EVENTS | VIEW CALENDAR
I’m not sure about this, but …
A thousand years ago, when I still worked in a laboratory, I had the pleasure of being the wet biochemist for two biophysicists, one of whom was so completely focused on mathematical gymnastics with Fourier transforms that he considered proteins merely an artifact of protein NMR—something to be tolerated.
One day, the two of us were talking in the lab—to this day, I don't know why that was allowed to happen—and he asked me what the average yield of a particular protein prep was. Very matter-of-factly, I told him the average yield was the total yield of all experiments, divided by the number of experiments performed to reach that total. He simply stared at me, unblinking for several minutes, while his brain tried to absorb what had to be the glibbest answer he'd ever heard in his life. And yet, I wasn't trying to be glib.
What the mathy-physicist could not grasp, because the "bio" in his title was something forced on him, was that unlike math—and to some extent, chemistry—biology is an art, not a science. It is not precise, but rather chaotic. And despite the best efforts of practitioners to make it fit the scientific method of, "do the same thing, you'll get the same result," anything that relies on a biological system is likely to give you a (hopefully only slightly) different answer every time you ask it the same question.
And as it was with my biophysicist boss, the basic imprecision is incredibly frustrating to biomedical researchers, physicians, patients and finance managers.
As I research clinical trials and preclinical research for my articles for Drug Discovery News or any of my other medical writing, I am always surprised at how much effort goes into the statistical analysis of a trial, as it seems any statistics that arise from a study become largely moot once the study is completed.
Now, please: This is not meant as a criticism or condemnation of clinical trial specialists or statisticians. These people do their damnedest to ensure that the trials are as fair and reflective of actual clinical practice as possible, but in an attempt to actually complete the study without too many confounding factors to accommodate, things get left out.
At the end of the day, you learn that in this study, 75 percent of patients on Drug X showed an efficacy endpoint, while only 53 percent of patients on placebo (or Drug Y) demonstrated that endpoint and that these findings were statistically relevant.
Congratulations, Drug X has been approved. Now let's put it into the real-world patient population—the equivalent of teaching your child to swim in a kiddie pool and then throwing her into the Colorado River during flood season.
Watch the rocks! We'll see you downstream … we hope.
Patient populations and patients are not the same thing. When you give an individual patient Drug X, he is unlikely to be 75-percent treated. He responds or he doesn't, or he inhabits that pharmacologic limbo somewhere between the two. Confounding factors such as comorbidities, compliance and lifestyle that were not part of the original trial very much complicate the real world.
The problem is, the patient expects to be better after taking the drug and comes out a little (or a lot) disappointed on the other end. He can't understand the imprecision of medicine because that's not how we've sold it.
Can we make medicine more precise? Yes. Personalized medicine is intended to add several layers of information into the treatment decision to make it more likely that a patient will perform as desired during treatment or perhaps more accurately, that we'll know how a patient will perform during treatment.
But personalized medicine is an incredible uphill battle, as many companies and organizations have seen over the last decade. Attempts to model human disease and drug response are fraught with knowledge gaps through which you could float a cruise ship, and many an organization has faded into the woodwork, exhausted from attempts to make these models practically useful. Sure, we get more data every day, but have you seen how big they're making cruise ships these days?
This is not an indictment of personalized medicine, or even a suggestion that it is a waste of time. I personally believe that to move in any other direction would be foolish and un-Hippocratic. This is, however, a warning that at the end of the day, even with the most precise models of the human condition, there will still likely be an unsettling degree of imprecision in our efforts to treat individuals. These are biological entities not electromagnetic pulses, and thus are prone to do what they will no matter how smart we think we are.
The sooner we understand that and make it part of our discussions with clinicians, patients and investors, the better it will be for everyone.
Randall C. Willis is the features editor of ddn. He has worked at both ends of the pharmaceutical industry, from basic research to marketing, and has written about biomedical science for almost two decades.