Sepsis is a high mortality syndrome characterized by organ dysfunction due to a severe and dysregulated acute inflammatory response to infection. Research into therapies for this syndrome has historically ended in failure, which has largely been attributed to the elevated levels of subject heterogeneity. What may have been previously attributed to variability in sepsis may be due to mechanistic differences between patients. Endotypes are distinct subtypes of disease, where underlying causes such as mechanistic or pathway related differences manifest into phenotypes of disease. The lack of mechanistic understanding of immune mediator dynamics and the responses they trigger necessitates a mathematical modeling approach to analyze its complexities. A transfer function model is proposed to describe and cluster the dynamics of key inflammatory mediators. Five sepsis endotypes were discovered and revealed motifs of overwhelming inflammation, various levels of immunosuppression, sustained inflammation, and immunodeficiency. An accurate clinical tool was proposed to classify subjects into endotypes using six-hour trajectories of clinical data. A physiological ordinary differential equation model of sepsis is proposed that characterizes the interactions of inflammatory signaling molecules, neutrophils, and macrophages across the bone, blood, and tissue compartments of the body. This model used to generate individual subject fits against human sepsis data. Population-level parameter analysis implicated macrophage cell death and cytokine half- dynamics in endotype-level differences. Several proof-of-concept statistical models were introduced to demonstrate that it is possible to estimate the pre-hospital time of sepsis subjects and to quantify their sepsis-induced systemic tissue damage. A nearest-neighbor-based method was verified against animal and human data and revealed that identifying infection time-zero of sepsis patients can be quickly estimated with high accuracy using commonly measured clinical features. A logistic regression ensemble model demonstrated revealed early organ dysfunction were significant contributors to systemic damage and mortality. Knowledge of time-zero and systemic damage levels, in combination with an endotype classifier, provides clinicians with a clear depiction of where a subject is located on their sepsis trajectory. Such a decision support system enables therapy timing, early organ support, and targeted therapies to guide personalized treatment and shift patients towards better outcomes in sepsis.
Roman M. Shapiro, Michelle P. Zeller, Theodore E. Warkentin
A 38-year-old woman was brought in to the emergency room (ER) because of altered level of consciousness. She had a known history of seizure disorder but was not on any prescription medications. She also had a history of intravenous (IV) drug abuse, including cocaine, methamphetamine, and heroin. She recently described herself as being “pill-sick” which according to her partner meant that she felt unwell after a recent administration of an illicit drug. In the ER, she was hypoxemic and was intubated. She was febrile, her blood pressure was 90 systolic, heart rate 120/min, and oxygen saturation undetectable by the digital probe. Admission hemoglobin was 11.6 g/dL (mean corpuscular volume [MCV] 86), white blood count 0.8 × 109/L (absolute neutrophil count [ANC], 0.2), and platelet count 97 × 109/L. Rare nucleated red blood cells were seen, but no myeloblasts or other primitive cells. Coagulation studies showed INR 3.1, activated partial thromboplastin time (aPTT) 81 s, and fibrinogen 1.9 g/L. The serum creatinine was 320 µmol/L (reference range, 50–98); serum lactate measured 13.0 mmol/L (reference range, 0.5–2.2). A chest X-ray showed multiple pulmonary infiltrates and right-sided pleural effusion. A CT head scan was negative for any acute abnormality. CT chest was suspicious for a right-sided empyema, but no vegetations were seen on the heart valves. The most striking finding on her admission blood work was leukopenia with near-absence of neutrophils in the peripheral blood. The clinical picture of hypotension, tachycardia, lactic acidosis, renal failure, pulmonary infiltrates, and pleural effusion in the setting of IV drug abuse, together with thrombocytopenia, coagulopathy, and normoblastemia, suggests septicemia in the setting of pneumonia or right-sided infective endocarditis, most likely complicated by disseminated intravascular coagulation (DIC)–although fibrin D-dimer levels would be helpful in supporting the last diagnosis. Severe sepsis can result in a transient leukopenia/neutropenia,1 although complete absence of circulating neutrophils is unusual and points to a possible de novo neutropenic disorder such as drug-induced agranulocytosis2, 3 (although the patient was not receiving any prescription medications). Current guidelines do not recommend starting this patient on G-CSF on admission due to lack of proven mortality benefit.4, 5 Circulating nucleated red blood cells (normoblastemia) portend a poor prognosis in a critically ill patient.6 She had a central venous catheter inserted into her right internal jugular vein and was transferred to the intensive care unit (ICU) on vasopressin and norepinephrine. Blood and urine cultures were sent and the patient was started on piperacillin-tazobactam and vancomycin. Repeat blood work showed progressive thrombocytopenia, with the platelet count measuring 9 × 109/L 20 h later, and with the leukocytes and ANC remaining profoundly reduced at 0.1 and 0, respectively. Repeat coagulation studies showed INR 3.9, aPTT >150, fibrinogen 2.0, and D-dimer > 20,000 μg/L fibrin equivalent units (reference range, <500), antithrombin activity 0.26 U/mL (reference range, 0.77–1.25), and protein C activity 0.14 U/mL (reference range, 0.70–1.80). Chemistry studies showed lactate dehydrogenase (LDH) 769 U/mL (reference range, 100–220), creatine kinase (CK) 3,800 U/mL (reference range, <168), total bilirubin 182 μmol/L (reference range, <21), conjugated bilirubin 117 μmol/L (reference range, <8.6), alanine aminotransferase (ALT) 67 U/L (reference range, <28), alkaline phosphatase 68 U/mL (reference range, 40–120), and creatinine (following initiation of continuous renal replacement therapy [CRRT]) had declined to 130 μmol/L. She was noted to have cold and dusky extremities and several necrotic ulcers were noted over the dorsum of her left hand. All peripheral pulses were intact, without peripheral cyanosis or other evidence of overt limb ischemia. She was started on IV heparin with dose adjusted by anti-factor Xa levels and given antithrombin concentrates (988 units, administered at 12-h intervals) as well as frozen plasma by infusion. HIV testing was ordered, along with hepatitis B and C serology. The fibrin D-dimer was measured to evaluate the presence of DIC. Although D-dimer levels are elevated in many or most hospital patients, greatly elevated levels as seen in this patient (>20,000) are consistent with a diagnosis of DIC.7 The elevated LDH, CK, bilirubin, and creatinine values likely reflect tissue (muscle) injury and organ (liver, renal) dysfunction.8, 9 The antithrombin and protein C activity levels were measured because of the potential role for acquired severe depletion of protein C and antithrombin activity in predisposing to ischemic limb necrosis/gangrene with pulses in critically ill patients who have DIC and acute ischemic hepatitis (“shock liver”).10 Though this patient had only mildly increased liver enzymes, the possibility of chronic liver disease due to viral hepatitis (frequently observed in IV drug abusers) or another etiology could represent an alternative at-risk scenario for limb ischemia.11 Although treatment for DIC-associated limb ischemia is uncertain, we followed a recent suggestion12 to treat with unfractionated heparin, antithrombin concentrates, and plasma infusion. In this patient, the presence of necrotic ulcers on the dorsum of her left hand along with cool, dusky extremities, and elevated LDH and CK levels (indicating possible tissue ischemia13), prompted her medical team to initiate preemptive heparin therapy to minimize the risk of developing acral limb ischemia/gangrene in the setting of liver dysfunction. Of note, with this therapy, the appearance of the limbs improved, becoming warm and of normal color, and there was no development of overt limb ischemia at any later time. On the fourth hospital day, the