Skip to main content
Premium Trial:

Request an Annual Quote

Q&A: Pfizer Researcher on the PSTC and Applying Bradford Hill's Principles to Biomarker Development


By Adam Bonislawski

Name: Josef Ozer
Position: Senior Principal Scientist, Pharmacokinetics, Dynamics, Metabolism at Pfizer
Background: Research Fellow, Safety Assessment at Merck; Assistant Professor at the Boston University School of Medicine

Josef Ozer is a senior principal scientist at Pfizer working in the company's pharmacokinetics, dynamics, and metabolism division. Previously, he was a research fellow in Merck's department of safety assessment.

At both companies he served as part of the Predictive Safety Testing Consortium, a public-private partnership launched by the Critical Path Institute in 2006 that includes 16 global pharmaceutical companies. He contributed to the consortium's liver biomarker work as well as its renal biomarker regulatory submission to the US Food & Drug Administration and the European Medicines Agency, coauthoring papers detailing the effort that were published in the May 2010 issue of Nature Biotechnology (PM 05/14/2010).

This August, Clinical Pharmacology and Therapeutics published a paper Ozer coauthored on the use of Bradford Hill's Principles of Causality Association in biomarker research.

Ozer spoke to ProteoMonitor about the Bradford Hill paper, how it was informed by his PSTC work, and what insights it might hold for biomarker research going forward.

The following is an edited version of the interview, which reflects Ozer’s personal views and not the positions of the companies or organizations discussed.

To what extent was the Bradford Hill paper a response to issues raised by the PSTC's FDA submission of biomarkers for drug-induced nephrotoxicity earlier this year?

The paper built upon the nephrotoxicity regulatory submission with EMEA and FDA, but in addition, the manuscript also looked towards new hepatotox markers that are currently being qualified. Many of the scientists on this paper were involved in both directives, the hepatotox and the renaltox, and we decided to come together as a group and put this concept paper together to look more closely at the causal associations and biomarkers.

The hepatotoxicity biomarkers are also part of the PSTC initiative?

That's correct. It still is and has been under the PSTC directive. Pfizer and Merck and AstraZeneca are companies that are making substantial efforts here in collaboration with other members. There's a huge dataset to analyze, all with companion histopathology anchoring in rat. [It's] actually much larger than the renal dataset, maybe two to three times the number of studies. So the team is still going through the process of data analysis and statistics and considerable data modeling. They're well along toward making an [FDA] submission, perhaps in 2011, but regulatory evaluation would follow on an independent timeline.

One issue with the PSTC's kidney biomarker submission was that FDA felt the histopathology readings should be done as blind evaluations, which the PSTC hadn't done. Has that issue been resolved with regard to the hepatotoxicity study?

Not yet to the best of my knowledge. It's an area of considerable discussion amongst the participants. Part of the timeline for the liver submission has been a focus on this issue. There are certain participants, including some FDA participants that felt very strongly that the first pathology readings should be blinded, which means that controls and the treated groups should be randomized and then you would perform pathology readings blinded in this way. When you look to the STP and leadership among members of the pathology community, their guidelines recommend controls be read separately and then treatment groups are randomized, to better determine a baseline control. Because we use different strains, different ages, and different sexes in our preclinical models your baseline could vary and some pathologists prefer to have that baseline before they can really say definitively that there's a treatment related effect. Now, in terms of severity rating, this approach still blinds the next stage to do the severity rating. But they (STP) do not recommend blinding in that first round. So this has been an issue that has produced a spirited debate. It's still an issue needing consensus and resolution.

Does it make sense to put forth a submission on the liver biomarkers before this question is resolved?

My view is that industry can submit the dataset and then the scientific package will wait for a regulatory decision on retrospective studies from the submission. I think it's always better for the science to move forward by packaging the available data for review, and decision might be, 'Well, from here on out we're going to do the pathology [the FDA's way], but the VXDS submission has taken nearly four years of effort (thus far), so the studies were performed using the best available technology and approaches from the participants.' I would think and hope that would be acceptable for peer review. I would also hope FDA would agree. This decision and approach was used with the renal submission. This issue probably isn't going to make or break whether the biomarkers are successful, but may affect the thresholds for decision making. The markers should ideally perform or not perform regardless of the pathology standard used, but others may not share this view. If the markers are affected by the standard, then they probably are not as robust as we seek for maximum impact.

What did you learn during the PSTC's kidney biomarker submission that you think has relevance to the liver biomarker work?

