Evaluating a screening procedure for a rare serious phenomenon, such as early-onset neonatal sepsis, is tricky. A perfect screening process would catch all of the cases that require treating at an early stage and would be completely specific, thus eliminating treatment of non-infected babies. Even ultra-rapid PCR of a blood sample, covering all known pathogens, would not be perfect, as it would probably identify many infants with transient low-level bacteremia…
Given that there will never be a perfect test, we should ask how many infants who are not ultimately found not to be infected are we happy to treat for each infant who has sepsis, and how many infants with sepsis are we prepared to allow to go without early treatment in order to avoid treating many uninfected babies?
This is a variant, I guess of the number needed to treat, and number needed to harm calculations that we are now used to.
What I mean is “are we prepared to miss one sepsis in order to avoid treating 100 healthy babies?” or should the number be 1000, or even more? Treating babies unnecessarily means painful investigations and treatment, potential complications of peripheral or central venous access, usually separation of mother and baby, as most hospitals don’t keep babies with IVs in routine post-partum rooms with the mothers, disturbance of the microbiome of the infant, leading to increased risks of late-onset sepsis and NEC among preterm infants, and possible increased obesity, increased allergic disease, and disturbed hypothalamo-pituitary-adrenal responses, in the long term among babies born at term.
On the other hand risks of mortality from unnecessary treatment are extremely low, compared to the risks of untreated sepsis.
There are a number of studies looking at large databases and noting the number of patients screened, and the number of truly infected cases found compared to the number of babies treated, in particular using new sepsis calculators. If early-onset neonatal sepsis is less than 1 case per 1000 among full-term infants (as it is in most parts of the developed world) then such results may not be able to answer the question about numbers if cases potentially missed with much confidence.
Another way to address the question is to come at it from the other direction and examine the history of babies who do indeed prove to be infected to ensure that the different approaches to screening would have required a screen and pre-symptomatic treatment. This new article attempts to do just that. (Morris R, et al. Comparison of the management recommendations of the Kaiser Permanente neonatal early-onset sepsis risk calculator (SRC) with NICE guideline CG149 in infants >/=34 weeks’ gestation who developed early-onset sepsis. Arch Dis Child Fetal Neonatal Ed. 2020:fetalneonatal-2019-317165). In 5 maternity units in the South-West of England and Wales, there were 70 confirmed early-onset sepsis cases in term and late preterm babies over various periods spanning 2008-2017. The total incidence was 0.5/1000 live births.
Of the babies with sepsis, 31 became symptomatic and were treated after 4 hours of age, and neither the Kaiser Permanente Sepsis Risk Calculator (SRC) nor the NICE (National Institute of Clinical Excellence) guidelines, largely followed in the UK, identified them as at-risk.
There were 27 babies for whom both the NICE guidelines and the SRC suggested screening and antibiotics.
The remaining 12 babies were recommended to have screening and treatment by NICE guidelines, but not by the SRC.
So if we imagine a region with 100,000 term and late preterm births per year, and 80 of those babies with early-onset sepsis. About 30 of them will have clinical signs early and will receive treatment under any scenario, all of the guidelines are directed at the other 50. From this study it seems that about 2/3 of that group do not have identifiable risk factors and are not treated either under NICE guidelines or using the SRC approach, they develop clinical signs later.
The remaining 16 babies (very roughly) have a risk profile that suggests treatment using NICE, but not when using SRC.
Previous guidelines (such as NICE) lead to about 20,000 of the babies being screened and treated, in order to cover those 16 babies during the asymptomatic period. Using the SRC reduces this number to about 4,000.
In other words, if I have my estimates somewhere near right, another 16,000 sepsis evaluations and antibiotic treatments are required to cover the 16 babies who are identified as being at risk by NICE but not by the SRC.
I think we should question whether 999 unnecessary sepsis screens and antibiotic courses are justified by 1 baby treated during the pre-clinical period; especially as there are twice as many babies who are asymptomatic who are not screened at all, using any standard, which means we must remain vigilant for the occurrence of sepsis in all newborn infants, and be ready to screen and treat when signs of sepsis develop.
Only one death was recorded in this study, among a baby who had clinical signs from birth, none of the babies treated after they developed signs at >4 hours of age died.
One thing that struck me was the enormous proportion of neonatal sepsis caused by Group B Streptococcus in this study, which was 90%. As a disease that has almost disappeared in North America, I don’t think the conclusions of this study, or the rough calculations that I did, can be translated directly to my practice. The total incidence of EOS in Canada is currently much lower than that reported in this paper, probably about 0.1/1000 among term and late preterm babies, about 50% of which are GBS (Sgro M, et al. Population-based study of early-onset neonatal sepsis in Canada. Paediatr Child Health. 2019;24(2):e66-e73). In other words, early-onset GBS disease is about 1/10 as common here as in the south-west of the UK. Time for universal GBS screening in pregnancy in the UK, anyone?
This study, although not directly applicable everywhere, does confirm that there will be a few babies with early-onset neonatal sepsis that develop clinical signs after the first few hours of life, that are not identified by current approaches before they become sick. But they are very few in number, which means that staying vigilant for signs of sepsis is vital for all clinicians caring for newborns. Dramatic reductions in unnecessary sepsis screens and treatments can be accomplished with a small risk that some truly infected babies will present and be treated later.
The NICE guidelines give an NNT of about 250 (20,000 screened for 50 babies with EOS without early clinical signs), compared to about 50 for the SRC. Each one of the extra 16 babies identified by NICE guidelines requires an extra 1,000 screens and treatments.
Which is a lot.
I’m very intrigued by the fact that all of the 31 babies who did not screen positive and were only treated after they developed symptoms survived. My understanding for why we tolerate using a screening test with such a low positive predictive value was that the downside of missing a baby and not starting treatment pre-symptomatically was so great, but these data don’t support that.