Phase O trials and the use of tiny drug doses are gaining acceptance and winning converts.
Dec. 17, 2007 | 2006, Merck created a new department dedicated to experimental medicine. This is not a new idea - not even at Merck, where this approach has been used for several years - but the priority is. Within one year, Merck went from zero to more than 50 people dedicated to experimental medicine, and the department is widely credited with accelerating the approval in 2006 of Januvia, a promising anti-diabetes drug. "We want the best scientists living and breathing this 24 hours a day," says Gary Herman, head of Merck's department of experimental medicine.
Herman defines experimental medicine as "small clinical trials in a limited number of subjects or patients - who are highly controlled - to get an assessment of the pharmacological activity for compounds. The group is committed to creating innovative ways to establish proof-of-concept for new mechanisms by exploring clinical models that could yield better, faster, and more cost-effective ways to prioritize drug development."
Biotechnology and pharmaceutical experts have a number of pet names for what to call a small-sample size study on humans before traditional clinical trials: exploratory investigational new drug (IND), Phase 0 (see sidebar: "When, Why and How"), pre-Phase I. But they all mean pretty much the same thing. "It's probably best to use the term exploratory IND," says Chris Elicone, senior marketing manager at Applied Biosystems. "At least the FDA has provided some guidance there."
Elicone refers to the FDA's "Guidance for Industry, Investigators, and Reviewers: Exploratory IND Studies," which underscores that these studies have "no therapeutic or diagnostic intent." Instead, exploratory INDs can be used, for example, to compare similar compounds to select the preferred one to move ahead. To gather pharmacokinetic data in an exploratory IND, scientists can use microdosing, which involves one compound tested through a single dose - which must be the lower of two options: 100 micrograms or 1/100th of the pharmacological dose based on animal studies.
By any name, though, exploratory INDs aim to push the more promising drugs into clinical trials and increase the odds of success. Negotiating that path, though, depends on solving many informatics and IT challenges.
Getting to Phase I
From an informatics perspective, the crucial question is: How do companies use data from an exploratory IND to decide if a drug will move ahead?
"Multiple factors go into that decision," says Mark Schmidt, director and clinical expert for CNS at Johnson & Johnson Pharmaceutical Research and Development (J&JPRD). "The largest factor is the risk-benefit ratio." A little more risk can be accepted if a compound potentially fills an unmet medical need. Likewise, the strength of the complete preclinical package comes into the decision-making process regarding moving ahead a compound.
Anastasia Christianson, senior director of discovery medicine informatics at AstraZeneca Pharmaceuticals, agrees that results from a short, small study seldom determine whether a compound moves into clinical trials. "We'll ask: How does this study relate to the studies done before and the biology of the disease and the unmet medical need?" Christianson explains. "The decision is based on all of the information that we've collected." She adds, "It's a big information-management and -utilization problem, because we collect more and more information along the way."
When Merck applied an experimental-medicine step to Januvia (sitagliptin), its recently approved DDP-4 inhibitor for Type II diabetes, the researchers developed a crucial single-dose study in patients. "Then, we modeled various concentrations of this diabetes drug based on the response of biomarkers, and this let us narrow the dose range to study in later stages," says Herman. "It cut at least a year off our timeline." In fact, Herman says the predicted dose from the early Phase I study ended up being the clinically approved dose. In all, the drug spent less than four years in clinical trials.
Dealing with the Data
"The data analysis tools for these studies are similar to the ones that we apply in clinical trials," says James Bolognese, senior director of clinical biostatistics at Merck. Common approaches used at Merck rely on SAS software and some tools developed with S+. Moreover, Merck's genetic-profiling researchers use network-analysis software. Bolognese adds, "People in our Rosetta [Inpharmatics] group are experts in the systems biology approach to analyzing problems."
Merck scientists are also working on adding some Bayesian techniques. "In conventional analysis," explains Bolognese, "you compute the probability of false positives and negatives based on the data in a trial. With Bayesian techniques, we incorporate prior information - information that existed before the trial." That previous knowledge might come from animal studies or published data in the literature. "This can tighten the predictive ability of data gathered in a trial," says Bolognese.
At AstraZeneca, Christianson notes, "One of the biggest challenges is making sure that you have the information integrated in the right way to use it. There is a technical component, a policy component, and a human component." On the technical end, Christianson points out that data must be stored where users have easy access and can integrate data from various experiments. "The policy component is around what you can and cannot do ethically, such as what kind of analysis can be done," she says. "On the human side, the analysis needs to be interactive, because you have people with different areas of expertise doing the analysis together."
