A group of nutrition researchers at the University of California, Davis has retracted their paper in the International Journal of Molecular Sciences for what they describe as a botched experiment involving mixed-up cultures.
The article, titled “Dextran Sulfate Sodium Inhibits Alanine Synthesis in Caco-2 Cells,” appeared in 2011 and was retracted in February 2012, although it just came to our attention.
According to the abstract:
To understand and characterize the pathogenic mechanisms of inflammatory bowel disease, dextran sulfate sodium (DSS) has been used to induce acute and chronic colitis in animal models by causing intestinal epithelium damage. The mechanism of action of DSS in producing this outcome is not well understood. In an effort to understand how DSS might impact epithelial cell metabolism, we studied the intestinal epithelial cell line Caco-2 incubated with 1% DSS over 56 hours using (1)H NMR spectroscopy. We observed no difference in cell viability as compared to control cultures, and an approximately 1.5-fold increase in IL-6 production upon incubation with 1% DSS. The effect on Caco-2 cell metabolism as measured through changes in the concentration of metabolites in the cell supernatant included a three-fold decrease in the concentration of alanine. Given that the concentrations of other amino acids in the cell culture supernatant were not different between treated and control cultures over 56 hours suggest that DSS inhibits alanine synthesis, specifically alanine aminotransferase, without affecting other key metabolic pathways. The importance of alanine aminotransferase in inflammatory bowel disease is discussed.
But as the notice explains:
It has been brought to our attention by the corresponding author that the results presented this article [1] are in error due to the fact that the media supplement glutaMAX was used in place of L-glutamine for culture of the control cells, while L-glutamine was used for culture of the treated cells. All authors have confirmed that the reported result could not be reproduced using the correct culture conditions. We would like to thank the authors for pointing out this error thereby upholding the ethics of scientific publication. The Editorial Team and Publisher have agreed with the authors that this manuscript should be retracted. We apologize for any inconvenience this may have caused.
The senior author is Carolyn Slupsky. We weren’t immediately able to reach her for comment.
We note that the journal is published by MDPI, which, as readers of this blog may recall, has been responsible for its share of doubt-inspiring articles, including this one in Molecules that bore “no relationship whatsoever” with the topic promised by the title of the journal. And its journal Life ran a paper on “gyres” that led the editor to ask how it was ever published.
Speaking of MDPI, I was alerted to the article “Cancer, Infant Mortality and Birth Sex-Ratio in Fallujah, Iraq 2005–2009” that appeared in MDPI’s International Journal of Environmental Research and Public Health in 2010. The paper’s author author claims extremely low p values, e.g., p < 0.00000001. I'm not a statistician and wonder if such values are really possible. The paper is here: http://www.mdpi.com/1660-4601/7/7/2828
Many things that are possible have no meaning. Such p-values have no meaning. Get a big sample, you get sometimes very small p-values.
p values are exactly defined by the risk ratio and the sample size – hence on a sample size of 4700 and a risk ratio on 4.2 will always have a very low p value.
You can have a play with this calculator
http://www.stat.ubc.ca/~rollin/stats/ssize/caco.html
enter 0.00000001 in the box for alpha value
for p0 enter the value you think they are using for the rate of cancer in their control population (in itself a big imponderable) and then enter the Risk Ratio, in this case 4.2 – and it will show you the sample size needed to see such an effect.
It really comes down to how good their data is and how valid are the assumptions behind the selection of their reference data – but if there data is good, the control is valid then a risk ratio of 4.2 in this size sample will have a p value is as low as they say.
I had a look through their publications some months ago – in a related paper there did seem some strong indications of iodine deficiency that might explain some of their results. Although I expect coating an urban area with uranium dust is not a very smart idea either. The birth ratio was interesting too – but it all comes down to how reliable their data collection is.
From the vendor:
“GIBCO® GlutaMAX™ media is our standard trusted cell culture media but it contains a stabilized form dipeptide from L-glutamine, L-alanyl-L-glutamine, that prevents degradation and ammonia build-up even during long-term cultures.
Extremely stable in aqueous solution, the L-alanyl-L-glutamine dipeptide will not degrade into ammonia in storage or incubation like L-glutamine.”
Alanine, you say…
Did they reproduce the experiment in L-glutamine/L-glutamine or in glutamax/glutamax? I imagine both would be distinct from each other, but at the end of the day the statistical difference between control and treatment would still be insignificant?