This three year observational study identified neonates who had clinical features of EONS and were born to mothers who had antenatal care in a refugee camp on the Thailand-Myanmar border. The objective of the study was to describe the epidemiology of clinically diagnosed EONS and to investigate possible bacterial causes and in particular GBS.
The incidence of EONS using a clinical definition was 44.8 episodes per 1,000 live births. It is not possible to compare this figure to that of other studies as clinically diagnosed EONS is not usually reported. This is unfortunate as we recognise that the true burden of neonatal sepsis will be greater than that which is simply based on positive blood cultures.
The majority of infants were diagnosed with EONS because they had a fever. Only four infants were diagnosed as having EONS because of the presence of other clinical signs. All temperatures were taken rectally and therefore accurately reflect core body temperature. Although correctly taking a rectal temperature does require training it is more straight forward to teach and to ensure accuracy (as the outcome is objective and not subjective) than the recognition of clinical signs. Hence it may be reasonable to use only the presence of fever as a diagnostic criterion for EONS in health care facilities run by staff with limited training. The incidence of EONS increased in March – May, corresponding to the hot season. This begs the question as to whether the infants who were diagnosed with EONS because of a fever in this time period, had a fever because of the high ambient temperature and not because they had true EONS.
Using clinical features alone may overestimate the incidence of EONS leading to the overuse of antibiotics. In our study only 86/165 (52.1%) infants who were clinically diagnosed as having EONS had a raised CRP at presentation. The National Institute for Health and Clinical Excellence has published guidelines for the use of antibiotics for EONS
. These guidelines recommend starting antibiotics in any neonate who has clinical signs of sepsis but to repeat a CRP measurement 18 – 24 hours after starting treatment, to determine those infants who truly do have sepsis. Adopting this strategy would increase the accuracy of calculating the incidence of neonatal sepsis in an environment where microbiological diagnosis is not possible.
The incidence of bacteriologically proven EONS was 0.7 per 1,000 live births (95% CI 0.1 – 2.1). This incidence is lower than that reported from other developing country settings. For example, in Kenya, Berkley et al. report a minimal rate for neonatal bacteraemia of 5.5 per 1,000 live births
. It was also notable that a significant proportion (10%) of infants who had an LP performed were found to have a raised CSF white cell count (>50 cells/mm3). Using a “normal” definition of < 20 white cells
 increases this proportion to 30%, yet a pathogen was not detected in any of these samples.
Escherichia coli was the only pathogen isolated from a blood culture and the only pathogen identified in CSF. The mothers of two of the three infants with a positive CSF E. coli PCR had received antibiotics in labour, potentially explaining the lack of bacterial growth. Specific PCRs were not available for all potentially relevant bacteria (e.g. Klebsiella spp.), thus we cannot exclude the possibility that other bacteria were implicated.
One explanation for the lack of positive microbiology is the high quality antenatal care that women living in Maela camp receive and the high rate of deliveries with skilled birth attendants, resulting in prompt antibiotic administration during labour when necessary. A significant proportion of women whose babies subsequently had clinical sepsis had prolonged rupture of membranes and 97% of these women had received antibiotics in labour.
One other significant factor is the use of un-prescribed antibiotic use in the community. Antimicrobials are readily available for purchase at many shops within Maela. A survey conducted in 2011 found that 25% of pregnant women had evidence of antimicrobial activity in their urine suggesting recent antimicrobial consumption (unpublished data). Maternal antibiotic use could suppress bacterial growth in samples collected from neonates, decreasing the conventional microbiological utility in determining aetiology.
It is conceivable that the use of specific PCRs against a larger number of potential pathogens (e.g. Klebsiella spp., additional viruses, etc.) might increase the yield of the CSF samples. Further work should be focused on this apparent burden of infection.
A further possible explanation for the low bacteraemia incidence is the amount of blood collected for each blood culture sample and the high bacterial contamination rate. However both values were similar to that reported by Berkley et al.
GBS is increasingly being reporting as a significant pathogen in the developing world, particularly in Southern Africa
[15, 18]. However GBS is unlikely to be an important pathogen in the current study described here. One explanation is that the proportion of women found to be colonised with GBS was relatively low. Additionally, there was significant use of IAP in this population, with nearly 20% of all women receiving IAP and 76.9% of women with significant risk factors for EOGBS (PROM, premature delivery and fever) receiving IAP. It is therefore conceivable that these antibiotics suppressed the growth of GBS in conventional cultures resulting in clinical disease with negative cultures. This suggests a potential role for molecular diagnostics and, although a GBS-specific PCR was possible for CSF samples (yielding no cases of GBS meningitis), it was not possible to do this on blood samples. The burden of EOGBS may therefore still be underestimated. It is also conceivable that the use of IPA could delay the onset of sepsis, however in the population described in this manuscript no neonate developed late onset sepsis from 7 – 28 days of age).
The majority of infants enrolled into the study were born at SMRU’s clinic. It has been argued that inborn infants are often over represented in studies on EONS as neonates who are born at home and develop EONS may die before reaching a health care facility
. This was not the case with our study, as all infants born to mothers who followed ANC at SMRU Maela were reviewed at 28 days of age in order to determine the neonatal mortality rate of the population. No infant was identified as having died from a potential septic episode at home.
No infant who was enrolled in this study died as a direct result of EONS. The only death from EONS that occurred at SMRU during the study period was in 2009, before the infant could be enrolled into the study. In total there were 55 early neonatal deaths over the study period. Using Lawn’s estimation that 26% of neonatal deaths are caused by sepsis, we could have expected to see 14 neonatal deaths due to EONS in this population in the time period studied. The explanation for this is multifactorial. Women received antibiotics when presenting with PROM and this may have reduced the infant’s risk of developing EONS. Additionally, medical staff were trained in the recognition of EONS and management protocols were clear.