Tuesday, September 25, 2012

The Innocence of John Timmer / Ars Technica

The Ars Technica Science Editor, John Timmer, has an article decrying the recent two-year study of rats exposed to Roundup(TM) and Roundup-Ready(TM) corn.  I came across Timmer's article when recommended in a tweet from someone that I generally like and admire; I was aghast that he was swayed by the arguments of the article.

I, myself, find that Timmer's article is mostly polemic: short on information, long on half-truths and, overall, misleading. I don't know if Timmer is intentionally misleading, so I'll give him the benefit of the doubt and simply refer to his "Innocence".  I will go on to show just how much Innocence that Timmer exhibits in his article.

Some Necessary Background

The basis of the study is a long-term, two-year, version of the current 90-day toxicological studies required by the European Food Safety Authority (EFSA).  A recent paper by several of the authors on improving GMO safety studies notes that the EFSA recommends performing long-term studies of toxicological effects.  The authors add that they agree with this:
"[W]e think that in order to protect large populations from unintended effects of novel food or feed, imported or cultivated crops on a large scale, chronic 2-year and reproductive and developmental tests are crucial. However, they have never been requested by EFSA for commercial edible crops."
The reason for the two-year study period is that this represents the entire lifetime of the rats used for the study.  This is certainly a more serious indicator of long-term effects.    

The authors also note that current tests concern either the effects of Roundup(TM) or of Roundup-Ready(TM) corn, but not the combination of the two.  Because the intent of Roundup-Ready(TM) corn is to be treated with Roundup(TM), it is clear that, in order to model the real world, the combination of the two must be tested.

The authors also note that almost all studies since 2004 have been produced by the GMO companies themselves.  In order to analyse some of the studies made by the companies and presented to the EFSA, the authors have had to resort to "court orders" in order to gain access to the study data.  

This is the background of the current science against which this study has been made.  This study is independent of the GMO companies, covers the rat's whole lifetime and concerns the real-world effects from the combination of the herbicide, Roundup(TM), and the Roundup-Ready(TM) GMO corn plant.

The Findings

An article in the New York Times sums up the findings well (without necessarily agreeing with them, mind you):
The study found that in groups that ate the engineered corn, up to 50 percent of the males and 70 percent of the females died before they would have from normal aging, compared with 30 percent of the males and 20 percent of the females in the control group.
Some 50 to 80 percent of the female rats developed tumors compared with only 30 percent of the controls. And there were several times as many cases of liver and kidney injury in the exposed rats.
Personally, I am not an expert on the statistical methods used by the study;  however, assuming the findings prove verifiable, I believe it evident that they are highly significant and require urgent attention.

The Embargo

A point that has been repeatedly hit upon about this study is the extra-restrictive NDA and embargo.  I personally believe that this is not such an important point and that, furthermore, it obviously has absolutely no bearing upon the veracity of the study itself.  It is, therefore, quite curious that Timmer makes it the centerpiece of his article.

In a nutshell:  (1)  Peer-reviewed journals require their studies to be approved by peers and never published previously, hence (2) Studies are distributed to reviewers and journalists with NDAs stipulating an Embargo date before which no mention can be made of the study and (3) Journals often provide press-releases that are communicated at the embargo date.

What was particular about the NDA of this study is that it forbid communicating the study to a third-party without prior approval of the third-party.  Journalists claimed, and rightly so, that this impeded third-party validation of the study prior to the embargo date; however, this in no way prevents third-party validation after the embargo is lifted.  This special NDA restriction was explained (in a radio interview on France Inter, in French) by the  fear that the study's publication could be blocked by industry interests if it were not kept secret prior to publication.  In fact, the entire two-year study was kept a strict secret, for fear of industry backlash.  Given the context, I believe that this is a quite reasonable fear to have.

The Innocence

Timmer weighs in early with "a completely cynical manipulation of the press" followed by "the offenders appear to be the scientists themselves" and "[t]he important checks provided by that [embargo] system have now been systematically undermined by a group of French researchers".  These are heavily charged statements that set an early bombastic, accusatory tone to the article.  This isn't Innocence, as it were, as much as simple exaggeration and polemic.  

More accusations follow:  
The manipulation didn't end there. .. a [press] release claiming that the research was supported by independent research organization, CRIIGEN ... neglected to note that the paper's lead author, Eric Seralini, is on the CRIIGEN board.
Now, folks, this is pure Innocence for at least two reasons:  (a)  The study itself identifies Séralini et al with CRIIGEN on its title page (b) One of the major complaints about some of those scientists that have been critical of this study is that these same scientists have affiliations with GMO industry (hence, conflicts of interest).  It is therefore quite spectacular that Timmer would so vehemently decry what can at worst be termed an oversight in the press release, entirely mitigated by the clear identification within the study itself, when the GMO industry itself is accused of systematically buying experts.  And, yet, Timmer uses the strong term .. manipulation .. a second time, implying devious intent.  Since we are giving Timmer the benefit of the doubt, this can only be explained by Innocence.

Lead on, John.
But these problems were only the beginning. As more critical reports began to appear and scientist/bloggers looked at the results, huge issues were made clear. The authors used a strain of rats that is prone to tumors late in life.
The implication of this statement is that the Séralini, et al study used a gamed rat, something different from the short-term studies.  Au contraire, Timmer!   Séralini, according the the aforemention France Inter radio program, used the same rat strain, Sprague-Dawley, as in the GMO-industry-sponsored studies presented to the EFSA.  The reason for this is obvious: provide a long-term study under conditions equivalent to the short-term study.  

This leaves me with a perplexing dilemma.  I would hate to think that Timmer would write this phrase in order to mislead his readers intentionally.  On the other hand, I would not want to suggest that Timmer was unaware that Sprague-Dawley was the one-and-only rat for the job, by definition.  Chalk it up again to Innocence.

Timmer continues:
 Every single experimental condition was compared to a single control group of only 10 rats.
(1) Materially incorrect: Ten rats of each sex, hence 20 control rats.  (2) That replicates again, if I understand correctly from the France Inter radio program, the GMO-industry studies.  So, if it's bad in the Séralini study, it must be bad in the GMO-industry studies, too.  Innocence.

The authors didn't use a standard statistical analysis to determine whether any of the experimental groups had significantly different health problems. 
I can't guess what Timmer considers standard, he being a smart guy in spite of an awful lot of Innocence, but the statistical analysis section of the  Séralini  study seems pretty hefty to me.  Is it too good to be Standard?
2.6. Statistical analysis
Biochemical data were treated by multivariate analysis with the SIMCA-P (V12) software (UMETRICS AB Umea, Sweden). The use of chemometrics tools, for example, principal component analysis (PCA), partial least-squares to latent structures (PLS), and orthogonal PLS (OPLS), are robust methods for modeling, analyzing and interpreting complex chemical and biological data. OPLS is a recent modification of the PLS method. PLS is a regression method used in order to find the relationship between two data tables referred to as X and Y. PLS regression (Eriksson et al.,2006b) analysis consists in calculating by means of successive iterations, linear combinations of the measured X-variables (predictor variables). These linear combinations of X-variables give PLS components (score vectors t). A PLS component can be thought of as a new variable – a latent variable – reflecting the information in the original X-variables that is of relevance for modeling and predicting the response Y-variable by means of the maximization of the square of covariance (Max cov (X,Y)). The number of components is determined by cross validation. SIMCA software uses the Nonlinear Iterative Partial Least Squares algorithm (NIPALS) for the PLS regression. Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was used in this study (Weljie et al., 2011; Wiklund et al., 2008). The purpose of Discriminant Analysis is to find a model that separates groups of observations on the basis of their X variables. The X matrix consists of the biochemical data. The Y matrix contains dummy variables which describe the group membership of each observation. Binary variables are used in order to encode a group identity. Discriminant analysis finds a discriminant plan in which the projected observations are well separated according to each group. The objective of OPLS is to divide the systematic variation in the X-block into two model parts, one linearly related to Y (in the case of a discriminant analysis, the group membership), and the other one unrelated (orthogonal) to Y. Components related to Y are called predictive, and those unrelated to Y are called orthogonal. This partitioning of the X data results in improved model transparency and interpretability (Eriksson et al., 2006a). Prior to analysis, variables were mean-centered and unit variance scaled.
Timmer continues:
One report quoted a scientist at UC Davis as saying, "There is very little scientific credibility to this paper. The flaws in the test are just incredible to me."
My friend Google makes it easy to find out to whom one can attribute this quote:  Martina Newell-McGloughlin, in DiscoveryNews.  We can attribute another quote to her:  her stance on the new Proposition 37 bill coming up for vote in California.  Proposition 37 would make it a requirement to label GM-based food products in California, hence giving consumers a choice to "opt-out" of GM foods.  Martina Newell-McGoulghlin takes a stand on that proposition:

“Mandatory labeling can only be scientifically justified when based on the characteristics of the food product, not on the processes used in their development,” said Martina Newell-McGloughlin, DSc. Executive Director Strategic Research Initiatives, University of California Davis. “But there are no material differences between crops that have been genetically modified using modern techniques and other crops, and they have routinely been found to be as safe. Unfortunately, it is easy to sell fear and doubt, which is exactly what the proponents are doing with this measure.”

It is, of course, admirable, that the good doctor from UC Davis takes a stand on such matters.  At the bottom of the same web-page, we can read, in a (perhaps rare) show of transparency:

Paid for by No on 37: Coalition Against the Deceptive Food Labeling Scheme, sponsored by Farmers, Food Producers, and Grocers. Major funding by Monsanto Company, E.I. DuPont de Nemours & Co., Grocery Manufacturers Association (GMA) and more than 40 food company members
Hmm.  The good doctor is quoted on paid political advertising by Monsanto?  Is that not a sign of "Conflict of Interest"?  Can we really trust the unsubstantiated doubts expressed by Martina Newell-McGloughlin and repeated by John Timmer in all Innocence?   From the article, what do those expressed doubts say exactly that is material?  Aren't they just a sort of FUD?  After all, the good doctor says, "it is easy to sell fear and doubt".

The rest of the Timmer article continues the Innocence without introducing anything new.  We have name calling ("sloppy") and insinuation ("manipulation") and minimizations ("small, specialized journal").  

Timmer ends his article on a note of insinuation that may border on lucid self-reflexion:
"In this case, however, the press wasn't reporting about science at all. It was simply being used as a tool for political ends."
I'm still nominally giving John Timmer the benefit of the doubt, but could this possibly be an exposure of his own "modus operandi"?  This is, for me at least, the End of the Innocence of John Timmer.


 Great film:  http://www.youtube.com/watch?v=Rml_k005tsU&feature=player_embedded