Good drugs can be held responsible for bad things happening to the liver if sponsors do not fully explore alternate explanations for liver toxicities that turn up in a clinical trial, cautions John Senior, associate director for science in CDER's Office of Surveillance and Epidemiology.

"If you want to get your drug off the hook, find another cause and prove it," Senior stressed. "If you don't exclude other possible causes, we may have to assume it was likely" drug-induced liver damage.

Senior, along with Mark Avigan, who is both associate director for Critical Path initiatives in the Office of Surveillance and Epidemiology and acting director of the Division of Pharmacovigilance I, and CDER Deputy Center Director for Clinical Science Robert Temple, discussed FDA's year-old ¹ guidance for industry on drug-induced liver injury during a July 27 Drug Information Association webinar.

Liver toxicity is the single most common cause for market withdrawal and a major reason for limitations on drug use, but serious toxicities - the kind that cause hospitalization, transplant or death - rarely turn up until after a drug is on the market.

Only the most overtly toxic drugs will come to light among the 1,000 to 3,000 subjects that typically populate the safety database submitted in an NDA. And even drugs that have been withdrawn for DILI have caused death or liver transplant in only one or fewer of every 10,000 people. Those figures point to the enormity of the challenge in extrapolating signals picked up in a drug trial to their meaning for real-world livers.

Predicting those post-market serious adverse events has become an increasingly important cause of non-approval, as the agency has learned to recognize signals in an NDA database, Temple said, adding that further improvement in "pre-market discovery" is the thrust of the 2009 guidance.

Toxic drugs with major benefits can still be approved, and drugs that are associated with liver injury that is rarely serious - like statins or the Alzheimer's drug tacrine ( Cognex , approved in 1987, which fell out of favor because of its adverse event profile, which includes hepatotoxicity) - can and will still be considered, Temple pointed out. But interpreting the data is key.

When they're reviewing an NDA, regulators need "more than just chemistries to define [the] severity" of toxicities, Senior stressed. A differential medical diagnosis is crucial to understanding "what's going on in the patient," he said. "The sad fact of the matter is [that] you can't make a diagnosis of drug-induced liver injury; you are left with it after you've excluded all other possible causes."

For that reason, Senior urged sponsors to have the narratives that are submitted with their NDAs prepared only by physicians. "Too many of the narratives we get are just inadequate," he said. "They're recitations of the clinical report forms."

Seeking Feedback On The New Guidance

The regulators, who largely shaped the guidance, were participants in the DIA webinar as much to solicit comments on how to improve the blueprint as they were to explain it as published.

Most of those comments likely will focus on clarifying the proposed rules for stopping treatment. The online audience appeared most anxious to talk about those suggestions.

The rules are based on a consensus formed from ideas that in some sense are contradictory, Avigan admitted. "We came to our best conclusion, but without an absolute precise way of doing it."

Regulators had to reconcile the need to protect and study patient safety with the idea that most patients who have isolated transaminase elevations (raised liver enzyme levels that indicate cell damage) "adapt and actually self-correct," Avigan said. Unfortunately, "we don't have biomarkers to indicate which patients will get better and which ones will have progressive injury."

"So, these stipulations, these numbers, are somewhat arbitrary. They're based upon a kind of reconciliation of these two counterpoised ideas," Avigan said.

Indeed, "that's why we're having this discussion," Senior added. "We would like to know how you think we should change [the guidance]."

Senior admitted that he had argued for more lenient rules that would allow patients to be dosed longer in order to learn more about the drug. "I was a minority dissenter," he said. "I'm not really worried about transaminase elevations unless they lead to something more worrisome, such as a loss of function."

"I think this is an open issue, and I think we do want to hear how this works in the world that you live in, where you are doing the clinical trials," he challenged.

One of the guidance's objectives is also to learn from the evaluation of drug-induced liver injury episodes. "We want to maximize what we can learn about DILI" in order to identify factors that influence DILI susceptibility, Avigan noted during his presentation.