I think the big success was the impact of getting the publications out in press and getting the science in a very general public forum with the Nature Biotechnology [papers]. I think that had a high impact and influence on the liver group – that you could spend all that time and effort and that in addition to getting regulatory approval for use pre-clinically, you could also get some high profile papers out and impact the science world more broadly. I was always an early advocate in the group to publish as you go, but there are differing views on that topic.

[ pagebreak ]

One issue you raise in the Bradford Hill paper is the need for prospective studies to demonstrate the usefulness of these kidney and liver safety biomarkers in humans. How might those studies be done?

In humans – you don't have histopathology readily available as a benchmark. This was an issue well described in the [Bradford Hill] article. When you get to the point where you go into humans, how does one demonstrate that these markers add value when you don't really have histopathology or biopsies available? Biopsies aren't typically available to qualify a biomarker in people. How does one experimentally indicate that your markers are really adding value when you don't have the benchmark of pathology? It's not a simple question to address. PSTC and others are planning to design studies in human to assess renal biomarkers, but they are being reviewed by stakeholders. PSTC is going to measure the biomarkers relative to serum creatinine in that situation.

The other option is you can go into a patient pool [renal impact] in a hospital setting. This is going on, too, in academic centers. There are several centers that are collecting urine and even biopsy samples from patients that come into the clinic presenting with acute kidney injury and creatinine changes or diabetic pools, and they're collecting urine and measuring these VXDS biomarkers. So they're looking at these disease pools of patients and saying, 'OK, do these markers add value over creatinine?' And clearly indications from several recent publications show there is value there.

It's a little bit hard to extrapolate from a disease pool back to a normal patient population in therapeutic trials [phase I]. It's addressing different issues. I think pharma is equally interested in looking at disease pools, because if we have therapeutics that improve diabetic nephropathy, you certainly have to look into a disease pool. You can't look in a normal patient pool with a treatment related biomarker response to tell me how a diabetic patient biomarker will similarly improve with therapeutic treatment. Pfizer and other companies are very interested in extending safety biomarkers to pharmacodynamic endpoints. We have datasets of representative preclinical models with some of these types of endpoints. So can we take these biomarkers and utilize them for efficacy endpoints, but it is still early days. Of course, the tricky thing is, if you're using safety markers and applying to efficacy environments, you probably need to have some multiplexes of markers and representative controls just to say, 'This makes sense, the data is consistent.' [That] is really what the Bradford Hill paper is all about. It's about taking the reported knowledge and making really sound conclusions based upon all the hypotheses about how things work biologically and then having a well thought out strategy map about what the information means. Eventually, the effort could be available for peer review publication or FDA purview or regulatory review. If you don't have this information organized in a manner that's systematic across companies, across projects, then the information management is not optimized for product development success. I think the paper is trying to bring some standardization in approach. It doesn't mean we all have to use the same format, but at least there should be some consistency among the ways we evaluate markers within a field in terms of their value.

Does the potential suitability of safety biomarkers for efficacy purposes provide an extra incentive for pharmaceutical companies to participate in consortia like the PSTC?

I think so. Absolutely. At Pfizer we're starting to link the knowledge across the portfolio to say, 'Oh, you can use this [biomarker] for solving one problem.' We have all this knowledge about a marker and how to use it, and now we’re discovering new applications to apply to another problem. It's still early days, but there's no question that the safety biomarkers have already percolated into pharmaceutical development across areas where people were not as familiar with safety biomarkers. Now they’re like, 'Oh, I can use this marker for this other application? Great, let's use it.' It's a whole different environment for development. You take the samples, you outsource them, you get an answer back in two weeks, whereas before it was like, 'Oh, my, gosh, we have to spend a year figuring out what marker to use.' It's really a totally different game plan now. And I think the Bradford Hill paper is intriguing because what I think we advocate is the knowledge dissemination and sharing so that these opportunities can manifest themselves to biologists in early discovery applications.

You can imagine if we had a collective biomarker database, people could search a huge string of potential biomarker candidates and pull out interesting connections about their marker and their program that nobody envisioned during the data submission.

So, the idea behind the Bradford Hill paper is that if you're going to do this you need some way to codify the standards for evaluating these biomarkers.

Yes, and that hasn't yet been achieved, but I think the paper points to that possibility and advocates this approach. I think you're going to see a lot more sharing across academia and pharma, more agreements to get drugs to market, more teamwork. And you see it happening now – companies doing shared marketing agreements, shared research agreements. Everything is headed in that direction. The consortia concept will build beyond safety issues, and with it the concept of Bradford Hill could apply to any kind of biomarker. It doesn't have to be a safety biomarker. It could be efficacy biomarkers, predictive biomarkers. It has a lot of potential.