Cross of Gold

I promised to write about clinical trials, so here goes. I'm just going to wing it so I don't know how many posts it will take to get this done. We'll find out.

Randomized controlled trials (RCTs) are considered the gold standard for determining the safety and effectiveness of what are broadly called interventions -- most often chemicals intended to treat or prevent disease, but it could be devices, programs intended to change people's behavior, surgery, you name it. The way the universe works, it's often difficult to prove that A causes B. There might be a lot of reasons why A happens and then B happens. For example, B might just happen a lot anyway. That's usually how superstitions develop. That's why physicians for centuries would bleed people or make them vomit or give them diarrhea. (Really.) Since most, or at least many people, eventually get better anyway, the doctors were convinced their interventions were the cause. Or, people who have A happen to them, for whatever reason, might also be people who are likely to have B happen to them, maybe for the same reason, or for some other reason. 

 

We might also deceive ourselves. If we believe, or want to believe, that A -> B, we may be more likely to perceive B after A, or come up with excuses when it doesn't happen. Or we might unconsciously choose people to expose to A who are more likely to experience B.  Or there might be component of A, something associated A, that is the real cause of B, not A itself or what we think is the active ingredient of A. You can probably think of a few more reasons why we can get fooled.

So, the double blind, randomized, controlled trial was developed to try to rule out these deceptions, at least to the extent possible. There are indeed some advantages of RCTs when it comes to causal inference, but there are also many limitations, common ways to go wrong, and conceptual traps associated with them. What I will do in following posts is to discuss the theoretically right ways to do them, why that is sometimes impossible, how it can go wrong when even when it is possible, and how we can end up being fooled anyway. This requires a bit of knowledge of the branch of mathematics called statistics, so I'll try to present that as accessibly as possible. Some of the takeaways I'm aiming at are:

• Why people often tout useless treatments and how they fool people into thinking they work;
• Why treatments that might in fact be useful get rejected or overlooked;
• Why treatments that work in RCTs often don't work so well in the real world;
• Why the expense of RCTs interacts with the profit motive to drive biomedical research down pathways that don't produce much if any benefit to humanity;
• Maybe a couple of other observations along the way.

And, let me know if you have any questions as we proceed.

 

A digression: Norton utilities is constantly trying to sell me upgrades and add-ons to my basic protection package. It's annoying, and if you don't want the constant pop-ups with sales pitches, you might want to try a different product. Anyway, the latest thing they're trying to do is tell me that people are selling my information on the "dark web" (horrors!) and they can get it scrubbed.  What is the information that puts me in such peril? 

1. My address (a PO Box) and two addresses I lived at 15 and 50 years ago, respectively.
2. My phone number. Well, actually no, two phone numbers, one of which is my old number from when I lived in Boston 15 years ago, and the other is the phone number of my parents when I was a child. My actual present phone number is not there.
   
3. My relatives, consisting of my dead parents, my brother, a guy who lives across the street to whom I am completely unrelated, and three women I never heard of. (Maybe I'd like to meet them.)

For much of my life, the Post Office used to deliver a book to my house containing the name, address, and phone number of everybody in town, or in the case of a small town, a lot of neighboring towns as well. If you wanted to get the info on somebody who lived anywhere in the United States, as long as you knew the town, you could dial 4-1-1 and get the information for free. You could also buy, for cheap, a so-called reverse phone directory that listed the info by address. It was already sorted alphabetically.  So go ahead, Dark Web. Sell my data. And I hope everyone will buy it.