What Procalcitonin and CRP Can and Cannot Decide About Antibiotics

Procalcitonin and C-reactive protein are supporting signals, not verdicts. The best trial evidence shows procalcitonin can help clinicians safely shorten a course of antibiotics in hospitalized patients, but its record for deciding whether to start antibiotics is weak, and CRP has performed worse still. Neither marker tells you, on its own, whether an infection is bacterial, and neither should overrule a sick patient in front of you. The useful question is not "what is the number" but "what decision is this number allowed to inform."

Two markers, two different jobs

Both procalcitonin (PCT) and CRP rise when the body mounts an inflammatory response, and both climb higher, on average, in bacterial infection than in viral illness or sterile inflammation. That average difference is real, and it is also where the trouble starts. A test that separates two groups on average can still misclassify a large share of individuals, because the distributions overlap. A pneumonia can present with a low procalcitonin. A bad viral infection, major surgery, trauma, or cardiogenic shock can drive it up. The marker reports on the size of an inflammatory signal, not on the organism causing it.

This is why the framing of "can and cannot decide" matters more than any single cutoff. The same molecule behaves differently depending on the decision you attach to it: starting therapy, or stopping it.

Stopping antibiotics: the stronger case

The most durable finding across the biomarker literature is that procalcitonin can help clinicians stop antibiotics sooner without an obvious safety cost, in the right setting. The ADAPT-Sepsis randomized trial, reported in JAMA, tested this directly in a large population: 2,760 adults with suspected sepsis across 41 UK intensive care units, assigned to a daily PCT-guided protocol, a daily CRP-guided protocol, or standard care. The efficacy outcome was total antibiotic duration through day 28; the safety outcome was 28-day mortality.

The PCT-guided protocol reduced mean antibiotic duration by roughly nine-tenths of a day compared with standard care (about 9.8 versus 10.7 days), and it met the trial's noninferiority threshold for mortality. That is a genuine result, and it is also a modest one. As the CIDRAP summary of the trial notes, this amounts to roughly a 10 percent reduction in antibiotic exposure, achieved on top of usual care that already included stewardship. The ADLM (formerly AACC) academy guidance reaches a compatible conclusion: in ICU patients with sepsis, procalcitonin can be used to reduce antibiotic duration, and in lower respiratory tract infection it may help reduce exposure, with less supporting evidence.

Two cautions travel with that finding. First, the effect size is small enough that the marker is buying you days at the margin, not transforming care. Second, the benefit depends on clinicians actually acting on the protocol; adherence in real practice varies widely, and a stop rule only works if someone stops.

Starting antibiotics: the weaker case

The decision to withhold or begin antibiotics is where these markers are asked to do more than they can. The ADLM guidance is direct: the data do not support using procalcitonin to guide antibiotic initiation in ICU patients, and in children it should not be used as a stand-alone test to diagnose sepsis or to decide on starting antibiotics. A meta-analysis in children cited in that guidance found procalcitonin sensitivity and specificity of roughly 55 percent and 85 percent for serious bacterial infection. A sensitivity near 55 percent means the test misses a large fraction of true infections. You cannot safely rule out a dangerous infection with a tool that misses that many.

The ProACT trial makes the same point from the emergency department. It randomized 1,656 adults with suspected lower respiratory tract infection to procalcitonin-guided care or usual care, and providing the procalcitonin result did not reduce antibiotic use. Contemporary trials in respiratory infection have repeatedly failed to reproduce the antibiotic reductions seen in earlier, smaller European studies, in part because the comparison arm is no longer naive; stewardship is now part of standard care, which shrinks the room a biomarker has to improve on it.

CRP deserves its own note. In ADAPT-Sepsis, the CRP-guided protocol produced no meaningful difference in antibiotic duration versus standard care. CRP is cheaper and more familiar, but familiarity is not evidence, and on the specific question this trial asked, it did not change the decision.

How to read the number

A few principles keep these markers useful and honest. Treat a value as one input into a decision that already has a clinical hypothesis behind it, not as the hypothesis. Anchor interpretation to the specific decision and setting, because a cutoff validated for stopping therapy in the ICU tells you little about starting therapy in clinic. Watch the trend more than the single value, since a falling procalcitonin in an improving patient carries different weight than an isolated reading. And respect the non-infectious causes of elevation, so a high number in a post-surgical or shocked patient is not read as proof of bacterial infection.

There is also a structural lesson for anyone evaluating diagnostic evidence. When a marker's supposed benefit shrinks as the control arm improves, the marker was partly measuring the gap in usual care, not adding independent value. That is a general trap in test appraisal, and procalcitonin is a clean example of it.

This article is educational and not medical advice. Decisions about antibiotics belong to a clinician who can see the whole patient, of which any laboratory value is only one part.