Does RoundUp cause cancer?

(Quick answer: probably not. See my update at the bottom of this post.)

For many years, environmental activists have been concerned about the herbicide glyphosate, which is the main ingredient in RoundUp®, the world's most widely-used weed killer. Since 1996, global usage of glyphosate has increased 15-fold, in part due to the widespread cultivation of "RoundUp Ready" crops, which are genetically modified to be resistant to RoundUp®. This allows farmers to use the herbicide freely, killing undesirable weeds without harming their crops.

RoundUp®'s manufacturer, Monsanto, has long claimed that glyphosate is safe, and they point to hundreds of studies that support their argument.

Nonetheless, a new study raises the question again.

First let's look briefly at another recent study. A bit more than a year ago, in November 2017, a large study in the Journal of the National Cancer Institute looked at nearly 45,000 glyphosate users (farmers and other agricultural workers who apply glyphosate to crops). These "users" have a much higher exposure to RoundUp® than ordinary people. That study concluded:
"no association was apparent between glyphosate and any solid tumors or lymphoid malignancies overall, including NHL [non-Hodgkin lymphoma]."
They did find, though, that there was a trend–not quite significant–towards an increased risk for one type of leukemia, AML. This trend appeared in users who had the highest exposure to RoundUp®.

In the new study, by a group of scientists from UC Berkeley, Mount Sinai School of Medicine, and the University of Washington, the authors (L. Zhang et al.) decided to focus exclusively on people with the highest exposures to glyphosate. They point out that including people with low exposure, who might have no increased risk of cancer, tends to dilute risk estimates. Statistically speaking, this is undeniably correct, but it also means that their results may only apply to people with high exposures, and not to ordinary consumers.

The punchline from the new study: people with the highest exposure to glyphosate had a 41% higher risk of non-Hodgkin lymphoma.

One caveat to this finding is that it's a meta-analysis, meaning the authors did not collect any new data. Instead, they merged the results from six earlier studies including over 65,000 people, and they focused on those with the highest exposure levels.

Meta-analyses can be prone to cherry-picking; that is, picking the studies that tend to support your hypothesis. However, I couldn't find any sign of that here. The authors include a frank assessment of all the limitations of their study, and they also point out that multiple previous studies had similar findings, although most found smaller increases in relative risk. In the end, they conclude:
"The overall evidence from human, animal, and mechanistic studies presented here supports a compelling link between exposures to GBHs [glyphosate-based herbicides] and increased risk for NHL."
A couple more caveats are important. First, this finding is all about relative risk. Non-Hodgkin lymphoma is one of the most common cancers in the U.S. and Europe, but the lifetime risk for most people, according to the American Cancer Society, is just 1 in 42 (2.4%) for men and 1 in 54 (1.9%) for women. A 41% increase in relative risk increases those numbers to 3.4% (men) and 2.6% (women).

Second, this higher risk only applies to people with very high exposure to glyphosate: primarily people who work in agriculture and apply RoundUp® to crops. Ordinary consumers (including people who eat "Roundup Ready" crops) have a far, far lower exposure, and dozens of studies have failed to show any increased risk of cancer for consumers. For most of us, then, this new study should not cause much concern, but for agricultural workers, it does raise a warning flag.

[Update 18 Feb 7:45pm] After I posted this article, the scientists at the Genetic Literacy Project pointed me to Geoffrey Kabat's piece about the Zhang et al. study. Kabat did a deep dive into the studies that Zhang et al.'s work is based on and uncovered a critical flaw in the study, one that I hadn't found. More than half of the "weight" of the meta-analysis by Zhang, and by far the largest number of cancer cases, come from a single study by Andreotti et al. published in 2018. That study reported risks for 4 different time points: 5, 10, 15, and 20 years. It turns out, as Kabat reports, that only the 20-year period showed any increase in risk of cancer. The relative risks of cancer at 5, 10, and 15 years were actually lower in the group exposed to glyphosate, and yet Zhang et al. didn't mention this fact.


Now, no one thinks that glyphosate lowers the risk of cancer, but Zhang et al. did not report that they had cherry-picked in this way. At a minimum, they should have reported what their findings would be if they used the other time periods. I suspect that they'd have found no increased risk of cancer–but this wouldn't make for such a catchy headline. This omission on their part is a serious flaw, indicating that they (and their results) might have been unscientifically biased.

The bottom line: even in those with very high exposures to glyphosate, the evidence that it causes any type of cancer is very weak. And for ordinary consumers, there's nothing to worry about.

The NY Times is far too worried about 23andMe's genetic test

The New York Times decided to publish an editorial this weekend warning people to "be careful about 23andMe's Health Test." What are they worried about?

Although the NY Times article is accurate, the warning suggests that 23andMe misleading its customers somehow. Is it? I decided to take a look.

 I'm a customer of 23andMe, and I'm also a researcher in genetics and genomics, so I know quite a bit about how their technology works and about what it can reveal. I've looked at 23andMe's latest genetic health reports, and they are remarkably clear and accurate.

Let me illustrate by revealing part of my own genetic test results. First I looked at my results for BRCA1 and BRCA2, the two genes that the NY Times article discusses.

I don't have any of the harmful mutations in the BRCA genes, which 23andMe reports like this:
Notice that they immediately provide the caveat that more than 1,000 other variants in these genes have been linked to cancer risk, and they make it abundantly clear that they didn't test any of those. Here's what the NY Times article said:
"The 23andMe test can miss other possible mutations.... there are more than 1,000 other BRCA mutations that contribute to your breast cancer risk. The 23andMe test doesn’t look for any of them."

A reader of the Times might think, upon reading this, that 23andMe somehow hides this fact. But the Times article's warning is little more than a paraphrase of what 23andMe's website states.

The Times editors also caution that
"Just because you test negative for the few mutations that 23andMe screens for doesn’t mean that you won’t get breast cancer."
Duh. 23andMe explains this as well, and much more, such as:
"This test does not take into account other risk factors for breast, ovarian, prostate, and other cancers, such as personal and family health history." [from 23andMe]
23andMe also provides a wealth of information about the BRCA genes, including links to the scientific papers describing the genes and their link to cancer. I was very impressed by how thorough they are.

The Times editors focused only on the BRCA genes, but 23andMe also tests a handful of others (9, in my case). I looked at the APOE gene report, which has a mutation that has been linked to Alzheimer's disease. The bad variant is called ε4, and fortunately I don't have it.

Once again, the 23andMe site was very clear about what this means, providing a detailed table showing the risks for people with and without the mutation. In my case, they tell me that:
"Studies estimate that, on average, a man of European descent has a 3% chance of developing late-onset Alzheimer's disease by age 75 and an 11% chance by age 85."
Looking at the detailed table, one learns that if you have one copy of APOE ε4 variants, your risk of developing Alzheimer's by age 75 is 4-7%, and if you have 2 APOE ε4 variants–the worst case–then your risk jumps to 28%. The website provide links to 10 scientific papers with far more detail, for those who want to know the basis of these numbers. This is far more than most people will want to know, and I couldn't find any flaws in 23andMe's description of the science.

The Times editorial concludes with this:
"23andMe has said that its health tests can raise awareness about various medical conditions and empower consumers to take charge of their health information. But doctors and geneticists say that the tests are still more parlor trick than medicine."
That last statement is the most egregious misrepresentation by the Times editors. Who are these geneticists who call DNA testing a "parlor trick"? The genetic tests run by 23andMe, which use technology that is run daily at thousands of labs around the world, are nothing of the sort. They are a highly accurate assay that has been repeated millions of times and validated by hundreds of peer-reviewed studies. Usually the NY Times is one of the most thorough and accurate sources in the media, but they really dropped the ball this time.

The fact is, genetics is not fate. Identical twins, who share identical DNA, rarely die of the same causes. Thus even if you knew your genetic risks perfectly, for every mutation in your DNA, you might not find anything to change in your behavior. At most, you might learn that you should get mammograms or colonoscopies slightly more often. It's legitimate to argue that you won't learn anything useful from the 23andMe tests, but you will learn something about genetics.

As far as changing your behavior to reduce health risks, you don't need a sophisticated genetic test for that. Just eat more leafy greens.

[Note: although I'm a customer of 23andMe, I have no financial relationship with the company and I'm neither an advisor to nor an investor in them.]