2 ms·

Are controlled experiments necessary for proving safety? Intuitively, there shouldn't be a placebo effect, and you could get twice as powerful a study by giving

Are controlled experiments necessary for proving safety? Intuitively, there shouldn't be a placebo effect, and you could get twice as powerful a study by giving twice as many people the drug and using statistics for the general population as the control group.

semi-extrinsic 7y ago
Why shouldn't there be a placebo effect? It's well documented for basically any drug, even for alcohol. Serve people non-alcoholic drinks but tell them otherwise, and they get drunk.
- dmurray 7y ago
  Because the trial wasn't about finding the expected effect (drunkenness) but the secondary effects. People don't get heart disease or cirrhosis from non-alcoholic drinks, or at least, I'd be interested in reading the study where they do.
  - semi-extrinsic 7y ago
    No, but they can get short-term secondary effects such as nausea, headaches the next day, etc. Probably this category of things is what they want to control for.
arkades 7y ago
No. For phase 1, it's generally sufficient to show that few or no people had life-threatening adverse effects, that none of the life threatening effects were plausibly related to the drug, and there were no meaningful hiccups in CNS, pulmonary, or cardiac function. In this case, they went for a placebo set-up because they were also trying to grab a little data to show whether 25 mg was significantly distinct from 10mg. The comparison to placebo was not needed, but is winning them a little extra publicity.
gus_massa 7y ago
What is general population? I guess the study exclude people with more than 95 year old and people with terminal cancer. Other exclusion factors are not so clear, so you need to know the results of no treatment in the same population. For this, you need to split the group at random. Note also that the study probably include some regular visits to the hospital. Some people may eat more healthy food a few days before a blood exam so the doctor gets happier. Some people may change other habits, like no binge drinking. So you need to measure the results in a group that follow the same explicit and implicit rules. Does the report of the symptoms use the same criteria in the study than in general population? What about dizziness? Does this produce dizziness? Does the general population report dizziness to their family doctor? Replace dizziness with chest pain, or constipation, or whatever symptom you are interested. Also, people lie to their doctor because they try to be nice. It's not a voluntary lie, but if there is a study about curing something, people want to be nice and give optimistic feedback. You must have a double blind control group so you can compensate this.

semi-extrinsic 7y ago

Why shouldn't there be a placebo effect? It's well documented for basically any drug, even for alcohol. Serve people non-alcoholic drinks but tell them otherwise, and they get drunk.

dmurray 7y ago
Because the trial wasn't about finding the expected effect (drunkenness) but the secondary effects. People don't get heart disease or cirrhosis from non-alcoholic drinks, or at least, I'd be interested in reading the study where they do.
- semi-extrinsic 7y ago
  No, but they can get short-term secondary effects such as nausea, headaches the next day, etc. Probably this category of things is what they want to control for.

arkades 7y ago

No. For phase 1, it's generally sufficient to show that few or no people had life-threatening adverse effects, that none of the life threatening effects were plausibly related to the drug, and there were no meaningful hiccups in CNS, pulmonary, or cardiac function. In this case, they went for a placebo set-up because they were also trying to grab a little data to show whether 25 mg was significantly distinct from 10mg. The comparison to placebo was not needed, but is winning them a little extra publicity.

gus_massa 7y ago

What is general population? I guess the study exclude people with more than 95 year old and people with terminal cancer. Other exclusion factors are not so clear, so you need to know the results of no treatment in the same population. For this, you need to split the group at random. Note also that the study probably include some regular visits to the hospital. Some people may eat more healthy food a few days before a blood exam so the doctor gets happier. Some people may change other habits, like no binge drinking. So you need to measure the results in a group that follow the same explicit and implicit rules. Does the report of the symptoms use the same criteria in the study than in general population? What about dizziness? Does this produce dizziness? Does the general population report dizziness to their family doctor? Replace dizziness with chest pain, or constipation, or whatever symptom you are interested. Also, people lie to their doctor because they try to be nice. It's not a voluntary lie, but if there is a study about curing something, people want to be nice and give optimistic feedback. You must have a double blind control group so you can compensate this.