In vino veritas — Part One
Within days of each other, two major news stories from Europe have reported sensational claims about alcohol and cancer. It appears no one has bothered to go directly to the studies behind the news and read them. Or, perhaps, media and governmental health ministries, don’t understand them or realize that not all studies are created equal. Just because something calls itself a study doesn’t mean it actually studied anything, was designed to soundly test an hypothesis, or that it actually found what is being claimed.
The first news story came from the French National Cancer Institute (INCA). It claimed that a single glass of wine or alcoholic drink a day doubles one’s risk of getting cancer. A single glass! “Small daily doses of alcohol are the most harmful,” the president of INCA told French consumers, adding: “There is no amount [of wine], however small, which is good for you.” With the INCA report, the French Ministry of Health published guidelines advising the public that “the consumption of alcohol, especially wine, is discouraged.”
The French government’s latest pronouncement suggesting that French citizens should stop drinking wine altogether has been met with considerable derision, as you can well imagine!
It’s not only oenophiles who argue the healthful benefits of wine, of course. Even the American Heart Association’s evidential reviews found that the body of evidence supports lower overall mortality associated with alcohol consumption. And, any science-minded, intelligent person knows that when someone is making a scary claim that there is “no safe exposure” and that any amount at all is dangerous just because extreme exposures may be unhealthful has fallen from sound reason. Such efforts to terrify people violate the most fundamental laws of science and biological plausibility: the dose makes the medicine or the poison.
INCA said in the press this week that “alcohol is now the second most avoidable cause of death after tobacco.” It was quick to add that the French’s love of red meat, fatty charcuterie and salt was also bad for them. So much for the French paradox, it would seem.
Going to the source
It turns out that the source of this news story, not surprisingly, was a “Communiqué de presse 17 février 2009” — a press release issued on February 17 by INCA. In French, it explained that at the request of the government, the National Nutrition and Health program was launched in 2001 to focus on food and physical activity among the population. Coordinating with a variety of food, preventive health and cancer stakeholders, in 2003, it had released a publication — Food, Nutrition and Cancer: truths, assumptions and misconceptions — summarizing the purported relationships between cancers and diet, obesity and lifestyle.
This week’s press release announced the re-release of this report. We’ll look at it in a moment. But before this latest report…
In November 2007, INCA had also published a report specifically on alcohol and cancer, titled “Alcool et risques de cancer: état des lieux des données scientifiques et recommandations de santé publique.” It reported that alcohol increases the risk of cancers in men and women, based on epidemiological studies it claimed provided a “convincing level of evidence.” It also asserted that the “scientific literature shows that the risk of cancer increases linearly with dose provided by ethanol alcohol beverages without a threshold effect. In other words,” it said, “even moderate alcohol consumption increases the risk of cancer.” In its recommendations, this report said that regular consumption of alcohol should be avoided because the increased risk is significant even with moderate drinking.
Few people, and certainly not public health policy officials, it appears, had bothered to read the report, let alone go directly to any of the studies referenced as providing evidence in support of its assertions. This report’s findings had been based on eight case-controlled (observational) studies, which it claimed had found outstanding increases in relative risks for cancer associated with alcohol, and five meta-analyses.
Reading the actual research. To give you an example of what this report considered “convincing” evidence to back its claims, we’ll look at the largest case-controlled study it referenced. This study, published in the International Journal of Cancer in 1999, had been designed and coordinated by the International Agency for Research on Cancer (IARC) in Lyon, France and led by cancer epidemiologists in Barcelona, Spain. It included five observational studies on a total of 830 esophageal cancer patients seen between 1986 and 1992 in South American hospitals in Argentina, Brazil, Paraguay and two in Uruguay. The cancer patients selected for this study were in their mid-60s, more than half had 3 or less years of schooling, 82.7 percent had smoked and most (78.4 percent) had drunk alcohol.
Already, you can see that these patients are not typical of citizens in most developed countries, nor were these patients well “matched” with the 1,779 controls. Not being randomized studies, the control groups chosen for comparison had higher education levels, were slightly younger and consisted of considerably fewer people who smoked or drank alcohol (40 percent fewer had ever had any alcohol, for example). As is readily apparent, there are immeasurable ways (confounding factors) the cancer patients probably differed from the control groups. Yet, in looking for correlations between cancer and drinking and smoking, the authors didn’t consider ANY other confounding factors.
Despite this allocation bias, the Spanish authors still found no correlation between the amounts of pure ethanol or years of drinking and esophageal cancer in women, only in the amount of alcohol consumed among the men. But rather than recognize that other factors may be at work here, alcohol was considered the culprit. To illustrate how uncredible and capricious the link attributable to alcohol itself and the men’s esophageal cancer, there was no correlation among the men between esophageal cancer and the number of years the men had been drinking alcohol. And men who waited until age 66 or older to quit drinking had no higher risk for cancer but those who stopped drinking between ages 48-65 doubled their relative risk. As we see daily — with studies finding something good for us one day and bad for us the next day — epidemiological studies can find all sorts of meaningless, contradictory and nonsensical correlations.
Compiling the five groups, the authors found untenable relative risks between esophageal cancer and any of those who had ever drank alcohol — which, of course, described nearly everyone in their study group. The relative risks (95 percent CL) were 1.3–3.3 in women and 2.7-4.9 in men. Given the weaknesses in the study design (clearly failing as a “fair test”), these correlations were untenable. And, in the actual published study, even the original Spanish authors had concluded: “Our data show that light-to-moderate drinking (i.e., 1 to 3 drinks per day) without smoking does not substantially increase the risk of esophageal cancer.”
This isn’t meant to frighten you about what government public health officials consider convincing evidence to make the exact opposite conclusions, but you really do deserve to know what the evidence really shows, not what they claim it shows.
So, this week, the new edition of the nutrition and cancer report from INCA was released, saying it was written to provide healthcare professionals with “updated, documented and recognized” scientific research linking alcohol, diet (namely salt, red meat and sausages) and weight to elevated risks for several cancers. It called for the need for health professionals to support interventions to get everyone to diet, exercise and stop drinking.
You won’t believe what the new, updated science was that necessitated a new edition of this booklet.
As INCA explained, since its first edition in 2003, the Expert Report of the World Cancer Research Fund and the American Institute for Cancer Research had been published in November 2007. In establishing the evidence of the relationship between diet and cancer, INCA said, the 2007 WCRF/AICR report had been based on 7,000 original scientific articles, systematic reviews and hundreds of meta-analysis conducted by nine international centers, and then subjected to an independent evaluation by 21 international experts. In light of its findings, INCA said it was necessary to publish an update of its own report, which was indistinguishable from the WCRF/AICR Expert Report.
In other words, this is another adaptation of the WCRF/AICR Expert Report, mentioned in a thousand news stories earlier this week! Yes, the same 2-year old report that had not only failed to provide what most scientists would consider convincing evidence (in fact, its authors had created their own unique method of claiming to prove causation from correlations), but its conclusions were repeatedly the opposite of the evidence. The original JFS examination [here] found that none of the 17 cancers examined in the WCRF/AICR Expert Report had found credible associations with foods (sugars or sugary drinks, fats, meats, fruits, vegetables, legumes, vitamins, alcohol or processed foods), nor was there a credible association between all cancers and obesity or BMI.
In this INCA report, the authors said that alcoholic beverages have recently been classified as a human carcinogen by the International Center Cancer Research.(!) Based on “relations considered convincing” in the meta-analysis of epidemiological studies in the WCRF/AICR report, the authors estimated an increased relative risk of cancer per glass of alcohol consumed each day of 168 percent for oral cancers of the mouth, pharynx and larynx, and of 28 percent for esophageal cancer (Table 1).
Remember, relative risks are the tool of agendas
These relative risks sound enormous and scary to lay people who don’t understand that these correlations are derived from epidemiological studies and aren’t even tenable — more than random chance, statistical error and due to confounding factors. [Tenable risks were explained here.]
But let’s pretend the relative risks these computer modelings found are real. They would translate to clinically meaningless actual (absolute) increased incidences.
According to the National Cancer Institute’s Surveillance, Epidemiology and End Results (SEER) Program, the foremost source for cancer statistics in the United States, age-adjusted incidence rates of all oral cancers combined among the U.S. population is 0.0108 percent, with a mortality of 0.0026 - 0.004 percent (and has shown a steady drop among all population demographics since 1975). So, the 168 percent increased relative risk associated with drinking a glass of alcohol every day would theoretically raise actual risks of getting an oral cancer by 0.016 percent and dying of it by 0.004 percent.
According to SEER, overall age-adjusted incidence rates of esophageal cancer in the United States is 0.004 percent — virtually identical since 1985. So, a 28 percent increase in relative risk associated with drinking a glass of alcohol every day would theoretically raise the actual incidence by 0.001 percent.
And if, as the WCRF/AICR new policy report released earlier this week claimed, “healthy” eating, weight loss and exercise could really lower the percentage of esophageal cancers by 69 percent (another untenable statistic from epidemiological correlations), that would equate to a reduction in actual risk of 0.002 percent.
These are all clinically meaningless values.
But reporting actual risks don’t have nearly the same scare value as relative risks, do they?
Based on alcohol consumption in France, the INCA report tallied all of the untenable relative risks for all of the cancers (science doesn’t really work that way, untenable numbers don’t become more credible by adding them up) to estimate that 10.8 percent of cancers in men and 4.5 percent of cancers in women were caused by alcohol. While alcohol consumption has been declining in France since the 1960s, the INCA report stated, the French still have among the highest alcohol consumptions in the world — sixth in the world and fourth largest in Europe, according to the World Health Organization.
The public health message was that the French are drinking too much wine and it’s bad for them — but did anyone bother to see if there is even any correlation between the amount of wine the French drink compared to other countries and cancer mortality rates? It turns out, based on World Health Statistics 2008 and World Statistics of per capita wine consumption by country**, there is no association at all. Without even a link, wine can’t possibly have a relevant causal role for population health.
The world’s highest cancer mortality rates, for example, are in Bolivia (256/100,000), with an average life expectancy of 66 years — yet it has about the lowest wine consumption (0.28 liter/year). Grenada’s cancer mortality rate is also high at 199/100,000, and its people have a life expectancy of only 68 years; yet they drink only 2.28 liters wine per year per person. [By comparison, most European and developed countries, including the U.S., share similar life expectancies of 78-82 years.] Contrast that with Switzerland’s low cancer mortality rate of 116/100,000 to its higher consumption level of nearly 40 liters of wine each year per person. Italians and Americans have the same rates of cancer mortality (134/100,000) — which are about the world averages — yet very different wine consumptions, with Italians drinking 48.16 liters of wine apiece each year compared to the mere 8.69 liters in the United States. The French and British have similar cancer mortality rates, at 142 and 143/100,000, yet wildly divergent wine consumptions, too. Individuals in the UK drink only 18.97 liters/year of wine compared to 55.85 liters among the French.
In other words, no one can credibly blame the wine!
Next, we’ll look at the other study in the news this week, telling women that if they don’t want to get cancer, they may not want to drink anything…
© 2009 Sandy Szwarc
** Vatican City, by the way, has the world’s highest wine consumption per capita, at 62.02 liters per year.