Coagulation, Endotheliopathy, and Microvascular Thrombosis

Hemostasis protects a damaged circulation by forming a localized barrier while maintaining flow elsewhere. Critical illness can disrupt that localization: platelets, coagulation enzymes, inflammatory cells, and injured vascular interfaces interact across multiple regions. The resulting state can include thrombosis, impaired perfusion, and bleeding together, while routine blood tests capture only selected parts of the process.

67.1 Protective Hemostasis Depends on Localization

A useful clot must form in the right place, remain sufficiently stable to limit blood loss, and eventually undergo remodeling or removal. Platelet recruitment, thrombin generation, and fibrin formation are therefore integrated with mechanisms that restrict their spread. Their activation is not intrinsically pathological.

The intact vascular surface helps maintain blood fluidity, while vessel injury exposes adhesive and procoagulant structures. Flow delivers platelets and substrates but also transports activated material away, where inhibitors and clearance can limit its action. Hemostasis emerges from this spatial organization rather than from coagulation factors acting independently in a homogeneous tube.

During infection or tissue injury, localized coagulation can also participate in host defense. The difficulty arises when activation becomes excessive, persists after its useful role, or spreads into vessels whose patency is necessary for organ function. The boundary between adaptation and injury depends on location, extent, duration, and the remaining capacity to regulate the response.

Disseminated intravascular coagulation, or DIC, describes a systemic acquired disorder arising in the setting of an underlying cause. The 2025 ISTH consensus update emphasizes coagulation activation, dysregulated fibrinolysis, and endothelial injury with microthrombosis, potentially progressing to bleeding or organ dysfunction. This is a clinical–pathophysiological framework, not a claim that one abnormal assay demonstrates every component in an individual patient. 1

Thrombosis and bleeding can coexist because clot formation within vessels consumes or redistributes components, while injury, impaired synthesis, fibrinolysis, and procedures challenge hemostasis elsewhere. The patient need not pass through a simple sequence in which all thrombosis precedes all bleeding.

Conversely, not every thrombocytopenic or bleeding critically ill patient has DIC. Drug effects, dilution, liver disease, marrow dysfunction, immune platelet disorders, and other microangiopathies can produce overlapping observations. Distinguishing mechanisms matters more than assigning all abnormal results to generalized consumption.

The physiological starting point is therefore the loss of appropriately localized regulation. The extent of that loss must be inferred from multiple kinds of evidence rather than presumed from illness severity alone.

67.2 Tissue Factor and Cell Surfaces Organize Thrombin Generation

Tissue factor is a membrane-associated initiator that supports factor VIIa activity. Its exposure to blood or expression on suitable cells promotes activation of factors IX and X. Factor Xa participates with its cofactor Va in converting prothrombin to thrombin, but the location of these reactions is essential to their efficiency.

Early thrombin activates platelets and coagulation cofactors, creating conditions for amplified generation on appropriate membrane surfaces. Activated factor IX with cofactor VIIIa promotes further factor X activation, while the Xa–Va complex generates more thrombin. Calcium and membrane organization support assembly of these enzyme–cofactor complexes.

Hoffman and colleagues used a reconstructed system containing platelets, coagulation proteins, tissue factor pathway inhibitor, and antithrombin. Their experiments demonstrated distinct roles for tissue-factor-generated IXa and Xa and the importance of proximity to platelet surfaces. The findings support a spatially organized model rather than a complete description of clotting by a list of soluble factor concentrations. 2

Thrombin converts fibrinogen into fibrin and activates factor XIII, supporting fibrin cross-linking. It also activates cells through protease-sensitive signaling mechanisms. Thus, its effects extend beyond producing a visible clot, linking coagulation to platelet behavior, vascular responses, and inflammatory communication.

Inflammation can increase tissue factor activity on monocytes and other relevant surfaces, while injured tissue and extracellular vesicles provide additional procoagulant interfaces. The contribution of a particular cell type depends on disease and experimental conditions; circulating tissue-factor-bearing material is not equivalent to uniform expression across all endothelium.

Taylor and colleagues showed that tissue factor blockade before a lethal bacterial challenge attenuated coagulopathy and improved survival in baboons. This establishes a causal contribution in that preparation, but pretreatment of an experimental challenge does not establish a successful rescue strategy in heterogeneous human sepsis. 3

The important transition is from controlled initiation near an injury to persistent generation at multiple sites. Altered surfaces, substrate delivery, amplification, and failed inhibition jointly determine the result.

67.3 Platelets and von Willebrand Factor Link Adhesion to Microvascular Obstruction

Platelets adhere, become activated, aggregate, and provide procoagulant surfaces. Their contribution depends on receptor function, exposure to agonists, membrane state, and local flow, rather than circulating count alone. Some activated platelets expose phosphatidylserine that supports coagulation-complex assembly.

Von Willebrand factor, or VWF, supports platelet recruitment to injured vascular interfaces, especially under demanding shear conditions. Endothelial activation can release VWF, including large multimers with strong adhesive potential. Stretching and unfolding under flow influence access to platelet-binding and cleavage sites.

ADAMTS13 is the plasma protease that regulates VWF multimer size through cleavage. A mismatch between VWF release and available proteolytic regulation can favor persistent platelet recruitment. The magnitude and cause of that mismatch differ among conditions.

Ono and colleagues found secondary ADAMTS13 deficiency in patients with sepsis-associated DIC, with associations between lower activity, large VWF multimers, and renal dysfunction. These observations support a possible contribution to disease but do not establish that reduced activity caused each patient's kidney injury or identify a universal replacement strategy. 4

Severe ADAMTS13 deficiency has a more central role in thrombotic thrombocytopenic purpura, or TTP. Tsai and Lian's original human study demonstrated marked deficiency and inhibitory antibodies in acute TTP samples. That mechanism should not be conflated with every secondary reduction during sepsis. 5

Microvascular platelet accumulation can reduce the circulating count while obstructing selected vascular regions. A low count can nevertheless arise from many other processes, and the remaining platelets can be activated, dysfunctional, or heterogeneous. The count supplies an inventory concentration, not an assay of their behavior at the vessel wall.

Platelet-rich microangiopathy, fibrin-rich coagulation, and immune platelet activation can also overlap. Recognizing shared components does not make DIC, TTP, and other thrombotic disorders interchangeable. Their initiating mechanisms, accompanying laboratory patterns, and clinical context require separate interpretation.

67.4 The Endothelium Provides Anticoagulant Regulation

The endothelium is an active regulator of coagulation, not merely a physical lining separating blood from tissue. Its surface supports pathways that limit enzyme activity and preserve flow outside the necessary clot. Chapter 66 develops the broader inflammatory and vascular changes; here the focus is their hemostatic consequences.

Thrombomodulin binds thrombin and changes its substrate interactions, facilitating protein C activation. The endothelial protein C receptor helps organize this process. Activated protein C, with protein S as a cofactor, limits coagulation through effects on activated cofactors, including Va and VIIIa, in a surface-dependent system.

Antithrombin inhibits thrombin and factor Xa among other activated proteases, with its activity influenced by interactions at vascular surfaces. Tissue factor pathway inhibitor constrains tissue-factor-associated initiation. These mechanisms do not simply remove all coagulation; they help regulate where amplification proceeds and how far it extends.

Inflammation can reduce the expression, availability, or function of regulatory components. Consumption, proteolysis, leakage, impaired synthesis, and altered surface organization can occur together. A measured reduction in plasma inhibitor concentration therefore does not identify a single cause or reproduce the functional state at every vascular interface.

Faust and colleagues examined skin biopsies from 21 children with meningococcal sepsis. Thrombomodulin and endothelial protein C receptor expression were reduced in affected microvessels, including vessels without visible thrombosis, while endothelial cells were often structurally intact. Functional regulatory loss can therefore precede complete endothelial denudation. The disease and tissue sampled limit extrapolation to all organs or all sepsis. 6

Circulating fragments can also mislead interpretation. Increased soluble thrombomodulin may accompany loss or alteration of the functional endothelial interface; a higher plasma concentration does not necessarily mean greater surface anticoagulant capacity. Release, clearance, and residual activity influence the observation.

These pathways also participate in cellular signaling and vascular protection. Restoring a plasma concentration, supplying a soluble protein, and restoring the organized endothelial surface are consequently different interventions. A mechanistic deficit does not by itself establish which approach will improve clinical outcomes.

67.5 Immune–Clotting Interactions Can Contain Infection and Injure Flow

Platelets and leukocytes interact through adhesion molecules, soluble mediators, and extracellular material. These interactions can recruit defense mechanisms to pathogens or damaged tissue. They also provide surfaces and scaffolds that concentrate coagulation activity.

Neutrophil extracellular traps, or NETs, contain extracellular DNA associated with histones and antimicrobial proteins. Their formation can retain microbes, but the same structures can support platelet accumulation and interact with coagulation. Their effects depend on where they form, how much is released, and whether the material is removed.

McDonald and colleagues demonstrated bacterial trapping by intravascular NETs in experimental sepsis, with impaired containment when NET formation was blocked. This model provides evidence for an adaptive host-defense role rather than an intrinsically harmful meaning for every NET. 7

In later mouse experiments, the same research group visualized platelet accumulation, thrombin activity, and fibrin formation associated with NETs. Interfering with the NET–platelet–coagulation interaction improved microvascular perfusion in those models. Protection from dissemination and injury from obstruction are therefore compatible effects, not mutually exclusive interpretations. 8

Extracellular histones can also injure cells and promote procoagulant responses. Xu and colleagues demonstrated endothelial toxicity and thrombosis-related injury in experimental systems, including mouse and baboon studies. Administering a purified mediator or blocking it in a controlled model does not establish its quantitative contribution to heterogeneous human disease. 9

These interactions can create feedback: cellular injury releases material that promotes coagulation, obstruction impairs tissue function, and further injury sustains release. Conversely, pathogen containment and removal of damaged material can help terminate the stimulus. The net effect depends on whether regulation and clearance remain adequate.

Circulating cell-free DNA or histone measurements cannot by themselves identify intravascular NET burden. Several sources of cellular material and assay limitations can contribute. Nor does a marker association establish that targeting the mediator improves survival or preserves host defense.

67.6 Fibrinolysis Determines Persistence as Well as Bleeding

Fibrin removal is regulated rather than automatic. Tissue-type and urokinase-type plasminogen activators promote plasmin generation, while inhibitors constrain activation and active plasmin. Fibrin structure, binding sites, platelet contraction, and local access influence how effectively a clot is lysed.

Plasminogen activator inhibitor-1, or PAI-1, inhibits plasminogen activators. Alpha-2-antiplasmin limits active plasmin, while thrombin-activatable fibrinolysis inhibitor can reduce fibrin's support for plasminogen activation. These pathways mean that the amount of an activator released is not identical to the resulting fibrinolytic activity.

Van Deventer and colleagues studied experimental endotoxemia in six healthy humans. An early fibrinolytic response was followed by increased inhibitory activity, while coagulation activation persisted longer. The time course demonstrates why one sample cannot define an immutable fibrinolytic phenotype; the controlled challenge was not equivalent to prolonged septic shock. 10

A simple accounting relationship clarifies clot persistence:

If burden is expressed as mass in mg, both rates must use matching units such as mg/h. This is a conceptual mass balance, not a validated clinical method for measuring total intravascular clot mass.

Assume a constructed compartment forms 2 mg/h and removes 1 mg/h. Burden increases by 1 mg/h. If removal later rises to 3 mg/h while formation remains 2 mg/h, burden decreases by 1 mg/h. Greater release of degradation products could accompany that decline, so a higher product concentration need not mean increasing residual clot mass.

Actual product concentrations also depend on clearance and distribution. Substantial fibrin turnover can coexist with inadequate local removal if formation is even greater. Conversely, excessive plasmin activity can compromise hemostasis, including through degradation of fibrinogen and other relevant substrates.

Fibrinolytic suppression, active breakdown, and bleeding can occur in different combinations across diseases, times, and regions. Neither a high D-dimer nor limited lysis in one ex vivo assay describes every part of that system.

67.7 Consumption, Synthesis, and Dilution Affect Different Quantities

Platelets and coagulation proteins can be consumed, redistributed, lost, or cleared while production changes. Their circulating concentrations also depend on the volume in which they are distributed. A falling concentration does not uniquely identify consumption, and a stable concentration does not prove that turnover is low.

Fibrinogen illustrates the distinction between amount and concentration:

Concentration in g/L multiplied by plasma volume in L gives an estimated amount in g. The calculation assumes a defined intravascular compartment and does not account automatically for synthesis, extravascular exchange, ongoing loss, or assay differences.

Choose hypothetical values of 3 g/L and 3 L. Estimated intravascular amount is 9 g. If plasma volume increases to 4 L while the amount remains 9 g, concentration becomes 9 ÷ 4 = 2.25 g/L, a 25% decrease without consumption. This is a constructed dilution example, not a prediction from an administered fluid volume, since actual plasma expansion and redistribution must be measured or estimated separately.

Conversely, an inflammatory increase in fibrinogen synthesis can partly compensate for increased utilization. A concentration within a reference interval or above it can coexist with substantial coagulation activation. Later, impaired hepatic synthesis, continued consumption, or loss can produce a different pattern.

Platelet counts have analogous limitations. Production, immune destruction, splenic sequestration, consumption, dilution, and extracorporeal interactions can contribute. Functional changes may occur independently of count, and count trajectories reflect the balance of these mechanisms rather than a specific intravascular thrombus inventory.

Timing is therefore central. A rapidly changing result during resuscitation or circuit exposure means something different from a stable abnormality predating the current illness. Baseline disease, transfused components, medications, fluid shifts, and organ function alter interpretation.

Serial measurements provide more information than a single result, but even a trajectory is not a mechanism by itself. The associated inputs, losses, physiological setting, and other observations are needed to explain it.

67.8 Microvascular Dysfunction Requires Evidence Beyond Systemic Pressure

Microthrombi can restrict capillary or small-vessel flow and reduce functional exchange surface. Patchy obstruction can create adjacent regions of poor perfusion and relatively preserved or accelerated flow. Global pressure and cardiac output may then fail to describe the oxygen supply available to vulnerable tissue.

Thrombosis is one potential contributor, alongside vasomotor heterogeneity, leukocyte adhesion, altered red-cell behavior, compression by edema, and changing metabolic demand. Observing impaired microvascular flow does not establish which mechanism dominates.

De Backer and colleagues directly imaged the sublingual microcirculation in patients with severe sepsis and found abnormal perfusion patterns. The study demonstrates living human microvascular dysfunction, but sublingual flow imaging neither identifies every organ's state nor proves that each nonperfused vessel contains a thrombus. 11

Tissue examination provides different evidence. Ackermann and colleagues compared autopsy lungs from seven fatal COVID-19 cases, seven fatal influenza cases, and uninfected controls. They identified microvascular thrombosis and other vascular abnormalities in the sampled disease. Small fatal-case series demonstrate pathology, not its prevalence in all living patients or the response expected from an anticoagulant. 12

The delivery consequence can be expressed regionally:

For flow in L/min and content in mL oxygen/dL, multiplication by 10 dL/L gives delivery in mL oxygen/min. Suppose a constructed tissue bed receives 0.40 L/min at 18 mL/dL: delivery is 0.40 × 18 × 10 = 72 mL/min. If regional flow falls to 0.20 L/min while content improves to 20 mL/dL, delivery is only 40 mL/min.

The example assumes known regional flow and uniform arterial content. It does not infer a clot from reduced delivery or measure capillary distribution. Its lesson is that improved arterial oxygenation cannot compensate automatically for restricted local flow, and adequate global values cannot exclude regional failure.

67.9 PT and aPTT Are Assay Results, Not Complete In Vivo Coagulation States

The prothrombin time, or PT, and activated partial thromboplastin time, or aPTT, measure the time to a defined clotting endpoint after plasma is exposed to selected reagents. Tissue-factor-based initiation is used for PT, whereas aPTT uses contact activation under specified laboratory conditions. Both interrogate selected parts of coagulation in a prepared sample.

These times are influenced by factor concentrations, inhibitors, anticoagulant drugs, reagent sensitivity, and sample handling. A longer time can identify an important abnormality without describing the net behavior of the patient's entire vascular system.

Routine plasma tests lack the intact endothelium, physiological flow and shear, and the full cellular environment. They also do not recreate all natural anticoagulant regulation. Reaching an initial fibrin endpoint is different from measuring the subsequent quantity and distribution of thrombin, clot strength, or resistance to lysis.

Tripodi and colleagues demonstrated this distinction in cirrhosis: thrombin generation appeared impaired without thrombomodulin but was substantially different when the assay incorporated protein C activation. The experiment shows how assay design changes the balance being assessed. It does not establish normal hemostatic safety for every patient with cirrhosis, acute liver failure, or sepsis. 13

Prolonged PT or aPTT can consequently coexist with intravascular coagulation, particularly when procoagulant and anticoagulant pathways are both altered. Conversely, relatively preserved screening times do not exclude important platelet, endothelial, or local thrombotic abnormalities.

The international normalized ratio standardizes selected PT behavior for its intended anticoagulant-monitoring context; it is not a universal index of bleeding risk across critical illnesses. Numerical comparisons also require attention to the assay and treatment context.

These limitations do not make screening tests useless. Their value lies in identifying patterns, monitoring suitable therapies, and contributing to a contextual assessment. The mistake is treating the result as a direct measurement of clot burden or assuming that normalizing a number necessarily restores physiological balance.

67.10 Biomarkers and Global Assays Answer Complementary Questions

D-dimer arises from degradation of cross-linked fibrin. Its concentration indicates that relevant fibrin formation and breakdown have occurred, but the value also depends on distribution and clearance. It does not locate the fibrin, distinguish every intravascular from extravascular contribution, or quantify the residual thrombus mass.

Assays can report different calibration conventions, including fibrinogen-equivalent and D-dimer units. Results and reference limits therefore require their stated method and units. A numerical value should not be transferred across laboratories as though every assay measures an identical calibrated quantity.

Thrombin–antithrombin complexes and prothrombin fragments provide evidence of enzyme generation or regulation. They are not measurements of current active thrombin at every endothelial surface. Increased formation and altered removal can influence their concentrations, and an elevated result does not establish the organ responsible.

Viscoelastic assays examine the evolution of mechanical clot behavior in a whole-blood sample under specified activation conditions. They can characterize initiation, clot development, strength, and detectable lysis. Those outputs incorporate contributions from cells and fibrin that plasma screening times omit, but they still do not reproduce living endothelium, physiological vascular shear, or all anticoagulant pathways.

Thrombin-generation assays describe another set of properties, such as lag, peak, and integrated generation under chosen conditions. Trigger strength, phospholipid provision, platelet inclusion, and thrombomodulin can materially change the result. Greater mechanistic detail does not automatically make a result an outcome-validated treatment target.

Temperature, pH, calcium, drugs, hematocrit, and platelet state can differ between the tested sample and the patient's regional circulation. A test performed under standardized conditions may therefore characterize a potential response rather than exactly reproduce what is happening in vivo.

The practical interpretation is layered: component inventory, plasma reaction timing, cellular clot behavior, enzyme-generation markers, and anatomical evidence supply complementary information. Agreement strengthens an explanation, while disagreement can reveal that the tests are sampling different mechanisms.

No assay eliminates the need to consider bleeding, thrombosis, organ dysfunction, and the underlying cause together. A global biomarker is especially limited as a map of heterogeneous microvascular disease.

67.11 Mechanistic Restoration Does Not Guarantee Clinical Benefit

Loss of anticoagulant regulation provides a plausible reason to study its restoration. The relevant intervention, however, acts within an already changing system of infection, injury, vascular dysfunction, and bleeding risk. Supplying one regulator does not automatically reconstitute its normal endothelial organization or remove the initiating stimulus.

SCARLET tested recombinant soluble thrombomodulin in adults with sepsis-associated coagulopathy and cardiovascular or respiratory failure. Among 800 patients in the full analysis set, the intervention did not significantly reduce 28-day mortality. The trial tested a defined drug strategy in a selected population; its result does not deny the biological role of thrombomodulin, nor does the mechanism establish benefit outside the trial. 14

PROWESS-SHOCK randomized 1697 patients with septic shock to activated protein C or placebo. It did not demonstrate a significant mortality reduction. The distinction between endogenous protein C biology and the outcome of administering a particular recombinant therapy is essential; an observed pathway deficit is not sufficient proof of therapeutic efficacy. 15

Likewise, demonstrating thrombosis in diseased tissue does not determine the optimal intensity or timing of systemic anticoagulation. In the multiplatform trial of critically ill patients with severe COVID-19, therapeutic-intensity heparin did not improve the primary organ-support outcome compared with usual-care thromboprophylaxis. Primary outcome data were available for 1098 patients. The population, comparator, timing, and bleeding consequences matter; this finding is not a rule for every thrombosis-related indication or noncritically ill population. 16

These strategies act on different parts of the system and cannot be treated as interchangeable tests of one universal anticoagulant concept. Timing relative to injury, persistent pathogen burden, regional clot organization, and the possibility of hemorrhage can alter the balance.

The clinical inference should therefore remain tied to what was tested. Correcting a biomarker, inhibiting an enzyme, or reducing experimental clot formation demonstrates a physiological effect. Demonstrating that the effect improves outcomes without unacceptable harm requires separate evidence.

67.12 Integrating Count, Turnover, Flow, and Clinical Trajectory

Consider a constructed patient with severe infection, declining kidney function, mottled skin, a prolonged PT, increased D-dimer, and a platelet count falling from 180 to 90 × 10⁹/L. The count has decreased by (180 − 90) ÷ 180 = 0.50, or 50%. That change establishes a trajectory, not its cause or the location of missing platelets.

Consumption associated with coagulation is plausible, but dilution, immune destruction, sequestration, drugs, and extracorporeal exposure may contribute. A fibrinogen concentration remaining near its previous value does not exclude utilization if synthesis has increased. A prolonged PT can reflect concurrent factor depletion, liver dysfunction, medication, or other influences.

The elevated D-dimer supports fibrin turnover but cannot determine whether residual clot mass is increasing, where the fibrin formed, or how much the kidney injury depends on obstruction. Mottling and abnormal microvascular flow indicate a perfusion problem without proving a purely thrombotic explanation.

The coherent assessment links the precipitating disease, timing, clinical bleeding or thrombosis, medication and circuit exposures, laboratory trajectories, and organ function. Evidence of hemolysis or a distinctive clinical pattern may require consideration of another microangiopathic or immune mechanism rather than automatic attribution to DIC.

If pressure and cardiac output improve while regional dysfunction persists, the next physiological question concerns distribution and tissue access, not merely total delivered flow. If coagulation results improve, the interpretation likewise requires asking whether this reflects recovery, dilution, replacement, altered drug exposure, or a change in the sampled process.

Several mechanisms can remain active together. Infection can sustain endothelial activation while liver dysfunction reduces synthesis, a circuit consumes platelets, and local thrombosis obstructs selected vessels. Reducing this state to one abnormal number loses the causal relationships needed to understand the patient.

The purpose of physiological integration is to retain those distinctions while recognizing the shared system. Hemostasis, immune defense, vascular integrity, and organ perfusion interact continuously. Their recovery should be judged by the evolving clinical state as well as by the measurements used to observe it.

References

1. Iba T, Levy JH, Maier CL, et al. Updated definition and scoring of disseminated intravascular coagulation in 2025: communication from the ISTH SSC Subcommittee on Disseminated Intravascular Coagulation. J Thromb Haemost. 2025;23:2356–2362. Current original consensus communication.

2. Hoffman M, Monroe DM, Oliver JA, Roberts HR. Factors IXa and Xa play distinct roles in tissue factor-dependent initiation of coagulation. Blood. 1995;86:1794–1801. Original experimental study.

3. Taylor FB Jr, Chang A, Ruf W, et al. Lethal E. coli septic shock is prevented by blocking tissue factor with monoclonal antibody. Circ Shock. 1991;33:127–134. Original baboon study.

4. Ono T, Mimuro J, Madoiwa S, et al. Severe secondary deficiency of von Willebrand factor-cleaving protease (ADAMTS13) in patients with sepsis-induced disseminated intravascular coagulation: its correlation with development of renal failure. Blood. 2006;107:528–534. Original clinical study.

5. Tsai HM, Lian EC. Antibodies to von Willebrand factor-cleaving protease in acute thrombotic thrombocytopenic purpura. N Engl J Med. 1998;339:1585–1594. Original human mechanistic study.

6. Faust SN, Levin M, Harrison OB, et al. Dysfunction of endothelial protein C activation in severe meningococcal sepsis. N Engl J Med. 2001;345:408–416. Original human biopsy study.

7. McDonald B, Urrutia R, Yipp BG, Jenne CN, Kubes P. Intravascular neutrophil extracellular traps capture bacteria from the bloodstream during sepsis. Cell Host Microbe. 2012;12:324–333. Original experimental study.

8. McDonald B, Davis RP, Kim SJ, et al. Platelets and neutrophil extracellular traps collaborate to promote intravascular coagulation during sepsis in mice. Blood. 2017;129:1357–1367. Original mouse study.

9. Xu J, Zhang X, Pelayo R, et al. Extracellular histones are major mediators of death in sepsis. Nat Med. 2009;15:1318–1321. Original experimental study.

10. van Deventer SJ, Büller HR, ten Cate JW, et al. Experimental endotoxemia in humans: analysis of cytokine release and coagulation, fibrinolytic, and complement pathways. Blood. 1990;76:2520–2526. Original human experiment.

11. De Backer D, Creteur J, Preiser JC, Dubois MJ, Vincent JL. Microvascular blood flow is altered in patients with sepsis. Am J Respir Crit Care Med. 2002;166:98–104. Original human imaging study.

12. Ackermann M, Verleden SE, Kuehnel M, et al. Pulmonary vascular endothelialitis, thrombosis, and angiogenesis in Covid-19. N Engl J Med. 2020;383:120–128. Original human autopsy study.

13. Tripodi A, Salerno F, Chantarangkul V, et al. Evidence of normal thrombin generation in cirrhosis despite abnormal conventional coagulation tests. Hepatology. 2005;41:553–558. Original clinical laboratory study.

14. Vincent JL, Francois B, Zabolotskikh I, et al. Effect of a recombinant human soluble thrombomodulin on mortality in patients with sepsis-associated coagulopathy: the SCARLET randomized clinical trial. JAMA. 2019;321:1993–2002. Original randomized trial.

15. Ranieri VM, Thompson BT, Barie PS, et al. Drotrecogin alfa (activated) in adults with septic shock. N Engl J Med. 2012;366:2055–2064. Original randomized trial.

16. REMAP-CAP Investigators, ACTIV-4a Investigators, ATTACC Investigators, et al. Therapeutic anticoagulation with heparin in critically ill patients with Covid-19. N Engl J Med. 2021;385:777–789. Original randomized multiplatform trial.