Impact of Sepsis on Neurodevelopmental Outcome in a Swiss National Cohort of Extremely Premature Infants
OBJECTIVE: Neonatal sepsis causes high mortality and morbidity in preterm infants, but less is known regarding the long-term outcome after sepsis. This study aimed to determine the impact of sepsis on neurodevelopment at 2 years' corrected age in extremely preterm infants.
PATIENTS AND METHODS: This was a multicenter Swiss cohort study on infants born between 2000 and 2007 at 24 to 27 weeks' gestational age. Neurodevelopmental outcome was assessed with the Bayley Scales of Infant Development–II. Neurodevelopmental impairment (NDI) was defined as a Mental or Psychomotor Developmental Index lower than 70, cerebral palsy (CP), or visual or auditory impairment.
RESULTS: Of 541 infants, 136 (25%) had proven sepsis, 169 (31%) had suspected sepsis, and 236 (44%) had no signs of infection. CP occurred in 14 of 136 (10%) infants with proven sepsis compared with 10 of 236 (4%) uninfected infants (odds ratio [OR]: 2.90 [95% confidence interval (CI): 1.22–6.89]; P = .016). NDI occurred in 46 of 134 (34%) infants with proven sepsis compared with 55 of 235 (23%) uninfected infants (OR: 1.85 [95% CI: 1.12–3.05]; P = .016). Multivariable analysis confirmed that proven sepsis independently increased the risk of CP (OR: 3.23 [95% CI: 1.23–8.48]; P = .017) and NDI (OR: 1.69 [95% CI: 0.96–2.98]; P = .067). In contrast, suspected sepsis was not associated with neurodevelopmental outcome (P > .05). The presence of bronchopulmonary dysplasia, pathologic brain ultrasonography, retinopathy, and sepsis predicted the risk of NDI (P < .0001).
CONCLUSIONS: Proven sepsis significantly contributes to NDI in extremely preterm infants, independent of other risk factors. Better strategies aimed at reducing the incidence of sepsis in this highly vulnerable population are needed.
WHAT'S KNOWN ON THIS SUBJECT:
Neonatal sepsis is responsible for high mortality and morbidity in extremely preterm infants. Infections may harm the developing brain, leading to sequelae such as periventricular leukomalacia. In an earlier study, neonatal infections were associated with poor neurodevelopmental outcomes.
WHAT THIS STUDY ADDS:
The results of this study, based on a contemporary cohort, confirm that proven sepsis has a major impact on neurodevelopmental outcome, independent of other risk factors. Better strategies are needed to reduce sepsis incidence in this highly vulnerable population.
Although advances in perinatal care have led to an improved survival of very low birth weight infants (VLBW) over the last decades, survival without major neonatal morbidity has not increased.1 Of concern remains the relatively high proportion of VLBW infants surviving with developmental impairments and cerebral palsy.2,3 Established risk factors for poor neurologic outcome include low gestational age; male gender; bronchopulmonary dysplasia (BPD); brain injuries, such as intraventricular hemorrhage or periventricular leukomalacia; and retinopathy of prematurity (ROP).4 In addition, 2 studies based predominantly on infants born in the 1990s demonstrated that neonatal infections were associated with an increased risk of poor neurodevelopmental outcomes.5,6
Bacterial infections continue to cause a high burden on neonatal morbidity and mortality.7 Approximately 1.5% of VLBW infants suffer from early-onset sepsis and up to 21% from late-onset sepsis,8,9 with mortality rates ranging from 10% to 30%. However, both the epidemiology and the approach to neonatal infection are subject to changes over time and differ regionally.10,–,12 Therefore, the Swiss Society of Neonatology conducted a Neonatal Infection Surveillance for preterm infants born earlier than 28 weeks' gestational age between January 1, 2000, and December 31, 2007. The aim of the present study was to describe the effect of proven or suspected sepsis on neurodevelopmental outcomes at 2 years of age in this recent national cohort of extremely premature infants.
PATIENTS AND METHODS
Patients and Data Acquisition
This study included extremely premature infants with a gestational ages between 24 and 27 weeks born in Switzerland between January 1, 2000, and December 31, 2007, who were admitted to the NICU. Infants with major congenital malformations were excluded. Only infants in whom neurodevelopment had been assessed using the Bayley's Scales of Infant Development (BSID) II13 were included. Clinical and follow-up data for this study were extracted from the prospective national database of the Swiss Neonatal Network and Follow-up Group. The data collection and evaluation for this study was approved by the institutional ethical review boards and by the Swiss Federal Commission for Privacy Protection in Medical Research. Participating centers were obliged to inform parents about the scientific use of anonymized data. Parents had the right to refuse participation and to refuse the use of data related to their child. Since 2000, the network database prospectively collects a large set of perinatal and follow-up data of VLBW infants in all 9 Swiss perinatal centers. Routine follow-up of infants born earlier than 28 weeks' gestational age has been recommended and performed by the Swiss Neonatal Follow-up Group since 2000. After the recommendation was issued to perform follow-up examinations using the BSID-II, not all centers were able to implement this immediately. Therefore, centers only were included in this study from the year the BSID-II was routinely performed.
Proven sepsis was defined as a positive result on 1 or more bacterial or fungal cultures obtained from blood or cerebrospinal fluid in an infant with clinical signs of infection (temperature instability, irritability, apathia, feeding difficulties, prolonged capillary refill, apnea, tachycardia, and tachypnea) treated with antibiotics for 5 or more days or until death. Pathogens that may represent contaminations (coagulase-negative staphylococci) were considered as proven sepsis only if laboratory signs of infection (such as elevated C-reactive protein, left shift, or leukopenia) were present and if the sepsis episode had been treated for 5 or more days or until death. Suspected sepsis was defined as an episode with clinical and/or laboratory signs of infection in the absence of a positive bacterial or fungal culture from a normally sterile site in an infant who received treatment with antibiotics for 5 or more days or until death. In addition, necrotizing enterocolitis (NEC) with pneumatosis intestinalis or pneumatosis vena portae (Bell's stage II or higher)14 without bacteremia was considered as suspected sepsis. The uninfected group consisted of infants who had not developed proven sepsis, proven NEC, or suspected sepsis, as defined above until a postconceptional age of 36 weeks.
Brain injury was defined as intraventricular hemorrhage grade 3 or higher, according to the classification of Papile et al15 and/or cystic periventricular leukomalacia.4 BPD was defined as requiring additional oxygen at 36 weeks' gestational age.16 ROP was defined as ROP stage 3 or higher.17z scores for weight, length, and head circumference were calculated at birth, based on the growth curves by Voigt et al,18 and at the age of 2 years, based on the growth curves by the World Health Organization (www.who.int/childgrowth/standards/en/).
Socioeconomic status was calculated by means of a score reflecting both maternal education and paternal occupation and was classified as upper (scores 2–5), middle (scores 6–8), or lower (scores 9–12) class.19,20
Neurodevelopmental examination was routinely performed by experienced developmental pediatricians at 18 to 24 months' corrected age at the neonatal follow-up centers. Neurodevelopment was assessed using the BSID-II.13 The mean BSID-II score is 100 for the Mental Developmental Index (MDI) and the Psychomotor Developmental Index (PDI), with an SD of 15. Therefore, MDI and/or PDI values below 70 indicate neurodevelopmental delay. If BSID-II testing was not feasible because of a lack of cooperation, the examination was repeated 3 to 6 months later. Infants who were so severely impaired that testing with the BSID-II could not be performed were assigned a MDI and PDI score of 49. Cerebral palsy was defined as a nonprogressive motor disorder characterized by abnormal tone in at least 1 extremity and abnormal control of movement and posture21 and was graded according to the classification by Palisano et al.22 Neurodevelopmental impairment (NDI) was defined as at least 1 of the following: presence of cerebral palsy; MDI score lower than 70; PDI score lower than 70; severe hearing loss requiring auditory amplification; or bilateral blindness.
NDI and cerebral palsy were defined a priori as primary outcomes. An MDI score lower than 70, a PDI score lower than 70, and an MDI or PDI score between 70 and 84 in the absence of cerebral palsy were secondary outcomes. We also calculated a prediction model for the combined variable, adverse outcome, which was defined as NDI or death after 36 weeks' postconceptional age.4,5 The χ2 tests was used to compare subgroups in nonadjusted comparisons. Linear regression was used for continuous outcome variables (MDI and PDI). Logistic regression was used for dichotomous outcome variables.
We first assessed which clinical and demographic variables are related to NDI and cerebral palsy using univariable logistic regression analysis. We used random-effects multilevel regression models with center-specific intercepts to allow for clustering on the study center level further adjusted for year of birth. We then performed multivariable regression analysis and included all explanatory variables that showed a weak association (P < .10) in the univariable analysis (BPD, brain injury, socioeconomic status class, ROP, and postnatal growth, defined as the z score of body weight at the follow-up examination) or that are known to act as a confounder (gender, gestational age, and birth weight). Sensitivity analyses were performed using an extended and a reduced set of covariables.
Associations are given as odds ratios (ORs) with 95% confidence intervals (CIs) and the 2 sided P value. All analyses were performed by using Stata 10 (Stata Corp, College Station, TX).
Description of the Study Population
During the study period, 975 infants born between 24 and 27 weeks' gestational age were admitted to the participating NICUs (Fig 1). Of these, 702 survived to the follow-up age and were included. BSID-II testing was not available for 161 (23%) infants for various reasons. Thus, 541 (77%) children were included in the study. Baseline characteristics between infants with BSID-II and those without BSID-II testing did not differ significantly, except that BPD was more frequent in infants with BSID-II testing (Table 1).
Of 541 infants, 236 (44%) had not experienced proven or suspected sepsis or NEC and constituted the uninfected group. Suspected sepsis occurred in 169 (31%) infants, including 6 NEC cases without bacteremia. Proven sepsis was diagnosed in 136 (25%) infants, including 7 NEC cases with bacteremia. Baseline characteristics of the 3 study groups are listed in Table 2. Infants with proven sepsis were born at a slightly lower gestational age, compared with uninfected infants, and were more likely to manifest pulmonary problems, such as longer duration of ventilation and a higher rate of BPD. A total of 152 proven sepsis episodes, including 3 with meningitis, were recorded, and 17 infants had developed multiple sepsis episodes. Pathogens recovered in blood cultures are listed in Table 3.
Neurodevelopmental Outcome in Infants With Proven Sepsis or Suspected Sepsis Versus Uninfected Infants
Neurodevelopmental examinations were performed at a median of 23 months' corrected age (interquartile range [IQR]: 21–25 months). Overall, the median MDI score was 87 (IQR: 74–99), and the median PDI score was 86 (IQR: 73–96). NDI was found in 151 (28%) infants and cerebral palsy in 41 (7.6%) infants. Median MDI scores were 88 (IQR: 74–101) in uninfected infants, 86 (IQR: 74–98) in infants with suspected sepsis, and 85 (IQR: 71–98) in infants with proven sepsis (Fig 2A). In linear regression analysis, the 3 groups did not differ significantly in relation to MDI (P > .1). Median PDI scores were 86 (IQR: 76–98) in uninfected infants, 86 (IQR: 72–97) in infants with suspected sepsis, and 84 (IQR: 69–94) in infants with proven sepsis (Fig 2B). Infants with proven sepsis had significantly lower PDI scores (−4.39 [95% CI: −0.56 to −8.22]; P = .024).
Logistic regression models were calculated to investigate the association of neonatal risk factors with cerebral palsy and NDI. Table 4 shows the univariable and multivariable contributions of risk factors included in the model. We then assessed the association of proven or suspected sepsis with predefined neurodevelopmental outcomes (Table 5). In univariable logistic regression, infants with proven sepsis had a significantly increased risk of cerebral palsy, NDI, and of PDI scores lower than 70. This association remained similar after adjustment for several potentially confounding neonatal variables: proven sepsis was associated with a two- to threefold increase in the risk of cerebral palsy, NDI, and of PDI scores lower than 70 in comparison with uninfected infants. In contrast, infants with suspected sepsis did not have an increased risk of poor neurodevelopmental outcome in multivariable analyses. Several sensitivity analyses using an extended or reduced set of variables has highly similar results (Table 6).
Influence of Pathogen Type on Cerebral Palsy and NDI
Infants with Gram-positive sepsis (n = 112) were significantly more likely to have cerebral palsy and NDI when compared with uninfected infants (Table 7). Multivariable analyses confirmed that Gram-positive sepsis was associated with a fourfold risk of cerebral palsy and an approximately twofold risk of NDI. This was largely attributed to coagulase-negative staphylococci sepsis (n = 77), which significantly increased the risk of cerebral palsy. Sepsis caused by Gram-positive organisms other than coagulase-negative staphylococci (n = 43) was associated with a twofold-increased risk of NDI. Infants with Gram-negative sepsis (n = 29) had an increased risk of NDI in univariable analysis compared with uninfected infants, but this trend was weakened after adjustment for confounders.
Prediction of Outcome
We then calculated a logistic regression model to predict adverse outcomes, defined as NDI or death after 36 weeks' postconceptional age. BPD, brain injury, proven sepsis, and ROP showed the strongest association with adverse outcomes. We therefore modeled the risk for adverse outcomes according to whether 1, 2, 3, or 4 of these major risk factors were present in a patient. Adverse outcomes occurred in 56 of 272 (21%) infants without any of these risk factors, in contrast to 67 of 190 (35%) infants with 1 risk factor, in 29 of 61 (48%) infants with 2 risk factors, and in 5 of 7 (71%) infants with 3 risk factors (P < .0001) (Fig 3). The risk of adverse outcomes was increased by an OR of 2.38 (95% CI: 1.52– 3.71; P < .001) in infants with 1 risk factor when compared with infants with no risk factors, by an OR of 4.04 (95% CI: 2.16–7.56; P < .001) if 2 risk factors were present, and by an OR of 11.14 (95% CI: 2.02–61.55; P < .01) if 3 risk factors were present.
In this national cohort study on extremely premature infants born earlier than 28 weeks' gestational age, proven sepsis independently increased the risk for poor neurodevelopmental outcomes at 2 years of age. The presence of sepsis, BPD, brain injury, and ROP was highly predictive of adverse outcomes.
In our study, children with proven sepsis had a threefold-increased risk of cerebral palsy and an almost twofold-increased risk of NDI compared with children who had not suffered from any infection. Because multivariable analyses were adjusted for several important confounders, our results indicate that proven sepsis is independently associated with poor neurodevelopmental outcomes. Sepsis was among the 4 main risk factors influencing long-term outcomes in this population, together with BPD, brain injury, and ROP, all of which had a greater impact on outcome than gestational age, birth weight, and gender. Our results, based on a contemporary Swiss cohort, confirm results from the previous studies by Stoll et al6 and Bassler et al5 performed from 1993 to 2001. Of note, the strength of association (ORs) found in our study was very similar to theirs, further supporting the relationship between sepsis and poor neurodevelopmental outcomes.
Sepsis may lead to poor neurodevelopmental outcomes by several mechanisms. First, bacterial products and the cytokine storm in the course of the systemic inflammatory response syndrome can directly damage the highly vulnerable premature brain23 and other organs, such as the lung and retina. Additional support for this hypothesis comes from magnetic resonance imaging studies demonstrating white-matter injury associated with bacterial infections and NEC in premature infants.24,25 In accordance, the recent study by Martin et al26 reported a higher risk of neurodevelopmental dysfunction and microcephaly in NEC infants with late bacteremia. Of note, postnatal infections resulted in increased white-matter injury in premature infants in the recent study by Chau et al,27 whereas histopathological chorioamnionitis did not. Second, arterial hypotension during sepsis may cause cerebral ischemia reperfusion injury. Because of varying institutional practices in the definition and treatment of hypotension,28 we could not assess this factor in the present study. Third, sepsis may be an indicator of disease severity in extremely preterm infants. For example, prolonged mechanical ventilation is associated with a higher sepsis risk.8 Finally, susceptibility to sepsis may reflect the biological immaturity of some infants, whose developing brain may be more vulnerable to injury.
