Example: Pharmaceuticals
Say a pharmaceutical company has developed a new drug. What do they need to do in order to sell it?
Permission from FDA - Food and Drug Administration.
How to get that?
1) Proof that drug is save. But what does save mean? - side effects. How does one test whether a drug is save?
2) Proof that drug works. But what does works mean?
How do you proof that a drug works?
Gold Standard: Clinical trial
Step 1) Find a group of people (patients) with the disease you try to cure
Step 2) Randomly divide them into two groups - treatment and control group
Step 3) Give drug to treatment group, give placebo to control group
Step 4) Wait a while
Step 5) Compare outcomes for the two groups
Question: is this ethical?
Consider a very famous clinical trial, the Salk Polio Vaccine trials
Consider what it means:
• using the vaccine the incidence rate seems to be 0.016%, or about 1 in 6250
• if all the 201,229 children in the control group had been vaccinated one would have expected that 0.016% (or about 32 of them) would get polio anyway
• instead 115 did
So this study design "condemned" about 83 (115-32) children to a life of illness (and early death)!
Was this ethical?
How about the following experimental design - offer the vaccine to everybody. Some parents will agree to vaccinate their children, others will not, so again we get a group of vaccinated children ("treatment group") and a group of unvaccinated ones ("control group"). After some time we compare the two groups, just like in a clinical trial. Will this work?
Problem (in the case of Polio)
• Parents with a high standard of living generally are more likely to vaccinate their children
• Children of parents with a high standard of living generally live in a more hygenic environment
• Childen in a cleaner environment generally don't get exposed to a mild form of the polio virus while still protected by the anti-bodies of the mother.
• Children who don't get the mild form are more likely to get the bad virus
Consequence? Under this setup the children in the treatment group are more likely to get polio than those in the control group because they tend to come from well-to-do families. On the other hand they are getting a vaccine that works. These two effects might have cancelled each other out, and the final (incorrect) conclusion could have been that the vaccine does not work!
The problem here is that the two groups differ not only in the treatment they receive (vaccine or placebo) but also in some other aspect (how rich the parents are). In Statistics this type of problem is called confounding. Again in this case the problem arises because it is not some random mechanism that assigns children to each group, but they (or their parents) assign themselves. This is called selection bias.
So here is why Statisticians recommend randomized trials: we never know all the facts influencing the outcome of an experiment, randomization is the only way we know of to avoid any biases.
Clinical Trial: Assignment to treatment or control group is done randomly.
Observational Study: Assignment to treatment or control group is done in some other way.
The conclusions we can draw from an observational study are never as good as those we get from a clinical trial.
Do no give anything less than the best treatment known (Aspirin study of British Doctors and Nurses, Women's Health Initiative and Hormone Replacement Therapy)
Another famous (infamous?) example: Mammery Artery Ligation
And another famous example: Stanley Milgrams Experiment
Ethical?
Note: no informed consent
but also: no real harm
One way to assure that experiments are ethical: Oversite Committees: