A link between science and hype? Not always!

Neal Beck points us to this news article by Aaron Carroll, “A Link Between Alcohol and Cancer? It’s Not Nearly as Scary as It Seems.” Here’s Carroll:

Citing evidence, the American Society of Clinical Oncology warned that even light drinking could increase the risk of cancer. . . .

It acknowledges that the greatest risks are with those who drink heavily, but it cautions that even modest drinking may increase the risk of cancer. In the United States, the announcement also notes, 3.5 percent of cancer deaths are attributable to alcohol.

Of course, this means that 96.5 percent of cancer deaths are not attributable to alcohol. If we eliminate heavy drinking, which no one endorses as healthy and where the association is surest, that number climbs. If we also eliminate those who smoke — smoking is believed to intensify the relationship between alcohol and cancer — the number of cancer deaths not attributable to alcohol approaches 100 percent. . . .

For breast cancer — which is the cancer that seems to be garnering the most headlines — light drinking was associated with a relative risk of 1.04 in the announcement. . . .

A 40-year-old woman has an absolute risk of 1.45 percent of developing breast cancer in the next 10 years. This announcement would argue that if she’s a light drinker, that risk would become 1.51 percent. This is an absolute risk increase of 0.06 percent. Using what’s known as the Number Needed to Harm, this could be interpreted such that if 1,667 40-year-old women became light drinkers, one additional person might develop breast cancer. The other 1,666 would see no difference.

Carroll’s not saying this doesn’t matter, just that it’s good to put risks in context, especially given that there’s other research out there:

If you accept the methodology of case-control and cohort studies, from which many of the links between alcohol and cancer arise, you have to accept the results of similar studies of other diseases. For instance, a cohort study of about 6,000 people found that those who drank at least once a week had better cognitive function in middle age than those who didn’t. . . . Randomized controlled trials of alcohol (they do exist) show that light to moderate drinking can lead to a reduction in risk factors for heart disease, diabetes and stroke. . . .

Carroll summarizes:

Once again, we’ve been told that something we eat or drink is going to kill us. Once again, we’re provided an opportunity: A more nuanced discussion of risk — and how we communicate it — can leave us much happier, and perhaps healthier. . . .

1) Consider the absolute risks. A 30 percent increase in risk sounds scary, but an increase from 1 percent to 1.3 percent absolute risk does not, though these are the same things. Likewise, we should be more concerned about a 5 percent risk increase to 21 percent from 20 percent than about a 30 percent increase to 1.3 percent from 1 percent.

2) Don’t give too much weight to observational data. This is especially true when causal data are available.

3) Don’t focus on any one disease while ignoring others. Something may be harmful regarding one disease while beneficial regarding another.

4) Don’t cherry-pick. That is, don’t focus only on some studies, or only on some results. Review all the evidence for the most holistic picture possible.

5) Acknowledge the harms, as well as the benefits, of recommendations. Consider both cost and joy.

These rules may not make for exciting headlines. They may, however, lead to happier, and perhaps healthier, lives.

It’s good to see some anti-hype in the newspaper.

14 thoughts on “A link between science and hype? Not always!

  1. As I have repeated elsewhere, some are just better diagnosticians than others. This would be true of researchers within every field.

    As an opinion,I would be far more concerned with the environmental pollutants implicated in cancer.

    • Oh now that’s a REAL finding. LOL

      In so far as food reseaarch goes, I would wager that eating small food portions & eliminating processed foods as an overall strategy is good: reducing metabolic syndrome & arterial plaque formation. Nutrient research has come a long way in my opinion despite the naysaing among epidemiologists. Some people have made health gains by experimenting with nutrients and food choices. It’s all in trial and error.

  2. Carroll wrote:

    “This is an absolute risk increase of 0.06 percent. Using what’s known as the Number Needed to Harm, this could be interpreted such that if 1,667 40-year-old women became light drinkers, one additional person might develop breast cancer. The other 1,666 would see no difference.”

    That should be “Number Needed to Treat,” NNT. The “Number Needed to Harm,” NNH, is calculated in a similar manner to the Number Needed to Treat, NNT; that is, NNH is the inverse of the absolute difference of the harms. In general, one has to dig diligently to find reference to NNH because it could be embarrassingly low.

    • Jonathan Baron: “It would also be nice if news sources (and medical journals, especially in abstracts) stopped emphasizing relative risk and relied more on risk differences”

      Relative Risk Reduction, RRR, looks more numerically impressive than Absolute Risk Reduction, ARR, so that is why the former gets emphasized by Big Pharma and news reporters. Many people believe that an even better figure of merit is the Number Needed to Treat, NNT, which is the inverse of ARR.

      • I agree when we’re talking about reporting on risks to the general public. That said, I suspect that relative risk is still more informative than absolute risk when it comes to trying to figure out the mechanisms of disease processes. Big effects are easier to study in the lab, where you can do things like artificially increase the baseline risk, or look for abnormalities at a more molecular level.

  3. Observational studies of what people eat or drink are so biased and confounded that I think it is always worth looking to see if there is a plausible mechanism for the effect claimed. For the case of alcohol causing cancer there does appear to be one (http://www.cancerresearchuk.org/about-cancer/causes-of-cancer/alcohol-and-cancer/how-alcohol-causes-cancer). Could this check be put on a more formal basis? Preferably a numerical one?

    Alas, the reported benefits of moderate alcohol use seen in observational studies may be due to confounders, such as income level – from https://www.bmj.com/content/357/bmj.j2755.full.print “Sadly, as with cardiovascular health, the reported benefits of light to moderate drinking on brain health are now looking decidedly shaky”

    I think you should consider that, just as a popular article might be written to attract readers who want to know the details of the latest scare story, a popular article might be written to attract readers who want to hear that their favourite pleasure isn’t a health risk after all. Both of these effects can introduce bias.

    • For those of us who have been super interested in nutrition and vitamin supplementation, we have noticed that medical researchers are woefully uninformed about both. While acknowledging that mainstream research has put in question as to their value in reducing risks for diseases, I think that physicians themselves learn a good deal from their patients who are using supplementation. Physicians themselves are also delving into non mainstream interventions given the harms that they have witnessed in the exercise of mainstream interventions. People have always experimented with concoctions. And maybe some small % have worked. I for example have unusually good blood glucose profile, even non fasting. It could be hereditary. Yes But
      I also don’t consume much sugar which has been implicated in metabolic syndrome. Some vitamin supplementation say Vitamin C and D3 seem to have rather good studies behind their efficacy in the body.

      In short what I’m suggesting is that perhaps some of these mainstream studies are inadequate to begin with b/c of conflicts of interests and research methods, as has been pointed out so routinely.

    • it is always worth looking to see if there is a plausible mechanism for the effect claimed

      This is a variant on the “is there a correlation?” error (Yes, there is.). There is a plausible mechanism for anything health related you can dream up and almost always already evidence in the literature for it. Take it as a principle.

  4. “For instance, a cohort study of about 6,000 people found that those who drank at least once a week had better cognitive function in middle age than those who didn’t. . . . Randomized controlled trials of alcohol (they do exist) show that light to moderate drinking can lead to a reduction in risk factors for heart disease, diabetes and stroke. . . .” — This kind of result is due to classifying non-drinkers and never-drinkers into the same comparison group.

    Details here, https://www.ncbi.nlm.nih.gov/pubmed/26997174

Leave a Reply to Jonathan Baron Cancel reply

Your email address will not be published. Required fields are marked *