July 12, 2013
Even More Studies You Should Ignore
Back when I was a medical student (in the Cretaceous Period) we were taught that someone once did a study comparing folic acid levels in the blood of cancer patients compared to the blood of healthy patients. The cancer patients had, on average, significantly lower folic acid levels. And the ones with the largest, fastest growing tumors tended to have the lowest folic acid levels. “Aha,” they thought. “Something about folic acid deficiency predisposes them to cancer. We should give folic acid to cancer patients.” Bad idea. A randomized trial showed that cancer patients given folic acid died sooner than those given placebo.
What happened? Low folic acid levels are a consequence, not a cause, of cancer. Folic acid is needed to synthesize DNA, and DNA synthesis is necessary for one cell to divide into two cells. So folic acid gets used up by rapidly dividing cells – like cancer cells. Giving cancer patient folic acid just gives their tumor a helping hand. (This roundabout insight led to medications that block folic acid metabolism which are used as chemotherapy to this day.)
The main lesson here is that correlation tells us almost nothing about causation. That means if A and B occur together, we say that they are correlated, but we have to be very careful not to assume that one causes the other. A might cause B, or B might cause A, or they may both be caused by some other factor that we’re not paying attention to. This is the cognitive error that the farmer makes when he notices that the daily number of deaths of his livestock correlates with sales of ice cream in the town’s ice cream parlor. He figures that some waste from the ice cream is contaminating his feed or water. He lobbies the legislature to ban ice cream sales around his farm. He doesn’t realize that the rise in ice cream sales and his livestock deaths are both caused by very hot days.
This week a study was published that encourages just such a mistake. It should be universally ignored, but a handful of patients have already emailed me about it, and it’s receiving a fair amount of confused media attention.
Before we look at the study, we need to learn a little about omega-3 polyunsaturated fatty acids. Omega-3 fatty acids are the predominant molecules in fish oil. They have been proven to lower triglycerides. They have gained popularity in the last few years, though the most recent trials (reviewed last year by The Medical Letter) have failed to show that they prevent stroke or heart attacks. There is certainly no convincing reason for the general population to be taking fish oil.
This week’s study, published in the Journal of the National Cancer Institute, compared blood levels of omega-3 fatty acids in patients with prostate cancer to those levels in healthy adults. Prostate cancer patients had higher levels of omega-3 fatty acids. Note that the study had nothing to do with fish oil supplements or diet. None of the study subjects were asked about supplements or how frequently they ate fish. The only comparison was blood levels of omega-3 fatty acids of people with prostate cancer to healthy people.
This should remind you of the ancient folic acid study.
The study authors conclude that this should make us worry about the risks of increasing omega-3 fatty acids in our diet, and some of the media coverage warns that fish oil supplements may increase the risk of prostate cancer, but the study proves no such thing. Other possibilities are that prostate cancer patients produce elevated levels of omega-3 fatty acids, or that some unknown metabolic defect both increases prostate cancer risk and elevates omega-3 fatty acid blood levels.
The only way to know for sure is to randomize lots of people to fish oil capsules or to placebo, follow them, and count the resultant prostate cancer. That study hasn't been done, but that study would deserve our attention.
So am I saying that fish oil is safe and everyone should resume taking it? No. It has no proven benefits (except possibly for elevated triglycerides). That the connection to prostate cancer is unproven isn't a reason to take it.
Who knows? Maybe after a randomized trial is done and the biological connection is meticulously worked out this might lead to a new prostate cancer test or treatment. Then medical students not yet born will learn about it as an example of the importance of not confusing correlation with causation.
Hold The Salmon: Omega-3 Fatty Acids Linked to Higher Risk of Cancer (Time)
Important legal mumbo jumbo: