FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2057185675> ?p ?o ?g. }

• W2057185675 endingPage "386" @default.
• W2057185675 startingPage "384" @default.
• W2057185675 abstract "ACAP, acetaminophen; ALF, acute liver failure; APACHE, Acute Physiology and Chronic Health Evaluation; KCH, King's College Hospital; MELD, Model for End-Stage Liver Disease; SIRS, systemic inflammatory response syndrome; SOFA, Sequential Organ Failure Assessment. Acute liver failure (ALF) has an unpredictable clinical course and often rapidly progresses to death within days or weeks, but recovery to full health can occur if the patient survives the crisis. Acetaminophen (ACAP) is the single most common cause of ALF in the United States, Europe, and Australia1; it accounts for approximately half of all cases and has an overall mortality rate of 30%.2 ACAP cases usually progress more rapidly and have higher rates of brain edema than non-ACAP cases3; early fatal complications such as cerebral herniation result in twice as many deaths for patients listed for ACAP-induced ALF versus patients with non–ACAP-induced ALF,3, 4 despite early organ availability.5 Liver transplantation is the only therapy with proven benefit, but making the decision to perform transplantation is difficult; these patients, who often have attempted suicide, cannot participate in decision making, frequently have psychiatric diseases, and often have drug or alcohol dependence, and decisions have to be made at the bedside, often with insufficient time for gathering all relevant information. A substantial fraction of patients have medical or psychosocial contraindications to transplantation and are triaged from further transplant consideration at the outset.3, 6 For potential transplant candidates, the accurate prediction of spontaneous survival is essential because patients who undergo transplantation for ALF have worse survival rates than spontaneous survivors,2, 3, 7 they have the lifelong medical burden of their transplant, and the liver could have saved the life of another listed patient. Predicting who is at risk of death allows early listing for transplantation and increases the likelihood of finding an organ in time. The King's College Hospital (KCH) model,8 the most widely used and validated prognostic model, has consistently been shown to have high specificity for identifying patients who will die without a transplant, yet the KCH criteria have 2 weaknesses: they have relatively poor sensitivity, and they can be used only late in the course of disease. Poor sensitivity means patients who will die without a transplant are not identified by the KCH criteria. For these reasons, Cholongitas et al.9 sought a better alternative to the KCH criteria: they systematically compared the Sequential Organ Failure Assessment (SOFA) score, the Model for End-Stage Liver Disease (MELD) score, the Acute Physiology and Chronic Health Evaluation II (APACHE II) score, and serum lactate. They found that the SOFA score had higher combined sensitivity and specificity (as measured by the area under the receiver operating characteristic curve) than the KCH, MELD, and APACHE scores, and although lactate could be used to discriminate between death or transplantation and spontaneous survival, it did not add additional discrimination to the examined models.9 This study expands on an earlier publication in which they first advocated the use of SOFA for patients with ALF.10 This article raises several issues. The first issue is statistical: how should the physician taking care of the patient at the bedside use these predictive models or tests, and should the SOFA score be used in preference to the KCH criteria? The second issue is biological: why does the SOFA score, which is a non–liver-specific assessment of organ failure, discriminate better than liver-specific models for ACAP-induced ALF, and are patients who are identified in this way salvageable? Third, should we constrain our use of these models to a patient's first entry to the intensive care unit after the initial resuscitation (as in this study), or can these models be used as the disease progresses? These findings will need to be corroborated. SOFA, however, has achieved some validation from a number of recent studies including its predictive accuracy for ACAP-induced ALF11-13 and non–ACAP-induced ALF11; it is said to be easier to perform than APACHE scoring. Studies of a variety of different models and blood tests have periodically shown that one or another test has greater diagnostic accuracy than the KCH criteria, but these findings have not stood the test of time. The KCH criteria are extremely well validated. Systematic reviews have consistently shown that the model has high specificity: it accurately predicts death without transplantation for patients with ACAP-induced ALF who meet the criteria. A recent meta-analysis of 14 of the best studies showed that the pooled specificity of the KCH criteria was 94.4%12; the specificity in the present study was 83% to 86%, and it was better than the specificity of the SOFA (74%-84%), MELD, and APACHE scores and serum lactate.9 As the authors indicate, the weakness of the KCH model is its relatively poor sensitivity (58.2% in the meta-analysis and 47% in this study), which leads to a relatively poor area under the receiver operating characteristic curve; therefore, the KCH criteria have poor overall predictivity. Poor sensitivity leads to deaths of omission: patients die because the model or test did not predict that they would die. The assertion that one test is better than another (eg, SOFA > KCH > MELD) needs to be treated with caution. In establishing the risk of death at the bedside, we use the notoriously nonintuitive logic of conditional probability (whether we recognize we are doing this or not). We are not limited to using any one test or model, and Bayes' rule informs us that maximizing