Translational medicine can reduce failure rates for drug targets in proof-of-concept studies.
Orest Hurko, MD, and J. Lynn Rutkowski, PhD
Hurko is assistant vice president, translational research,
and Rutkowski is senior director, translational development
at Wyeth Research, Collegeville, Pa.
As science becomes more sophisticated, it often acquires depth at the expense of breadth. Both academic and industrial scientists run the risk of learning more and more about less and less. Nowhere else were the silos deeper and the cultures more divided than between clinical and bench discovery research, both in universities and in the pharmaceutical industry. In both settings, basic and clinical research coexisted side by side but interacted only indirectly. Over the last quarter century, however, the situation has begun to change. These boundaries have become increasingly blurred. In the last decade, dedicated efforts to bridge the remaining gaps between bench and bedside have been called translational medicine.
Important gaps between discovery and development continue to be a defining feature of the culture of the pharmaceutical industry. In discovery, the emphasis is on innovation, while in development it is on speed and process. Both are scientific endeavors, but the methods are very different and communication is often unidirectional and ineffective. These different cultures notwithstanding, this process has served the industry well for many decades. There was little need for translational medicine to bridge this gap. Most drug discovery and development involved making chemical improvements on existing compounds directed against a little less than 500 well-validated targets that had already proven their efficacy in the clinic [J. Drews, In Quest of Tomorrow’s Medicines, Springer, New York (1998)]. A slow, steady flow of development candidates could be handled efficiently in familiar and well-charted paths.
This balance was disrupted by the adoption of new technologies by discovery organizations. Combinatorial chemistry, high-throughput screening, and the availability of the human genome sequence had two major consequences. First, they dramatically increased the number of targets for discovery. [R. Archer, Nature Biotechnology, vol. 17, p. 834 (1999)] The number of candidates available for clinical development increased several fold in many companies. And this at a time when the unit costs of clinical trials had already strained the resources of development organizations! Second, these technologies opened the doors to novel targets never before pharmacologically tested in humans. No matter how well these targets responded in animal models, there was no longer any assurance they would be significant in humans.
Detailed studies of failure rates bear out these concerns. The break point in the pipeline has shifted to the transition between discovery and development. The highest rate of failure (50.4% between 1981 and 1992) is in phase II proof-of-concept studies [J. A. Dimasi, Clin. Pharmacol. Ther., vol. 69, pp. 297-307 (2001)]. This failure rate has doubled in the last decade. As might have been predicted, most of the new compounds that work very well in laboratory animals don’t work in humans. Cost-benefit analyses of existing development protocols no longer apply—at least not until we know that the drug works in people. In contrast, phase III failure rates have remained fairly static in the last decade. Reducing failure rates in phase II proof-of-concept studies is the primary mission of translational medicine.
Scope of translational medicine—the four key questions
The primary mission of translational medicine is to enable cost-effective determination of efficacy and safety through the use of biomarkers and experimental studies in
click the image to enlarge
Each biomarker should address one of four cardinal questions. The technolgies underlying these biomarkers range from imaging, to biochemical assays, to the more modern ‘omic technologies. (Source: Wyeth Research)
humans. For high-risk, novel targets it would be useful to have an earlier and cheaper indicator of human efficacy than is afforded by traditional phase IIa studies that use registrable clinical endpoints on the population intended as the market.
If a target has never before been pharmacologically proven in humans, we believe it may be cost-effective to look for signals of efficacy in humans in a preliminary study. In such a study, the test population, dosing, and endpoints are selected very narrowly to give the compound the best chance possible to demonstrate a positive effect. Only those compounds that demonstrate some efficacy in these admittedly artificial conditions would be permitted to advance to traditional phase II proof-of-concept studies using registrable clinical end points with a dosing regimen and patient population specified in a target product profile.
Criticisms of such an approach include concerns that such an additional study: will slow things down, will require filing of an extra investigational new drug (IND) submission, will add more cost, and will take eyes off the finish line.
These are legitimate concerns for some programs, but not for others. The higher the residual risk, the more cost-effective translational studies are. Delays can be eliminated or reduced by beginning translational activities such as biomarker discovery while the project is still in discovery phase. The adoption of exploratory INDs would increase the range of activities that can be cost-effectively undertaken in discovery phase. Costs will be reduced if negative results of translational study are agreed in advance to result in a no-go decision, and the translational study is less expensive and time-consuming than a traditional phase IIa study. Optimizing such a test requires answering the four key questions that define translational medicine.
(1) Which subjects are most likely to respond?
Are there clinical subgroups that will respond differentially to treatment? Some examples to date have been selection of patients for targeted chemotherapy [J. Pittman et al., Proc. Natl. Acad. Sci. USA, vol. 101, pp. 8431-8436 (2004)]. Examples include Her2/neu in breast cancer, EGFR mutations in non-small-cell lung cancer, and the Philadelphia chromosome in chronic myelogenous leukemia.
Marker positive patients show better responses to targeted chemotherapy than do patients with identical histological diagnoses but lacking these biomarkers. Direct measurements in biopsies of human tumors of the presence and/or activity of putative drug targets will extend the availability of biomarkers. These include histochemical, immunohistochemical, transcriptional profiling, and proteomic biomarkers. Although most successes have been in the definition of subgroups in cancer, there are limited data indicating that other clinical diagnoses include a number of definable subgroups with differential responses to medication. Defining these subgroups is a major responsibility of translational medicine.
