A large data set of highly biased, uncontrolled data is still useless. You can't model yourself out of the factors that have not been recorded. And believe me, the data you'll get on patients that try experimental medication will be very, very incomplete.
It also won't be a large dataset. How many people have squamous cell carcinoma between now and the moment of approval and are willing and wealthy enough to buy this particular medicine? 100 seems too much already.
> The mechanisms that were in place at the time did their job.
Barely, according to wikipedia.
> Second, the response to Thalidomide was overblown and indeed downright hysterical. "Better to let 100,000 people die of neglect than allow one person to suffer an awful drug reaction" is not rational policy.
Thalidomide was never going to save 100,000 people. It did cause 2500 birth defects in West Germany alone, though, plus an unknown number of abortions. Given that it was legal in 46 countries, the number of people with birth defects must be well over 10,000. This drug isn't going to save 100,000 people either, certainly not in the period until admission.
> A large data set of highly biased, uncontrolled data is still useless. You can't model yourself out of the factors that have not been recorded. And believe me, the data you'll get on patients that try experimental medication will be very, very incomplete.
There are many safety issues that placebo-controlled and double-blinded studies didn't catch, but postmarketing surveillance later did. Large data sets are more powerful than you give them credit for --- and, moreover, in the real world drugs are not always used the way they are in clinical trial settings. In the real world, people skip doses, double up on doses, drink grapefruit juice, etc.
If you want a full picture of "factors that have not been recorded," you want as large a dataset as possible, as close to real world conditions as possible. It may not be as clean as you'd like, but it'll give you enough to draw every inference.
> How many people have squamous cell carcinoma between now and the moment of approval and are willing and wealthy enough to buy this particular medicine? 100 seems too much already.
When the alternative is to allow the drug to spend a decade or longer in development hell, I'll take poor data and lives potentially saved over "great data" (not really) and a pile of dead bodies.
Besides, when drugs are potentially curative, and the disease otherwise so invariably fatal, even 50 datapoints should be sufficient. This is trivial to model mathematically.
> Barely, according to wikipedia.
Oh come on. Besides, teratogenicity is now an absolutely basic component of safety tests. You don't need the FDA's failed paradigm of extensive efficacy trials to prevent "Thalidomide part II."
> Thalidomide was never going to save 100,000 people.
The exaggerated response to Thalidomide, exemplified in your first comment, has absolutely condemned more than 100,000 people.
A large data set of highly biased, uncontrolled data is still useless. You can't model yourself out of the factors that have not been recorded. And believe me, the data you'll get on patients that try experimental medication will be very, very incomplete.
It also won't be a large dataset. How many people have squamous cell carcinoma between now and the moment of approval and are willing and wealthy enough to buy this particular medicine? 100 seems too much already.
> The mechanisms that were in place at the time did their job.
Barely, according to wikipedia.
> Second, the response to Thalidomide was overblown and indeed downright hysterical. "Better to let 100,000 people die of neglect than allow one person to suffer an awful drug reaction" is not rational policy.
Thalidomide was never going to save 100,000 people. It did cause 2500 birth defects in West Germany alone, though, plus an unknown number of abortions. Given that it was legal in 46 countries, the number of people with birth defects must be well over 10,000. This drug isn't going to save 100,000 people either, certainly not in the period until admission.