Sepsis and Septic Shock Bundle

In depth

The one-page summary above is what the PDF prints. These sections go further.

Pathophysiology: dysregulated host response

In sepsis, pathogen molecules (endotoxin) and danger signals from injured cells (HMGB1, mitochondrial DNA) bind toll-like receptors, releasing TNF-α, IL-1β and IL-6, activating complement and expressing tissue factor. Thrombin generation outruns protein C, antithrombin and fibrinolysis, so microthrombi form and platelets fall. Within days many patients swing to immunoparalysis (lymphocyte apoptosis, low monocyte HLA-DR) and late secondary infections.

Vasoplegia is the core of septic shock. Inducible NO synthase raises cGMP and relaxes vascular smooth muscle; ATP-sensitive potassium channels open as ATP falls and acidosis rises, hyperpolarizing the cell and closing calcium channels. Adrenergic receptors downregulate, and vasopressin stores deplete after an early surge (Landry 1997: about 3 pg/mL in septic shock against 23 in cardiogenic shock), the rationale for fixed-dose vasopressin.

Sublingual microscopy shows stopped capillaries beside fast shunts (De Backer 2002), so a MAP of 65 does not guarantee tissue flow. The endothelial glycocalyx, a 0.5–2 µm gel of syndecan-1, heparan sulfate and hyaluronan, limits leak and cell adhesion; TNF, heparanase and hypervolemia itself (through atrial natriuretic peptide) shed it. Under the revised Starling principle most capillaries do not reabsorb fluid at the venous end, so large positive balances cause edema without restoring volume.

After filling, warm shock shows high output, low SVR and often a normal or high ScvO₂, because shunting and mitochondrial dysfunction leave oxygen unused. Reversible septic cardiomyopathy is reported in 20–60%, depending on definition. Lactate also rises from β₂-driven aerobic glycolysis and reduced hepatic clearance, so it marks stress and risk, not only oxygen debt.

Sepsis-3 definitions and SOFA

Sepsis-3 (Singer 2016) defines sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection, measured as an acute SOFA rise of 2 or more from a baseline assumed to be 0 unless prior dysfunction is known; that rise carried about 10% in-hospital mortality. Septic shock needs both a vasopressor for MAP 65 or more and lactate above 2 mmol/L despite adequate fluid, a combination with mortality above 40% (Shankar-Hari 2016). SIRS left the definition because it is nonspecific and still misses cases: 1 in 8 ICU patients with infection and new organ failure met fewer than 2 criteria (Kaukonen 2015). A revised SOFA-2 appeared in 2025; Sepsis-3 and the exams use the original.

Screening tools and their accuracy

Screening aims at sensitivity, then confirms at the bedside. qSOFA was derived to predict death, not to detect sepsis: in Seymour 2016 its AUROC for in-hospital mortality was 0.81 outside the ICU (SOFA 0.79, SIRS 0.76) but 0.66 inside it (SOFA 0.74). SSC 2026 strongly prefers NEWS, NEWS2, MEWS or SIRS over qSOFA as a single tool.

NEWS2 of 5 or more, or 3 in one parameter, prompts urgent review, and 7 or more an emergency response (RCP 2017). The Epic model alerted on 18% of all inpatients yet missed two thirds of sepsis (Wong 2021).

Hour-1 bundle element by element

Time zero is ED triage, or recognition elsewhere. Seymour 2017 (49,331 New York patients) found each hour to antibiotics raised the odds of in-hospital death by about 4%, while time to complete the fluid bolus was not associated with mortality; Kumar 2006 reported survival falling about 7.6% per hour of delay after hypotension began. Observational timing data cannot separate delay from atypical presentations, so SSC 2026 allows 3 h of rapid workup when sepsis is only possible.

Underfilled bottles are the commonest cause of false-negative cultures. Give the beta-lactam first: cefepime and meropenem can be pushed over 3–5 min where policy allows, while vancomycin runs 1–2 h. Loading doses are full regardless of renal function because sepsis expands the volume of distribution. Epinephrine, metformin, liver failure and seizures raise lactate without sepsis; lactated Ringer's does not.

Fluids: choice and volume

Saline (154 mmol/L chloride) causes hyperchloremic acidosis and renal vasoconstriction. SMART found fewer major kidney events with balanced fluid, and its 1,641 sepsis patients had lower 30-day mortality (26.3% vs 31.2%). The blinded BaSICS and PLUS trials found no mortality difference, though many patients had fluid before enrollment. Saline stays preferred in traumatic brain injury because LR is mildly hypotonic.

The 30 mL/kg figure comes from pre-randomization volumes in the early goal-directed therapy trials. Rivers 2001 (mortality 30.5% vs 46.5%) was not reproduced by ProCESS, ARISE and ProMISe (2014–2015). After the first liters, CLASSIC (ICU septic shock) and CLOVERS (ED hypotension after 1–3 L, earlier norepinephrine) each gave about 2 L less with equal safety, so the next liter is an individual decision. Without ICU support, fluid protocols raised mortality (FEAST 2011, children; Andrews 2017, Zambian adults).

Fluid responsiveness and perfusion targets

A patient is fluid responsive if stroke volume or cardiac output rises 10–15% after a bolus; only about half of unstable ICU patients are (Michard and Teboul 2002). CVP does not predict it (AUROC 0.56, Marik 2008). Give a responder fluid only for hypoperfusion and only while tolerated (no rising oxygen need, B-lines or venous congestion).

PLR shifts about 300 mL and reverses when lowered. Start semirecumbent, not supine, and judge by cardiac output, VTI or end-tidal CO₂, not blood pressure (Monnet 2016: sensitivity 85%, specificity 91%). PPV fails with spontaneous effort, arrhythmia, low tidal volume or RV failure.

ANDROMEDA-SHOCK (2019, 424 patients in septic shock) targeted capillary refill every 30 min against lactate normalization or a 20% fall every 2 h for 8 h. Mortality at 28 days was 34.9% vs 43.4% (HR 0.75, p=0.06), with less organ dysfunction at 72 h. CRT can normalize within hours while lactate stays high for non-hypoxic reasons, so chasing lactate alone risks over-resuscitation.

Technique: press a glass slide on the pad of the distal index finger until it blanches, hold 10 s, release and time color return with a stopwatch; normal is 3 s or less. Supporting signs: mottling, ScvO₂ below 70%, venous–arterial PCO₂ gap above 6 mm Hg.

Vasopressors and corticosteroids

Norepinephrine (α₁ with modest β₁) raises arterial tone and venous return; a typical start is 0.05–0.1 mcg/kg/min. Early start helps: CENSER (2019, 310 patients) controlled shock by 6 h more often (76% vs 48%). Dopamine was no better and more arrhythmogenic (SOAP II 2010: 24% vs 12%). SEPSISPAM (2014) found no mortality difference with MAP 80–85 vs 65–70, and the 65 trial (2020, about 2,600 patients aged 65 or older) found permissive 60–65 numerically better (90-day death 41.0% vs 43.8%, significant only after adjustment), the basis of the 2026 age-specific target.

Vasopressin acts on V₁ receptors and is fixed at 0.03 units/min, not titrated. VASST (2008, 778 patients on NE ≥5 mcg/min) found no overall mortality difference (35.4% vs 39.3%) but a lower rate in less severe shock (26.5% vs 35.7%). VANISH (2016) found no change in kidney failure–free days but less RRT. Watch for digital ischemia. Many protocols wean NE before vasopressin because stopping vasopressin first caused more hypotension in an observational study (Bauer 2010); practice varies.

Epinephrine is the third agent, or the first with cardiac dysfunction; expect lactate and glucose to rise. Angiotensin II (ATHOS-3, 2017, 321 patients on NE-equivalent >0.2 mcg/kg/min) raised MAP at 3 h in 69.9% vs 23.4%, with 28-day mortality 46% vs 54% (ns).

Hydrocortisone restores vascular responsiveness to catecholamines; a cosyntropin test no longer selects patients. ADRENAL (2018, 3,658 ventilated patients, 200 mg/day infusion for 7 days) found no 90-day mortality difference (27.9% vs 28.8%) but faster shock resolution and shorter ventilation. APROCCHSS (2018, 1,241 patients) gave 50 mg every 6 h plus fludrocortisone 50 mcg enterally for 7 days and lowered 90-day mortality (43.0% vs 49.1%). Tapering practice varies; monitor glucose and sodium.

