OHDSI Home | Forums | Wiki | Github

Phenotype Phebrurary 2023 - P6 - Acute Hepatic Failure

Thank you @Christian_Reich for volunteering to perform an evaluation on the candidate cohort definitions for Acute Hepatic Failure. Your evaluation is due on February 8th 2023.

Briefly - we are aiming to phenotype Acute Hepatic (liver) Failure. According to EASL (EASL Clinical Practical Guidelines on the management of acute (fulminant) liver failure - PubMed), Acute Liver Failure (ALF) refers to a highly specific and rare syndrome, characterized by an acute abnormality of liver blood tests in an individual without underlying chronic liver disease (such as cirrhosis).

To perform the evaluation, @Christian_Reich will be provided the following:

  1. Clinical description (see below)
  2. Cohort definitions
  3. Output of Cohort Diagnostics available at data.ohdsi.org/PhenotypeLibrary

Clinical Description:
Acute hepatic failure refers to the development of severe acute liver injury with encephalopathy in a patient without cirrhosis or preexisting chronic liver disease. Clinically, the presence of encephalopathy is the key differentiator of acute hepatic failure from other forms of liver injury, i.e., the mere presence of coagulopathy, impaired liver synthetic function or liver injury in the absence of encephalopathy is not considered acute hepatic failure. Encephalopathy may vary from altered mental status to coma. Symptoms include fatigue/malaise, lethargy, anorexia, nausea and/or vomiting, right upper quadrant pain, pruritus, jaundice, and abdominal distension. Signs on physical exam may include orthostatic hypotension, coagulopathy, vesicular skin changes, ascites, right upper quadrant tenderness, jaundice, and hepatomegaly. Common assessments include liver enzymes, bilirubin, coagulation/INR/ prothrombin time (PT), hepatitis viral serology. Look for Acetaminophen exposure. The condition is managed in specialized center, hospitals with liver transplant program. N-acetylcysteine is used for the treatment of acetaminophen toxicity. Supportive care includes bleeding prevention, infection prevention, nutritional support, monitoring for ammonia levels, encephalopathy, hypoglycemia, hypokalemia, hypomagnesemia, and hypophosphatemia. Prognosis depends on severity/underlying causes and need for liver transplantation. An episode of acute liver failure may be considered to have ended if person is discharged from hospital, receives liver transplantation, ceases to have encephalopathy.

Cohort Definitions:

  1. Cohort Id 67: Hepatic Failure with inpatient.
    • Logic description: Persons with hepatic failure, necrosis, encephalopathy or comma and managed in a inpatient or emergency setting.
  1. Cohort Id 292: Hepatic Failure. Logic description: Persons with hepatic failure or encephalopathy, but not chronic hepatic failure.

    • Logic description: Persons with hepatic failure or encephalopathy, that do not also have chronic hepatic failure. This exclusion to because according to EASL definition persons with prior liver disease are not considered eligible to have acute hepatic failure.
  2. Cohort id 295: Acute Hepatic Failure in persons with no pre-existing liver disease. Logic description: Persons with hepatic failure but not Chronic Hepatic Failure and many other liver related diseases.

    • Logic description: Persons with hepatic failure or encephalopathy that do not also have chronic hepatic failure and other liver diseases including Viral Hepatitis, Liver Cirrhosis, Liver Fibrosis, Toxic Liver Disease, Hepatic Necrosis, Alcohol use disorder, Alcoholic Hepatitis, Alcoholic ,Fatty Liver, Malignancies that spread to liver (breast, esophageal, stomach, pancreatic, lung, kidney, melanoma), Chronic Hepatic Failure, Other Chronic Liver Disease, Ascites, Portal Hypertension, History of Liver Transplant, Cholestasis, Non-Alcoholic Fatty Liver Disease, Autoimmune Hepatitis, Primary Biliary Cholangitis.

Output of Cohort Diagnostics: for the above cohort are available at data.ohdsi.org/PhenotypeLibrary .

typo in definition: comma-> coma ?

I am working on implementing transplant datamart on OMOP CDM and interested in helping with this Phenotype group for experience if I can get access to the shared folder link

Thanks, @Gowtham_Rao. This is a wonderful starting point. But since I keep nagging I can’t just take it. I need to build it up the way I am proposing how I think this should be done. :slight_smile:

Clinical Description

Semantics: “Hepatic” and “liver” are synonymous. “Injury” means damage, “failure” means cessation of function. The injury comes before the failure. Sometimes, the condition is considered “injury” only before the onset of encephalopathy, while “failure” comes afterwards. But mostly the terms are used synonymously.