patient's blood work yielded a hemoglobin 6.7 g/dL, MCV 85.6, WBC 0.2 × 109/L (with ANC 0), platelet count 8 × 109/L, INR 1.2, aPTT 64, anti-factor Xa level 0.18 U/mL (therapeutic range, 0.35–0.70), fibrinogen 3.3, D-dimer 11,500, total bilirubin 270 μmol/L. Blood cultures and pleural fluid cultures were positive for methicillin-sensitive Staphylococcus aureus (MSSA), while hepatitis B and C serology came back negative. Antinuclear antibody (ANA) was sent as part of an autoimmune screen. Bone marrow aspirate and biopsy were attempted in order to better characterize the cause of her persistent profound neutropenia but the patient was too hemodynamically unstable to have the procedure done. She was started on G-CSF (300 μg/day). At this point, the differential diagnosis for neutropenia included septic shock, chronic liver disease, HIV, autoimmune disease, or drug-induced neutropenia/agranulocytosis. As she was receiving appropriate antibiotics since admission, it seemed less likely that infection was the cause of ongoing neutropenia. Severe sepsis is associated with bone marrow dysfunction, but more often as a result (not a cause) of underlying bone marrow suppression.1, 14, 15 Certain infections are known to cause neutropenia, including tuberculosis, HIV, cytomegalovirus, Epstein-Barr virus, and any of the hepatitis viruses.16-19 Although this patient had risk factors for HIV and viral hepatitis, serological studies ruled out these infections. A common association of infection with profound neutropenia occurs following systemic antineoplastic chemotherapy (“febrile neutropenia”). For patients who develop severe sepsis with persisting neutropenia, the pathophysiology is believed sometimes to be hematopoietic stem cell “exhaustion.”14 However, this tends to occur in infants and the elderly where the bone marrow reserve of granulocytic precursors is smaller than in adults.14, 15 The finding of an ANC of zero on presentation to hospital also argues against sepsis-induced bone marrow exhaustion. As persisting absence of circulating neutrophils strongly indicated the possibility of drug-induced agranulocytosis, a thorough re-review of any medications the patient might have received prior to and following admission was undertaken. Although both vancomycin and piperacillin-tazobactam may cause neutropenia,3 these antibiotics were only started after the patient already had an ANC of zero; and because only 4 days had elapsed following admission, the timeline was too soon for superimposed antibiotic-induced neutrophil-reactive antibodies to be plausible. Accordingly, further consideration for a medication or other drug taken prior to admission was given. Of the illicit IV drugs she used, cocaine has been associated with agranulocytosis due to the presence of levamisole as an adulterant.20, 21 On the sixth day of admission, her bloodwork yielded a hemoglobin 8.1 g/dL, WBC 0.6 × 109/L with ANC 0, platelet count 19 × 109/L, INR 1.2, aPTT 46. Her HIV test as well as ANA came back negative. Repeat blood cultures from the fourth day of hospital admission also came back negative. She continued to be hemodynamically unstable requiring increasing vasopressor and inotropic support. Blood culture obtained on the sixth day of hospitalization from the dialysis catheter site yielded budding yeast cells and pseudohyphae (after 28-h incubation), and the infectious disease consultant diagnosed superimposed Candida albicans infection in this patient with persisting profound neutropenia. Despite adding anidulafungin, she died on the eighth day due to multiorgan failure as a result of septic shock. Her ANC remained at 0 throughout her hospitalization. Figure 1 summarizes the patient's clinical course, including serial platelet count and ANC values (panel A), fibrinogen and INR values (panel B), aPTT and anti-factor Xa levels (panel C), with evidence for initially disturbed “procoagulant-anticoagulant” balance,10, 12 with greatly elevated fibrin D-dimer levels and markedly reduced antithrombin and protein C activity levels (panel D), which improved during treatment with heparin, antithrombin concentrates, and frozen plasma infusion. Summary of the patient's clinical course. (A) Serial platelet count and absolute neutrophil