Anne Kapuscinski is a professor of sustainability science at Dartmouth College, who recently led a team of 53 scientists in writing a book on the subject of risk assessment science as applied to genetically modified fish.
I spoke with professor Kapuscinski about AquAdvantage salmon, a fast-growing genetically modified fish poised to become the first GM animal approved for human consumption. Under the proposal being considered by FDA, AquAdvantage eggs would be produced on Prince Edward Island, Canada, and grown to market size at a small, inland facility in Panama. None of the production steps would take place in the U.S.
All told, this approval process has dragged on for nearly two decades. One recent event that is referred to several times in this interview was a public meeting of the FDA’s Veterinary Medicine Advisory Committee (VMAC) in Sept. 2010. The VMAC’s purpose was to discuss the human health and environmental safety threats posed by AquAdvantage salmon. The environmental issues had been described in an environmental assessment (EA) that FDA had released a few weeks earlier.
Following the VMAC meeting, FDA was to make a decision: either move forward with that EA, presumably with input from the VMAC meeting and elsewhere, or, the agency would instead produce an environmental impact statement, which would take a more comprehensive look at the environmental risks posed by the fish. The FDA ultimately chose to stick with the EA, a new draft of which was released on Dec. 21, 2012. A public comment period began then, and will remain open until Feb. 25.
Note: This interview was conducted by phone. Small edits were made in the interests of readability and organization, with the subject’s approval. Also: I had hoped to place this interview in a news publication, but was turned down by every one that I pitched it to. So here it is.
Update: On February 13, the FDA’s comment period was extended by 60 days.
Update: Professor Kapuscinski’s comments to the 2010 VMAC, referred to in the interview, can be found here.
Update: On March 9, 2013, the New York Times ran an Op-Ed by Emily Anthes in support of AquAdvantage salmon. The story’s first sentence built a case that AquAdvantage’s maker, AquaBounty, is “the most virtuous entity on the planet.” So we could see where this was going. Further down, Anthes asserts:
“But scientists, including the F.D.A.’s experts, have concluded that the fish is just as safe to eat as conventional salmon and that, raised in isolated tanks, it poses little risk to wild populations.”
Professor Kapuscinski: “I want to start by saying my role in this public discussion, as a scientist, is to advocate for scientific integrity, scientific quality, and scientific reliability of the methods used for risk assessment and risk management. What I bring to this discussion is expertise in risk assessment science. I’m not here to advocate for or against the fish, but for the quality of the science in the FDA’s draft environmental assessment, which is especially important in this case because this is a precedent-setting case.”
Question: What is your biggest concern, should this EA be approved?
Risk assessment normally has three steps. One, you identify what the hazard is. Two, you figure out what the consequences would be to the environment if the hazard were to be realized. And the third step is, you say, “well, can we somehow manage the risk, can we do something to prevent the consequences from happening?”
What FDA has done, in both the 2010 draft and 2012 draft EA, is essentially skip to step three, the risk management step. Both versions are saying: We have so many different kinds of confinement, so we are concluding that the chances of any of these fish escaping and being able to establish a reproducing population is virtually zero.
The FDA has not changed much about that, compared to the 2010 EA, except for a few details. They’re still hanging their whole conclusion on risk management—that is, multiple confinement systems for the fish. But what they still haven’t done is what Dr. Sundström and I asked for, which is a quantitative failure mode analysis. If multiple confinement is what their entire conclusion really rests on, and if they are correct that the levels of confinement are so great all in combination, a quantitative failure mode analysis should actually come to that same conclusion. And it would be a much more scientifically reliable way of substantiating the conclusion. As said before in our written comments, it’s a standard practice in risk assessment and risk management to do a failure mode analysis, and it should be as quantitative as possible.
A key part of step two is a consequence assessment. That’s where one is asking, “if the fish did escape, what would happen, could that harm the environment?” In the 2012 draft EA, the FDA changed basically nothing from what was laid out as a consequence assessment in the 2010 EA, and it’s just as weak and just as scientifically unacceptable as it was in the 2010 draft. Dr. Sundström and I fully explained the scientific weaknesses in our 2010 comments. I recommend that they cut out the consequence assessment, unless they are willing to make all of the changes that Dr. Sundström and I had recommended two years ago.
Question: In addition to the consequence assessment, you also mentioned the failure mode analysis—of the confinement measures— that is missing. Could you talk more about failure mode analysis?
If you look at how the overwhelming majority of salmon are farmed in the world today, they are not farmed in highly confined facilities with all of these layers of confinement. So unless there is going to be an enormous transformation of the industry, it’s just hard to see how the company will be able to sell enough eggs to enough salmon farmers who would all be able to raise them at such high confinement.
The company’s facility in Panama is quite small; it’s not really at commercial profit scale, it’s at a scale to show proof of concept of the commercial viability of this. Once the company scales up to selling millions and millions of eggs, the fish will be farmed by producers with all kinds of facilities.
Maybe they’ll sell them to people who will farm them in net cages in China, or maybe even on land in China, but how do we know what the quality of the confinement will be? And you could argue, “well, it doesn’t matter, that’s a problem for China.” That, of course, would be a narrow way of thinking about environmental consequences.
Future farming of this fish is probably not going to happen in facilities that are as confined as the one in Panama. And even if future farmers try to have a lot of confinement, it gets harder and harder to make that work when you have larger-scale fish farms and tens or hundreds of fish farms around the world. So that’s why having a method of failure mode analysis in this precedent-setting document is so important.
Because I have not been to the Panama facility, I have to trust on paper what they’re saying about the confinement measures. And on paper, all of those different confinement measures look like the best that they could do for this situation. And like I did in 2010, I commend the company, AquaBounty, for having done multiple confinement measures. And I think it’s smart to use geographic confinement for one of their measures. It makes a lot more sense to try to grow out these fish in Panama than it would in a high-mountain area in the Sierra Nevada, California, where if the fish got into the river they would end up where there is suitable habitat for salmon.
That’s why I still urge them to do a quantitative failure mode analysis. If there was a quantitative failure mode analysis and it was laid out in the EA, then independent scientists could actually review that, sink their teeth into its results, and say yes, the analysis confirms the measures you have in place are the best you could possibly have.
Question: Beyond the positives you just mentioned that this application has going for it, do you see any improvements to the 2012 EA, compared to the earlier version you commented on in 2010?
In the 2012 draft, the FDA did make available more information to the public about possible storm surges and other questions that had been raised about vulnerability of the hatchery on Prince Edward Island. And for the Panama facility they gave more information about the conditions in the ocean immediately beyond the mouth of the river. They made a better case than they did before that, if the fish ever made it to the mouth of that river, the conditions are really inhospitable to salmon and it’s very unlikely that salmon would survive if they got to the ocean.
Beyond that, there were a number of small additions of information that strengthened some parts of the EA. The 2012 draft admits that there had been an outbreak of the virus ISA [infectious salmon anemia] at the Prince Edward Island hatchery. In the 2010 draft EA, there was no mention of that. Since that came to light, I think AquaBounty did a good job of showing that the facility is disease-free now.
That said, everything that is in this EA that looks at possible consequences of a fish escape is unacceptable to me. It’s very poorly done in terms of basic scientific standards in multiple ways. And if they just don’t want to do the work for a more scientifically acceptable consequence assessment, then I would recommend that the FDA simply deletes all of that from this environmental assessment.
I’d rather have them say, “We base our entire conclusion on multiple levels of confinement of the fish and we’re therefore concluding that the hazard is really not going to happen, the fish aren’t going to escape. We don’t have the data, we don’t have the studies needed for an environmental consequence assessment, and therefore we’re not going to do a consequence assessment.” And just leave that out. Because what’s in here is scientifically unacceptable.
Question: FDA has stated that other fish farms that wish to grow AquAdvantage salmon and sell it to the U.S. will apply for permission via a supplemental application to the current one, with details filled in from the fish farm in question.
One concern I have is that FDA is not required to share future applications with the public. The agency perhaps made a strategic decision that it would be in its interest to release some information to the public about this first application. We have no guarantee that applications for future, commercial-scale operations, which might have weaker confinement and be more risky than this one, will be shared with the public. Therefore if this application’s weak scientific standards for assessing consequences of fish escaping confinement are accepted, staff inside the agency could assume that this standard is OK. As an ecological risk assessment scientist, I don’t think that it’s OK.
