In my initial foray in this series looking for facts behind the rhetoric on transgenic food, I reasoned that, since we can’t eliminate risk, we’d better have the right regulatory safety net in place.

As soon as I asked this, however, I was met with an apparent contradiction: Critics of genetic engineering say the industry is not required to do any safety testing. The developers of GE crops say they have to do a ridiculous amount of safety testing. Both are correct.

If you try to cross-check the claims of people on either side of the GM debate, you run into problems, because these warring clans speak different dialects. Their foundational assumptions point them in opposite directions, facing different landscapes and talking past each other. This can leave outsiders feeling that someone is lying. But often the miscommunication comes down to a difference in perspectives: One side says, “We’re right!” and the other — instead of explaining how that specific claim might actually be wrong — answers, “That’s poppycock. You’re left!”

Grist thanks its sponsors. Become one.

When I began reporting on regulation of GE crops, I thought it would be easy to fact-check the assertion that there’s no required safety testing for transgenic crops. Surely, I thought, it would take no more than one phone call. Actually, it took quite a bit more work than that — but I return successful from my quest, and I think I can sort out this debate once and for all.

First, some quick background: In 1974, Paul Berg, often called the father of genetic engineering, persuaded other molecular biologists to lay out the potential hazards of their work. These scientists imposed a moratorium on their research while they formed guidelines for the science going forward. Those guidelines were eventually translated into a set of bureaucratic hurdles. A genetic engineer has to get a permit from the USDA for field tests, and, after several years of trials, petition for deregulation of the crop. If the crop is pest-resistant, the EPA regulates it as pesticide, and demands more data. Finally, the FDA evaluates the plant for safety and nutritional content.

Grist thanks its sponsors. Become one.

All this sounds reassuring. Then I spoke with Michael Hansen, a critic of GM foods at the Consumers Union.

“There’s absolutely no requirement for any safety testing,” Hansen said.

I told him that didn’t make any sense: I’d just learned about all kinds of testing.

“Well, the companies are supposed to do a voluntary consultation,” he said, “but it’s voluntary. Look at what the FDA says when they approve a food: ‘It is our understanding that Monsanto has concluded this is safe.’ They just rubber-stamp it.”

This legalese is patently ridiculous, but I wanted to stay focused long enough to resolve this question of regulation. “But is it really voluntary in practice?” I asked. “Do you know of a case where a company did no safety testing?”

Hansen didn’t. In practice, companies generally do voluntary testing, he said, but there’s no guarantee that they are doing the right sort of testing. Hansen would like to see more long-term feeding studies that use the actual crop that humans are going to eat. Often, he said, instead of feeding rats the food, researchers will just feed them proteins from the genes they are adding. Furthermore, he added, the companies can do these studies over and over until they get results they like, and only show those favorable results to the government.

When I began talking to people on the other side of the debate, they seemed baffled by the assertion that testing was voluntary. They didn’t know what I (or Hansen) was talking about. Finally, I reached Val Giddings at the Information Technology and Innovation Foundation. He was once a biotech products regulator at USDA and worked closely with the FDA while it was enshrining the rules to govern transgenic foods, and understood how all this worked. The murkiness arises from a consultation process laid out in 1992, he said.

Here’s the deal: The process is technically voluntary, but in practice, absolutely everyone does “volunteer.” That’s because the FDA can stop any food (GM or otherwise) from going to market. It would be incredibly foolish for a company to spend a lot of money on a new breed of plant while thumbing its nose at the agency, Giddings said.

In other words, it’s voluntary — but if you don’t volunteer, the FDA is standing there to whack you with a very big stick. Moreover, regulatory review from the USDA and the EPA is mandatory in every sense — there’s nothing even legalistically voluntary about that. But it’s the FDA that’s most responsible for insuring that the food doesn’t harm people.

“In my opinion it is misleading to and past the point of dishonesty to claim that FDA does not require safety testing,” Giddings emailed me. “A Jesuit would blush at the rhetorical convolutions to which the activist opponents resort to make it seem otherwise.”

I pushed him on this. What if a company decided just to opt out?

The FDA, Giddings wrote, “frankly cannot really envision any circumstances under which anybody placing a ‘bioengineered’ food on the market would have the temerity NOT to consult with them.”

Hansen had left for a Codex Alimentarius meeting in Rome, but when I suggested by email that the FDA regulations were mandatory in practice, he shot back a response, disagreeing hotly. He restated his objections and added a salient piece of evidence: A report by former EPA scientist Doug Gurian-Sherman showing that companies had refused to hand over data that the FDA had requested. If FDA oversight was truly mandatory in practice, the agency should be able to get the information it wants, Hansen pointed out.

I read the report [PDF]. Gurian-Sherman had gotten documents relating to the FDA’s consultation process through the Freedom of Information Act. In six cases the FDA had asked for more data, and, half the time (three of six), the companies had said no. In two of those cases the FDA had asked for more information about the genetically engineered corn’s stalks (that is, the agency was satisfied with the safety of the corn itself, but it also wanted some data on the leaves, which are sometimes fed to livestock).

