By Larry Marsh, Kansas City Star Midwest Voices columnist 2009
The aim of a great deal of medical research is to get approval from the FDA for some drug or medical procedure. Other considerations such as determining the drug's or procedure's appropriateness for individuals are often ignored or "left for future research" and never done. Many lives could be spared, not to mention vast sums of money saved, if the research were done correctly and completely.
An old joke defines a statistician as a person who can put one hand in boiling water and the other hand in freezing water and say "On average I feel just fine." Getting FDA approval requires presenting convincing evidence that a drug or procedure is effective for the average patient and not dangerous for the vast majority of patients. The problem is that the average patient may not exist at all. Even when such a patient does exist, that "Jill" or "Joe" average might not be you.
To avoid systematic bias in medical studies patients are randomly and secretly assigned to a treatment group to take the real drug or a control group to take a placebo drug that looks real but is really just a dummy drug that does nothing. The best of such randomized studies are double-blind which means that neither the doctor administering the drug nor the patient taking the drug know if the drug is real or placebo.
It is not possible to carry out a randomized trial when the treatment cannot be randomly assigned. For example, it is not feasible to randomly assign some people to be smokers and others to be nonsmokers. In that case an observational study is done instead. Systematic bias could distort the results if nonsmokers were more likely to work out at the gym or eat healthier food than smokers. Consequently, observational studies require including a host of control variables such as age, gender, exercise and nutrition to extract out their effect on your probability of getting cancer. This is important because your chances of getting cancer might be quite different from that of "Jill" or "Joe" average.
Including control variables also allows researchers to check for interaction effects. For example, the effect of age on the probability of a heart attack or stroke depends on a person's gender. A drug that might be very important for a man might be less important for a premenopausal woman of the same age. In other words the effect of age cannot be determined in isolation, gender and other factors may be important in finding out the impact of age on heart attacks.
Randomized clinical trials are the considered the gold standard of medical research, but, more often than not, they are used to avoid determining the effect of a drug or procedure on an individual patient. They typically don't bother with control variables. Since patients are randomly assigned to their groups, the average values of the control variables tend to be same for the two groups. For purposes of convincing the FDA to accept the drug, the researchers can claim a balanced study that not only uses two groups with equal observable characteristics but presumably has equal unobservable or even unknown characteristics as well.
Why bother with control variables in a randomized clinical trial? Two reasons that have nothing to do with FDA approval jump out at you immediately. Each individual can get much more appropriate and effective treatment if control variables are used to determine the effect of the drug or procedure on them as individuals. Lots of money can be saved if drugs and procedures are just used on people who are helped by them and not by people on whom they have no or little effect.
Not all medical studies are done poorly. On Tuesday, November 17 (yesterday) I attended a research presentation sponsored by the Biostatistics Department at the KU Medical Center. Professor Seng-jaw Soong gave a talk "Multivariate Modeling of Melanoma Prognosis and its Web-based Applications." His study was a randomized clinical trial. He not only included important control variables, but he also checked for interaction effects among the variables. To make sure that the results of his work would be used appropriately for individuals, he created a web site where you can plug in your personal characteristics to get results that are appropriate for you as an individual and not just those for "Jill" or "Joe" average.
The FDA should change the rules of the game. Not only should new drugs and procedures have to be effective for the average patient and be harmless for the vast majority of patients, but control variables must be included and a web site must be created where patients and their doctors can check the appropriateness of the drug or procedure for that specific patient. Checking for interaction effects must be part of the analysis.
It is time to stop one-size-fits-all medicine. Write to Congress to demand that the FDA wake up to its responsibilities to individual Americans and stop wasting our lives and our money.
. . .
Also see these health links:
Reprogram your subconscious mind to commit terrorism or lose weight
Kennedy's death reminds us that men in their 70s must be defensive players
Is calorie restriction a good defense against cancer?
No-Eat-Day Diet: A good strategy or bad advice?
Why cavemen didn't have weak bones
. . .
Follow Larry on Twitter.
. . .
Delicious
Digg
The average person
Actually, Larry, the joke I like is that the average patient has one testicle and one ovary.
There's also the story about the doctor who used one of these web-based computer models to calculate the risk for a particular patient. She plugged in all the relevant information, and the model produced an estimate saying that there was a -0.2 chance that this patient would live. The doctor went back to the statistician and asked what was going on. I understand what a survival probability of zero would be. That's the patient who should do as the Tim McGraw song says and go skydiving and Rocky Mountain climbing. But what does a NEGATIVE survival probability mean. The statistician replied, "Well, that's a fate worse than death."
Seriously, there is a lot if difficulty doing what you propose, which goes under the title of individualized medicine. Not only do you have to take all of the risk factors into account, but you have to also take into account the patient's value system. How much pain and suffering would you endure for an extra month of life? How much would it bother you if a treatment that could cure your disease would also make you sterile? Would you endure a treatment if one of the side effects was impotence?
The Simpsons had a clear take on the need to balance individual perceptions of benefits and risks. Marge Simpson was taking a pregnancy test, a Krusty the Clown pregnancy test, which is pretty funny all by itself. But even funnier was the warning right below the smiling clown face that read: "May cause birth defects."
Back on a serious note, a worthwhile treatise on the fact that we can't treat everyone as if they were average is Stephen Jay Gould's "The Median Isn't the Message" which is reproduced at
http://cancerguide.org/median_not_msg.html
Steve Simon