(© Carmen Segovia) 

Biomedical Information— A Matter of Trust A Q&A With Howard R. Asher

Howard R. Asher, president and chief executive officer of Global Life Sciences in San Diego, Calif., not only watched, but participated in, the evolution of information technology (IT). He started in product development at Pfizer, then Baxter and Bayer before founding a series of his own companies—now doing so for 30 years. During those years, Asher found that many technical advances depend on trust. Here, Worldview talks to Asher about trust in biomedical IT and how it might be enhanced. This is an edited excerpt of that interview.

What sparked your interest in IT?

» My Interest In It, essentially, occurred in the early eighties. In 1978, I founded Advanced Bioresearch Associates, ABA. We were helping a number of different types of companies with a number of devices. One was the first human artificial heart, which came out of Stanford. That artificial heart was actually in machine code. So I had to figure out how to comfort the FDA with software and hardware as it related to an artificial heart, which is a pretty significant product as trust goes. So it really kind of brought me out of the closet of IT and really got me into the concepts of things like software validation. In helping the FDA accept such technologies, I depended on a very simple word that's guided me forevermore, and that's trust. How do you trust the IT to do exactly what it claims it will, and—more important—that it wouldn't harm somebody or cause a problem?

What range of issues gets impacted by trust in today's computational information from biotechnology?

» If We Take A Global Perspective and look at the global biotechnology centers, they develop around medical universities that spill out information and technology and IP, intellectual property. So much of that IP is coming out of academic institutions that have used computational tools to characterize some of the IP. The issue there becomes interoperability if you will. As we take the intellectual property that has been developed by computational means, is it something that we—in the big sense of "general public"—can trust?

Then, as we take that computational process and we move it into, say, an industrial environment that may actually start applying other computational processes to that core IP, we eventually get to a point where that is going to be overviewed by a regulator. Of the 50 nations, we have 49 various views of: How do I trust the primary information, be it information that characterized a compound or information that was gathered in the nonclinical studies that proved safety or demonstrated efficacy, and how do I trust the IT and statistical assessments and all of the data management that has gone through not only the preclinical but including the clinical stages? From a regulator's perspective, we want to trust that those are well-engineered, they are validated systems, and they are trustworthy in all respects.

With so many stages where IT is involved, you get a pretty big chain reaction in the requirement of trust.

» That's Absolutely True. As I've been flying around the world and visiting with companies in different countries, the issue of trust makes me ask: When have we—as the general public—experienced this before? Many years ago, we'd take money to our branch of our bank and deposit our cash. They would put it in their vault, and if we needed to remove some of that money, we had to go back to that branch and that vault, from which the money was provided back to us. But now, we have the global ATM. The only IT that the general public—as a world public—will trust is the ATM. They do not trust the telephone IT. They don't trust the credit-card IT. They don't trust many, many other billing mechanisms and other IT stuff, but they do trust the ATM.

How can the technology behind a money machine improve biotechnology?

» Let's Go Through A Scenario of what this might look like in the future. What if right beside our money set our entire health record? So we have the automated teller machine, ATM, and next to that we have the automated telemedicine machine, the new ATM of our medical information from birth. What if we uploaded into our healthcare information our genotype and then—as we experience the environment of life ongoing—we could add our phenotype? Then, every dental X-ray, every medical record, every drug, everything we take is uploaded into our medical record. Then let's imagine that we have hundreds of millions of people around the world with their medical records uploaded into their "ATM." What then could theoretically occur is that—just like we can check a box to say we are an organ donor—we could say that we are a genetic-information donor. Then, we as an industry can benefit from real human genomic and therapeutic information related to disease types. We could structure and stratify that data. It would be real human data at the genetic level. We could then look at how different drug therapeutics affected different phenotypes and so on. We could look at biomarkers and start harvesting real information associated with real disease. All of a sudden, we would have a goldmine of information that we would trust more when applying it to therapeutic compounds.

Do you say 'trust more' in this case partly—or almost entirely—because the sample size would be big enough to be really trustworthy?

» Exactly. We would now be dealing with biostatistical significance that healthcare professionals must truly trust. Not only would we understand the therapeutic dynamics, but we would also understand the biomarker outliers.

If we’re at the molecular level and the cellular level, might we be able to harvest out the information of where something will therapeutically be effective, what is the mechanism of action, and where might we get a bad outcome or unanticipated effect that we're not wanting?

We hear a lot about the medical and pharmaceutical communities wanting to do things like personalized medicine or gene therapy, but doing this effectively seems to depend on—or even require— the kind of database of information that you are describing.

» That Is Really The Essence Of The Point. The biotech industry is really at an embryonic state with IT. Part of that is that when we are dealing with data that are at the cellular and genomic level, we very quickly start looking at petabytes of information and terabytes of data. One simplistic concept many times escapes us, and it is: We want to go from data to information and then to knowledge. What's happening is that the interpretation of volumes of data into truly meaningful and trustworthy information is not yet complete. We haven't quite figured out how to make sure that we have adequate data sample sizes to make an informational set of facts that we as a people can trust. Then, as we convert the information and set of facts that we can trust—based on the adequacy of data—then we start getting into the knowledge that we as an industry so desperately need. Vioxx and many other drugs serve as troubling examples where we have not known what we did not know when we put a drug or therapeutic to the market.
 
The current mechanism, the current model, does not work. We know that fact. We know that the animal data give us much misinformation. We know that some of the Phase I, Phase II clinical studies add to the misinformation, partially because the inclusion and exclusion criteria have become so stringent. Then all of a sudden, we're into a Phase III, not really getting the data that we really need, except at the organ level. We're still in the art form of medicine. We're not in the science of medicine.

We have to take some harsh looks at what is happening to drug development and therapeutic development, and it's not very pretty. We're seeing about a 99.9 percent failure rate. So for, say, every 10,000 compounds targeted, we’re seeing one really get through the process and get market approvals. That’s horrifying when we look at the economics.

Given those bad odds, how can biotechnology put your ATM analogy into operation to improve the situation?

» We Could Harvest the genomic and cellular data of humans, associate that with a disease, and look at the therapeutic potentials with computational predictive modeling. Then, we could determine in the modeling what biomarkers are expressed, where a potential therapeutic would be most effective in which population and equally important—if not more important—the populations we should avoid and actually contraindicate—meaning that if you have a specific biomarker, you should not have this drug because it will be a bad outcome.

Let'’s talk about the toxicological information that we gain when we go through our current paradigm. We are actually taking histology and toxicological information from animals. We are trying to associate that with humans. We are saying that the animal must be pure, a laboratory controlled animal—not anything reflective of a human, who would be eating all kinds of different diets, consuming all kinds of supplemental products, and environmentally they are exposed to everything under the sun. So right there, just look at the contrast between the human patient and that animal from which we're gaining toxicological knowledge. The information provided by that kind of data is just wrong.

It's hard to miss the international potential behind your idea of a medical ATM. If health information from around the world could be made available, a pharmaceutical company could access data from most anywhere. Likewise, this could bring better healthcare to developing countries if more information were available about medical histories of their people.

» Absolutely True. If we are going to build a successful therapeutic, it is nothing but naïve if we think that we only need to conduct our clinical assessments and our development within, let's say, the United States. That is very nearsighted. We have to be global.