I could see why a company might resist pressure to test its corn stover, and why the FDA might let it off the hook: It’s hard to argue that leaves and stalks are a threat to human health. But the final case was more perplexing: The FDA had asked for more numbers on the BtCry1F protein. The company (Dow AgroSciences) had reported the average level, and the FDA suggested it could also provide the minimum and maximum amounts it had found in individual samples.

Why had Dow AgroSciences said no? It should be easy to look up the low and high values from the tests, and send those in. And it might be germane because an average could hide a few dangerously high maximums.

When I asked Giddings about this, he said that a company would only refuse the FDA if they thought the request was truly unreasonable. “These people are terrified of the FDA,” he said. “Usually their approach during these consultations is just to close their eyes and think of England.” (Dow AgroScience did not respond to my request for comment.)

But BtCry1F, he said, is utterly safe. It turns into the pesticide Bt that’s used in large quantities by organic farmers. It attacks structures in the insect gut that aren’t present in humans. In this case, the FDA must have simply agreed with the company that, no matter what the maximum, there was nothing to worry about, Giddings guessed.

He was right. When I asked Jason Dietz, a policy analyst at the FDA, about this case, he filled in the details:

“We said, you could improve on the submission by doing this [adding min and max numbers]. The company said, we understand, we see what you are suggesting, but we think that looking at the means [i.e. averages] would get at the same end point. And we said, yes, we agree, that rationale makes sense to us.”

If a company refuses to hand over information that the FDA really wants, Dietz said, the agency will ask more forcefully. And if it’s not satisfied with the data it gets, the FDA won’t sign off on the crop, he said. But that’s never happened. When the FDA insists, companies always have handed over the data.

“We haven’t been in a position of needing to write a letter saying that we think your product is unsafe and unlawful,” Dietz said. “And I think that speaks to the strength of the consultation process.”

I also asked Dietz about Hansen’s indignation over the FDA’s “rubber-stamp” way of saying that the safety determination is made by the biotech company rather than the agency. Dietz responded that this is simply a reflection of the process: It was set up so that the companies would do the testing and pay for it.

“It’s the manufacturer’s responsibility to insure that the product is safe,” he said. “It’s their safety assessment. What we do when we evaluate their safety assessment is we look to make sure the safety and regulatory questions have adequately been resolved, and then we say we have no questions about your safety assessment.”

The bottom line is that the FDA requires all foods to be safe, he said. Insuring safety has meant that every firm bringing a genetically engineered food to market has gone through the consultation process, and has done safety testing.

After I hung up with Dietz, the phone rang again. It was Doug Gurian-Sherman, the author of the report criticizing FDA regulations. (Gurian-Sherman is now with the Union of Concerned Scientists.)

“I gotta say,” I told him, “I really am beginning to think that it’s just incorrect to say that companies aren’t required to do any safety testing.”

This launched a conversation that lasted nearly two hours. I’ll spare you the rehashing of FDA policy and cut straight to the conclusion.

The issue isn’t in how much authority the FDA can exert — it certainly can, and for all practical purposes does, require testing. The issue is in how much authority the FDA chooses to exert. Since the requirements for safety testing aren’t spelled out in regulations, but instead determined on a case-by-case basis, Gurian-Sherman says it’s correct to say there’s no required safety testing.

The difference between Gurian-Sherman and Giddings on this comes down to their respective assessments of the risk. Giddings thinks genetically engineered foods are as safe as, or safer than, conventional crops, so the FDA testing looks onerous, and it seems obvious that it’s obligatory. Gurian-Sherman thinks that there’s a small, but very serious, chance that some future GE food could cause harm, so he’s troubled by the fact that there aren’t specific requirements written into the regulations.

So: Are people like Hansen and Gurian-Sherman right to say there’s no required safety testing? Or is Giddings right that the testing is de facto mandatory?

In a very real way, both are right. Still, after doing this reporting, I won’t be saying that there’s no required safety testing: saying so implies that GE food is unregulated and untested, and that simply isn’t the case. The FDA does a thorough job; it just focuses tightly on a few potential dangers, looking for allergens, amplification of toxins that naturally exist in the plant, and changes in nutritional composition.

When I asked Dietz why they didn’t cast a wide net, and ask for a rundown of all the compounds in each GE crop, he said: “An analysis like that would be very hard to interpret because it would capture everything, and it would be hard to identify those things that would be most important to safety. What we use is a focused approach that targets those things that we know, if they were changed, might be a problem.”

This seems eminently reasonable — if genetic engineering is just an extension of conventional breeding. But if GE foods are fundamentally different, the FDA’s approach appears, as Gurian-Sherman put it, “so thin that it would never raise a red flag if something were dangerous.” There are significant questions even about the problems the FDA focuses on, he said. (There’s a good overview of those concerns here.)

This understanding leads us right back to the question my last post left us with: Is there something fundamentally different about GE food? In my next post, I’ll be looking at the ways genetic engineering is similar and distinct from other forms of plant breeding.

P.S. I’m reading your comments (the thoughtful ones, at least), and I will, indeed, delve into many of the issues raised there. Gurian-Sherman told me that he is concerned about health risks from GE foods, but that’s last on his list, behind concerns about the environment, monopolies, and the direction of agriculture. I’ll get to all that stuff.

More in this series: