differ in their predicting of prethreshold/threshold ROP. Furthermore, the incidence of stage 3 to 5 ROP was increased in the presence of any neonatal bacteremia.
The relative scarcity of information about neonatal sepsis and ROP risk was the motivation for this study. In the past decade, studies37- 40 of the relationship between sepsis and ROP have focused on fungal sepsis. Still, several studies4- 7,10,11,41- 43 have reported sepsis (not restricted to fungi) as an independent risk factor for ROP. This group of articles is highly heterogeneous, making it difficult to compare one study to another.
One problem is the inconsistent definition of sepsis. For example, of the 7 articles cited in the previous paragraph, 3 did not define sepsis4,10,42; 1 study defined it as a positive blood culture, not differentiating fungal or bacterial11; 1 defined sepsis as a positive blood culture (again without fungal or bacterial differentiation but with the additional criterion of the presence of clinical symptoms)7; 1 study defined sepsis as “diagnosed clinically” and required changes in leukocyte count, C-reactive protein level, or positive blood culture findings41; and the final study diagnosed sepsis by clinical and hematologic findings but did not provide any other details.5
Most of the previously mentioned studies reported sepsis as an independent risk factor for ROP in univariable and multivariable analyses.4,5,7,10,41,42 In 2 of the studies,9,11 sepsis was called an independent risk factor; however, no multivariable analysis was included in the report. In 3 studies,6,17,43 sepsis was reported as a risk factor only during univariate analysis, and the significant relationship was lost in the multivariate analysis.
One of the central themes of the ELGAN study is that mediators of inflammation might damage the brain and other organs.44 ELGAN study investigators18,45 previously hypothesized and reported that perinatal inflammation plays a role in visual morbidities in preterm infants.
One of the normal stimuli for retinal blood vessel growth is physiologic local hypoxia, which prompts the release of vascular endothelial growth factor and other growth factors.3,46 These proteins not only induce blood vessel growth but also indicate where the growth should occur. When the growing blood vessels can deliver adequate amounts of oxygen, the stimuli for blood vessel growth are downregulated.
The pathogenesis of ROP seems to occur in 2 phases. The first phase is suppression of the synthesis and release of proteins that normally promote blood vessel growth. This results in the cessation of normal growth and the development of new vessels.3,46 The second phase is characterized by abnormal vessel growth and proliferation that is believed, in part, to be related to a consequence of the first phase.3,14,15,46,47 This abnormal development can lead to vessels growing into the vitreous rather than into the retina, leaving the eye at increased risk for the hallmarks of ROP, including hemorrhage, fibrous scarring, contractions, retinal detachment and blindness.3,14
The finding of a relationship between late bacteremia and increased severe ROP risk is biologically reasonable. The emerging concept of neutrophil-dependent, vascular endothelial growth factor–mediated inflammatory angiogenesis48 suggests that infection/inflammation might promote the second (vasoproliferative) phase of ROP. In experimental models of pathologic retinal angiogenesis, ω-3-polyunsaturated fatty acids were found to be protective, at least in part, through a downregulation of proinflammatory tumor necrosis factor.49,50
This study has several strengths. The large sample size makes it unlikely that we have missed important associations owing to lack of statistical power. Infants were selected on the basis of gestational age, not on birth weight, to minimize confounding due to factors related to fetal growth restriction.51 All the data were collected prospectively. Examiners were unaware of the medical histories of the infants, thereby minimizing diagnostic suspicion bias.52
The weaknesses of this study are, first, those of all observational studies. Second, the definitions of early (first postnatal week) and late (thereafter) bacteremia are slightly different from the generally accepted definitions.
In conclusion, these results suggest that neonatal bacteremia might, indeed, be an independent risk factor for severe ROP. Presumed or definite late neonatal bacteremia seems to be an especially important independent risk factor for prethreshold/threshold ROP and plus disease.
Submitted for Publication: February 10, 2011; accepted March 29, 2011.
Correspondence: Olaf Dammann, MD, MS, Division of Newborn Medicine, Floating Hospital for Children, Tufts Medical Center, 800 Washington St, PO Box 854, Boston, MA 02111-1526 (firstname.lastname@example.org).
Author Contributions: Drs Leviton and Dammann contributed equally to this article. Ms Allred had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None reported.
Funding/Support: This study was supported by cooperative agreement 5U01NS040069-05 with the National Institute of Neurological Diseases and Stroke, grant 5R21EY019253-02 from the National Eye Institute, program project grant 5P30HD018655-28 from the National Institute of Child Health and Human Development, and the Richard Saltonstall Charitable Foundation.
ELGAN Study Investigators:Baystate Medical Center, Springfield, Massachusetts: Bavesh Shah, Patrick O’Grady, Solveig Pflueger, and William Seefield. Beth Israel Deaconess Medical Center, Boston, Massachusetts: Camilia R. Martin, Bruce Cohen, Jonathon Hecht, and Deborah Vanderveen. Brigham and Women's Hospital, Boston: Linda J. Van Marter, Thomas F. McElrath, and Andrew B. Onderdonk. Massachusetts General Hospital, Boston: Robert M. Insoft, Laura Riley, Drucilla J. Roberts, and Tony Fraioli. Floating Hospital for Children at Tufts Medical Center, Boston: Cynthia Cole, John M. Fiascone, Sabrina Craigo, Terri Marino, Ina Bhan, and Caroline Baumal. UMass Memorial Health Care, Worcester: Francis Bednarek, Ellen Delpapa, Bo Xu, and Robert Gise. Yale University School of Medicine, New Haven, Connecticut: Richard Ehrenkranz, Keith P. Williams, Miguel Reyes-Múgica, Eduardo Zambrano, and Kathleen Stoessel. Wake Forest University Baptist Medical Center and Forsyth Medical Center, Winston-Salem, North Carolina: T. Michael O’Shea, Maggie Harper, Dennis W. Roth, and Grey Weaver. University Health System of Eastern Carolina, Greenville, North Carolina: Stephen C. Engelke, Hamid Hadi, John D. Christie, and Elaine Price Schwartz. North Carolina Children's Hospital, Chapel Hill: Carl Bose, Kim Boggess, Chad Livasy, and David Wallace. Helen DeVos Children's Hospital, Grand Rapids, Michigan: Mariel Poortenga, Curtis R. Cook, Barbara J. Doss, and Patrick Droste. Sparrow Hospital, Lansing, Michigan: I. Nicholas Olomu, Steve Roth, Gabriel Chamyan, and Linda Angell. Michigan State University, East Lansing: Nigel Paneth and Padmani Karna. University of Chicago Medical Center, Chicago, IL: Michael D. Schreiber, Mahmoud Ismail, Aliya N. Hussain, Ahmed Abdelsalam, and Kourous Rezaei. William Beaumont Hospital, Royal Oak, Michigan: Daniel Batton, Robert Lorenz, Chung-ho Chang, Michael Trese, and Antonio Capone Jr.
Previous Presentations: This study was presented at the Annual Meeting of the New England Perinatal Society; March 12-14, 2010; Newport, Rhode Island; and at the Annual Meeting of the Pediatric Academic Societies; May 1-4, 2010; Vancouver, British Columbia, Canada.