We have explored differences in outcome according to sepsis pathogens. Although the results need to be interpreted with caution because of the small numbers in the subgroups, infants with Gram-positive sepsis had the poorest neurodevelopmental outcomes. It is important to note, however, that the follow-up group was biased insofar as mortality was highest for children with Gram-negative sepsis (OR for death: 1.83 [95% CI: 1.10–3.07]) (additional details not shown). This finding supports similar findings by Stoll et al,8 who have shown that sepsis mortality rates are pathogen dependent and increase in the order of coagulase-negative staphylococci, other Gram-positive, Gram-negative, to fungal sepsis. Our results thus suggest that any episode of proven sepsis needs to be considered as a significant risk for both mortality and NDI.
The number of extremely premature infants surviving with significant neurologic disabilities remains high. A total of 8% of infants surviving to follow-up in the present study suffered from cerebral palsy and 28% from NDI. These rates are in agreement with previous studies3,6 from Europe and the United States. It is therefore crucial that every attempt is made to minimize potential contributing factors interfering with optimal brain development. Although low gestational age, low birth weight, and gender are nonmodifiable risk factors, strategies aimed at reducing the incidence of intraventricular hemorrhage and BPD have so far largely failed.16,29 In contrast, sepsis represents a partially preventable disease. Studies in adults, children, and neonates have shown that institutional protocols, such as care bundles, if applied correctly, can significantly reduce nosocomial infections.30 Institutional practices in central catheter handling, for example, have a highly significant impact on bloodstream infections in premature neonates.31 NICU guidelines aimed at minimizing central-catheter use, prompt removal of central catheters when unnecessary, meticulous skin care, early enteral feeding protocols, and weaning protocols have the potential to reduce sepsis further.32,–,34 Prospective multicenter surveillance studies are needed to identify points of vantage and to improve best-practice guidelines.
Several limitations of this study need to be addressed. First, 23% of infants were not tested with the BSID-II and were thus excluded. However, the rates of proven and suspected sepsis were very similar in included and excluded infants, making it highly unlikely that this biased the observed relationship between sepsis and poor neurodevelopmental outcomes. Our follow-up rate is comparable with that in the study by Stoll et al.6 Although the follow-up rate in the study by Bassler et al5 was higher, their cohort was based on a selected population from a randomized trial. Second, assessing neurodevelopment at 2 years of age may underestimate the full spectrum of cognitive and neuromotor outcomes, such as specific learning difficulties or milder motor dysfunctions.35,36 Third, a higher risk of poor neurodevelopment has been reported in infants with NEC and with meningitis.6,37,38 Because the incidence of NEC and meningitis was small in our cohort, we did not analyze these separately, but a sensitivity analysis excluding infants with NEC has very similar results.
This study has several strengths. Our data are based on a national cohort of extremely premature infants born after the year 2000 and thus represent the current situation in Swiss NICUs. Only infants who had been tested with the BSID-II by an experienced neuropediatrician were included because evaluation by Griffith Motor Scales may underestimate neurodevelopmental delay.39 The definition of sepsis required culture-proven bacteremia, which represents the gold standard for infection.40,41 Because postnatal growth has an independent effect on neurodevelopment,42 multivariable analyses were adjusted for postnatal growth.
The results of this study indicate that proven sepsis is independently associated with NDI in extremely preterm infants. We hope that a better understanding of risk factors associated with poor neurodevelopment may provide the basis for improved preventive and therapeutic strategies. In addition, these data may provide valuable tools in the counseling of parents and aid in the estimation of long-term prognoses. In view of the results of this study, it seems urgent to develop protocols aimed at reducing sepsis incidence in this highly vulnerable group.
Dr Aebischer received financial support from NycoMed (Zürich, Switzerland) for travel expenses. Dr Natalucci received financial support from the Swiss National Science Foundation (grant 33CM30-124101/1).
The Swiss Neonatal Network and Follow-up Group members: Aarau: Cantonal Hospital Aarau, Children's Clinic, Department of Neonatology (G. Zeilinger), and Department of Neuropediatrics (A. Capone); Basel: University Children's Hospital Basel, Department of Neonatology (R. Glanzmann), Department of Neuropediatrics, and Developmental Medicine (P. Weber); Berne: University Hospital Berne, Department of Neonatology (M. Nelle), and Department of Neuropediatrics (M. Steinlin and S. Grunt); Chur: Children's Hospital Chur, Department of Neonatology (W. Bär), and Department of Neuropediatrics (E. Keller and Ch. Killer); Lausanne: University Hospital, Department of Neonatology (J.-F. Tolsa), and Department of Child Development (M. Bickle-Graz); Geneva, Switzerland: Department of Child and Adolescent, University Hospital, Division of Neonatology (R. E. Pfister), and Division of Development and Growth (P. S. Huppi and C. Borradori-Tolsa); Lucerne: Children's Hospital of Lucerne, Neonatal and Pediatric Intensive Care Units (T. M. Berger), and Department of Neuropediatrics (T. Schmitt-Mechelke); St Gallen: Cantonal Hospital St Gallen, Department of Neonatology (A. Malzacher), Children's Hospital St Gallen, Neonatal and Pediatric Intensive Care Units (J. P. Micallef), and Department of Child Development (F. Steiner); Zurich: University Hospital Zurich, Department of Neonatology (R. Arlettaz Mieth), and University Children's Hospital Zurich, Child Development Centre (B. Latal).
We thank Luciano Molinari, MD, University Children's Hospital Zurich, for providing growth chart calculations. We also thank Bastian Lindauer, University of Basel; Monica Morend-Freschi, University Hospital Geneva; Claudia Schad, University Hospital Berne; Florian Bauder, Children's Hospital Lucerne; Jelena Buetler, University Hospital Zurich; and Christina Schaefer, Children's Hospital St Gallen, Switzerland, for help with data acquisition.
- Accepted April 14, 2011.
- Address correspondence to Luregn J. Schlapbach, MD, NICU and PICU, Department of Pediatrics, Inselspital, University of Bern, 3010 Bern, Switzerland. E-mail:
Dr Schlapbach had primary responsibility for the study design, data acquisition, analysis, and writing of the manuscript; Mrs Aebischer and Mr Adams were involved in study design, data acquisition, analysis, and writing of the manuscript; Drs Natalucci, Bonhoeffer, Nelle, and Bucher were involved in the study design, data acquisition, and writing of the manuscript. Dr Latzin was responsible for the statistical expertise, performed all final statistical analyses, and contributed to the writing of the revised manuscript; and Dr Latal supervised the design and execution of the study and data analyses and contributed to the writing of the manuscript.
FINANCIAL DISCLOSURE: The authors have indicated they have no personal financial relationships relevant to this article to disclose.
- VLBW —
- very low birth weight
- BPD —
- bronchopulmonary dysplasia
- ROP —
- retinopathy of prematurity
- BSID —
- Bayley Scales of Infant Development
- NEC —
- necrotizing enterocolitis
- MDI —
- Mental Developmental Index
- PDI —
- Psychomotor Developmental Index
- NDI —
- neurodevelopmental impairment
- OR —
- odds ratio
- CI —
- confidence interval
- IQR —
- interquartile range
- Fanaroff AA,
- Stoll BJ,
- Wright LL,
- et al
- Vohr BR,
- Wright LL,
- Poole WK,
- McDonald SA
- Schmidt B,
- Asztalos EV,
- Roberts RS,
- Robertson CM,
- Sauve RS,
- Whitfield MF
- Bassler D,
- Stoll BJ,
- Schmidt B,
- et al
- Stoll BJ,
- Hansen N,
- Fanaroff AA,
- et al
- Fernando AM,
- Heath PT,
- Menson EN
- Bayley N
- Chen ML,
- Guo L,
- Smith LE,
- Dammann CE,
- Dammann O
- Glass HC,
- Bonifacio SL,
- Chau V,
- et al
- Barrington K
- Aly H,
- Herson V,
- Duncan A,
- et al
- Horbar JD,
- Rogowski J,
- Plsek PE,
- et al
- Costello JM,
- Morrow DF,
- Graham DA,
- Potter-Bynoe G,
- Sandora TJ,
- Laussen PC
- Hack M,
- Taylor HG,
- Drotar D,
- et al
- Hintz SR,
- Kendrick DE,
- Stoll BJ,
- et al
- Buttery JP
- Modi N,
- Dore CJ,
- Saraswatula A,
- et al
- Copyright © 2011 by the American Academy of Pediatrics