Drug Experiment: FDA Eyes Simulated Studies
11 CommentsBy Ed Silverman // December 16th, 2008 // 7:09 am
The agency plans to use new computer technology to simulate how some drugs in development are supposed to work, in hopes of identifying safety and effectiveness issues before late-stage clinical trials are completed, The Wall Street Journal writes.
Entelos, the biotech that developed the technology, claims this will allow researchers to obtain computer-generated test results in a matter of days or weeks, compared with years required for most major clinical trials. Far more “simulated patients” also can be tested than in conventional human trials.
Under an agreement with the FDA, three drugs now being studied for heart-related conditions in large human trials will be tested by the simulation technology, the paper writes, although the drugs involved in the project are not being disclosed (here is the Entelos statement).
The agreement is part of a broader effort by the FDA to get a bead on safety and efficacy, and to identify characteristics of patients that lead some to benefit from a drug while others are at risk of a bad outcome. A particular aim of the project is to see if simulated studies might indicate if the drugs could be linked to heart attacks and other cardiovascular events in cases too rare to be picked up in conventional clinical trials, the paper writes.
“What this study is about is trying to anticipate bad scenarios before they occur,” Robert Powell, associate director in the office of translational sciences in the FDA’s Center for Drug Evaluation and Research, tells the Journal.
He adds that regulators “wouldn’t make a decision to kill a drug based on a simulation,” but the findings could be used in discussions with drugmakers to influence decisions such as the design of clinical trials. Eventually such info also could affect prescribing information included in drug labels.
Both the FDA and drugmakers have used computer simulation in drug development, but have typically been limited to analyzing data generated in actual clinical trials, he notes. This project is different because it will evaluate drug data by plugging it into a virtual model of disease.
The effort follows controversies over Vioxx and the Avandia diabetes pill that were linked to increased heart-attack risk long after they were approved for the market. Merck withdrew the painkiller in 2004, while Glaxo’s Avandia remains on the market, the paper points out.
Entelos says its technology, called the Cardiovascular PhysioLab, uses a mathematical model to simulate how cholesterol functions in the human body and how deposits of fatty material called plaque develop in the artery walls and become prone to rupture, the Journal writes. By running certain chemical characteristics of drugs through the model, the hope is to be able to predict whether the compounds might cause cardiovascular problems and, if so, in what types of patients.
The program is “a flight simulator” for drug testing, Entelos ceo Jim Karis tells the paper. “What we’ve done is created underlying technology to mathematically represent human biology, and then you can do things to it to see what happens.” And guess what? He believes the program, which was finished in 2007, could have detected the issues that ultimately doomed Vioxx.
The Country Doc
Computer simulations are still programmed by people with limited ability to predict unpredictable side effects. It would seem to be hubris to put too much stock in whatever a computer simulation concluded. While controlled studies are better than computer simulations they are still controlled as well. I’m afraid there will never be a replacement for real life empirical everyday usage and wait and see.
Salmon
Having been intimately involved in these simulations at FDA. It is my firm opinion that they are rigged to a degree especially as to safety. Bob Powell worked at Pfizer and Pharsight prior to coming to FDA. At Pharsight he was VP of strategic marketing and entered into agreements with FDA to use and evaluate these Pharsight programs. He then received a severence package and several months later was showing up at FDA internal meetings even though he wasn’t even on the FDA payroll. He then joined FDA to head up the group evaluating these programs before the 2 year statutory requirement was over. (Check SEC filings and his stated responses in FDA public meeting meetings to when he joined FDA.)
Wehn Bob Powell joined FDA he made statements to the effect that clinical pharmacology (to which he was only the associate director) would be turned into a modeling and simulation group.
When I and other reviewers could clearly see mutiple structural signals for hepatotoxicity, we were told that the Entelos program didn’t detect any. When I asked for any information at all about the Entelos program, and wanted to tell the company they should be aware of the signals and it would be better to address it early my comments were intentionally not sent (documented by emails) and I was set up for retaliation.
When I have questioned Bob Powell’s people about anything including why they simply simulate 4 points for the medians of each quartile and then always plot straight lines and why won’t they give me plots of raw data. I’ve been told that I should just be quiet and cut and paste whatever they tell me into my review like other reviewers do.
Bob Powell’s assistant has gone into reviewer’s offices and made implied threats that visa’s would be revoked.
I have overheard nonmedical reviewers working with Bob Powell telling medical reviewers that they are not allowed to write their medical opinion in reviews.
To justify their group Bob Powell and his helpers published papers of survey’s claiming how much they helped labeling decisions etc. Yet some of these drugs and indications were turned down when lethal toxicities were found in children and other reviewers found fundamental problems in by the modeling group under estimating drug exposures by nearly 90% (i.e. they were really 8 fold higher in children as compared to adults).
Salmon
Salmon
As drug disease state modeling Powell presented an example for a cancer drug in slides at a very small nonpublicized public meeting. Pharsight photographed the slides and incorporated it into a drug-disease model for cancer that they are now selling. However, similar data is now being redacted from published FDA reviews which any one can find in the future as they’re being called trade secrets even though FDA developed the model and did the analyses and thus should be public domain if it had been presented in the right way in the review.
