Stagnating pharma research, adverse events, need for tailored therapeutics, and how linked health data can help
In its December print edition, Wired magazine featured interesting insights about "Trials and Errors: Why Science Is Failing Us". The article describes how mistaken assumptions of causation hamper modern science. Wired author Jonah Lehrer argues that scientists think they understand how certain diseases are triggered and therefore how they can be cured, but that we are actually having trouble understanding a system as complex as the human body. As medical research is targeting increasingly complex diseases, more drug research projects that look promising in theory lead into a dead end. The discovery of a new compound today costs 100 times more than in 1950s (inflation adjusted); the cost per approved molecule are projected to reach $3.8bn in 2015. The article also provides several examples where procedures or FDA-approved drugs were later shown to offer little or no benefit, including MRIs for back pain (there seems to be little correlation between MRI-discoverable tissue problems and actual back pain); biomarkers that were supposed to indicate cardiovascular problems, infectious diseases and the genetic risk of cancer; and hormone replacement therapy to reduce the risk of heart attack in women.
Even if research produces an drug that is shown to be effective in clinical trials, that doesn't mean that the drug is working for all patients. At last year's Sage Congress, Food and Drug Administration (FDA)'s Senior Advisor Vicki Seyfert-Margolis discussed at length the importance of tailored therapeutics because many major drugs are ineffective for many patients. FDA statistics show that different anti-depressants don't work for 20-50% of patients, statins (cholesterol drugs) don't work for 30-70% of patients, and asthma drugs are ineffective for 40-70%. As a result, many patients with chronic conditions have to go through a period of trial and error until the right drug or combination of drugs can be established. During that time the disease will progress and the patient will likely suffer from side effects of the different drugs. And according to Seyfert-Margolis, about 100,000 (!) people die from adverse events each year in the US alone (making it the 6th leading cause of death). Reason enough to call for improvements in targeting treatments to individuals.
As we struggle in the discovery of new drugs and identifying effective drugs for individuals, a key to the solution are better and broader health data. The fact that smoking causes lung cancer was established based on vital registration data. Successful drugs from vaccines to statins were developed based on targeted research and clinical trials. However, research is targeting increasingly complex diseases and looking for more specific and targeted drugs and interventions. This requires more and more granular data. A lot of those data are already being collected. Health records, pharmacy records, emergency service records, records of vital events, censuses, surveys, clinical trials, and increasingly social health networks (i.e. social networks where individuals or patients share and discuss information about their own health or disease) compile very detailed data on personal health history, diagnoses, treatment plans, prescribed drugs, treatment results, and general health outcomes. Additionally, more patients record their experience with treatments and drugs in databases, social networks and personal health records. The potential for getting additional insight from all these sources for drugs and treatments is huge. Even detailed studies after the launch of a drug or devide (aftermarket study) could be based on those existing data. Tim Vaughan, Senior Scientist at social health network PatientsLikeMe, provides a fabulous example (presentation for download) for social health network research. When a rumor surfaced in their ALS community that Lithium may be slowing the progression of the disease, the PatientsLikeMe database contained enough ALS patients taking Lithium, controls, and data on their disease progression to determine that Lithium (unfortunately) does not slow down ALS. However, their results were published before an official clinical trial was even started.
However, most health data are currently scattered in various databases of governments, providers, payers, producers and others, and are often inaccessible for broader research. While those data allow unique insights for the respective data holders, the full potential of health data can only be realized when they are combined and linked. One option are Personal Health Records, which let patients collect all their health data in one spot. Ideally, they would then be able to offer up for research those linked data without direct personal identifiers (like name, address, social security number etc.). However, all data holders in the health data field have a responsibility to maximize the value of health data by sharing them for research (within the limits of patients' consent to such data sharing and reasonable protection of their privacy). More on the pros and cons of privacy vs. access here. With more data available, there are huge opportunities to inform research for new drugs, improve the targeting of drugs to individuals, increase prevention, and improve outcomes while reducing cost. It's a long road, but it's well worth taking.