Neurodevelopmental Outcomes After Suspected or Proven Sepsis: Secondary Analysis of INIS Trial Database

Rationale Neonatal sepsis is an important determinant of adverse neurodevelopmental outcome [Adams-Chapman 2006, Stoll 2004]. A recent meta-analysis estimated that neonatal sepsis more than doubled the risk of adverse neurodevelopmental outcome in very-low-birth-weight (VLBW) infants (OR 2.09 [95%CI 1.65-2.65]), including cerebral palsy (OR 2.09 [95%CI 1.78-2.45]).[Alshaikh 2013] When a baby with signs suggestive of sepsis, it is common practice to commence intravenous antibiotic treatment as soon as possible after cultures from blood and relevant additional sites have been obtained. However, cultures remain negative in an important subgroup of babies treated for sepsis, despite convincing clinical and/or laboratory signs of sepsis being present. Although there is some evidence to suggest that long-term outcomes following neonatal sepsis may differ according to whether or not the diagnosis was confirmed by culture,[Stoll 2004, Schlapbach 2011] to the best of the investigators' knowledge a direct comparison between these groups has never been made. Greater understanding of the impact of culture-positivity on long-term outcomes in the setting of clinical neonatal sepsis is essential to better inform parents about the future prospects of their child and to guide patient follow-up.

Objective Investigate differences in neurodevelopmental outcome at two years after neonatal sepsis according to whether or not the diagnosis was confirmed by culture.

Research questions Primary Are neurodevelopmental outcomes at two years different between babies with suspected versus culture-proven sepsis? Secondary Are neurodevelopmental outcomes at two years different between babies with suspected versus culture-proven sepsis when considered according to the underlying pathogen?

Design Secondary cohort analysis of all babies included in the International Neonatal Immunotherapy Study (INIS) randomized controlled trial (RCT).[INIS Study Collaborative Group 2008]

Study population The study population consists of all babies included in the INIS trial. The INIS trial was an RCT of intravenous immunoglobulin (two infusions of 500 mg/kg body weight polyvalent IgG) versus placebo in the treatment of suspected or proven neonatal sepsis.[INIS Study Collaborative Group 2008] 3493 infants were included from 113 hospitals in nine countries.[INIS Collaborative Group 2011] There was no difference between the treatment groups in the primary outcome of death or major disability at the age of two years (RR 1.00 [95%CI 0.92-1.08]), or in any of the secondary outcomes. Neither was an effect of immunoglobulin demonstrated in any of the pre-specified subgroup analyses. As the intervention under study did not have an impact on any of the outcomes, secondary observational analyses within the dataset are feasible.

Exposure For the primary analyses two groups of babies with neonatal sepsis will be considered: those with proven sepsis and those with suspected sepsis. Sepsis will be considered proven when pathogenic organisms (i.e. bacteria or fungi) were cultured from blood and/or cerebrospinal fluid. All babies with clinical neonatal sepsis in whom cultures remained negative are considered to have suspected sepsis. In a secondary analysis, neurodevelopmental outcomes of babies with proven sepsis will be compared to those with suspected sepsis according to causative organism grouped as follows: coagulase-negative staphylococci (CONS); other Gram-positive bacteria; Gram-negative bacteria; fungi. Babies will be classified according to the sepsis episode at the time of inclusion in the INIS trail. The impact of any additional episodes of sepsis will be investigated in sensitivity analyses (see below).

Outcomes The primary outcome of interest is death or major disability at two years of age. Secondary outcomes (all at two years of age) include: death; disability; and components of disability: neuromotor disability; disability due to seizures; hearing disability; visual disability; communicative disability; cognitive disability. All types of disability are categorised into major/non-major/no disability. The investigators used two questionnaires; the Health Status Questionnaire (HSQ) was completed by a pediatrician and the Parent Questionnaire (PQ) was completed by a parent or carer. A third questionnaire, the Short Health Status Questionnaire (SHSQ), was developed by the Project Management Group in order to capture the primary outcome for difficult to find children. These questionnaires were sent to general practitioners or health visitors.

The disability classification for each domain (with the exception of the cognitive domain) was calculated from the answers to questions on the HSQ or the PQ [see INIS Collaborative Group 2011 (online supplement) for more detail]. If one questionnaire was missing, the calculation used only the answer on the other questionnaire. If the domain classification was unknown and a SHSQ had been received, the classification from that questionnaire was used for that domain only. Cognitive function was assessed in INIS by parents using the revised Parent Report of Children's Abilities questionnaire (PARCA-R1 [Saudino 1998]), as described previously [INIS Collaborative Group 2011 (online supplement)].

