Good data (and good conclusions) on (not much) arsenic in CA wines

Short: A new article in the American Journal of Enology and Viticulture publishes data showing that California wines contain honest-to-goodness insignificant amounts of arsenic. Even if you can’t call the authors completely disinterested, they’ve published their data openly in an excellent scientific journal (and that’s a lot better than you can say for the arsenic fearmongers).


At the risk of bringing up what must surely be everyone’s least favorite wine story from 2015, I’m writing once again about arsenic in wine. I hope you’ll agree with me that, given the reason, it’s justified. A group of reasonably trustworthy researchers has published a scholarly research article demonstrating that California wines don’t have an arsenic problem, and while their work won’t put the arsenic-wine baby to bed once and for all, it goes a long way in the right direction.

My denigration of the perpetrators a year before the scandal broke is here, and my response to the scandal itself is here.

Two problems have lain at the bottom of all of the talk about arsenic in wine. One is that no one – not the US government, not the World Health Organization, and not expert toxicologists talking amongst themselves in the peer-reviewed toxicology literature – is sure how much arsenic is too much arsenic. The other is that the people talking about arsenic in wine haven’t been transparent with their data. This article doesn’t help with the first problem, but it does help with the second.

In short, two independent wine testing labs ran arsenic stats for 28 wines named in the arsenic-related lawsuit brought against California wineries, plus 73 other California wines bought essentially at random* from stores in Pennsylvania and New York. The results:

Wine category                               Mean arsenic concentration                      Standard deviation

Reds                                                6.75 µg/L**                                                    7.33 µg/L

Whites                                           10.9 µg/L                                                         11.0 µg/L

Rosés                                             27.2 µg/L                                                        16.9 µg/L

The first and most important thing to observe about these numbers is that they’re all dramatically lower than the 100 µg/L guideline Health Canada has set for wine and even more dramatically lower than the 200 µg/L the International Organization of Vine and Wine (OIV) has set for its members (of which the United States is not one, but which includes nearly every other significant winemaking country around the globe). (The United States hasn’t regulated arsenic in wine.) Even the wine with the highest concentration of arsenic in this study clocked in at 68.4 µg/L. Yes. Some of these numbers are higher than the 10 µg/L the US Environmental Protection Agency allows in drinking water. But do you drink wine the same way you drink drinking water? Compare wine to food, not to water.

The second thing to note is that, generally speaking, cheaper wines had higher arsenic concentrations. That’s both interesting and not unexpected. (Higher concentrations in rosés and whites than reds, incidentally, isn’t a factor of price. The white >> red phenomenon has been observed in other studies and is assumed to have something to do with processing, though we’re unsure what.)

The third thing to know is that the wines cited in the lawsuit had higher arsenic concentrations (25.6 µg/L) than the random wines (7.42 µg/L), but even the lawsuit-associated wines were still dramatically lower than the 100 µg/L guideline. The wines in the lawsuit were cheaper, generally speaking, than the average wines.

And the fourth thing to note, once again, is that you don’t need to be worried about arsenic poisoning from wine. These researchers did something I think is silly and calculated “average daily” arsenic exposure from average wine consumption data rather than calculating exposure for a heavy-imbiber. To generalize just a bit from their data, the average American drinking two five-ounce glasses of Californian wine a week, every week, will get no more than 10% of the arsenic in her diet from wine, even if she’s drinking the cheap stuff. Arsenic is naturally present in the environment. It’s in water, rice, meat, beer, and bread. If you smoke, you’re getting arsenic in tobacco.

I’m not willing to call this research “unbiased.” Research is never unbiased; scientists can’t escape having a point of view any more than you or I, and it colors the scientific process from the questions they ask to the methods they choose, the way they read their instruments, and the conclusions they draw even if they’re trying in all earnestness to be objective. The labs behind this paper have also run tests and provided evidence for the wine industry in the past, and it’s presumably in the authors’ best interest to conclude in the wine industry’s favor. For that matter, it’s also in the best interest of the publishers of the journal, the American Society for Enology and Viticulture, to side with California’s wineries. All of that said, everyone’s long-term reputations here are best-served by being fair and transparent and there’s every good reason to think that they’re doing just that. And that’s a lot more than can be said for the arsenic fearmongers BeverageGrades.

*The researchers used a “convenience sample,” which in this case means that instead of trying to devise some strategy for obtaining a “representative” sample, they used what they could find. Convenience samples are sometimes a real problem in research (when a psychologist uses the freshmen in his psych 101 class to represent the American public, for example), but here this strategy makes a lot of sense: these are wines you’re going to commonly find as a consumer.

**µg/L = micrograms of arsenic per liter of wine

Why science may not have “the truth” about red wine and health

Vox published an article by Julia Belluz last week entitled “The Truth About Red Wine’s Health Benefits,” and while Belluz did a credible job with the topic, the topic was critically flawed from the outset. Titling an article “The truth about X” is gutsy. First, you’re saying that there is a truth about X, just one of them in fact. Second, you’re saying that you know what that truth is. Or that your sources know it. Getting anywhere with the wine-and-health story means trusting that the scientific studies and researchers she consulted could provide “the truth of the matter.” They can’t.

There are two major problems with scientific studies on “the truth” about wine and health. The first is about definitions. The second is about measurements. Both are about how science works and the kinds of questions scientists ask.

Definitions are important because where you start ultimately determines where you can end up. The scientific method depends on asking questions and then answering them carefully and systematically. As even non-scientists know from personal experience, the kinds of questions you ask are crucial for influencing the kind of answers you get.

“Hey, honey, would it be possible for you to take the baby sometime this week?”

“Hey, honey, will it be easier for you to take the baby Wednesday or Thursday? I need to schedule a meeting.”

When an article this year on alcoholic beverage labeling asked whether labeling has “improved,” they had to start out with some definition of what “improvement” meant, and the idea that alcoholic beverages are drugs about which the consumer needs to be warned in the first place. This idea that wine is a drug shows up in social research, and it shows up in wine health research. It’s no wonder. The United States regulates wine and other alcoholic beverages through the Alcohol and Tobacco Tax and Trade Bureau, the TTB, not the Food and Drug Administration (FDA). To the US government, wine and beer are more like guns and cigarettes than like food. That affects what kind of funding researchers can obtain for alcohol-related research.

Research often starts with this idea of wine as a drug and ask: when is wine a useful drug? When is wine a harmful drug? If we start off thinking of wine as food instead of as medicine, that approach doesn’t really make sense.

Belluz quotes a Harvard physician-scientist to the effect that wine is only useful for heart disease and diabetes. When drugs go up for approval with the Food and Drug Administration, they’re approved to treat specific conditions under limited circumstances. You may have heard about “off-label” drug use, when a physician prescribes a drug for a condition it hasn’t been approved to treat. In our Western medical system, drugs treat conditions, not people.

This is a silly way to think about drugs, but it’s an especially silly way to think about food. When you ingest something, you may be thinking about its effect on your breathing, or a skin rash, or your ability to think clearly, but in most cases what you care about at the end of the day is how your body functions as a whole. We’ve devised piles of drugs with nasty side-effects, in part because we fix disease rather than promoting health. Are some of these drugs wonderful? Do fixing disease and promoting health sometimes overlap? Of course. But the way we think about how drugs work and what they’re good for means that the focus is on fixing one broken thing. When research imagines wine as a drug, it thinks about its ability to fix one broken thing: heart disease, or diabetes, or colon cancer. If wine is a food, not a drug, the questions change.

Then there’s the measurement problem. As Belluz explains, since researchers can’t ask people to follow a proscribed drinking regimen for weeks or months, they have to make do either with very short-term studies or with observational studies of people who are already drinking or not drinking of their own volition. Something she doesn’t mention is that studies often also employ surveys in which people are asked to report their past drinking behaviors. As you might imagine, that strategy comes with it’s own set of issues. But in all of these situations, even for the experimental studies, scientists have to decide at the outset of the experiment what the relevant variables are and how to divide them into categories.

Alcohol-health research has tended to focus on how much people are drinking, maybe with some sort of questionnaire asking people to report how many glasses of wine or bottles of beer they drink on an average week. Those sorts of questions assume that all alcohol consumption is the same. What if it’s not? Medicine has recognized that binge drinking is a problem, and so we see recommendations that women consume no more than one drink per day on average, or two for men. According to the National Institute on Alcohol Abuse and Alcoholism, “low-risk drinking” means having no more than three drinks for women, four for men, in one sitting. Current science classifies the “risk” of alcohol as the risk of low, moderate, or heavy consumption. No doubt; amount is clearly important.

