Epidemiology – A Life Saver

April 18, 2008 at 8:17 pm (Bad Science, Good Science) (, , , , , , , , , )

This post might be more for the casual reader than skeptics and Bad Science bloggers. I’m sure they will already know far more about epidemiology than I do. Here we go anyway:

From time to time, I see criticism of epidemiology and the use of statistics. Usually on the basis that epidemiology “can’t prove anything” or is “not real evidence”. Sometimes a seemingly opposite tack is used: “oh, but you can use statistics to prove anything”. The criticism usually comes from someone whose own particular worldview is being challenged. Like, for instance, Dave Hitt. Dave Hitt features in this post on the Apathy Sketchpad blog. Or for another example of the statistically-challenged, there’s Gus from the JABS forum. [“Why is it you never listen to the evidence (the autistic children) and are only interested in the science and epidemiology provided by tabloid gutter press as it was hardly hard to see where the research had come from?”]

So what’s epidemiology ever done for us? There’s a couple of chaps I’d like you to meet first: Richard Doll and Austin Bradford Hill. Now quite apart from anything else Bradford-Hill did, he encouraged use of controlled trials (PDF) – something important in its own right. Bradford-Hill also helped to show, along with Doll, the part that tobacco smoking played in lung cancer. The original Doll/Bradford-Hill paper is available via Pubmed here and here as a PDF [which might take a while to download]. The authors concluded that “smoking is an important factor in the cause of carcinoma of the lung.” It’s now accepted (almost universally) that smoking is harmful – but how long might it have taken without the work of Doll and Bradford-Hill? Can you imagine what might have been if Doll and Bradford-Hill’s work had been ignored by a nation of Dave Hitts?

Another example of the usefulness of epidemiology and statistics is in the epidemiological approach to another notable condition. In Lange’s Medical Epidemiology (Third Edition – the Fourth Edition is available*), there is a description of a young man with no obvious underlying causes of immune dysfunction who is suffering from three concurrent infections. Within the preceding six months, three other patients with similar symptoms had been referred to the UCLA Medical Center. Other, similar reports were received by public health authorities and the CDC set up a task force to collect detailed information on those affected. Within months, the disease was named the acquired immune deficiency syndrome (AIDS).

Epidemiological methods were used: to monitor the patterns of the occurrence of AIDS; to measure the rapidity of occurrence; and to search for causes by identifying risk factors. They were also used in determining case fatality, survival time and prognostic factors. Which I would have thought were all important things to know. Lange’s Medical Epidemiology tells us that “medical progress often is best advanced when the sciences that focus on subcellular and molecular basic research work in tandem with the population-oriented science of epidemiology. For example, as bench scientists were struggling to characterize the molecular properties of HIV, epidemiologists already determined that AIDS is a contagious disease that is spread through certain interpersonal behaviours. As the painstaking search continues for improved treatment, or even a cure or vaccine, public health professionals have recommended measures to prevent the spread of HIV by reducing the frequency of the fololwing high-risk practices: (1) casual, unprotected sex and (2) sharing needles among drug users.”

Notes:
Cardiff University has a page with links to lists of Bradford-Hill’s principal publications and literature related to Bradford-Hill.
*McGraw-Hill Medical Publishing Division seems to be here, and the Fourth Edition of Medical Epidemiology is available here.

12 Comments

  1. dvnutrix said,

    John Snow, of course, brilliant bit of detection. Janet Lane-Claypon is outstanding for many reasons but notably for providing the first documented case of using statistics to analyse epidemiological data, and being one of the early people in what we now call epidemiology. Many of her findings still stand today with comparatively minor refinements.

    Epidemiology has many problems but relevance is not one of them.

  2. David Colquhoun said,

    Of course you are right that many important things have come from epidemiology, But also, I suspect , so have many unnecessary scares. The problem of attributing causality is enormous in non-randomised cohort studies. Epidemiologists are very aware of this problem, but their methods of solving it are fragile. To take one example, when the papers were full of headlines saying that “one sausage a day increases your risk of cancer by 20%”, I went into the data that lies behind this statement. It isn;t baseless, but the evidence for causality is far more flimsy than the headlines suggest in my view,

  3. dvnutrix said,

    And last week we heard and read lots of ‘informed’ criticism attacking the Cochrane Review of antioxidants because it insisted on RCTs and excluded non-randomised cohort studies.

    Maybe my last sentence shouid have read: “Well-grounded epidemiological studies that adhere to agreed and transparent protocols and use appropriate statistical analyses have greater practical application than comparatively shoddy or unfounded ones”.

    Of course, the message that a sausage a day is not that harmful has no ready market in our various media that depends upon novelty cycles rather than fulfilling a public education remit. [/whinge]

  4. AndyD said,

    I think you mean Acquired Immune DEFICIENCY Syndrome.

  5. jdc325 said,

    AndyD – you are quite right. Thanks for pointing out the error. Fixed it now.

    dvnutrix – thanks, again, for providing useful links and info. As well as checking out Janet Lane-Claypon, I looked up John Snow and Cholera. There’s a decent collection of articles on Snow here: http://www.ph.ucla.edu/epi/snow.html – some of which should ring a few bells for those who did GCSE Schools History Project (ah, learning about 19th Century Public Health – happy days).

    David – thanks for your comment. I find the misrepresentation/exaggeration of scientific studies fascinating. And this kind of thing – “the evidence for causality is far more flimsy than the headlines suggest in my view” – is very common in the British media. I tended to assume that this was mostly the fault of the media, but the fault seems to lie variously with: the scientists doing the study; press officers presenting the study; and journalists reporting on the study.

  6. She-Liger said,

    Sorry for some off-topic.

    http://postpostscriptum.wordpress.com/
    About quackery courses in universities.

  7. Dave Hitt said,

    When discussing someone’s work it’s usually a good idea to go directly to the source, rather than depend on a temper tantrum by someone who thinks calling someone a “twat” is intellectual debate.

    Andrew (Apathy) was upset about the site The Facts, which discusses epidemiology and statistics and how they’re used to vilify smokers with the claim that second hand smoke is deadly. It’s not hard to find – if you Google “Second Hand Smoke” it usually shows up on the first page. You can get to it here:

    http://www.davehitt.com/facts/index.html

    The first chapter of any primmer on statistics drills home the fact that *correlation does not prove causation.* Virtually all heroin addicts drink milk, but that doesn’t mean milk consumption leads to heroin addiction. The Facts gives a detailed account of correlation gone wrong: For years it was believed that having many children reduced a woman’s chance of breast cancer. All the studies confirmed it. It was considered a fact, “proved” by epidemiology. But eventually research showed that the effect was caused by having children at an early age, and had nothing to do with the number of kids a woman has. The correlation was there (since woman with many kids usually start young) but the causation was wrong.

    Epidemiology is a science of probabilities. What is the probability that A causes B? Because correlation can not prove causation, epidemiology can’t *prove* anything. Instead, used correctly, it is a useful guidepost – it provides subsequent researchers a plausible place to look, and helps spare them time wasted darting down blind alleys. In the case you cite epidemiology *indicated* how AIDS was spread, but it had to be confirmed in a lab before it became a fact. The epidemiology told them where to look, but in and of itself didn’t prove anything.

    Andy was really annoyed with my article “Name Three,” where I wrote to many prominent anti-smokers and asked them to name three people who died of second hand smoke. Most refused the request. The few who did reply lied. It was an entertaining exercise, and I enjoyed heaping well-deserved ridicule on them. Andy couldn’t quite handle that, though.

    In the future, if you’d like to discuss something of mine, please do me the courtesy of contacting me directly – I’m very easy to find – instead of relying on the hissy fit of some semi-literate punk who can’t handle a debate.

  8. jdc325 said,

    “When discussing someone’s work it’s usually a good idea to go directly to the source, rather than depend on a temper tantrum by someone who thinks calling someone a “twat” is intellectual debate.”
    Uh, the very first link on this post was to your work Dave.

    For more info on Dave Hitt, simply click one of the links in this comment to Dave Hitt.

  9. Dave Hitt said,

    Thank you JDC. If there was any question about your integrity, you just answered it.

    And I see you are unable to answer any of the points I made.

    If anyone would like to see what I *really* said, please stop by davehitt.com, and feel free to drop me a line.

  10. jdc325 said,

    Dave – I linked to your work in the very first link on this post. To suggest otherwise is disingenuous. Thank you for casting aspersions on my integrity anyway, though.

    As to your other points – where did I claim that correlation equals causation or that epidemiology can provide irrefutable proof of anything? My point was that epidemiology is often unfairly castigated and that it has not simply proved useful, but has saved lives. The AIDS example illustrates the way that preventative measures that can be put in place to protect people from serious disease by identifying risk factors and the lung cancer example demonstrates the usefulness of epidemiology in identifying risky behaviours that were previously assumed to be safe.

  11. Hobbes said,

    I think the big reason people are critical of statistics is that it can be misleading and is used quite often to mislead or misrepresent. And when it is not misrepresentation it is ignorance of the methods and lack of knowledge to interpret the stats. Same is with epidemiology. Nothing is absolute, everything is realtive and has to interpreted with caution. People are so much stuck to p-values and significance that they fail to see beyond. People have to understand that at best epi and stats help us to find patterns and just the like the court of law everything is considered innocent unitl proven guilty. Not being proven guilty does not mean that the person was innocent.

  12. Unsung Heroes « Stuff And Nonsense said,

    […] I wrote about epidemiology, I mentioned Richard Doll and Austin Bradford Hill as examples of prominent figures in the field. […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: