This came out before I started my blog, but think it is very important. I was reminded that I had never mentioned it on this blog when I met Joseph Randall Moorman at the airport on the way back from Istanbul. I think it is important because there is very little in recent years which has been shown in well-designed trials to reduce mortality in the NICU, but this innovative monitoring technique did.
I say innovative, but in fact people have been studying heart rate variability for years as an early indicator of various things, in this particular instance a reduction of baseline heart rate variability with an increase in transient decelerations as an early indicator of sepsis. Dr Moorman and his collaborators have been investigating this for years; they have developed ways of quantifying the changes which occur and expressing them as a relative increase in risk of developing sepsis in the next 24 hours.
In this study, (Moorman JR, Carlo WA, Kattwinkel J, Schelonka RL, Porcelli PJ, Navarrete CT, et al. Mortality Reduction by Heart Rate Characteristic Monitoring in Very Low Birth Weight Neonates: A Randomized Trial. The Journal of Pediatrics. 2011;159(6):900-6.e1) 3000 babies less than 1500 grams birth weight in 9 NICUs were put on a new monitor which records the index of heart rate characteristics (HRC). Half of the babies were randomized to have the information made available to the clinical staff, and the other half had the information hidden. The staff were trained to understand that when the HRC increased that was a potential indicator of early stages of sepsis. The use of an HRC index which is a relative risk, i.e. when the HRC is 2 the risk of sepsis is doubled, makes it very easy to understand and use as a part of clinical evaluation of the baby. A baby who otherwise looks perfectly fine and has a low pre-test probability of sepsis could just be evaluated and observed; for another baby at increased risk and some may-be signs of sepsis a full evaluation and starting antibiotics might well be appropriate.
The primary outcome variable for the trial was an increase in ventilator free days during the first 120 days of life. I am not sure why. The trial was powered for a 2 day difference in ventilator free days, they actually found a difference of 2.3 but it wasn’t “statistically significant” (see my previous post!) I presume the variability in this outcome was greater than anticipated. There are a lot of things other than sepsis that affect ventilator free days, so choosing this for the primary outcome could really be questioned, it was probably a compromise between the statisticians, trialists and neonatologists. I would urge all of you to read the article, it is a model of a scientific report, concise, clear, with all the essentials, and a very nice discussion section of the pathophysiology. The one lack is that there is no discussion of why they chose this primary outcome.
In the control group 10.2% of the babies died. In the group with the HRC visible 8,1% died. This difference was statistically significant. The effect seemed more marked in the smaller babies.
The advantage of being in the HRC visible group was mostly among those who did indeed have an episode of culture proven sepsis. “The mortality rate in the 30 days after the first episode of proven sepsis was 10.0% in the infants whose HRC monitoring results were
displayed compared with 16.1% in the control infants,” the intervention group did get a few more sepsis work ups, and a slight increase in antibiotic use.
Now what? I think this technology deserves to become widespread. I think that this particular algorithm is now better supported than any other monitoring technique in the NICU. I think that the company should licence the algorithm to other monitor manufacturers, who should buy the licence and make it available with software upgrades and new monitors at a reasonable price. I think that if anyone else wants to develop and try to sell their own algorithm, we should all be very wary, and recognize that this is the only HRC algorithm shown to reduce mortality, and anyone else needs to prove their own algorithms in a new large simple well-performed RCT. Like this one.