With this blog I am leaving my beaten path: neonatal sepsis and probiotics. But so does this double-blind placebo-controlled study published today in Nature. To me, probiotics are still “something promising since 25 years”, without ever having substantiated that promise (like Ajax and the Chicago Cubs, until recently). In fact, colleagues of mine once led a study in which probiotics apparently killed patients with acute pancreatitis. This new study may change my view completely.
Let’s review the paper by critically appraising its validity, precision, relevance and generalizability.
The problem: “Sepsis in early infancy results in one million annual deaths worldwide, most of them in developing countries.” Sometimes a single sentence (the 1st of the paper) says more than 1,000 words.
This study is exceptional in all its aspects: A community-based, double-blind, placebo-controlled randomized trial in 149 randomly chosen villages in rural India, with 4,556 newborns enrolled hat all received the synbiotic (or placebo) orally for 7 days beginning on day 2–4 of life. The symbiotic contained Lactobacillus plantarum ATCC strain 202195 in combination with fructooligosaccharide. With an impressive study infra-structure community health volunteers identified 704 infants with suspected sepsis that were referred to the study hospital for further evaluation by a paediatrician. Upon clinical examination 196 infants were sent home without any treatment, 189 were treated in an ambulatory clinic (hospital outpatients) and 319 were recorded as cases of sepsis (WHO criteria) and blood cultures were obtained in 189 (blood cultures were not obtained in case of lower respiratory tract infection (LRTI)). All 319 were treated with iv antibiotics for at least 5 days.
The results: The occurrence of sepsis (in all variants) was significantly reduced in the treatment arm, with risk reductions of 82% for Gram-positive infections, 75% for Gram-negative infections, 47% for culture-negative sepsis and 34% for LRTI.
The primary endpoint was sepsis and/or death during the 8-week follow-up, which was 9% in placebo and 5.4% in treatment arm (40% reduction, number needed to treat 27). Yet, this effect was entirely due to the impact on sepsis; 10 infants died, 6 in treatment and 4 in placebo arm.
My appraisal: no questions on the validity of the study and the precision of the reported results. Both are excellent and undisputable (unless I have overlooked something in my first reading of the paper. Yes, count this as a disclaimer). Then generalizability. The study domain contained healthy newborns in India. In “developed” countries, neonatal sepsis mostly affects prematurely born infants in neonatal ICUs. If similar effects can be obtained in that population remains to be determined.
Finally, what about the relevance of the intervention. In this study, sepsis reduction did not translate (not even a bit) into less mortality. Fortunately, mortality rates were low in both groups, and the study was therefore underpowered to say anything meaningful on that outcome. Yet, I don’t think anyone will question that sepsis can kill infants (and other humans). Yet, I also expect the lay press will put emphasis on the least relevant part of the combined endpoint. Still, successful sepsis prevention is also important, especially if the costs of the intervention is 1US$ (for 7 days!). The authors suggest that “maintenance of a cold-chain” is needed, which could limit its feasibility in non-study settings.
Most interesting, though, is the diarrhea of hypotheses produced by these findings. How does it work? How to explain the effects on LRTI which are presumably of viral origin? I am sure that these findings will motivate many (more) to dive into host–probiotic interactions with the gastrointestinal mucosal and systemic host immunity. It is, therefore, of paramount importance that these results are reproduced.
That’s easier said than done. Enrolling 4,556 newborns and detecting 319 sepsis episodes in the community in India, resembles (and may even surpass) enrolling 84.495 healthy senior adults and detecting 139 vaccine-type pneumococcal pneumonia episodes in the Netherlands. Count on 10 years of work!