Hy's Law Is An Alert To Find Out Why

The gold standard predictor of potentially dangerous liver injury is Hy's Law, named for researcher Hyman Zimmerman, who observed in 1978 that a patient with both increased liver enzymes (indicative of cellular injury) and jaundice (meaning the liver is so damaged it is not secreting bilirubin) had as much as a 50 percent chance of death. Today that dour prediction includes transplantation.

Hy's law was defined by FDA in 2000 to mean, briefly, that liver enzymes are elevated to three times or more the upper limit of normal combined with bilirubin levels at least twice ULN. If those conditions are associated with an investigational drug, there will likely be drug-induced liver injury in the post-market population, the agency reasons.

But it is not a perfect regulatory measure, because Hy's Law cases are often not related to the drug in question. "If you're looking at a population of diabetics who are elderly obese people, there's an awful lot of other stuff going on ... that has to be excluded," Senior said. Finding a case of Hy's Law is only the beginning, he argued. "It's an alert that says you have to find out why."

"Remember, a genuine, absolutely clear Hy's law case or two is enough to turn down a drug" that isn't in a high-risk category, Temple stressed. "It's important to follow these people up so the drug doesn't get a reputation that it doesn't deserve."

Among the alternative causes of toxicity to look for are viral hepatitis A, B, C or even E, a strain formerly associated with Asia and Africa that has now been seen in Europe, Senior said. Also on the list are blockages, fatty liver disease, autoimmune or alcoholic hepatitis, biliary tract disorders, concomitant medications and liver injury related to heart failure or shock.

Additionally, a drug can be well-tolerated by most people but dangerous to some, perhaps the trickiest part of predicting for post-market populations. So, those idiosyncratic reactions deserve extra scrutiny, Avigan pointed out.

If there is a liver signal in the analysis submitted with an NDA, "it's really important to exclude all those other causes when you tally up cases. If you don't exclude them, you get a very muddy water, and it becomes very difficult for FDA to sort through what are the true cases of interest," Avigan said.

Stopping Rules Aimed At Striking A Balance

The guidance on evaluation of DILI episodes is there both to guard against runaway hepatotoxicity and to maximize what can be learned about drug-induced liver injury, Avignon said. Managing the risk for serious outcomes in individual study subjects based on predictive considerations is part of that balance.

The regulators also stressed that trial design depends on the indication. "You are prepared to tolerate with drugs for oncologic treatments, toxicity that you wouldn't tolerate in your cold pill," Avigan said. "You have to think it through."

For example, because transient enzyme fluctuations are common, and progression to severe DILI or acute liver failure is uncommon, automatic discontinuation of a trial drug based on a finding of liver enzymes three times ULN "may be unnecessary," according to the guidance.

In fact, the first of four criteria to be considered before discontinuing a drug is whether liver enzymes are at levels greater than eight times ULN. "The 3x number is contaminated by a lot of non-drug causes," Avigan said, so "we felt relatively comfortable elevating the bar from the 3x number."

However, Avigan was quick to point out the second recommendation, that a drug be stopped after enzymes are elevated to five times ULN for more than two weeks. "We didn't want to miss an opportunity to intervene at a point before injury is accelerated and perhaps in very rare cases irretrievable," he said.

Asked why a single observation of eight times ULN is grounds to stop the drug while levels can go on for two weeks at five times ULN, Avigan agreed it might be prudent to get another value the next day after the 8x result to eliminate testing errors before stopping the drug. But, he cautioned, the rule is based on the idea that "if you caused that much enzyme to leak out of the cells, you may be doing a lot of damage, and the patient is your first priority."

The third rule is enzymes greater than three times ULN with total bilirubin greater than twice ULN or international normalized ratio (which measures the liver's ability to produce coagulating factors) of greater than 1.5. And the fourth is enzyme levels greater than three times ULN with the appearance of fatigue, nausea, vomiting, right upper quadrant pain or tenderness, fever, rash, and/or eosinophilia (the level of reactive eosinophils in the blood greater than 5 percent).

- Shirley Haley ( ² s.haley@elsevier.com )

FDA Open To Revising Guidance On Drug-Induced Liver Injury

More from Archive

More from Pink Sheet