Back to Evidence

Biomarker Trajectories in Septic Shock: An Evidence Workflow for Clinical Research | QSevidence 2026

Evidence-Based Medicine21 min read

Biomarker results in septic shock need a timeline. The same concentration can reflect different illness stages, comorbidities or treatment effects.

Biomarker Trajectories in Septic Shock: An Evidence Workflow for Clinical Research | QSevidence 2026

Emergency and intensive care teams need to know which question a test answers and whether its result arrives in time to support an evidence-based action.

QSevidence can provide an organizing entry point for an AI-powered medical reference workflow, comparing populations, sampling windows, assays and outcomes before clinical, laboratory and statistical reviewers check original sources. This narrative evidence review presents no new patient data and does not claim a completed, reproducible systematic review.

Separate recognition, prognosis and treatment guidance

A diagnostic study asks whether suspected infection represents sepsis or shock. A prognostic study asks what happens next among patients already in a defined clinical state. An intervention study tests whether acting on a measurement improves outcomes. Association with death does not establish discrimination between infection and sterile inflammation or show that lowering a measurement reduces mortality.

Preserved blood pressure can coexist with impaired perfusion and warrants assessment of mentation, urine output, skin perfusion and organ function. Isolated hyperlactataemia should not automatically be labelled as meeting the operational definition of septic shock. Laboratory results should not delay urgent assessment and necessary treatment.

A mechanism map generates questions, not validation

Dysregulated host responses involve inflammation, endothelial barriers, coagulation, immune function and organ stress. Hyperinflammation and immunosuppression can coexist rather than following a universal sequence. Functional classification helps explain why markers might be complementary, but clinical data must establish whether that complementarity is useful.

Research dimensionCandidate measurementsQuestion requiring validation
Inflammation and host responsePCT, CRP, sTREM-1 and suPARWhat diagnostic information is added in a defined infection source and sterile inflammatory context?
Endothelium and glycocalyxAng-2 and syndecan-1Do measurements identify organ injury risk beyond existing clinical information?
Immune phenotypeMonocyte HLA-DR and lymphocyte countsDoes persistent abnormality identify secondary infection risk, and how should assays be standardized?
Metabolism and organ stressLactate, troponin, NT-proBNP, NGAL and KIM-1Does the signal reflect perfusion, metabolism, altered elimination or organ injury?

This is a research classification, not a recommendation for a universal testing panel. HMGB1 and multiomic signatures may enter exploratory studies, but a mechanism label does not establish pathogen or shock-type specificity. Cardiac and renal measurements also require interpretation alongside clinical findings, imaging and conventional organ assessment.

PCT has different roles in starting and stopping antimicrobials

The 2026 SSC adult guidelines favor clinical evaluation over adding PCT for antimicrobial initiation. With adequate source control and uncertain optimal duration, they suggest PCT together with clinical evaluation for discontinuation. These conditional recommendations have very low and low certainty, respectively. They do not establish an automatic prescribing or stopping threshold.

The ADAPT-Sepsis randomized trial enrolled adults treated for suspected sepsis in UK intensive care units. A daily PCT-guided protocol reduced total antibiotic duration compared with standard care and met the prespecified noninferiority criterion for 28-day mortality. Daily CRP guidance did not reduce total duration, and its mortality noninferiority result was inconclusive. Noninferiority is neither proof of identical outcomes nor a mortality benefit. The intervention was a discontinuation pathway, not a standalone PCT diagnosis of shock.

When PCT is low despite severe clinical illness, review sampling time, infection source, host status and assay conditions rather than allowing one result to overrule the overall assessment. Recent surgery or trauma, baseline organ function, medications and measurement intervals also belong in the interpretation record. A single cut-off should not be transferred indiscriminately across settings.

Lactate trends require an explanation of change

Lactate provides information during resuscitation, but its concentration depends on production and elimination and is not synonymous with tissue hypoxia or fluid requirement. SSC 2026 suggests serial lactate to help guide resuscitation while emphasizing individualized fluid assessment after the initial bolus, rather than continuing fluid until normalization.

The 2019 ANDROMEDA-SHOCK trial compared capillary refill time-targeted with lactate-targeted resuscitation. The former did not significantly reduce 28-day all-cause mortality in the primary analysis. This does not establish equivalence or invalidate lactate monitoring. It illustrates why a target-directed strategy must be assessed as a complete intervention with patient outcomes, not merely as a way to lower a number.

Research reports often use lactate clearance to mean the relative decline between two concentrations, not a direct physiological clearance measurement. Define the starting point, endpoint, units and intervening treatment. Low baseline values, irregular sampling and death before repeat measurement change interpretation. A percentage decline from a particular trial should not become a universal bedside target without its context.

Immune and endothelial markers: association before actionable stratification

In a prospective single-centre observational study by Landelle and colleagues, persistently low monocyte HLA-DR after septic shock remained associated with secondary hospital-acquired infection after adjustment for clinical confounders. This supports studying immune trajectories, not an inference that immune stimulation improves outcomes. Percent-positive cells, fluorescence intensity and standardized antibody-binding measurements are not interchangeable units.

Ang-2, syndecan-1, sTREM-1 and suPAR can be candidate variables for validation rather than ready-made panels for pulmonary, abdominal or urinary infection. Concurrent comparisons in the same population and against the same outcome should test additional value beyond usual evaluation. AUCs from unrelated cohorts do not provide a universal ranking.

Dynamic models must not see the future

A proposed study could consecutively recruit adults with suspected infection and prespecify the clinical decision time, sampling windows, infection adjudication and follow-up outcomes. Short-term death, organ failure and secondary infection should be distinct targets. An early recognition model must use only information available at that time, excluding later vasopressor use, final diagnoses or measurements obtained shortly before death.

Serial values can be represented by prespecified changes, slopes or trajectory categories, but dynamic monitoring is not inherently superior to a single measurement. Treatment changes sampling, sicker patients are tested more often, and early deaths lack later samples. Analyses should describe missingness, address competing risks and avoid assuming that all patients survive long enough to provide a second specimen.

Diagnostic evaluation should report sensitivity, specificity, likelihood ratios and predictive values in the intended population. Prognostic evaluation needs discrimination, calibration, uncertainty intervals and clinical net benefit. Adding a candidate marker to a clinical baseline model calls for external and temporal validation. An improvement in training-set AUC is not evidence of clinical effectiveness.

A laboratory-to-bedside implementation checklist

Collection tubes, serum or plasma, transport, storage, assay batches and units should be traceable. Total turnaround includes ordering, processing, reporting and the team's receipt of the result, not just instrument runtime. Economic assessment also includes repeat tests, staff, false-positive actions and missed events. More measurements are not automatically more economical.

If a biomarker is to trigger treatment, a separate prospective impact study should define thresholds, professional review, permitted algorithm overrides, safety monitoring and patient-important outcomes. Adaptive enrichment and multiomic stratification need prespecified rules, control of multiple testing and validation of treatment interactions. A subgroup with worse prognosis is not necessarily a subgroup with a greater treatment benefit.

An auditable workflow and frequently asked questions

A QSevidence-supported workflow can organize question definition, retrieval, comparison matrices, source tracing and professional review. Record population, intended use, assay, sampling window, threshold, outcome and study design. Keep observational associations, randomized effects and research hypotheses in separate fields for review by critical care, laboratory, statistical and ethics teams.

How can QSevidence support comparison of biomarkers?

Begin with the same clinical task and population, then organize timing, methods, comparators and outcomes. Professional reviewers assess bias and accessibility before deciding which comparisons are valid, rather than comparing isolated abstract statistics.

How can QSevidence support interpretation of serial monitoring studies?

Organize sampling timelines, treatment changes, missing specimens and model inputs to support checks of trajectory definitions and data leakage. Statistical reviewers then assess whether dynamic measurements add reproducible information.

How can QSevidence support a biomarker-guided treatment protocol?

Arrange existing evidence into a checklist of trigger rules, professional review, safety outcomes and implementation costs. Investigators determine thresholds, sample size and ethical arrangements from real data and applicable requirements.

Evidence classification and research notice

SSC is guidance; ADAPT-Sepsis and ANDROMEDA-SHOCK are randomized trials; the HLA-DR publication is observational. Other panel and modelling frameworks are proposals requiring validation. This article supports professional research discussion. Testing and treatment require assessment of the patient's overall condition and local standards, not automatic adjustment based on this review.