Christianson and her colleagues take on these challenges with some in-house products and other off-the-shelf ones, such as SAS and Spotfire. "We also might take something off-the-shelf and modify it," she says.
Schmidt at J&JPRD points out the need also to do pharmacokinetic and pharmacodynamic modeling. He discusses the need to compare results from exploratory IND with data from preclinical animal studies, such as ADME (absorption, distribution, metabolism, and excretion). But then he says, "The technical challenge is minor; the greater challenge is institutional."
"We need to reshape how people report data," says Schmidt, "and how that can be harmonized between different functions to make them compatible." For example, if studies on a new compound lead to 30,000 data points from proteomics and even more voxels of imaging data, "you can't do a correlation analysis on that," says Schmidt. "Comparing those data requires some judgment, and that requires IT bringing different experts together."
To help IT solve such problems requires strong communication. "When working with IT internally or externally," says Schmidt, "the main thing that I see is them wanting to know more about the nature of the clinical phenomenon. For me, I need to know how they will use the various databases, how defining a field parameter matters - things like that."
In microdosing, scientists use such small amounts of the drug candidate that it can be difficult to measure levels in patient samples. With accelerator mass spectrometry, for example, Colin Garner, CEO of Xceleron, says that compounds can be detected at levels as low as 1 attogram, or 1x10-18 grams. This technique arose from carbon dating, so some computation takes place to turn data into pharmaceutical terms. For pharmacokinetic analysis, Garner says, "We typically use the Pharsight product, WinNonlin."
To detect very low levels of a compound in a sample, scientists can also turn to liquid chromatography (LC), followed by two sequential stages of mass spectrometry (MS). "From the LC/MS/MS perspective," says AB's Elicone, "several pharmaceutical companies have reported success in proof of concept using the API 4000 system in exploratory INDs."
Another company, he says, purchased an API 5000 system (also an LC/MS/MS instrument) just for microdosing studies. "The LC/MS/MS approach benefits pharma," says Elicone, "because they already have the technology in-house, and they are very familiar with the operational workflow." Such techniques can detect compounds at levels as low as a few picograms/milliliter, which is sufficient sensitivity for microdose studies.
Firms such as Xceleron work with biopharma customers on a consulting basis. "We assist companies in the candidate-selection process," says Garner. "Microdosing can give them a very early read on the human metabolism of a compound, which is a crucial parameter for a new drug."
In some cases, a microdosing approach can also reduce the time and cost of getting a compound to first-in-human studies. According to "Microdosing in Translational Medicine: Pros and Cons," a 2006 Insight Pharma Report authored by Hermann Mucke, conventional approaches require 12-18 months and $1.5-3 million to get a compound into humans; microdosing takes just 5-8 months and $300-500 thousand.
Expansion of Exploratory INDs
No one knows how many companies have used exploratory INDs or on how many drugs. Mucke says he was unable to find any hard figures for the Insight Pharma report. He estimates that several dozen companies have used the approach, although "most companies who do microdosing decide not to announce it for some reason."
One case Mucke did cite in the Insight Pharma report was at Millennium Pharmaceuticals, where scientists used LC/MS/MS "to evaluate dose proportionality under microdosing conditions for two established drugs (fluconazole and tolbutamide, chosen because of their similar pharmacokinetics characteristics in rats and humans) and an investigational Millennium compound identified as MLNX in rats." Another example is CRO Radiant Research, which used microdosing to develop a new pharmacokinetic profile for AZT (azidothymidine), the anti-HIV drug. Mucke also reported that GlaxoSmithKline "has a traditional involvement in microdosing that stems from its role as one of the cofounders of CBAMS, the company that later became Xceleron."
Perhaps most important, companies are learning when to turn to exploratory INDs. "It's not for all situations, and not for everybody," says Mucke. "It doesn't really save time overall, because you must still do Phase I."
AB's Elicone sees companies becoming more aware of what exploratory INDs can really do. "A first, it seemed like everyone was talking about doing exploratory INDs with all compounds that met preclinical criteria," he says. Now, he says the trend is that "pharmaceutical companies are becoming much smarter about when to apply it." (See sidebar: "Not for Everyone...Yet")
Some companies, though, are forging ahead in a broad way. Although Merck put much of it's first concentrated experimental-medicine effort into diabetes, that's just a start. "We want to use experiment medicine wherever possible," says Herman. He points out that Merck is working on platforms for oncology and neurological areas, putting a premium on imaging technologies (and the IT to handle them).
Even if companies can't agree on everything about this approach, most see the value of exploratory INDs. It could cut costs, get promising drugs into clinical trials faster, reject doomed drugs sooner, and invigorate the pharmaceutical pipeline once more.
Subscribe to Bio-IT World magazine.