count (ANC) levels; timing and dose of intravenous heparin infusion rate is also shown. (B) Serial INR (international normalized ratio) and fibrinogen levels. (C) Serial partial thromboplastin time (aPTT) and anti-factor Xa levels. (D) Serial D-dimer, antithrombin (AT) and protein C (PC) activity levels; timing and dosing of AT concentrates and of plasma infusion are also shown. The patient died 7 days following admission Persistence of agranulocytosis unresponsive to G-CSF suggested an ongoing profound bone marrow suppression effect. Levamisole-induced agranulocytosis appeared likely given this patient's known habitual use of cocaine: it is estimated that 70%–80% of all cocaine in North America contains levamisole as an adulterant.21 Studies on levamisole identified long-term exposure to the drug as a significant risk factor for development of agranulocytosis.21, 22 Drug clearance is normally rapid with normally functioning kidneys, but this patient was anuric. Furthermore, levamisole is not cleared by dialysis, suggesting that plasma levels would not be lowered by CRRT.23 As this patient was injecting cocaine in the dorsum of her hand and had several necrotic lesions located at prior injection sites, the possibility of a persistent depot of levamisole was also considered to explain sustained agranulocytosis due to prolonged exposure to the drug or a relevant metabolite. Neutropenia is a common phenomenon frequently encountered in patients on antineoplastic chemotherapy, as an infection-related event, in certain autoimmune conditions, and in nutritional deficiency.2, 3 In contrast, agranulocytosis is a rare immune-mediated phenomenon triggered by certain medications3 or IV drugs.20, 21 Several medications have a known association with agranulocytosis, with the more common mechanism presumed to be antibodies to a neoepitope formed by a combination of the medication (or metabolite) and a receptor on the neutrophil membrane.24, 25 However, definitive proof is often lacking because detection of such antibodies is challenging and rarely performed. On some occasions, drug-dependent anti-neutrophil antibodies have been detected in the serum of patients with agranulocytosis.25, 26 A recent review of drug-induced agranulocytosis identified medications with the greatest likelihood of inducing agranulocytosis (Table 1), including clozapine, propylthiouracil, carbamazepine, trimethoprim-sulfamethoxazole, beta-lactams, and levamisole, among others.3 Among these, clozapine has the highest frequency of triggering agranulocytosis, estimated at 1 in 125 patients.24, 25 Beta-lactams Vancomycin Dapsone Trimethoprim-sulfamethoxazole 25 days24 82 days27 39 days24 UNKa 0.42/106 per year28 UNK UNK 5.6/106 per year29 Immune-mediated25 Immune-mediated25 Reactive metabolite induces apoptosis25 Immune-mediated29 Diclofenac Sulfasalazine UNK 42 days24 0.14/106 per year28 1:1600 in IBD29 1:6100 in RA29 Inhibit myelopoiesis25 Direct toxicity, immune-mediated30 Methimazole Propylthiouracil 42 days24 36 days24 0.25/106 per year28 0.2-0.5% 31 Immune-mediated32 Reactive metabolite induces apoptosis25, 33, immune-mediated29 Clozapine Chlorpromazine 56 days24 45 days24 0.8% 29 UNK Reactive metabolite induces apoptosis25, 33, inhibit myelopoiesis25 Inhibit myelopoiesis29 Phenytoin Carbamazepine 14 days33 49 days24 UNK 0.09/106 per year28 Reactive metabolite induces apoptosis25 Inhibit myelopoiesis29 Procainamide Captopril Ticlopidine 47 days24 32 days24 39 days24 UNK UNK 0.39/106 per year28 Reactive metabolite induces apoptosis25 Immune-mediated25 Reactive metabolite Induces apoptosis25 Levamisole Deferiprone Rituximabb 60 days24 UNK 4 cycles24 UNK 0.2-0.4/100 patient years29 UNK Immune-mediated25 UNK Immune-mediated25 Among IV drug users, levamisole is becoming a more recognized etiological agent due to its near-ubiquitous presence in cocaine distributed in North America.21, 22 Its association with agranulocytosis was first recognized in patients who received levamisole for treatment of rheumatoid arthritis during in the 1990s.22 Due to its relatively high risk for causing agranulocytosis, levamisole was withdrawn from the market in 2000, but it continues to find use in the veterinary world as an antihelminthic.21, 34 Its physical resemblance to cocaine and the relative ease of acquisition at a low price explain its common use as an adulterant in the manufacture of cocaine.23 Pharmacokinetic studies of levamisole injected into the muscle and fat of animals yielded a short plasma half-life and rapid elimination via the urinary tract within 24 hours.35, 36 However, repeated exposure to low doses of the drug over a period of months can induce the formation of antibodies implicated in agranulocytosis.26 The clinical presentation of patients who develop agranulocytosis during exposure to levamisole-adulterated cocaine is variable, with many patients either asymptomatic or developing a nonspecific flu-like illness. Around 50% of patients may initially present to the ER with mouth ulcers or odynophagia.21 A vasculitic rash has been associated with levamisole exposure that develops in the setting of an occlusive thrombotic vasculopathy. The latter most frequently involves the cheeks and ears and tends to be ANCA-positive.37 However, as there are cases of levamisole-induced agranulocytosis not associated with positive ANA or ANCA, the absence of these autoimmune markers does not rule out levamisole as the culprit drug38 (note: we did not measure ANCA in our patient case). In addition, case reports of patients who develop an autoimmune marker such as ANA, c-ANCA, p-ANCA, or antiphospholipid antibody in the setting of known levamisole exposure describe the disappearance of this marker within 14 months of discontinuation of drug exposure.21 It is not yet clear why a relatively high proportion (approximately 2%–4%) of patients who habitually use cocaine and who are likely chronically exposed to levamisole develop agranulocytosis.21 In previous studies of rheumatoid arthritis patients, risk factors for levamisole-induced agranulocytosis included female sex, HLA-B27, and frequency of drug exposure.21, 39 The expected time for recovery of drug-induced agranulocytosis after stopping the offending drug, without the addition of growth factor support, is a median of 9 days;40 the addition of G-CSF reduces the time for neutrophil recovery to approximately 4-7 days.40 When severe agranulocytosis persists beyond this period, unusual circumstances have been reported. For example, some patients with persisting agranulocytosis unresponsive to G-CSF have parvovirus B19 infection, and recover after administration of IVIgG (implicating removal of circulating virus by IgG anti-parvovirus antibody, reversing virus-induced inhibition of granulopoiesis).41 The typical course of levamisole-associated agranulocytosis usually involves improvement in the neutrophil count beginning several days after discontinuation of the drug, with recombinant G-CSF hastening recovery.21 Given that levamisole is normally metabolized in the liver and secreted via the renal system, the presence of significant liver and/or kidney dysfunction is likely to significantly prolong its half-life.21, 23 Furthermore, hemodialysis does not clear the drug.23 When injected intramuscularly or into adipose tissue a depot of the drug may develop that results in continuous long-term exposure. Given our had an ANC of zero that persisted for 8 days following admission (until death) without any other clear precipitant or cause identified and despite starting G-CSF, we hypothesize that the combination of an intramuscular or subcutaneous depot of levamisole, together with absent renal clearance and hepatic dysfunction, resulted in ongoing exposure to levamisole (or a metabolite) that potentiated pathological effects of levamisole-dependent anti-neutrophil antibodies. The recommended management of levamisole-induced agranulocytosis is drug withdrawal and supportive care.21 Reported experience, mainly based on cocaine users who were not critically ill, typically show rapid recovery within a few days of drug discontinuation. There is a paucity of data in patients who are critically ill with multiorgan dysfunction; indeed, to our knowledge, presentation as septic shock (due to MSSA septicemia) in the setting of agranulocytosis presumed to be caused by levamisole-adulterated cocaine has not been previously reported. The authors have no conflicts of interest to report.