But what if how much you drink isn’t the only important consideration? Research is only just beginning to pay attention to the difference in long-term health effects between imbibing with a meal and having a habitual after-work drink or two in-between eating. Likewise, when studies ask whether red wine is better than white wine, they’re assuming that those are the most relevant categories. They also assume that, for health purposes, all red wines are more or less the same and distinctly different from all white wines, which again can be treated as more or less the same. What if the way the grapes were grown mattered? Or the age of the wine? Or the kind of food you chose to accompany it? Or the mood you were in while you were drinking? As its been designed, research simply wouldn’t be able to capture that information.

In other words, the best way to understand the roles wine plays in health and disease may not be the kinds of questions science is asking and the kinds of measurements science is making. The seeming contradictions among research findings suggests that they’re not. When research findings conflict, one or both of two things might be happening. Either scientists are making experimental mistakes, or experiments aren’t paying attention to the right factors, or both. An experiment can only control for and measure so many things. One or more of the factors being ignored may in fact be important to the outcomes of the experiment. But if it’s not being measured, that connection can’t be made. Instead, the effect of mystery factor X may be lumped into “unexplained variation:” results that the scientist can’t explain using the data she chose to collect. The seemingly endless parade of studies telling us first one thing and then the opposite about wine and health suggest that maybe scientists haven’t yet hit on measuring the important variables to measure.

And then there’s a third problem, not about the truth of wine health research, but about what it’s good for. What if research conclusively demonstrated that weekly consumption of three to five glasses of wine with food and up to two additional glasses without food decreased the risk of heart attacks for women over forty of European descent, but that any more than three glasses of wine per week increased the risk of colon cancer in European women over forty who have a family history of colon cancer. I’m making up the details, but these really are the kinds of conclusions epidemiological research on wine and health produce.

If you were a Polish-Irish woman over forty, what would you do? Would you keep those guidelines in mind, along with all of the other guidelines about diet and exercise and other lifestyle behaviors related to all of the other chronic diseases and cancers you might get? We know that too much stress isn’t good for mental or physical health, either, and managing all of those guidelines sure sounds stressful. Moreover, most cancers are random as far as we understand them; you may be able to increase or decrease your risk, but you’ll also certainly change the kind of lifestyle you choose to have. I hope that, instead, you would you try to eat moderately, exercise, and drink moderately – unless you had a specific reason to abstain, like a family history of alcoholism or religious belief or sheer personal preference? Or maybe you would choose to live an immoderate life because that’s your thing; moderation and prioritizing quantity of life isn’t for everyone. As Belluz notes, the bottom line on which all of the research tends to converge is that moderate drinking is okay and maybe even good for most people.

Wine is usually about reducing stress, not making more of it, about enjoying good food and good company. The Harvard physician-scientist Belluz quotes recommended (in Belluz’s words) “that exercise might well be a better way to get heart benefits compared with moderate drinking.” Sure. But since when has that been the point?

Why writing about wine and health is a dead-end

My August article for Palate Press is a brief update on some new research about wine and cancer. It’s a tricky subject. Trying to determine the relationship between two highly variable things is always tricky, and cancer and drinking are highly variable. Cancer comes in a lot of different forms – all breast cancers or colon cancers aren’t the same – and affects a lot of different kinds of people, and we don’t even know about all of the different factors that influence when and how it progresses. Meanwhile, people’s drinking habits are a lot more complex than those abstinence-light-moderate-heavy drinker scales make it seem. Do you drink wine, beer, spirits, or a combination? What kinds? Do you drink with meals, or alone? If you drink with food, what are you eating? Do you have a drink a day all week, or seven drinks all in one setting, and is your “drink” anything like my “drink?” Are you happy while you’re drinking, or sad?

The wine-and-health story, or the wine-and-cancer story, consequently has to be a lot more complicated than “drinking good” or “drinking bad.” As I point out in the Palate Press article, this is a good thing. We’re understanding enough about disease and lifestyle to stop doing the lifestyle modification equivalent of treating all ailments with leeches, recommending that everyone stop drinking because drinking is bad, and to start asking why and when drinking might be a bad idea.

Here’s my problem. Every time I write about wine and health, I find myself wanting to shorten the entire 1200-ish word article to one sentence: “Drink moderately, especially with food; don’t go overboard, and don’t worry too much about the whole thing.” Continue reading