Preservatives, Cancer, and the Need for Courage

Remember Chicken Little?

One day an acorn lands on her head. She decides the sky is falling, so she rushes out to warn her friends. Each friend tells another friend. Soon everyone’s afraid.

Chicken Little wasn’t trying to deceive anyone, and nor were her friends. They genuinely believed the sky was falling. Given that belief, their fear was understandable.

This week I read the national news coverage of a French study linking six common food preservatives to cancer. Then I read the study.

What I found was a Chicken Little scenario: Researchers claiming that the six preservatives are dangerous, journalists echoing their claims, and no reason to doubt anyone’s sincerity. And yet, when I looked at the data, I found little cause for concern. At most an acorn, not atmospheric collapse.

Ultimately, we want to know whether our diets are healthy. At the end of this newsletter I discuss some of the health risks associated with food preservatives, and I offer some advice. But I think it’s also crucial to tease out what went wrong with this study.

Why? Because in it we find statistical problem that has fueled the so-called replication crisis in the social sciences and biomedicine. It’s a scientific concern that impacts what you might choose to eat – or not eat – for your next meal.

The new study

The new study was published on January 7 in the BMJ, one of the world’s oldest and most prestigious medical journals.

Briefly, Anaïs Hasenböhler (Sorbonne Paris North University) and her colleagues examined the development of cancer among 105,260 adult participants enrolled in the ongoing NutriNet-Santé project.

Back in 2009, when the study began, all participants were cancer-free. Hasenböhler and colleagues looked at whether intake of dozens of food preservatives predicted which participants would develop cancer (4,226 people did), while controlling for the effects of demographics, health, physical activity, smoking, diet, and more.

Methodologically, NutriNet-Santé is stronger than most large prospective nutrition studies. One reason is the frequency of food intake measurement – an average of 21 times per participant over a roughly 14-year period. Each time, people reported exactly what they’d eaten over the previous 24 hours, right down to specific brands of food products.

Data analysis focused on the 17 preservatives that participants consumed most. 11 were found to be unrelated to cancer; the table below shows the six that did appear to be carcinogenic.

You can find all of these names on product labels, as per FDA requirements, though more inclusive terms like “sulfites” may be used, and products containing vinegar rarely mention acetic acid.

But before you start checking labels, let’s look at little closer at the findings.

Participants were divided into three groups according to amount of each preservative consumed. In most cases, this meant the top vs. the middle vs. the lowest third.

Increases in risk of cancer were typically moderate. For instance, among the one-third of people who consumed the most acetic acid, rates of cancer were 12.4% higher than among those who consumed the least.

I like vinegar, as well as foods pickled in it, but that 12.4% statistic doesn’t worry me. Here’s why not:

Researcher degrees of freedom

Why did Hasenböhler and colleagues divide participants into three groups? Why not four? Or six? Why not treat preservative intake as a continuous quantity (1 mg/day, 2 mg/day, 3 mg/day, etc.)? The researchers provide no rationale.

In fact, there’s no scientific, statistical, or practical reason to compare three groups. The researchers could’ve just as well compared the highest vs. lowest 20% of participants.

Scientists make countless choices about how to gather and analyze their data. Any one of these choices may influence whether or not the results are significant.

When the choices are made arbitrarily, or in a way that favors significant outcomes, we have the “researcher degrees of freedom” problem.

One consequence is that a number of findings in social science and biomedicine have not been replicated – i.e., repeated – since attention to the problem began to intensify in the 2010s. The findings can’t be replicated because they stem from choices that artificially created significant effects.

In this study, a principled approach to analyzing preservative intake might’ve caused some or all of the significant effects to disappear. We have no way of knowing.

Digging deeper

Chicken Little sounds the alarm and her friends repeat it, without examining what fell on her head.

A close look at Hasenböhler and colleagues’ data reveals inconsistencies. In several cases, it’s the middle group, not the group highest in preservative intake, that shows the greatest risk. Biologically, this doesn’t make sense.

For instance, for potassium metabisulfite, total nitrites, and sodium nitrite, the highest rates of breast cancer were seen among those who consumed moderate amounts of the preservative, not those who eat the most. (You can see this in the screenshot below; for each preservative, the Hazard Ratio – a measure of relative risk – is higher for group 2 than for group 3.)

Arguably, this isn’t even the most serious “researcher degrees of freedom” problem in the study.

The elephant in the room

If you want to read food labels and change your diet, the new study provides little guidance.

Why not? Because the researchers’ estimates of preservative intake were crude and mostly concealed.

