I have railed against the use of C-Reactive Protein, CRP, on this blog previously, it was my analysis that the CRP is sensitive, but with very poor specificity, both for early-onset sepsis, and for late-onset sepsis. A new systematic review in JAMA Pediatrics (Brown JVE, et al. Assessment of C-Reactive Protein Diagnostic Test Accuracy for Late-Onset Infection in Newborn Infants: A Systematic Review and Meta-analysis. JAMA Pediatr. 2020)
suggests that I was wrong (gasp!), CRP is not very sensitive either.
Analyzing 22 publications including over 2000 infants using CRP to diagnose culture-positive sepsis among mostly preterm infants after 72 hours of age. Among the infants in the studies who presented with clinical signs suggestive of sepsis, the systematic review overall included articles where positive cultures were found in 40%.
The results show a test that is of virtually no value at all, whatever threshold was used for deciding that a CRP result was positive. After analysis of the results, they found: “At the reported median specificity (0.74), sensitivity was 0.62 (95% CI, 0.50-0.72); at the reported lower quartile specificity (0.61), sensitivity was 0.76 (95% CI, 0.66-0.83); at the reported upper quartile specificity (0.84), sensitivity was 0.45 (95% CI, 0.34-0.57)”.
Sensitivity and specificity refer to the performance of the test, the meaning and usefulness of a test depend on the prevalence of the condition among those tested, which will then lead to the positive and negative predictive values. This is illustrated by the great editorial, with a Barrington-esque subtitle, which accompanies the systematic review, (Cantey JB, Bultmann CR. C-Reactive Protein Testing in Late-Onset Neonatal Sepsis: Hazardous Waste. JAMA Pediatr. 2020)
If a baby presents with signs suggestive of sepsis you could do one of 2 things, send 0.4 mL of precious blood to the lab for a CRP when you do the blood culture, or save the baby’s blood (or add it to the blood in the culture bottle to increase the yield) and flip a coin instead. This table from the editorial shows the relative value of a CRP test and a coin flip.
Flipping a coin saves blood, saves money, and is just as useful as performing a CRP!
One adverse consequence of measuring CRP is that there is sometimes an assumption that, even when the culture is negative, if the baby had an elevated CRP they must have “culture-negative sepsis” and they then receive multiple days of unnecessary antibiotics. I think the argument on rounds that we should continue the antibiotics “because it was a ‘heads'” would be laughed at, we should do the same thing when someone says we should continue antibiotics because the CRP was elevated.
Measuring CRP in the evaluation of late-onset sepsis should be abandoned. The big question to answer now is whether we consider ‘heads’ or ‘tails’ to be a positive test!
I have never been a believer for CRP, procalcitonin, white count, and I/T ratio. Recently I have seen normal procalcitonin but extremely high CRP but end up with positive culture for bacteria. I also saw some of my previous colleagues keep the healthy looking babies in nursery for antibiotic treatment, simply due to CRP remained abnormal, although the blood culture has been negative for 5 days. How many times my fellows or residents determined to start antibiotics due mainly to neutropenia ignoring the mother had severe pre-eclampia. Some also use increased I/T ratio as the reason which in fact has a very low inter-rater reliability. It needs experience to judge when and how to treat when we suspect a late sepsis.
Hi Dr. Barrington! I also have been calling them « CRaP levels » for years. As you mention, sensitivity and specificity should never be looked at in isolation. A quick in-head calculation that I call the « Elmo characteristic » or « the Sesame Street test » is to just add sensitivity and specificity together. The minimum value is 1 (coin flip) and the maximum value is 2 (perfect discriminator). It’s a mini version of the ROC curve. Like mentioned above, I’ve seen antibiotic treatment withheld or extended solely based on CRP values. I moved into a hospital that uses mg/L, which inflates the values I was used to seeing by 10x as well. I’ve also seen antibiotics started because the CRP went up over the first 48 hours, which is also normal after birth. Like bilirubin nomograms, it’s the desire for a simple « don’t think about it » answer and the Achilles heel of evidence based medicine (« the easy answer »). I agree that there is no substitute for experience.
I may not have enough room in this comment to do the argument justice, but here I go.
The value of a single CRP, at the initiation of illness, has little value. The biology of CRP, released after stimulation of interleukin, takes time, at least 12 hours. The value increases with obtaining serial CRPs. However, in late onset sepsis, where the patient probably does not have antibiotic in the body yet, certainly there is less value, especially at the initial evaluation.
But, in early onset, perinatally acquired disease, I feel CRP is valuable. In the population that we manage, many (most?) will have been exposed to maternal antibiotics. In my unit, in the last 80 infants born to mothers with chorioamnionitis or PPROM, our blood cultures were obtained on average at 40 hours after the first maternal antibiotic administration. Having a blood culture with no growth is not surprising, and in fact, probably had a role in the poor correlation of CRP to positive blood cultures. But, blood is only one organ. Barton’s pathological report in the 1990s showed that the presumed clinical cause of death was frequently respiratory, but the post mortem cause was infection, usually pneumonia, even though the blood culture at autopsy was no growth.
In adults, sepsis is defined by organ system response, and there is no requirement for a positive blood culture. In other words, they believe in “culture negative sepsis.” Are our patients immune from culture negative sepsis?
So what does that have to do with CRP? In our early onset patients, in whom blood cultures were obtained during antibiotic exposure, obtaining SERIAL CRPs may give clinicians confidence in discontinuing antibiotics if the CRP trend is low (negative predictive value). In those with an elevated CRP, especially with non-specific clinical findings, prolonging antibiotics until CRP normalizes may be prudent. I feel that this approach is a “both-and” solution. It allows for discontinuation of antibiotics in the likely unaffected baby (stewardship), while reserving antibiotics for the patient more likely in need.
So while I agree somewhat with the lack of usefulness in late onset disease, I don’t think CRP measures are “CRAP” in early onset disease in the common face of maternal antibiotic exposure.
Thanks for the comment and your thoughtful (as always) analysis, David. The systematic review was about late-onset sepsis only, and I agree therefore with some of your concerns. Doing a CRP at the time of a workup for late-onset sepsis is of little value, partly, I think because it does take some time for the acute phase reactants to lead to an elevated CRP. I think, almost always, if you are concerned enough about a risk of sepsis to poke the baby for a blood culture, you should start antibiotics, rather than decide based on another test: be it CRP, procalcitonin, or the direction the wind is blowing. With modern blood culture methods, and enough blood in the bottle, you can stop at 36 hours, with no risk of deterioration, the potential exceptions being ‘focal infections’ such as pneumonia or NEC.
For early-onset infections, in particular, when the mother has received antibiotics, you should probably add the nuance that you really need to be sure that the blood culture bottle had enough blood, and there may be a real chance of “culture-negative sepsis”. I say there “may” be a real chance of that, because I am not sure. If the antibiotics the mother received were effective enough that there was less than 1 cfu/mL in the babies blood, then, even if the baby has a systemic inflammatory response accompanied by an increase in CRP, they will not necessarily benefit from antibiotics.
I don’t, therefore, think the question is whether or not we “believe” in culture-negative sepsis, but whether a systemic inflammatory response, caused at some point by micro-organisms, benefits from antibiotic therapy.
I well remember a pair of 28 week twins with an almost identical presentation from a mother with florid chorioamnionitis, one of whom had cultures that were positive, and the other where they were negative. They both had cardiovascular collapse and respiratory failure. You could certainly refer to twin B as having culture-negative sepsis, but you could also refer to him as having a systemic inflammatory response with negative cultures. I have no idea if that second diagnosis needs antibiotics, but I tend to think that they are not necessary, and may have adverse impacts on the microbiome for no benefit.
In adults, “culture negative sepsis” is still primarily bacterial in origin, it’s just that the bacteria are in one of the many other organs that we humans have.
In EOS evaluations when the mother received antibiotics, an analogy can be made like this. Suppose you drew blood cultures and started antibiotics, and then 12 hours later the lab calls and states that the culture is growing gram (-) rids. You then redraw cultures to assure the blood is being cleared, and that second culture does not grow in 36-48 hours. Are you then going to stop antibiotics? Likewise, the fetus receiving antibiotics can clear bacteria by the time we obtain cultures, and then we are left trying to decide if the baby ever was infected or not. If the CRP remains normal, then the fetus/baby most likely was never infected. If the CRP goes to 40, 50, 60 mg/L, I am not comfortable saying that baby was not infected.
The CRP had to be driven up by something, and in the context that we are discussing (premature baby who’s mother received antibiotics for chorio or PPROM), it most likely is bacterial infection.
I’m not aware of anyone who used CRP to rule IN sepsis. I’ve only used it – and ever heard of people using it – to rule OUT sepsis. In the study you cite, it has a NPV of 95%+ which confirms its usefulness for this purpose. I use it to confirm stopping antibiotics that I’ve already decided to start, not to prove that I should have started them.
The systematic review shows that such use of CRP in late-onset sepsis is really not evidence-based. CRP is so non-specific that you would be as well off to toss a coin. If the blood culture is negative at 36 hours, just stop the antibiotics without worrying about the CRP.
Pingback: Neonatal Updates | Neonatal Research