Salmon
FDAer
In Spring 2006 an Office meeting of the Office of Clinical Pharmacology was held and a number of reviewers complained about the treatment they were receiving at the hands of Bob Powell’s pharmacometrics group. A short while later a reviewer stated in a division meeting (1 of 5 in the office) of about 2 dozen reviewers that Bob Powell stated that he wanted to be able to fire reviewers. By the following Fall every reviewer who had said anything at the Spring meeting was being harassed and several of them left FDA around that time.
As far as I know only 2 of the 7 reviewers who said anything at the Spring 2006 meeting are still at FDA. 1 has been fired, 1 forced to retire, 3 left FDA due to harassment. Of the remaining two still at FDA one is an expert in computer modeling and has been forced out of the group and mainly works from home and is terrified, and the other reviewer is being harassed and is also terrified.
Rich
This seems a bit like fantasy to me. I don’t think that we can claim to even understand how statins prevent heart disease (in light of the recent ezetimibe data). We have only hand-waving to suggest why Vioxx might cause heart disease. Sure, everybody has a theory, and any one of these theories, if correct, would suggest some change in how we treat disease.
The problem is that we don’t know which, if any, of our theories are correct. So, how can you model it?
I could see correllating biomarkers with drug risk profiles, and using software to perform the math. However, most pharma companies are already doing this - if some commercial product with decent accuracy suggested a drug was likely to cause problems it wouldn’t even make it to clinical trials. Nobody takes a $200M gamble on a molecule that was identified in a software screen before it was even synthesized. A smart company wouldn’t even waste the time testing it except as a proof-of-concept.
I can see this technology slowing down adoption of lifesaving medicines, and being used in defense of drugs that actually are harmful. Bad data is sometimes worse than no data.
However, the concept of testing out models and comparing them to real world results is sound.
An FDA Reviewer
I don’t know why Bob Powell is so interested in seeing if computer models can predict cardiovascular toxicity. Every single drug I’ve seen actually had cases during development that were simply dismissed.
Remember this is a guy who likely worked on the safety database for Paxil, which hid toxicities.
He’s also been working on the new computer safety database and reporting system for FDA.
As Bob Powell himself says, “The best way to predict the future is to create it.”
FDA Insider
Rich makes and excellent point. “I can see this technology slowing down adoption of lifesaving medicines, and being used in defense of drugs that actually are harmful. Bad data is sometimes worse than no data.”
Imagine a scenario where a drug under review has been placed in limbo (the approval is way past the PDUFA due date) because there’s a real concern that there’s cardiovascular toxicity.
Even if you know the mechanism you may not have appropriate data to model it. Suppose that there’s a toxic metabolite. What if the company didn’t provide the binding data on the metabolite. What if the company did the metabolism studies and didn’t report over 80% or 90% of the metabolites in blood like happened with Zyprexa. What if there’s insufficient information on drug interactions with drugs likely to be used in combination that induce the metabolite so you can’t predict how much will actually be formed under real life conditions. What if other drug interactions prevent the elimination of the toxic metabolite. What if other drugs like SSRI’s will increase the risk due to a synergistic mechanism. What if it’s going to be used in children but you don’t have good exposure data in children. What if underlying disease states like atherosclerosis in the elderly make them have toxicity in just over a year whereas children might take several years or more. What if there is a huge amount of marketing to children. What if the clinical studies only tended to study the drug for a few months so you don’t have good data. What if you could do cardiac ultrasound and get a better idea but didn’t. What if not just one but several metabolites out of 3 dozen or more cause it but the experiments haven’t been done. What if you don’t even know all the metabolites.
Even if other drugs in the class cause it by the same mechanism you would need all this information and even then the potency or dose etc. may be totally different or even clinical data for the cardiovascular toxicity hasn’t been collected yet in the relevant population.
I believe there is way too much opportunity for hanky panky.
Plus this is the same guy (ex-industry VP in Office of Translational Sciences) who Rep. Rosa DeLauro sent Von Eschenbach a letter last Nov having about undue industry influence in the Reagan Udall Drug Development and thus potentially undue influence on FDA science and approvals without adequate controls.
Defiant
Sounds like sour grapes are in the air. Anyone who has worked in any large organization, or small for that matter, knows animosities build for personal reasons. No one is liked by all, and often times someone is hated by everyone.
This is no reason for attacking someone, faceless and nameless on the internet. If you have a bone to pick with someone, and feel you have legitimate reasons to throw mud at them, then do it through work channels.
But, back to this article. Do you not think that Entelos has not tried this computer model on dozens of drugs that had been approved, then taken off the market? Starting from thalidomide down to Bextra, Vioxx, etc. Maybe this was addressed in another article, but regardless, how stupid would it be for Entelos to not use positive controls in their sales pitches? If someone knows they have not, I would like to know. Then we can question this beyond just reading a general article.
Salmon
Defiant,
Thank you so much for the heads up on Thalidomide.
Salmon
Defiant
Salmon,
bite me
Defiant
Salmon,
bite me