Source control, antibiotics and adjuncts

Control the source within 6 h when feasible (SSC 2026 Rec 24) by the least invasive effective route: percutaneous drainage of an abscess, ERCP for cholangitis, a stent or nephrostomy for obstructed pyelonephritis, emergency debridement for necrotizing soft tissue infection. Remove a possibly infected intravascular device once other access works.

Septic patients have large volumes of distribution and sometimes augmented renal clearance: give full loading doses, dose vancomycin to an AUC of 400–600, and use prolonged beta-lactam infusions (SSC 2021; BLING III 2024, 7,031 patients, 90-day death 24.9% vs 26.8%, p=0.08). Narrow to the organism at 48–72 h and keep courses short: BALANCE (2025, 3,608 bacteremic patients) found 7 days noninferior to 14.

Procalcitonin rises within about 4–6 h of bacterial infection, peaks at 12–24 h and roughly halves daily as infection resolves. SSC 2021 recommends against using it to start antibiotics but suggests it, with clinical judgment, to stop them. In SAPS (2016, 1,575 ICU patients), stopping at PCT below 0.5 ng/mL or an 80% fall from peak shortened therapy (5 vs 7 days) and lowered 28-day mortality (20% vs 25%). Surgery, trauma, burns and pancreatitis raise it without infection.

Vitamin C rose on a before-after study (Marik 2017) and fell in randomized trials: VICTAS (2021, 501 patients) found no gain in vasopressor- and ventilator-free days, and LOVIT (2022, 872 patients, 50 mg/kg every 6 h for 96 h) found more death or persistent organ dysfunction at 28 days (44.5% vs 38.5%, RR 1.21). Other adjuncts without benefit: polymyxin B hemoperfusion (EUPHRATES 2018), intensive insulin to 81–108 mg/dL (NICE-SUGAR 2009, more deaths) and transfusion above Hb 7 g/dL (TRISS 2014).

CMS SEP-1, monitoring and documentation

SEP-1 is an all-or-none, chart-abstracted measure built on Sepsis-2: severe sepsis is suspected infection plus 2 SIRS criteria plus 1 organ dysfunction within 6 h. Organ dysfunction includes SBP <90 or MAP <65, creatinine >2.0, bilirubin >2, platelets <100,000, INR >1.5, lactate >2 or new ventilation. One missing element or time fails the case. IDSA (2021) asked CMS to limit it to septic shock; CMS kept it in Hospital Value-Based Purchasing for FY2027.

SEP-1 accepts less than 30 mL/kg when the provider documents why (heart failure, dialysis, overload). Because it keeps 65 as its MAP threshold, a patient aged 65 or older managed at 60–65 can look like untreated persistent hypotension, so the target and its reason belong in the note.

• Vitals every 15 min during resuscitation; trend MAP; arterial line leveled at the phlebostatic axis
• Hourly urine output (≥0.5 mL/kg/h), mentation, CRT and mottling; lactate every 2–4 h until normal
• Lungs, SpO₂ and oxygen need after each bolus; stop and notify for intolerance
• Peripheral NE site checked at least hourly; for extravasation stop, aspirate, phentolamine per protocol
• Document times of lactate, cultures and first antibiotic, fluid volume and weight used, each reassessment and the MAP target

Exam pearls

• Sepsis-3: infection plus SOFA rise ≥2. Septic shock: pressors for MAP ≥65 and lactate >2 after adequate fluid
• qSOFA predicts death; it is not a diagnostic or single screening tool
• Balanced crystalloid over saline (saline in TBI); no starches; albumin is not routine
• PLR with a real-time cardiac output rise ≥10% predicts fluid response; CVP does not
• Warm septic shock: high cardiac output, low SVR, normal or high ScvO₂, wide pulse pressure
• Norepinephrine, then vasopressin 0.03 units/min fixed, then epinephrine; dobutamine for low output
• Hydrocortisone 200 mg/day for a persistent pressor need (classic trigger NE ≥0.25 mcg/kg/min for 4 h)
• Capillary refill ≤3 s is an accepted resuscitation target (ANDROMEDA-SHOCK)
• Severe sepsis, 3- and 6-hour bundles and a documented reassessment mean SEP-1