positive and negative predictivity (for death in this case) depends on the prior or pretest probability of death for the individual and on the characteristics of the test or model (sensitivity and specificity). The choice of the most efficient test or model depends on the pretest probability of the diagnosis (death) and on the wish to maximize positive (death) or negative predictivity (spontaneous survival). Generally, tests with high specificity increase discrimination more if the pretest probability (or prevalence in populations studies) is relatively high, whereas high sensitivity provides greater discrimination if the pretest probability is low. A patient with characteristics similar to those in this study has a pretest probability of death of 0.46 (the prevalence of death in the study). Using the KCH model's specificity of 0.83 from this study yields a posttest probability of death of 0.7 for patients meeting the KCH criteria. The application of SOFA with this new prior probability increases the risk to nearly 0.9. If the patient does not meet the KCH criteria, the probability of death is no longer 0.46 but drops to 0.35. The subsequent application of SOFA shows that patients not meeting the criteria have a reduced probability of death of 0.16. The probabilities are identical, whichever model is used first. As long as the models or tests are independent, any number can be applied to an individual to refine the probabilities of death or spontaneous survival. We can use available independent models to serially refine the assessment of probability, although we will obtain the greatest delta in positive predictivity from highly specific tests for patients with a high pretest probability of death (maximized positive predictivity) and in negative predictivity from tests with high sensitivity for patients with a low pretest probability of death. When multiple models or tests are being used, they need to be independent (eg, we should not apply SOFA and then APACHE because the model parameters overlap). Bernal et al.,6 for example, showed that among 28 nonsurvivors who did not meet the KCH criteria (all met 2 criteria but not 3 criteria), an APACHE III score > 60 identified 16 patients who died, and they demonstrated the principle of improving predictive accuracy by using conditional probability with serial models. We have to accept that no single test will ever be perfect, and despite the proliferation of predictive tests and models, we have not yet found the magic bullet. Future studies should address ways of combining information in predictive tests or models to increase the clinician's level of certainty when he is making the excruciatingly difficult decision to list a patient for transplantation or not. What do we know about the timing of a KCH or SOFA assessment? Traditionally, the SOFA score and the KCH criteria are determined at the time of a patient's entry to the intensive care unit because this is the time at which they have been studied. However, this is somewhat random because of the variability of the times and conditions of patients at the point of admission. Cholongitas et al.9 interestingly found no differences in outcomes for patients with different pre-admission courses. Recent studies have shed additional light. When the SOFA score was determined not according to the admission time but from the time of ACAP ingestion, it yielded highly valuable predictive information: 48, 72, and 96 hours after the overdose (not admission), SOFA scores > 4, > 6, and > 7, respectively, had high sensitivity and specificity for death. Moreover, 48, 72, and 96 hours after the overdose, the areas under the receiver operating characteristic curve were 0.80, 0.90, and 0.91, respectively. A SOFA score > 7 during the first 96 hours predicted death or transplantation with a sensitivity of 95.0 (95% confidence interval = 78.5-99.1) and a specificity of 70.5 (95% confidence interval = 66.3-71.6).13 It may not always be possible to know the time of ingestion, but the earlier identification of patients at risk may prove invaluable in reducing the number of deaths of patients listed for ACAP-induced ALF. This study by Cholongitas et al.9 also raises intriguing questions about the biology of ALF. Why do non–liver-specific SOFA and APACHE scores have superior predictivity? The answer may be that they occur in the setting of an uncontrolled inflammatory response because high SOFA and APACHE scores are strongly associated with the presence of systemic inflammatory response syndrome (SIRS) and blood lactate levels.13, 14 Thus, high SOFA scores appear to occur in concert with the massive cytokine release that is associated with SIRS, which in turn is increasingly being recognized as the harbinger of progressive liver failure, raised intracranial pressure, renal failure, and death.13-16 The absence of SIRS has been shown to have strong negative predictivity for death: the mortality rate was 0% for 30 patients who failed to develop SIRS.13 In Cholongitas et al.'s study,9 the cardiovascular and respiratory organ failure components correlated strongly with poor outcomes. This study by Cholongitas et al.9 provides additional ammunition for improving our ability to accurately identify patients who will die without transplantation; it confirms the high sensitivity, specificity, and predictive accuracy of SOFA. The implication that SOFA should be used instead of the KCH criteria, the MELD score, or any other model or parameter, however, should be rejected. A simple conditional probability analysis (Bayes' rule) using independent models or parameters will be more effective at maximizing the positive and negative predictivity of death. At the end of the day, probabilities apply to populations; the outcome for an individual is never certain, and we have to continue to rely on our clinical judgment, which is informed by studies such as this one." @default.