The FDA doesn’t have to share any of this with the public because the law used to exercise its authority over the commercialization of genetically modified animals is the Food, Drug, and Cosmetic Act. And they’re using the drug provision, which has strict requirements for the agency to keep everything confidential. They’re not required to share information with the public. In this case, AquaBounty had to give FDA an OK to share some information with the public.
But without the applicant’s permission, the FDA can’t even divulge if it has received an application. There could be another applicant out there that has submitted another application for the commercialization of a genetically engineered animal, and we wouldn’t even know.
If FDA does approve these fish, the final environmental assessment is going to be the standard, it’s going to set the precedent for future approvals. So it absolutely has to have the best scientific reliability and quality, especially given that future applications may not be shared with the public.
Question: How has the EA changed between the one you commented on for the VMAC meeting and the version released on Dec. 21, 2012?
They didn’t change very much from the company’s version of a draft environmental assessment that FDA asked the Veterinary Medicine Advisory Committee to discuss in Sept. 2010. Dr. Fred Sundström and I submitted extensive scientific comments to the FDA for that committee’s meeting. The FDA didn’t change the basic structure of the argument from the 2010 version. Its conclusion of “no significant impact” hinges on multiple confinement measures to prevent fish from escaping or reproducing in nature.
What FDA did do was make a key strategic change. The agency came out clearly saying that the National Environmental Policy Act does not require it to assess effects on environments of other countries. So FDA is limiting the assessment to the question: “Could the production of these fish in the hatchery in Prince Edward Island and the small grow-out facility in Panama have effects on the environment of the United States?”
Question: You were not invited to be on the committee, your input was not sought, and your sole avenue to participate in the discussion was getting in the queue to comment as a member of the general public at the VMAC hearing in Rockville, MD. Given your two decades of work in the field of environmental risk assessment for GM fish, does this bother you?
I didn’t spend much time worrying about that because the VMAC has little power in this case. That committee was convened primarily so that the FDA would have a venue to share the then-current EA with the public. It would have been much more scientifically credible to have someone on the committee who has been working on the cutting edge of methods of ecological risk assessment science. Although I would have been one option, there are others they could have considered.
What I am more frustrated by is that Dr. Sundström, who does research on ecological risks of GM salmon, and I worked very hard to write scientifically thorough comments and submit them by the deadline. There was very little time from the day FDA released the notice until the deadline to submit written comments, just 13 working days I think. We worked very hard to write the most comprehensive and scientifically strong set of comments. What’s really disappointing is that, reading this new environmental assessment, it looks like either they didn’t read our comments or they just decided to ignore them. They changed so very little in the EA.
They didn’t do anything to the consequence assessment section, which is really what the bulk of our comments were about.
And they completely ignored our call for a failure mode analysis. They are still refusing to pay attention to the updating of ecological risk assessment science that’s all pulled together in a book published by a large group of scientists in 2007. Throughout both the 2010 and 2012 EA, the text cited two important publications that I led back in the early 1990s, one of which I was lead author and the other by a working group I chaired.
The FDA is hinging most of its scientific approach on the consequence assessment on those two reports. And yet I myself am now saying that they’ve been replaced by better methods.
Back in 1991 and 1995, those two reports were the best thinking about what would be the sets of questions we should be asking and how to go about getting information for environmental risk assessment. But the science has advanced tremendously since then, so much that we felt it was important to bring together all of the key scientific advances in a book published in 2007. That book went through really rigorous peer review. It was blind peer-reviewed by reviewers from around the world. The scientifically honest way to do this consequence assessment now would be to look at the best advances and draw on the best science.
Question: Your work from the 1990s is cited 14 times in this document, yet the FDA isn’t taking your recommendations. How do you feel about that?
Ph.D. students are required to write a dissertation proposal and defend it before a committee. If a student cited literature in the way it was done in this report, we would fail them.
Students would get into serious trouble if they were citing really old methods, and there had been huge advances in the methods since then and they ignored that. That would be a reason to fail them.