(2) What is the optimal dosing regimen?
Ideally, dosing for initial tests of efficacy would be done intravenously to minimize confounding from absorption and first-pass metabolism. The optimal dosing regimen should be based on the duration and degree of receptor occupancy in the tissue compartment relevant for the disease. For targets never before tested in humans, receptor occupancy can be measured directly in dose-response animal efficacy models. Minimally invasive methods, such as ligand-displacement positron emission tomography (PET), are used to ensure that optimal levels of receptor occupancy are achieved in humans.
A target should not be dismissed from further consideration if the test compound failed to achieve sufficient receptor occupancy for a sufficient duration in the clinical trial. Rather, initial attention should turn to improving receptor occupancy. Direct measurements of receptor occupancy in humans are preferable to monitoring of blood levels or simple allometric scaling [R. Frank, R. Hargreaves, Nature Reviews Drug Discovery, vol. 2, pp. 566-580 (2003)]. This is particularly the case for drug with targets in the central nervous system or in tumors. However, PET ligands are available for only several dozen receptors. Development of novel PET ligands or other imaging biomarkers is an important translational activity and is becoming a major focus in many companies. Failing direct measurement, target occupancy by the drug can be assessed by measurement of downstream consequences through the discovery and development of pharmacodynamic biomarkers. Much work needs to be done to ensure that the pharmacodynamic effects are in a relevant tissue compartment and that their quantitative relationship to target occupancy and/or effects on the relevant pathophysiological pathway are understood.
(3) What are the earliest and most sensitive markers of altered pathophysiology?
Biomarkers detecting amelioration of a pathophysiological cascade may be more sensitive and less variable markers of efficacy than are clinical end points. For many diseases, clinical efficacy is only evident after prolonged treatment or observation. However, favorable biochemical alterations in the pathophysiologic cascade should begin concurrently with the onset of treatment. A biomarker that quantitates such biochemical changes would be evident long before clinical improvement is evident. Clinical outcomes are affected by many parameters unrelated to treatment. Detection of efficacy by clinical outcome requires a large sample size in order to detect the treatment signal by balancing out these other variables. Direct measurements of relevant biochemical changes can be less variable and thus detectable in smaller patient populations. Clinical outcomes are problematic when the desired outcome is disease modification rather than simple symptomatic relief. In principle, this distinction can be made by complicated and prolonged clinical trials that examine end points throughout the course of treatment. Independent measurements of altered pathophysiology would facilitate this distinction in shorter trials.
(4) Which patients are likely to experience adverse effects?
Ideally, patients at high risk for toxicity would be identified before dosing and excluded from initial tests of efficacy. An important translational activity is the collection of baseline, predosing blood samples for retrospective analysis by transcriptional profiling, proteomics, and other modalities in a search for such predictive biomarkers. Mechanism-based toxicity during the initial efficacy trial should be monitored by concurrent monitoring of pharmacokinetic measurements. Rare idiosyncratic toxicities are unlikely to be encountered in the limited populations used in early development. However, banking of DNA prior to treatment of larger populations may prove useful in retrospective surveys for vulnerable genotypes. Prospective banking would avoid the bias that would confound retrospective collections.
Indications and targets that benefit from translational medicine
Translational medicine does not have equal utility for all targets and indications. As a general rule, Those drug targets that have not yet been pharmacologically proven in humans, and thus without precedent in clinical development, are those most amenable to translational support because innovation and risk level are likely to be high. Also, indications for chronic diseases are more likely to benefit from efficacy biomarkers than those that offer a simple acute readout.
For example, an analgesic for acute pain or an oral contraceptive are best assessed by standard clinical measures. On the other hand, initial assessments for efficacy in chronic degenerative disorders such as osteoarthritis or Alzheimer’s disease would benefit from sensitive early biomarkers.
Future of translational medicine and R&D
We limited our discussion here to the current status of translational medicine. All of the studies discussed the discovery and use of biomarkers for internal decision-making, not as substitutes for clinically meaningful endpoints for registration studies.
Even so, such biomarkers and experimental studies have the potential of reducing costs by filtering out compounds that fail to show human efficacy in an idealized test setting. The US Food and Drug Administration recently issued a guidance document for exploratory INDs proposing a limited toxicology package before limited exposure of human subjects to pharmacological doses of novel compounds. This as well as earlier guidances for micro-dosing will increase the possibilities for translational activities while the compounds are still in discovery phase.
The Holy Grail of biomarker research is elevation to the status of true surrogates. Thus far this status has only been granted to a handful of biomarkers. For example, it is now possible to register drugs intended ultimately to reduce the incidence of cardiovascular disease and stroke by demonstrating a beneficial effect on the surrogate biomarkers of hypertension or cholesterol [J. A. Tobert, Nature Reviews Drug Discovery, vol. 2, pp. 517-526 (2003)]. This has had a dramatic beneficial impact not only on drug development but also on clinical practice. Surrogate status will only be achieved by many years, if not decades, of confirmatory research.
Such an effort will require research by multiple independent investigators both from industry and the academy. Progress would be enhanced greatly by sharing of information as well as of risks and costs, as is currently being done by the Alzheimer’s Disease Neuroimaging Initiative consortium. Industry should consider the relative risks and benefits of making surrogate research a precompetitive activity.
Filed Under: Drug Discovery