Analyses The following characteristics will be compared between babies with proven and suspected sepsis in tabular form: baseline characteristics, clinical characteristics at presentation with sepsis, neonatal outcome at discharge; outcome at two years. Primary and secondary outcomes at two years will be compared between the two groups in univariable analyses using Chi square test. Multivariable adjustment for potential confounders around birth and at presentation with sepsis (including country and INIS study drug) will be performed using separate logistic regression models for each outcome. The logistic regression models will be considered the primary analyses. Separate models will be constructed for major disability and for any disability (major+non-major).

Subgroup analyses Two sets of subgroup analyses will be undertaken. First, given the difference in aetiology between early onset (\<72 hours after birth) sepsis and late onset (>=72 hours after birth) sepsis, the data will be re-analysed according to whether the sepsis episode was early or late onset.

Second, as gestational age is a key determinant of adverse neurodevelopmental outcome, the investigators will undertake a subgroup analysis to investigate the impact of proven versus suspected sepsis on neurodevelopmental outcomes according to whether gestational age at birth was \< or ≥ 30 weeks.

Sensitivity analyses Missing values are present for a number of covariates to be included in the multivariable adjustment. By default, logistic regression analysis will be based on babies with complete data only on all variables in the model. The investigators will thus undertake a sensitivity analysis after multiple imputation of missing data on all covariates included in the primary model, creating and analysing five unique datasets.

For the primary analysis of this study, only the neonatal sepsis episode at presentation (i.e. at inclusion in the INIS trial) will be considered to classify babies as having suspected or proven sepsis. An important subgroup of babies however, experienced one or more additional episodes of sepsis following inclusion in the study, which may affect the findings depending on whether or not these were culture-proven. The investigators will therefore undertake two additional sensitivity analyses to assess the impact of subsequent episodes of sepsis on the difference in neurodevelopmental outcome according to whether or not proven sepsis was present. In a first sensitivity analysis the investigators will exclude from the cohort all babies in whom additional episodes of neonatal sepsis were present following the initial presentation with sepsis. This will facilitate a fair comparison of babies with single episodes of proven versus suspected sepsis. In a second sensitivity analysis of the full cohort, the investigators will re-classify as proven sepsis all babies labeled as suspected sepsis based on the initial presentation, in whom subsequent episodes of proven sepsis were present. The investigators will then compare neurodevelopmental outcomes according to whether or not proven sepsis was present at any time point during initial hospital admission.

Sample size The sample size of the cohort is based on the original RCT design.[INIS Study Collaborative Group 2008, INIS Collaborative Group 2011] For the purpose of this secondary analysis the cohort may thus be considered a convenience sample. To the best of the investigators' knowledge this cohort is the second largest to evaluate neurodevelopmental outcomes following neonatal sepsis.[Alshaikh 2013] In multivariate analyses of 6093 extremely-low-birth-weight (ELBW) infants from the National Institute of Child Health and Human Development (NICHD) cohort followed up at two years of age, Stoll and colleagues were able to demonstrate clinically relevant differences (OR ≥1.3) in key neurodevelopmental outcomes between uninfected babies and different categories of babies with neonatal sepsis with statistical significance.[Stoll 2004] The same accounted for their subgroup analyses according to causative organism.[Stoll 2004] In much smaller studies, others have also demonstrated statistically significant impact of neonatal sepsis on neurodevelopmental outcome, as reviewed elsewhere.[Alshaikh 2013] The investigators therefore expect the INIS cohort to have sufficient power to detect clinically relevant differences (OR ≥1.3) in neurodevelopmental outcomes between VLBW infants with suspected versus proven neonatal sepsis.

Ethical considerations The original RCT was approved by the national and local ethical committees of each participating hospital. Anonymised data were used for the purpose of this secondary analysis. According to UK and Dutch regulations, no formal ethical assessment is required for secondary analyses of anonymised data.

Dissemination Findings will be reported in a manuscript to be submitted for publication with a leading medical journal in an appropriate field (i.e. paediatrics, infection/immunology, neurology), and may be presented at international clinical and research meetings.