Food labels rarely note preservative quantities, and manufacturers rarely make the information public (because they’re not required to, so long as they meet FDA and/or international standards). As a result, Hasenböhler and colleagues had to cobble together estimates from a limited number of assays, combined with generic information gleaned from publicly available databases. No details are provided on how this was done.

How do you go from a database stating that most Chardonnays contain 80 to 150 ppm of sulfites, to the specific estimates used for each Chardonnay drinker in this study? Hasenböhler and colleagues don’t say. What estimates were used exactly? The researchers don’t tell us that either. The door is wide open here to researcher degrees of freedom – and error.

To put it in practical terms, we don’t really know how much of each preservative people consumed.

If we could trust the study, the only practical takeaway is that we should eat less of those six preservatives. How much less is unclear.

So, should we be examining food labels and adjusting our diets?

Food preservatives: Some practical advice

The CDC’s updated Dietary Guidelines for Americans, released earlier this month, contains exactly one statement about food preservatives:

“Limit foods and beverages that include artificial flavors, petroleum-based dyes, artificial preservatives, and low-calorie, non-nutritive sweeteners.”

Coincidentally, this is the same vague advice we might glean from the new study. And, somewhat coincidentally, it might even be reasonable.

Given limited scientific consensus on links between food preservatives and health, and given that food labels only list preservatives but don’t give quantities, we might just want to play it safe and avoid them when possible.

This feels sensible to me, with two qualifications:

1. The DGA’s focus on artificial preservatives is too narrow.

The natural vs. artificial distinction doesn’t apply to all preservatives, because chemically speaking they may be, or become, the same.

A critical example is that meat can be processed with synthetic nitrites or a natural substance (often, celery powder), but in either case, once we fry the bacon, the nitrites react with amines in proteins to form nitrosamines that are known to be carcinogenic.

You can’t get around this by purchasing “uncured” bacon, even though the labels on these products say things like “No nitrates or nitrites added.” This is a marketing ploy. It just means that a natural substance like celery powder was used to inhibit bacterial growth. Data suggest that nitrosamine levels end up about the same regardless of whether synthetic nitrite or celery powder was used.

Higher heat, as well as burning or charring, increase nitrosamine levels. You might keep that in mind when cooking meat – especially the processed kind. (Here’s a mnemonic: The crispier, the riskier.) But other compounds in processed meats are carcinogenic too, so maybe we should just eat less of these foods rather than focusing narrowly on cooking temperature or their preservative content. This leads me to one further qualification:

2. Most preservatives suspected of being carcinogenic are found in foods and beverages that contain other carcinogens too.

In the new study, most of the preservatives linked to cancer are found mainly in alcohol or processed meats, all of which contain other carcinogens (e.g., alcohol contains sulfites as well as acetaldehyde).

Other potentially carcinogenic preservatives include sodium benzoate, which forms benzene when exposed to ascorbic or citric acid and heated. Sodium benzoate is commonly found in sodas (particularly diet soda), and fruit drinks. But these products also contain sugars or artificial sweeteners, which, consumed in excess, contribute directly or indirectly to the development of cancer too.

In sum, the healthiest strategy might be to minimize consumption of processed meats, sodas, and alcohol, without stressing about how to define “minimize” (since we don’t know yet) or worrying that one particular constituent, such as a preservative, is an especially bad guy (since we don’t know yet).

Final thought

The CDC’s new Dietary Guidelines have been criticized – rightly so, I believe – for changing the recommendations on alcohol. Instead of advising adults to take no more than 1 or 2 drinks per day, we’re now told to drink “less”, and to “limit” our drinking. No quantities are specified.

This is unhelpful, because the science is clear that above a certain amount (roughly 2 drinks per day), the more alcohol a person consumes, the greater their risk of a variety of health problems. Most of the scientific debates, which I discuss here and here, focus on whether or not lower levels of intake can be harmful.

The Dietary Guidelines should’ve kept the original advice – 1 to 2 drinks per day, at most ­­­– while adding the qualification that according to some experts, no amount is safe.

As for preservatives, however, I think the vagueness of the Guidelines is suitable. The science is evolving, and until we know more, it’s probably wise to limit consumption of foods that contain preservatives – particularly, but not exclusively, those that are synthetic – even though we can’t be certain about which preservatives to treat cautiously or exactly how much to “limit” consumption.

It takes courage to acknowledge when we’re uncertain. We need to sidestep the overconfidence of TikTok influencers and quacks, and accept some degree of uncertainty around our food choices. Ideally we can do so without the fearfulness of Chicken Little and her friends. So, here’s my new mantra: Eat healthy, be brave.

Thanks for reading!