���On 28 Dec 2009 Chris Green wrote: >There's nothing surprisingly egregious about this particular >article, is there?
In response to which Stephen Black replied: >I've never seen a university press release, which should >have been vetted by the authors and presumably ran with >their approval, hide the fact that the research was in animals. Chris Green responded: >I'm still surprised. Without actually going back a checking press >releases, I can recall the case of the "moderate drinking causes >breast cancer" announcement in Britain earlier this year, in which >it seemed pretty clear that the scientists had sexed it up for the >university press team, who had then re-sexed it up for the new >media, who had then re-re-sexed it up for public (when in fact >the actual increase in the breast cancer rate was something like >2 in 10,000, and there was little reason to believe that alcohol, >rather than the billion or so things correlated with increased >alcohol consumption, was responsible even for this tiny increase). Let's all agree that there is much dismal reporting of scientific findings (especially in the field of health) in the media. But Chris's response to Stephen does not directly answer his challenge. Moreover his supposedly "just as bad" example turns out, on investigation, not to live up to Chris's assertions (at least as far as the British press is concerned). I though it might be interesting to investigate the specific example Chris gives concerning the study which was reported as saying that moderate drinking increases the risk of (not "causes") breast cancer. My conclusion, at least in relation to the British press, is that the reporting was nowhere near as bad as Chris asserts, and that he understates the claimed increase of breast cancer rate for moderate drinking by a factor of about 5. First the study by the University of Oxford's Cancer Epidemiology Unit: "Moderate Alcohol Intake and Cancer Incidence in Women", Allen N. E. et al, : Journal of the National Cancer Institute, Volume 101, Number 5, 4 March 2009 , pp. 296-305(10). From the Abstract's "Conclusion" (relating to alcohol): "Low to moderate alcohol consumption in women increases the risk of certain cancers. For every additional drink regularly consumed per day, the increase in incidence up to age 75 years per 1000 for women in developed countries is estimated to be about 11 for breast cancer…" http://tinyurl.com/yc6esev What is evidently the University press release is on the BMJ Evidence Centre: http://tinyurl.com/yjeq7hm The first Guardian report (24 Feb 2009) is a reprinting of the press release: http://tinyurl.com/bmveew I could find surprisingly few reports from the British press on this study. A later article in the Guardian (http://tinyurl.com/d8hu7v) and the report of the study in The Times did not go beyond anything stated in the press release (in other words, they were not in the least "sexed up"). Moreover the Times report carried some caveats: "The study, being conducted by the University of Oxford's highly respected Cancer Epidemiology Unit, isn't without its limitations. Strictly speaking, its findings apply only to middle-aged women. Certain types of people might also be more likely to volunteer for such studies, which might make the sample unrepresentative. Since most of the information is “self-reported” rather than collected objectively, can we really believe what people say about themselves? Critics point out that women embarrassed about their alcohol consumption are likely to write down that they are drinking significantly less than they really are. "And could the rise in cancers among women who drink be attributable not to alcohol itself but some other characteristic associated with people who drink, that they generally live less healthy lives than non-drinkers for example? It's possible, but the researchers have statistically weighted their findings to take account of such possible confounding factors…." http://tinyurl.com/c3sk25 (N.B. The NHS report on the study also included similar caveats: http://tinyurl.com/cqsxbf) There were shorter reports in Daily Telegraph and the Daily Express, neither of which went beyond the statements made in the press release: http://tinyurl.com/yegxknx and http://tinyurl.com/y8g5lk3 So let's see how Chris's assertions stack up: >…the scientists had sexed it up for the university press team, > who had then re-sexed it up for the new media, who had then > re-re-sexed it up for public… Judging from the Abstract of the paper, I see no evidence that the University press release sexed up the study (it helpfully provides statistics from which one can judge the general statistical claims), nor that the reports in the British press sexed up the press release. Chris again: >when in fact the actual increase in the breast cancer rate was >something like 2 in 10,000, By my reckoning it was actually around 1 in 1000. Here are the statistics: http://tinyurl.com/yjeq7hm "The risk was most increased for breast cancer. In developed countries like the UK, the chance of having had breast cancer by the age of 75 is 9.5 in 100. According to the study, for every extra daily unit of alcohol (over 2 a week), that risk increases by 1.1 per 100. So if you had a roughly 9.5 percent chance of getting breast cancer by the age of 75, but you drank one glass of wine a day, that risk would go up to 10.6 percent. If you drank two glasses of wine a day, that would increase to 11.7 percent." My calculation gives: Chance of getting breast cancer up to age 75 is approximately 1 in 10 Moderate drinking gives 1% increase, i.e., 1% of 10% = 0.1% increase = 1 in 1000 This is considerably higher than Chris's 2 in 10,000. N.B. The BBC News online report seems to give a much higher rate than this, which I (slightly) paraphrase as follows: "The rates for individual cancers varied, with one drink a day causing a 12% rise in the risk of breast cancer… On a population scale, this would mean 11 extra cases of breast cancer diagnosed for every 1,000 women…" http://news.bbc.co.uk/1/hi/health/7906355.stm Can any TIPSters check out the BBC report to help me reconcile this latter statement with my conclusion of a 1 in 1000 increase? Chris adds: >and there was little reason to believe that alcohol, rather than >the billion or so things correlated with increased alcohol consumption, >was responsible even for this tiny increase. As the NHS report indicates, the researchers did allow for some possible confounding factors. Of course this kind of problem applies to all epidemiological studies, but I think it would be more fruitful if Chris were to suggest some other possible confounding factors rather than indulging in unspecific hyperbole. Conclusion: Chris's assertions about the reporting of the study in question (at least as far as the British press is concerned) are considerable overstatements. The reports in question are not at all comparable to the one cited by Stephen. Just one final point. No TIPSter is likely to dispute that there is frequently poor reporting of health studies in the press (though this may be exaggerated by highlighting the worst reports and ignoring the more responsible ones), but to be fair to the journalists in regard to one highly relevant issue, they are not alone in frequently failing to grasp the significance of statistics. Questioning of doctors in recent years reveals that a goodly proportion are far from adept at making sense of health statistics, for instance on the question of false positives and false negatives in screening tests. Allen Esterson Former lecturer, Science Department Southwark College, London http://www.esterson.org ---------------------------------------- From: Christopher D. Green <chri...@yorku.ca> Subject: Re: Cannabis damages young brains Date: Mon, 28 Dec 2009 10:19:37 -0500 sbl...@ubishops.ca wrote: >>There's nothing surprisingly >>egregious about this particular article, is there? >Yes. I've never seen a university press release, which should >have been vetted by the authors and presumably ran with their >approval, hide the fact that the research was in animals. I'm still surprised. Without actually going back a checking press releases, I can recall the case of the "moderate drinking causes breast cancer" announcement in Britain earlier this year, in which it seemed pretty clear that the scientists had sexed it up for the university press team, who had then re-sexed it up for the new media, who had then re-re-sexed it up for public (when in fact the actual increase in the breast cancer rate was something like 2 in 10,000, and there was little reason to believe that alcohol, rather than the billion or so things correlated with increased alcohol consumption, was responsible even for this tiny increase). Sorry to be so blase about the whole thing, but far from being unusual, it is endemic Chris -- Christopher D. Green Department of Psychology York University Toronto, ON M3J 1P3 Canada --- To make changes to your subscription contact: Bill Southerly (bsouthe...@frostburg.edu)
