Sepsis and septic shock
Sepsis is organ dysfunction caused by the body's own dysregulated response to infection, and septic shock is the subset in which the circulation fails. Both are common and deadly, and the damage comes less from the pathogen than from the host's inflammation and immunosuppression, which injure organs that are scored and tracked as they fail. In 2020 they were treated by time-bound care bundles whose antibiotic and fluid elements were contested, deaths were hard to attribute to them, survivors often faced lasting harm, and research was turning to patient subsets and artificial intelligence.
Definition
A small percentage of the many people who acquire infections go on to develop sepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. As the definition emphasises, a patient with sepsis is threatened not by a particularly dangerous microbe but by their own disproportionate response: the principal difference between infection and sepsis lies not in the pathogen but in the excessive response of the host, which results in multiple organ dysfunction. About a third of patients with sepsis go on to septic shock, a subset in which particularly profound circulatory, cellular and metabolic abnormalities carry a greater risk than sepsis alone. Septic shock features circulatory failure and so severely reduced oxygen delivery to the tissues, and it needs vasopressors to raise arterial pressure. As for every patient in shock, it needs rapid aggressive treatment to raise blood pressure enough to keep the organs working. Patients with sepsis need hospital care, generally in an intensive care unit. The two conditions are complex, heterogeneous, plagued by comorbidities, incompletely understood and beset with therapeutic difficulties, and despite their prevalence they are not well known.
Incidence, mortality and cost
Infections progress to sepsis in about 1.7 million people a year in the United States. In high-income countries mortality is about 10% for sepsis and 40% for septic shock, and it is estimated to be considerably higher in low- and middle-income countries; in-hospital death is common, its rate depending on the patient's age and comorbidities and on hospital quality. Together sepsis and septic shock account for about 35% of all in-hospital deaths in the United States, and if sepsis were on the US list of causes of death it would rank third, with 270,000 deaths a year before COVID-19. The annual cost in the United States alone is estimated at more than $20 billion.
Causative organisms
Which organisms cause sepsis is a difficult question to answer. Only 60% to 80% of cultures from septic patients yield a presumptive causative organism, most often a bacterium. Several things complicate the picture further: the primary infected site may be relatively isolated from the circulation, samples are sometimes taken only after antimicrobial treatment has begun, cultures may turn positive only after days, later cultures may reflect secondary infections acquired in hospital after sepsis-induced immunosuppression, and detecting viruses needs special testing that is not often done. Bacteria, fungi, parasites and viruses can all induce sepsis, bacteria most commonly; some studies suggest viruses are rarely causative, but adequate testing shows them often to be viable candidates, and by 2020 severe COVID-19 was recognised as viral sepsis. Fungal infections have become more common, particularly in the immunosuppressed, and cause about 10% of cases. About 90% of US cases are acquired in the community, but as many as half are acquired in hospital in some countries.
The host response
Microbes carry pathogen-associated molecular patterns (PAMPs) that bind pattern-recognition receptors on innate immune cells and endothelial cells and start an inflammatory response, and host cells damaged during inflammation release damage-associated molecular patterns (DAMPs) that bind the same receptors and keep it going. Activation leads to the production and systemic spread of cytokines and other mediators that promote inflammation, and at the vascular endothelium its consequences include vascular injury and activation of the coagulation and complement cascades, while exposure to PAMPs and DAMPs potently attracts a variety of innate immune cells. Ordinarily these responses are necessary and beneficial, but in sepsis a dysregulated response of largely unknown origin produces multiple organ dysfunction. Pathogens seldom reach several organs, so most failing organs are not themselves infected: their dysfunction comes from unresolved hyperinflammation and its systemic mediators, the "cytokine storm", which probably accounts for early deaths with multiple organ dysfunction.
That storm was once considered the hallmark of sepsis, but it turns out to be invariably paired with profound immunosuppression, also of largely unknown origin, the two varying in timing and size. The immunosuppressed state includes exhaustion and drastic loss of lymphocytes (lymphopenia); more poorly functional immune cells that inhibit immunity — immature neutrophils, myeloid-derived suppressor cells and regulatory T cells; and macrophages and dendritic cells less able to activate other immune cells. All of these predict poor outcomes and death: the depth of lymphopenia on admission to intensive care is associated with initial severity, and its persistence with secondary infections and death. Patients therefore fall prey to opportunistic organisms in hospital and after discharge, and infections may smoulder rather than resolve. Some patients' immune cells show increased surface markers of lymphocyte exhaustion like those seen in many cancers, which may be a therapeutic opportunity.
Organ dysfunction and its scoring
The systemic barrage of inflammatory mediators is enough in itself to induce organ dysfunction, and three other features of sepsis promote it: complement activation, vascular endothelial dysfunction and coagulation. Complement activation, immediately after exposure to PAMPs and DAMPs, releases cytokines and chemokines that markedly dilate blood vessels; the vascular endothelium then becomes extremely leaky, moving plasma and blood proteins into the interstitium, and together vasodilation and leakage across the microcirculation reduce tissue perfusion and damage organs, creating the conditions for shock. Finally, sepsis is hypercoagulable: microthrombi decrease microvascular perfusion, often as disseminated intravascular coagulation, in which clots form throughout the bloodstream and block small vessels. Organ dysfunction is presumed to come from hypoxia arising from poor perfusion and hypercoagulability, combined with the release of inflammatory mediators into the circulation. Any organ can be affected, and single-organ dysfunction is rare. Because outcomes are affected by the specific set of failing organs, it is essential to establish which organs are dysfunctional and how far.
The Sequential Organ Failure Assessment (SOFA) score assesses this by averaging individual scores of 1 to 4 for six organ systems: a cardiovascular score rising as mean arterial pressure falls and increasingly needs pharmacological support; a renal score rising with serum creatinine or as urine output falls; a respiratory score rising as, roughly, the oxygen in arterial blood falls relative to room air; a central nervous system score rising as the Glasgow Coma Score falls; a coagulation score rising as platelets fall; and a liver score rising with serum bilirubin. A change in SOFA score of 2 or more reflects a mortality risk of 10% and can be followed over time to track the patient's progress, and it is the change rather than the absolute score that is relevant, because comorbidities are frequent in patients with sepsis and their SOFA score is thus often 0 before they develop sepsis.
In the cardiovascular system, increased endothelial permeability, vasodilation and cytokine output reduce intravascular volume and cause hypotension and cardiac depression, which increase mortality. Kidney dysfunction progressing to failure is a major cause of morbidity: more than half of patients develop acute kidney injury, with a 62% higher risk of in-hospital death. Damage to the lung alveoli causes oedema and hypoxaemia, raising the respiratory rate and tiring the respiratory muscles, and acute respiratory distress syndrome occurs in 40%, increasing mortality. Insufficient brain perfusion alters mental state — lethargy, disorientation, confusion, delirium or coma — increasing mortality and causing prolonged cognitive dysfunction in survivors. Excess platelet consumption and depleted coagulation factors cause thrombocytopenia, disseminated intravascular coagulation or both in about 60%, increasing mortality. Haemodynamic changes, including endothelial dysfunction and microthrombi, reduce liver perfusion, leading to dysfunction and higher mortality in 50%.
The gut and its microbiome are counted as a seventh organ. In sepsis the microbiome is severely perturbed — less diverse, overgrown by opportunistic pathogens and releasing toxic mediators that damage distant organs. The microbiome also affects the efficacy of the immunotherapy used in cancer that was by 2020 being tested in sepsis, and because both sepsis and antibiotics alter it, both may compromise that therapy.
Septic shock
In every form of shock, markedly impaired oxygen delivery cannot meet the cells' demand, and organ dysfunction becomes profound and quickly irreversible. Septic shock is classified as distributive: systemic vascular resistance falls through vasodilation and the loss of intravascular fluid to the interstitium, until blood pressure and volume are too low to keep vital organs working. Intravenous fluid improves tissue perfusion quickly, but it always leaves the vasculature within a few hours, and faster through a leaky endothelium, so these patients also need a vasopressor, which constricts vessels and improves the rate and strength of the heart's contraction to raise blood pressure.
Guidelines, care bundles and SEP-1
Every four years the International Guidelines for the Management of Sepsis and Septic Shock issue recommendations from an expert panel, which grades each as strong or weak and its evidence from high to very low quality. The guidelines suggest that most clinicians should follow strong recommendations most of the time, but that a patient's clinical characteristics can make one inapplicable: "a strong recommendation does not imply standard of care". Care bundles condense the lengthy guidelines into what should be done quickly. The one-hour bundle current in 2020 was to measure lactate and remeasure it if it was above 2 mmol/L (weak recommendation, low-quality evidence); obtain blood cultures before giving antibiotics (a best-practice statement); give broad-spectrum antibiotics (strong, moderate); rapidly give 30 mL/kg of crystalloid if the patient is hypotensive or lactate is 4 mmol/L or more (strong, low); and give vasopressors if the patient is hypotensive during or after fluid resuscitation, to keep mean arterial pressure at 65 mmHg or more (strong, moderate).
The one-hour bundle is made of recommendations, but the similar elements of the SEP-1 quality measure of the US Centers for Medicare and Medicaid Services (CMS) are closer to requirements. US hospitals must report their compliance rates, the measure "mandates" that patients meeting its criteria receive the bundle, and CMS measures adherence rather than mortality, so a hospital with poor adherence risks penalties and loss of accreditation however its patients fare. Bundles like SEP-1 improve survival, but many physicians argue that its elements should be flexible guidelines leaving room for physician judgment without bureaucratic oversight, and the debate, often heated, centres on antibiotics and intravenous fluids.
Antibiotics
The first task when a patient with sepsis arrives is to establish whether there is an infection, which can be difficult: there may be no definitive sign, and even fever is often absent, especially in elderly patients. A patient in shock is an emergency in which the type of shock has to be found quickly, and for septic shock antibiotics are widely agreed to be standard of care, because proof of infection may never come and every hour's delay increases mortality. For sepsis without shock, opinions differ on the requirement to give broad-spectrum antibiotics within one hour. The Infectious Diseases Society of America (IDSA) and others hold that SEP-1's one-hour requirement promotes antibiotic overuse, since the presumed infection is sometimes wrong or not bacterial. Antibiotic overuse promotes the development of antibiotic-resistant bacteria. One warning, drawn from the dire prospects of pneumococcal pneumonia before penicillin and of severe COVID-19, in both of which no specific antimicrobial was available, is that the rise of resistance from inappropriate antibiotic use could make bacterial sepsis as deadly as it was before antibiotics. Antibiotics are not harmless: apart from their better-known risks, they may show mitochondrial toxicity, which promotes organ failure, and alter the gut microbiome. CMS replied that without compelling evidence that SEP-1 had caused excessive and unwarranted antibiotic use it saw no reason to alter it, and that the burden of proof lay with the IDSA — a demand one reviewer of the field finds extreme where clinicians must sometimes follow recommendations resting on low-quality evidence.
Intravenous fluids
Intravenous fluid is needed in hypotensive patients to raise blood pressure and prevent the inadequate oxygen and nutrient delivery that leads to irreversible organ failure and death, but about two-thirds of an infusion moves into the interstitium within 60 minutes, so infusions are often repeated. The resulting oedema is generally well tolerated, but less so in the lungs, where it compromises oxygen delivery. Giving fluid to a hypovolaemic patient can be very dangerous, particularly in older patients, because the therapeutic window is narrow: under-resuscitation leaves hypovolaemia unresolved and raises mortality and morbidity, while over-resuscitation reduces oxygen transport and causes oedema, and fluid beyond what corrects hypovolaemia is associated with longer stays, more morbidity and more deaths. Many physicians therefore object to SEP-1's 30 mL/kg for every patient with sepsis or septic shock, one review calling its continued enforcement "scientifically, morally, and ethically unacceptable". Help in deciding how much to give and when to stop should come from dynamic assessments, such as the change in stroke volume on passive leg raising and careful assessment of capillary refill time. These improve on static assessments such as the change in mean arterial pressure, which may not even change with fluid, and periodic measurement of serum lactate.
Source control, vasopressors and respiratory support
The other treatment basics are less contentious. Source control — draining abscesses, removing necrotic material and removing infected devices such as catheters — is a core treatment where possible; vasopressors, most commonly norepinephrine, are needed in septic shock when fluid cannot sufficiently restore blood pressure, to raise it and so organ perfusion; and respiratory support, by supplemental oxygen or mechanical ventilation, is often required.
Attributing deaths to sepsis
Many patients die with sepsis, but whether they died of it is not straightforward. Sepsis is not precisely defined, particularly when infection cannot be shown, and in one study incidence varied more than threefold between four methods of assessment. Most patients are elderly, frail and have serious comorbidities such as cancer — 40% in one study had a hospice-qualifying condition on admission — and it has been suggested that their deaths should not be attributed to sepsis. Patients with sepsis often develop acute respiratory distress syndrome or acute kidney injury, either of which can be taken as the cause of death, and coding an illness as sepsis may have favourable financial consequences for the hospital.
Life after sepsis
Discharge is often less than a cure. Many survivors remain chronically ill and many die within two years. Of patients with septic shock, half recover, a third die in the following year and a sixth have severe persistent impairment, including functional limitation, a threefold increase in moderate to severe cognitive impairment, and much anxiety, depression and post-traumatic stress disorder. Within three months, 40% of patients are back in hospital. Only a minority return to a fully functional life, most going to long-term nursing or rehabilitation facilities. Patients often suffer from post-ICU syndrome — insomnia, depression and loss of cognitive function — which may last indefinitely. Apart from the residue of sepsis-induced damage, their cause may be a chronic critical illness with a compromised and immunosuppressive immune system, named persistent inflammation, immunosuppression and catabolism syndrome.
Research directions
By 2020 there had been great advances in understanding the molecular details of sepsis and septic shock, which were expected to profoundly influence the development of future therapies, perhaps soon. A vast amount of preclinical and clinical research was under way and the number of prospective treatments was growing swiftly, with the sense, as in cancer immunotherapy and targeted therapy, that big things might happen any day. Diagnosis and management are so difficult that physicians should ideally have abundant specific experience. Because not all physicians have that training, protocolised care is of great value in making every physician and team effective. Even so, morbidity and mortality remained high. One path to improvement may be artificial intelligence, not to replace physicians but to draw on the volume of patient data streaming from monitors and entered by hand. One tool, the Artificial Intelligence Clinician, had outperformed physicians in decisions on fluid resuscitation, and many others were being developed. Another use of artificial intelligence is in dealing with the crucial problem of patient heterogeneity, arising from variables such as pathogen, age, comorbidities and history. Heterogeneity in the patient's evolving immunophenotype, which makes a patient hyperinflamed or immunosuppressed, is sure to be especially important, because some treatments that failed in trials of heterogeneous patients would likely have shown benefit in selected, more homogeneous subsets. Separating patients into more homogeneous groups before treatment was being very actively pursued, both pathophysiologically and immunologically. Marked improvements in outcomes were expected from existing and future tools, especially as the known unknowns are reduced.