In a renewed quest for medical breakthroughs, the Canadian government recently committed $67.5-million for research into "personalized medicine." This bold initiative promises faster, smarter and cheaper drug development by uncovering new targets, identifying patients likely to benefit from experimental therapies, and using molecular markers of disease to predict drug response.
Yet, personalized medicine follows on decades of efforts at plying molecular genetics toward drug development. To be sure, these approaches have borne fruit. But the process of developing cures remains painfully slow, error prone and costly. Despite steady increases in industry and public investment in research, the number of new drugs approved each year continues to decline.
Personalized medicine may very well fix a science that seems to be failing drug development. But perhaps we should be asking how drug development is failing science. Medical innovation involves a peculiar mix of seemingly contradictory motivations. Scientists and sponsors are driven by the pursuit of knowledge and a desire to relieve human suffering. But they also seek fame and fortune. Medical journals want to foster progress as well, but they sell more subscriptions when they report breakthroughs.
With the right balance of incentives, these often parochial motivations can work together and propel the best science toward the clinic. But countless failures in drug development – and their burdens for patients and health-care systems – should prompt a hard look at whether we're striking that balance properly.
Consider the tension between truth and compassion. Research in psychology suggests that scientists, like the rest of us, often have difficulty distinguishing what they want to see and what they actually see.
Over the past several decades, medical researchers have built in numerous safeguards to prevent desire from shaping observations. For instance, researchers often intentionally keep themselves in the dark about which drug a patient is receiving in a trial so their evaluation of the patient isn't influenced by beliefs about the drug. Surprisingly, however, many simple procedures used to control bias in clinical trials have yet to permeate studies of new diagnostics or animal tests of new drugs, leaving them susceptible to inflated claims of medical promise.
Truth and fortune can also work at cross purposes. Publication is the main vehicle through which experimental observations are absorbed within the larger body of medical knowledge. But publication also undermines strategic advantage, because others can ride free off a researcher's hard-won advances.
In what seems a concession to the private interests of drug developers, we now tolerate a culture in which the results of animal experiments and clinical trials are published selectively, with important negative findings languishing deep inside filing cabinets. It's not surprising, then, that drugs that seem to work miracles in animals and in small drug trials choke when put to tests that can't be hidden from the public.
Then there's the reward structure of medical research, where, to paraphrase playwright David Mamet, "first prize is a Cadillac Eldorado, second prize is a set of steak knives, and third prize is you're fired." Journals, funders and the media reward researchers who report the most favourable – and first – results, rather than the (often later) accurate ones.
That explains why, for example, a vastly disproportionate number of cancer prognostic marker studies show statistically significant effects. Or why strong gene-disease links reported in initial publications often "disappear" in subsequent publications. Or why almost 10 per cent of published randomized trials testing targeted cancer drugs fail to say whether drug-related deaths occurred. Or why scientists at two major drug companies recently reported they were unable to reproduce animal experiments proclaiming the promise of new therapies.
How might these problems be fixed?
One place to begin would be guidelines and careful peer review of animal studies before human trials are initiated. Another would be the creation of public registries for studies of new diagnostics; medical centres should condition ethical approval of such investigations on registration in advance of study conduct. Funding agencies, universities and journals need to find better ways of rewarding scientists who report negative findings, who deposit data sets in public repositories, who synthesize evidence rather than create it, or who participate in peer review. And regulators such as Health Canada should use their authority to incentivize transparency and good research. They should demand more of animal studies intended to support human trials. Above all, they should compel drug companies to publicly report all trial results – even when drugs or diagnostics are never approved.
Physicians, patients, payers and public health programs depend on the research enterprise to supply a steady stream of medical evidence. The process of creating this social good, however, is driven by a mix of parochial interests. Personalized medicine – and other ways policy-makers are trying to prime medical innovation – will only deliver on its full potential if policies bring these motives into alignment with the goal of generating reliable and relevant medical evidence.
Jonathan Kimmelman is a genetics researcher and bioethicist at McGill University.