What CRP level is considered mildly vs severely elevated, and how?

The short version: there is no single universal cutoff for "mildly" versus "severely" elevated CRP. The categories depend on the clinical context and the specific assay used.

The evidence here does not give a single authoritative table of mild versus severe cutoffs. What it does show is that different studies use different thresholds for different purposes. In community-acquired pneumonia, a CRP above 80 mg/L was used to distinguish bacterial from viral infection, but it performed poorly—only 52% sensitivity [1]. In the same condition, a CRP above 100 mg/L on day 1 was a strong independent predictor of 30-day mortality, with an odds ratio of 42.9 [11]. In pyogenic vertebral osteomyelitis, a CRP above 54.5 mg/L at admission predicted adverse outcomes [19]. In cirrhosis with variceal bleeding, a CRP of 13 mg/L (1.3 mg/dL) was the cutoff for high 6-week mortality in Child-Pugh class C patients [17].

What these numbers tell you is that "mild" and "severe" are not fixed labels. A CRP of 10 mg/L is mildly elevated for a healthy person but severely elevated for someone with cirrhosis. A CRP of 100 mg/L is severely elevated in almost any context. The table below shows the cutoffs used in the retrieved studies and what they predicted.

The evidence does not define a single "mild" versus "severe" boundary because CRP is a continuous marker that must be interpreted in context. A CRP above 10 mg/L is abnormal. Above 50 mg/L suggests significant inflammation. Above 100 mg/L is consistently associated with worse outcomes across multiple conditions. But the exact cutoff that matters depends on the disease, the patient, and the outcome being predicted.

My call: there is no universal mild-versus-severe cutoff. In practice, CRP below 10 mg/L is generally mild elevation, 10-100 mg/L is moderate, and above 100 mg/L is severe, but these are rough guides, not fixed rules. Confidence: moderate, because the evidence supports context-dependent cutoffs but does not provide a single authoritative classification.