It’s really quite amazing–SARS, West Nile Virus, Small Pox, Ebola, swine flu, etc… I’ll never forget the dude I saw in the airport about a month ago wearing a surgeon’s face mask–very creepy looking. Dude probably thought he was cool because he was flew a lot for his cool job and thus needed to wear the mask. The reality is that dude has a greater chance of getting hit by lightning twice in the next month.

The problem with all of the misleading media is often times the underlying news is legitimate, and perhaps even of the very type of sensationalist nature that the media loves. In addition, by design the media gravitaties toward publishing misleading stories on items/issues that are inherently difficult for the public to question. Like health stuff–disbelieve an article in the New York Times about red meat being very dangerous is not right–the risk of me falsely disbelieving it is that I continue to eat too much read meat and die. Plus, it just sounds bad to question stuff like universal health care, global warming, etc. And indeed, sometimes the headlines and the news stories are unbiased and accurate. It almost sounds funny just to type that, but it does happen. The swine flu is an excellent example: apparently it is a more legitimate health concern than SARS and all the other viruses of the month that the media shoved down our throats. However, I simply do not know what to think about it–I am unable to determine from the articles I read if that is a serious thing. Sure, I know there is a chance, albeit small, that it spreads like wildfire and kills everyone.

I do know this though, I can improve my health and chances for a long life thousand times more by exercising, not eating at McDonald’s, washing my hands regularly, driving safely, minimizing stress, and so on.

Once I do those things better, maybe then I’ll start worry about getting short, bit by a shark or dying from second hand smoke.

Crap, all that stuff takes a lot of willpower though. So much stuff out there is bad for us, and we are all doomed anyways–we will run out of oil soon, plus global warming will cause the sea level to rise way too much–in light of such crazy things out there, worry about watching my weight seems trivial.

It won’t matter if I’m in good shape when China overtakes us–apparently their current growth rates can be assumed to continue in perpetuity.

As for the headline writer thing, everyone realizes that papers go for “catchy” over accuracy. First off, that’s just the point: putting out misleading headlines and articles in a newspaper about health is irresponsible and unethical. Such a practice is yet another result of an industry focusing entirely on short-term profits instead of sustainable value. In addition, headlines have a larger font and thus are more readily noticeable by the public. Just because we can all take the fine print doesn’t mean it’s okay to install widespread fear.

Misleading and biased journalism is slowly running its course–it’s not a coincidence that we have witnessed many large news conglomerates go bankrupt lately. Most people either realize how misleading media health reports can be. A lot of people are stupid though, and they love such stories because it provides a distraction from their stupidity. Deep down they know, or at least should know, that nothing within their control poses as much of a threat as their lack of exercise, or their smoking habit, or their affinity for Fil-0-Fish sandwiches at McDonald’s.

These stupid stories are getting old every single day. It’s amazing how far it went though: the shark attacks on the TV news networks I will never forget (all 54 of them that year). My roommate one year in college used to really get into it–he was literally scared he was going to get killed by a shark, yet we were over 1,000 miles away from the nearest ocean.

Here are a couple references to look at to see that the RCT should not be worshipped. First: Colagiuri, Morley, Boakes, and Haber, Psychotherapy and Psychosoamatics 2009 v78 pp 167-171: “Expectancy in Randomized Controlled Trials.” Short story: a priori hypothesis failed: no effect for drug, vs. placebo, intended to assist quitting alcohol; however, post hoc analysis: for participants in both the drug and placebo group, those who BELIEVED that they were getting active drug not placebo reported lower avg cravings for alcohol, and lower number of drinks per day. Thus, a technically ideal study suffers mightily from bias. You could argue: well, that was a post hoc analysis. To which I might argue: yes, and darn glad we have it, when our RCT research is filled with many such biases never detected by any analysis, regardless of Latin name. In this example, expectanciy could have been reduced by an active placebo – just to show that it is not the design per se that yields the truth.

Another example: Concato, Shah, and Horwitz, NEJM 2000 v. 242 pp 1887-1892: “Randomized, Controlled Trials, Observational Studies, and the Hierarchy of Research Designs.” Short story: across a range of clinical topics, the obtained point estimates for the efficacy of some intervention (prevent stroke by lowering hypertension, etc.) varied greatly across a handful of RTCs, but did not vary much across a handful of observational studies; conclusion is that the inclusion/exclusion/subject recruitment and selection process of an RCT leads each time to a different set of participants who are biased in various unknown ways away from their respectful populations.

All this to say that no study in itself is ideal, and we need to review Bradford Hill; journalists would do well to read this classic, and live by its wisdom: Hill AB. 1965. The Environment and Disease: Association or Causation? Proceedings of the Royal Society of Medicine, 58, 295-300.

… explains many sociological phenomena, (I’ll take the prevalence of disinformation and superstition as demonstrative proof) but it’s a rather sad thing to accept such basic fallibility in human nature.

And you win!

www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0006103

Ideally, although much of it is based on ‘common sense’ recommendations given to the media need to be evidence-based otherwise we fail to reciprocate what science and evidence-based medicine is all about.

Even a huge study would have low power if we set the type 1 error rate at a ridiculously low level. For example, suppose we set it at ZERO. That is, we refuse to make a type 1 error, so we do not reject the null hypothesis, NO MATTER WHAT THE DATA SAY. Then the type 2 error rate is 100%, because we cannot possibly reject the hypothesis when it is false. And if the type 2 error rate is 100%, then the power is zero.

The way we get away from this push-pull of the error rates is to set one of them arbitrarily, and that is the type 1 error rate, usually at 5%. Then we look to see what the power is for a given sample size. (And, this is often overlooked in these discussions, for a given effect size.)

This is how the first trials when a new drug is given to humans work for eg. and why the repeat of the Jesse Gelsinger incident (gene therapy caused cancer) indicated that the transgene was doing something funky (it was prerentially inserting next to a known risk gene).

“if we didn’t know this and we tossed a coin four times and got three heads and one tail, we might conclude that getting heads was more likely than tails. But this chance finding would be wrong. If we tossed the coin 500 times – gave the experiment more “power” – we’d be much more likely to get an even number of heads and tails”
The probability of exactly equal heads and tails for 4 coins is 4C2/2^4 = 6/16
The probability of exactly equal heads and tails for 500 coins is
500C250/2^250 = well, I can’t work that out but it is much less than 6/16

As investigative journalists, detectives, and everyone else who wants to know what is going on have always said:

“Follow the money”

This is not necessarily so. Studies may indeed give misleading results because of a lack statistical power and making the sample size bigger does reduce this risk. However (observational) studies of any size can also give misleading results because of confounding factors. Increasing the sample size does not, in itself, do anything to reduce the risk of confounding. Of course if you have a large sample size then you may have more opportunity to gather the kind of data that would eliminate confounders, but you have to ask the right questions, and you may not know what they are. So even really large studies, like the Million Woman Study may need to be treated with a certain degree of caution.

(raw is this www.badscience.net/2009/09/how-to-read-articles-about-health-by-dr-alicia-white/www.nhs.uk/news)

the link behind the link…

Thank you for another excellent article.

Your link entitled “Behind the Headlines” goes to another of your blog entries rather than the Behind the Headlines website.