PSP Shows Higher Prognostic AUC Than CRP and PCT in Sepsis

Key Takeaways

• Among adults with sepsis, septic shock, or acute infection-related illness in emergency or critical care, PSP showed significantly greater prognostic discrimination than CRP in paired comparisons.
• PSP also showed significantly greater prognostic discrimination than PCT in paired comparisons.
• Between-cohort variation was substantial in the CRP comparison, while none was detected in the PCT comparison; the latter does not establish truly homogeneous effects.
• Greater discrimination did not establish that PSP-guided decisions improve clinical outcomes, according to the investigators.

Severe infection can progress to organ failure or death, yet a biomarker measured at presentation may not reliably distinguish who will deteriorate. Pancreatic stone protein (PSP), C-reactive protein (CRP), and procalcitonin (PCT) offer potential prognostic signals, but differences between patient groups can complicate comparisons across separate studies. Patients with PSP and either comparator measured can be assessed within the same cohort rather than compared across different patient groups.

Investigators conducting the paired-biomarker meta-analysis in Intensive Care Medicine Experimental identified seven studies involving 803 unique adults and 21 paired comparisons. They searched PubMed, EMBASE, and Scopus from inception through March 13, 2026, for English-language human research. Eligible emergency-care or critically ill adults had sepsis, septic shock, or acute infection-related illness; pediatric and purely diagnostic cohorts were excluded. Eligible outcomes included mortality, organ failure, intensive care unit readmission, therapeutic escalation, and subsequent sepsis development. Baseline meant the earliest measurement at hospital or intensive care unit admission or sepsis diagnosis. PSP and CRP and/or PCT had to be measured in the same people. Investigators defined the area under the receiver operating characteristic curve (AUC) and calculated the paired difference (ΔAUC) as PSP AUC minus comparator AUC, so positive values favored PSP. They pooled the comparators separately using random-effects models and prioritized one baseline outcome per independent cohort and comparator, selecting mortality before other eligible outcomes.

Across eight independent cohorts, the principal PSP–CRP comparison yielded a ΔAUC of 0.179 (95% confidence interval [CI] 0.095–0.262). That pooled comparison drew on eligible prognostic outcomes rather than mortality alone.

Across six independent cohorts, the principal PSP–PCT comparison yielded a ΔAUC of 0.153 (95% CI 0.114–0.193). Every cohort in this principal PCT analysis reported a mortality outcome.

Both principal estimates were statistically significant (p < 0.001 in each comparison). Between-cohort heterogeneity was I² = 82% for CRP and I² = 0% for PCT. The baseline mortality-only PCT analysis was numerically identical to the principal PCT analysis, not an independent replication. Excluding the intra-abdominal infection cohort from the CRP analysis left precisely the same cohorts as its baseline mortality-only analysis, despite slightly different rounding of the estimates; those analyses likewise did not independently replicate one another. Alternative assumptions about within-cohort AUC correlation preserved the direction and approximate size of the pooled differences.

The authors noted that few independent cohorts contributed and that the CRP comparison varied across populations, endpoint definitions, sampling, and assays. No detected PCT heterogeneity with few cohorts does not prove that effects are homogeneous. Because empirical covariance between paired AUCs was unavailable, variance estimates relied on an assumed within-cohort correlation; supportive analyses retaining repeated comparisons from a cohort may have overstated precision. Comparative risk-of-bias judgments were mixed, with some high and many unclear. Protocol changes broadened eligibility to acute infection-related illness and added multi-organ failure. The resulting mix of outcomes and selected populations limits a mortality-only or universal-sepsis interpretation. AUC measures discrimination, not whether PSP-guided choices improve clinical-model classification or patient-centered outcomes.

The investigators concluded that paired comparisons favored PSP over both conventional biomarkers, while the limited independent cohorts and variation in the CRP comparison left the clinical relevance of those differences unresolved.

Clinician Questions

Does PSP prognostic evidence in sepsis include patients without sepsis at enrollment?

Yes. The pancreatic stone protein (PSP) versus C-reactive protein (CRP) meta-analysis included an intra-abdominal infection cohort whose participants did not have to have established sepsis or septic shock at enrollment; some developed sepsis during hospitalization. That cohort extended the sampled population beyond adults already diagnosed with sepsis.

How does PSP assay access compare with CRP and PCT in sepsis risk assessment?

According to the investigators, pancreatic stone protein (PSP) testing remains less accessible than C-reactive protein (CRP) and procalcitonin (PCT) in many healthcare systems. Assay availability, turnaround time, laboratory integration, and cost-effectiveness remain unresolved practical considerations for sepsis risk assessment.