Etiology: In brief, when the liver fails, metabolic functions of blood homeostasis are impaired: Synthesis of albumin and plasma proteins for coagulation, and clearance of substances other organs dump into the blood stream and the kidney can’t take care of. That causes the symptoms: jaundice (uncleared hemoglobin metabolites), coagulopathy (lack of clotting factors), encephalopathy (poison to the brain). There are others, like splenomegaly, resulting from changes to the structure of the organ and blocked blood passage.

Causes: viruses, drugs (especially acetaminophen), illicit drugs, vascular (ischemia, venous), pregnancy, autoimmune, malignant infiltration, hereditary, 30% no visible cause. Viral, drug-related, vascular, autoimmune and malignant can also cause chronic liver problems. Alcohol is mostly causing chronic liver problems. Post-transplantation failure is not considered here.

Onset: Liver failure can take vastly different time to develop. Here, we are looking for the acute one: there are different definitions, but roughly up to 12 weeks from the first symptom (usually jaundice) to encephalopathy. Chronic can take years. The acute and the chronic are typically caused by different conditions, which could be used to distinguish if we don’t get good timing. However, this not 100%, and tricky if we want to study something with acute failure as an outcome, the logic can become circular. There is also an acute on top of a chronic. It is difficult to deal with this one, because the acute piece could be the continuation of the chronic or a truly different process. In other words, the chronic alcoholic can have the liver finally give up, but could also get an unrelated ALF.

Outcome: The liver recovers spontaneously or it goes into organ failure, which means the patient dies if not transplanted quickly (hence databases like @George_Oliver has). Life without liver function is not possible.

Diagnosis: symptoms, imaging, but largely liver function tests. When liver cells die, the content gets into the blood. That is not a problem, but enzyme activity can be measured that shouldn’t be there.

Treatment: underlying cause, supportive and liver transplantation (only if the underlying cause is addressed).

Relevance for clinical research: ALF is very dramatic (time for finding a transplant is short), and often caused by very common OTC drugs.

Cohort definitions

This is tricky, particularly in the absence of the scientific question. Because:

  • We need failure, not just injury. We could trust the description (and mappings), or insist on encephalopathy, or take encephalopathy alone as a sign of a failure.
  • We need acute. Which, again, we can rely on the concepts or we exclude any patient with liver trouble earlier than 3 months before.
  • We need to decide if we allow the acute on top of chronic scenario, or if it is acute only.
  • We need an index date. Is that already when the liver is injured, or only when it failed?
  • For a conceptset, we might just pick 4026032 “Acute hepatic failure” and descendants. @Gowtham_Rao went one up the hierarchy to 4245975 “Hepatic failure”, apparently assuming the “acute” is implicit. Which it is in some ICD10 mappings (need to be fixed). We’d also correctly gain fulminant hepatic failure, subacute and subfulminant failure (which really is the same thing), but have to exclude chronic failure, decompensated cirrhosis and end stage liver disease (usually assumed to have many prior “stages”, making this a chronic disease)
  • We need to exclude these not only in the conceptset but also using a separate exclusion criterion, because not all of those are pre-coordinated, and some concepts are explicitly not necessarily acute.
  • We need to decide if we want to consider the cause, indicated in many concepts. If we do drug outcome research, an “acute liver failure due to pregnancy” should not be included. If however we are looking at outcomes of the treatment of AFL we certainly should.
  • We need to figure out if we have a specificity or sensitivity problem or both, which we could then specifically address.

These could be combined in any variation, but since we need to have something concrete:

  1. Basic (target and characterization flavor): index criterion only.
  2. Broad and precise onset (outcome flavor): new hepatic injury, with hepatic failure OR hepatic encephalopathy within 3 month, and no chronic liver disease any time before
  3. No cause (drug outcome flavor): same as 2 MINUS failure by cause

If you haven’t already decided, I’d suggest that the initial development of this phenotype focus on acute liver failure solely (as opposed to acute on chronic or chronic). There certainly may be instances when you’d want to identify a cohort of patients with chronic liver failure, particularly if one was investigating how a drug may be tolerated in that population. But from a postmarket safety surveillance perspective, “acute” is the operative word. Drug-induced liver injury/failure is a diagnosis of exclusion, so I concur that the many causes listed above also NOT be present. Now I need to figure out how to look at the cohort diagnostics… :slight_smile:

Hi @JudyRac thank you for engaging on this thread! Agree with you on the focus on acute hepatic failure (as opposed to acute on chronic or chronic) - a phenotype that is more applicable for the use case of post marketing surveillance.

We have several candidate cohort definitions that have been developed (but not evaluated) and are available here https://data.ohdsi.org/PhenotypePhebruary2023_p6_AcuteLiverFailure/

In the interest of resurrecting this thread - we can set up a workgroup call this week. In this call - we can have a group discussion on this phenotype and get it to completion. What do you all think @Christian_Reich, @Patrick_Ryan @JudyRac

Stop stop stop. Where and when did we declare that phenotypes are for that purpose? Did I miss something? I am not saying we cannot do that, but then we have to say it! Somewhere. The phenotype library is for defining outcomes of drugs.

I keep saying and whining and complaining and pledging that the purpose of the phenotype (the flavor, the use case, the type of cohort) has to be part of it. Explicitly. Right now, we are not doing that. We are creating phenotypes that just state a condition. Like “Acute hepatic failure” or “Anaphylaxis”.

I tried to do that. I made a basic one, which could be useful as a target phenotype (or part of the target that defines the “clinical setting”, something we need to define as well). It is based on the diagnosis of “acute hepatic failure”, but we know that 50% of them are acute on chronic. But, as @JudyRac said, that could be a good target to study drugs, or transplantations, or therapies, or natural history, or whatever.

And I made one that is explicitly acute and excludes the chronic, plus I fixed the index, because you need a precise index for an outcome. It says all that in the name.

And I made another one that is acute and not chronic, and excludes those acute failures that are due to some cause, because those cannot be the outcome of another cause. But I don’t believe that one is much of use, because those causes are not captured.

Speaking off: Now we have a tone of overlapping and not well defined phenotypes, We need to reduce them. Can you, @Gowtham_Rao?

The clinical concept of Acute hepatic failure can be defined as ‘new onset of Acute hepatic failure for the first time in person’s with no known liver disease’. This clinical concept is what is surveilled for in safety surveillance. By extension, persons with chronic liver failure or history of liver disease are not eligible to have acute hepatic failure. These assertions maybe made before looking at Atlas or cohort diagnostics. ie this assertion is independent of Cohort definition logic.

Now we are modeling acute hepatic failure. The presence of evidence that there are person’s with chronic liver disease is a sign of specificity errors. Similarly, cohort diagnostics may show signs of sensitivity errors and index date misclassification errors. Ie we are thinking in terms of operating characteristics of 1 or more Cohort definitions.

Operating characteristics are attributes of the combination of Cohort definition * data source. A study may pick a data source * cohort definition combination based on what operating characteristic is most important to it.

The ability to choose cohort definitions based on operating characteristic that is important to a study is what helps select between various overlapping cohort definitions.


I am confused. Are you agreeing with me or not?

  1. Are we only considering outcome phenotypes for drug surveillance?
  2. Is the purpose of a phenotype part of the definition?
  3. Is sensitivity, specificity and index date optimization part of the definition?
  4. Will these things be explicitly mentioned in a structured form? Or as part of a lengthy “clinical description”, where the library user has to figure out what the use case must be, given the textbook prose?

Obviously, as a general statement that is wrong. In 50% of the cases clinicians don’t see it that way. What you want to say is “when studying ALF as an outcome of drug exposure, you want to exclude any other cause, and in this case you want to exclude even a predisposition.”

Even the guideline talks about acute-on-chronic as something to consider. And a drug can set off an acute liver failure in a chronic patient no problem. Probably the same mechanisms. In fact, we could run a study about that!!! And see if there is a difference.

:slight_smile: Since the success of this thread is to develop and evaluate Acute Hepatic Failure. There are currently 9 cohort definitions that are attempting to model Acute Hepatic Failure waiting for an evaluation, followed by peer review (@Christian_Reich @Patrick_Ryan )

But this is an important debate. Let us move the debate to here . I will respond there