Before ordering CRP
Ask what the result could change. CRP can support a defined diagnostic or monitoring pathway, but it cannot identify the source of inflammation, diagnose sepsis or decide admission on its own. If the patient is unstable, assess oxygenation, perfusion and organ dysfunction first; do not wait for CRP before indicated treatment. SSC 2026.
Why the result lags behind the patient
Inflammatory cytokines, particularly IL-6, stimulate hepatic CRP synthesis. CRP therefore measures part of the host response, not the pathogen itself. The response can follow infection or sterile tissue injury. Human hepatocyte study.
CRP commonly starts rising about 6–12 hours after an acute inflammatory stimulus and peaks around 24–48 hours, but those are approximate kinetics, not a diagnostic clock. Laboratory interpretation. A low early value may precede a substantial inflammatory response. There is no symptom-duration boundary after which a normal CRP safely excludes infection. The reported onset is also an uncertain starting point: inflammation may begin before the patient notices symptoms.
Human tracer studies found a plasma half-life of about 19 hours. Concentration reflects ongoing production as well as clearance, so CRP can remain high after the patient starts improving. Conversely, a fall means less circulating CRP, not necessarily eradication of infection. Primary turnover study.
Bacterial, viral or noninfectious?
These groups overlap. In a retrospective cohort of 839 adults with CRP above 100 mg/L, infection accounted for 55.1%; rheumatologic disease, other inflammation, malignancy and drug reactions also occurred. This was a selected population with high CRP, not a validated diagnostic threshold for undifferentiated ED patients. Landry et al..
A serial slope and an estimated velocity are different measurements. A measured slope is (second CRP − first CRP) divided by elapsed hours. The 2022 estimated-velocity study instead divided admission CRP by hours since reported symptom onset. It does not measure the patient’s actual change between samples.
The 2022 cross-sectional study included 181 confirmed bacterial and 83 confirmed viral infections. Estimated velocity differed between groups, but selected confirmed cases and uncertain onset times limit application to undifferentiated illness. Its proposed thresholds should not become antibiotic or discharge rules. Original estimated-velocity study.
A separate retrospective study examined 136 adults with initial CRP ≤31.9 mg/L and a second sample within 24 hours. Bacterial cases were blood-culture positive; all patients received antibiotics between samples. A steeper measured rise was associated with bacterial infection, but exclusions and selection limit generalization. Neither study established the safety of withholding treatment. Original serial-CRP study.
Check the units before interpreting the number
1 mg/dL = 10 mg/L. A result of 10 mg/dL is 100 mg/L. Compare the result with the local laboratory’s units and reference interval, and use the same units for serial calculations.
High-sensitivity CRP measures the same protein at low concentrations; cardiovascular-risk categories are not acute-infection cutoffs. A value above the reference interval identifies an inflammatory signal, not its cause or urgency. Assay information.
When the inflammatory response is muted
Check medication exposure before treating a low or falling CRP as reassuring. In a randomized pneumonia-trial analysis, prednisone lowered CRP during treatment. This is a biological effect, not an assay correction problem. Prednisone trial analysis.
IL-6 blockade, including tocilizumab, can suppress acute-phase responses even during infection. Product labeling warns that inflammatory signs may be reduced. A concerning history, examination or deterioration still requires investigation. Tocilizumab product warning.
Do not apply a numerical “correction factor” for medication, age or baseline inflammation. Review previous values when available, but do not dismiss a new clinical change because the patient usually has an abnormal CRP.
Use the pathway, not a universal cutoff
The useful question is whether CRP changes a decision that remains uncertain after assessment. Repeating it solely because the previous value was high adds little if no plausible result would alter management.
Possible sepsis: seek infection with acute, life-threatening organ dysfunction and consider alternatives. CRP does not measure perfusion or organ failure. Clinical deterioration requires reassessment and treatment, even when CRP is low or falling. SSC 2026 explicitly advises against diagnosing or excluding sepsis with one biomarker. Guideline.
Uncertain lower respiratory tract infection: NICE NG237 applies to over-16s at first presentation. When antibiotic need remains unclear after clinical assessment, it suggests point-of-care CRP: below 20 mg/L, do not routinely offer antibiotics; 20–100 mg/L, consider a back-up prescription; above 100 mg/L, offer immediate antibiotics. These are pathway-specific management bands, not proof of bacterial pneumonia. Recommendation 1.3.4.
Apply that respiratory pathway only to its intended setting. A low CRP does not override hypoxemia, hemodynamic instability or a convincing focal infection, and those bands should not be transferred to children or undifferentiated sepsis.
Focal symptoms: persistent abdominal tenderness, focal neurologic findings or another concerning examination needs the appropriate diagnostic pathway. CRP cannot resolve equivocal imaging or replace reassessment.
A discordant result should prompt reassessment
When the examination and CRP disagree, recheck timing, units, prior values and medications. Then ask whether the suspected diagnosis still explains the patient’s physiology. A low CRP should not terminate an otherwise indicated cardiac, neurologic or surgical evaluation.
After surgery or trauma, sterile tissue injury can raise CRP. Interpret a persistent rise with the operation, elapsed time, symptoms and examination. There is no universal postoperative day or concentration that diagnoses infection.
When infection and an inflammatory disease flare are both plausible, pursue the findings that distinguish them. A change in CRP alone cannot settle that decision, particularly when immunomodulating treatment changes the response.
Three avoidable errors
Do not label a high value “bacterial” without a compatible clinical picture. Do not use an early low result to close a concerning differential. Do not repeat CRP instead of examining a patient whose vital signs or symptoms are worsening.
Monitoring and disposition
Before repeating CRP, specify the question and the action each possible trend would trigger. Clinical improvement, source control, microbiology and treatment tolerance matter more than achieving a normal number.
The ADAPT-Sepsis trial randomized 2,760 critically ill adults in 41 UK ICUs. A daily CRP-guided protocol did not reduce total antibiotic duration to day 28 versus standard care; its mortality noninferiority result was inconclusive. This does not establish a routine CRP-based antibiotic stopping rule, and it does not test ED discharge safety. Trial report.
Discharge depends on the diagnosis, stable physiology, adequate function and support, and a workable follow-up plan. Explain which worsening symptoms should prompt return. Neither an isolated normal value nor a falling trend establishes safe discharge.
At the bedside
Interpret CRP with three things beside the number: the clinical question, the time course and the patient’s physiology. Use it when it can refine a decision; when the patient and the marker disagree, reassess the patient.
Selected evidence
Vigushin et al., 1993: human CRP turnover; Castell et al., 1990: hepatic acute-phase response.
Largman-Chalamish et al., 2022: estimated CRP velocity; Bernstein et al., 2021: serial CRP in a selected low-initial-CRP cohort.
Landry et al., 2017: markedly elevated CRP cohort; Raess et al., 2021: prednisone and pneumonia biomarkers.
Dark et al., 2025: ADAPT-Sepsis randomized trial; SSC 2026 adult sepsis guideline; NICE NG237 respiratory guidance.
