Imagine you're a 50-year-old male in good shape, with a normal weight, with decent blood pressure, but with a family history of cardiovascular disease. You are considering taking cardiovascular Drug X. You look at the clinical trial results and find that 50 percent of patients taking Drug X improved while only 25 percent improved with a placebo. You look at the patient segmentation (results by characteristic) to see what someone with your characteristics can expect. Unfortunately, the patient stratification isn't perfect, but you believe that 60 percent of patients who are similar to you improved. However, you remember that you take an aspirin a day, exercise vigorously, are Asian, and eat a fatty diet. The results don't cover these finer breakdowns.
One way to get to the goal of personalized medicine would be to increase the number of patients in each clinical trial by a factor of 10 or 100 or 1,000 and clearly specify every attribute of every patient. At some point, perhaps, there would be enough stratification so that you could view the results for someone like yourself. But the clinical trial costs would go through the roof and, with such high costs, fewer new drugs would be developed. There might not be enough qualified patients willing to enroll in these big clinical trials, which is a problem drug companies face today, even with smaller trials.
Even with larger clinical trials, two huge problems still remain: (1) Those clinical trial results apply to you (Asian, male, 50-years-old, etc.) only if you are like those tested. However, you may never know if some key attribute is different (e.g., triglyceride or creatinine levels or genetic differences or diet). (2) Even with such a fine level of segmentation, perhaps only 70 percent of the patients in your group improved. You still don't know if Drug X will help you or not. You could get lucky and end up in the 70-percent group or be unlucky and end up in the 30-percent group.
What's a patient (and physician) to do? Easy. Do what people have done since the beginning of time: Try taking some Drug X and see how it works for you. Did you have an adverse reaction? Did you achieve your therapeutic goal? Is Drug X well tolerated and convenient enough to make it worthwhile to take chronically? Is it affordable to you?
Did you notice something? You could have done the same thing without those clinical trial results that showed how 70 percent of patients benefited. What matters is whether Drug X works for you, not for some people with similar characteristics. If 99 percent benefit from Drug X but you don't, you'll stop taking it anyway. And if only 1 percent benefit, but you are in that 1 percent, you'll keep taking it. Those clinical trial results that you, your doctor, and the FDA reviewed were neither necessary nor sufficient for you to learn how Drug X works for you.
This is one more reason to consider shifting the FDA's approach away from pre-marketing gatekeeping to post-marketing surveillance--the Consumer Reports model presented by Alex Tabarrok.