• W2057185675 created "2016-06-24" @default.
• W2057185675 creator A5079803718 @default.
• W2057185675 date "2012-03-29" @default.
• W2057185675 modified "2023-10-16" @default.
• W2057185675 title "Predicting death in patients with acetaminophen-induced acute liver failure: The King's college hospital model is on the SOFA, not the mat" @default.
• W2057185675 cites W1608215573 @default.
• W2057185675 cites W1894298692 @default.
• W2057185675 cites W1978113689 @default.
• W2057185675 cites W1984934101 @default.
• W2057185675 cites W1992410639 @default.
• W2057185675 cites W2001087077 @default.
• W2057185675 cites W2027908374 @default.
• W2057185675 cites W2037201775 @default.
• W2057185675 cites W2065261900 @default.
• W2057185675 cites W2073345523 @default.
• W2057185675 cites W2104461353 @default.
• W2057185675 cites W2117912365 @default.
• W2057185675 cites W2129595649 @default.
• W2057185675 cites W2150256003 @default.
• W2057185675 doi "https://doi.org/10.1002/lt.23386" @default.
• W2057185675 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/22253086" @default.
• W2057185675 hasPublicationYear "2012" @default.
• W2057185675 type Work @default.
• W2057185675 sameAs 2057185675 @default.
• W2057185675 citedByCount "2" @default.
• W2057185675 countsByYear W20571856752012 @default.
• W2057185675 countsByYear W20571856752014 @default.
• W2057185675 crossrefType "journal-article" @default.
• W2057185675 hasAuthorship W2057185675A5079803718 @default.
• W2057185675 hasConcept C126322002 @default.
• W2057185675 hasConcept C177713679 @default.
• W2057185675 hasConcept C194828623 @default.
• W2057185675 hasConcept C2778722691 @default.
• W2057185675 hasConcept C2779609443 @default.
• W2057185675 hasConcept C2911091166 @default.
• W2057185675 hasConcept C2994427840 @default.
• W2057185675 hasConcept C42219234 @default.
• W2057185675 hasConcept C71924100 @default.
• W2057185675 hasConceptScore W2057185675C126322002 @default.
• W2057185675 hasConceptScore W2057185675C177713679 @default.
• W2057185675 hasConceptScore W2057185675C194828623 @default.
• W2057185675 hasConceptScore W2057185675C2778722691 @default.
• W2057185675 hasConceptScore W2057185675C2779609443 @default.
• W2057185675 hasConceptScore W2057185675C2911091166 @default.
• W2057185675 hasConceptScore W2057185675C2994427840 @default.
• W2057185675 hasConceptScore W2057185675C42219234 @default.
• W2057185675 hasConceptScore W2057185675C71924100 @default.
• W2057185675 hasIssue "4" @default.
• W2057185675 hasLocation W20571856751 @default.
• W2057185675 hasLocation W20571856752 @default.
• W2057185675 hasOpenAccess W2057185675 @default.
• W2057185675 hasPrimaryLocation W20571856751 @default.
• W2057185675 hasRelatedWork W2057185675 @default.
• W2057185675 hasRelatedWork W2537698388 @default.
• W2057185675 hasRelatedWork W2626043347 @default.
• W2057185675 hasRelatedWork W2788947572 @default.
• W2057185675 hasRelatedWork W2982569037 @default.
• W2057185675 hasRelatedWork W3022607854 @default.
• W2057185675 hasRelatedWork W3170886946 @default.
• W2057185675 hasRelatedWork W3193609807 @default.
• W2057185675 hasRelatedWork W4239712546 @default.
• W2057185675 hasRelatedWork W4301854616 @default.
• W2057185675 hasVolume "18" @default.
• W2057185675 isParatext "false" @default.
• W2057185675 isRetracted "false" @default.
• W2057185675 magId "2057185675" @default.
• W2057185675 workType "article" @default.