Statistical Analysis of the Extensive Bioassay Data on Glyphosate

Glyphosate is a famous phosphonomethyl amino acid herbicide used extensively throughout the world in weed control; its fame lies in the quantity of hysterical claims about how carcinogenic it is vs alternative claims that it does not cause cancer. The chemical structure [below] is small and actually quite simple:

In 2015, after a review of multitudes of scientific evidence, the International Agency for Research on Cancer (IARC) concluded that “the evidence for the carcinogenicity of glyphosate was limited in humans” but “sufficiently strong in experimental rat and mouse studies”. IARC further concluded that glyphosate was a “probable carcinogen: in humans. Other agencies, however, have not concurred with IARC’s conclusions (EFSA, 2015; U.S. EPA, 2019; WHO, 2016). Whether or not glyphosate presents a cancer risk — therefore — remains controversial.

In 2015, IARC reviewed ten rodent bioassays and reported on five tumors in three of these bioassays (interpreted as showing evidence of carcinogenicity). The animal carcinogenicity data on glyphosate are unusually extensive; at least 24 studies are discussed [see attached article]. Each bioassay was conducted in both sexes, each sex potentially having 40–60 unique tumor types — resulting in >1,000 potential statistical tests [which could easily result in many putative significant (P <0.05) tumor increases occurring by chance alone, i.e. roughly 5% of them would be expected to provide a P-value of <0.05 simply by chance (i.e. “false positives”) — even if exposure had no effect on carcinogenicity]. Thus, in evaluating such a large data base, it is not sufficient to identify sites in individual bioassays in which a statistical test is significant. One must also take into account the large number of statistical tests performed and the attendant possibility that statistically significant findings could be due to random chance alone. Authors [see attached article] therefore analyzed ten glyphosate rodent bioassays — including those in which IARC found evidence of carcinogenicity, using a multiresponse permutation procedure — which adjusts for the large number of tumors eligible for statistical testing and also exposes valid false-positive probabilities. The test statistics for these permutation tests are functions of P-values from a standard test for a “dose-response trend”, applied to each specific type of tumor. Authors evaluated three permutation tests, using (as test statistics) the smallest P-value from a standard statistical test for “dose-response trend”, and the number of such tests for which the P-value is less than or equal to 0.05 or 0.01. The false-positive probabilities obtained from two implementations of these three permutation tests are: smallest P-value: 0.26, 0.17; P-values <0.05: 0.08, 0.12; and P-values <0.01: 0.06, 0.08. In addition, authors found more evidence for negative “dose-response trends” than for positive. Thus, these careful statistical analyses found no strong evidence that glyphosate is an animal carcinogen. The main cause for the discrepancy between IARC’s finding and the present study appears to be that IARC: did not account for the large number of tumor responses analyzed, and the increased likelihood that several of these tumor appearances would show statistical significance simply by chance. This impressive study provides the most comprehensive analysis of the animal carcinogenicity data for this important herbicide than previously available. 😊 DwN Toxicol Sci Jun 2020; 175: 156-167

This entry was posted in Center for Environmental Genetics. Bookmark the permalink.