I recently was assessing a lengthy review of the evidence on environmental exposures and breast cancer risk, and as I read, red flags started popping up. While I may not know the evidence base in this area extremely well, I knew it well enough to recognize that the authors were making statements I was pretty sure were not supported by the evidence — or at least not to the extent the review suggested.
I’ve written before about the need to check citations and stats cited in studies. In the example above, I also would closely examine some of the statements and citations in the piece to determine their reliability.
I write a lot about diethylstilbestrol (DES), a drug given to a generation of pregnant women that led to serious health problems, including increased cancer risk, in their children, called DES daughters and DES sons. Because of my strong familiarity with DES research, this sentence on page 6 of the review jumped out: “It now appears that granddaughters of women prescribed DES during pregnancy are also experiencing an elevated incidence of breast cancer [100].”
There has been concern about intergenerational (epigenetic) effects of DES exposure, but I’d just read a new study on DES granddaughters that mentioned all known evidence in this area and breast cancer risk wasn’t mentioned. I’d never seen other studies suggesting it either. So I looked up the reference 100 study. It’s not a clinical study, but a research review as well, and stated in its abstract, “Recent findings in a preclinical model suggest that not only are women exposed to DES in utero at an increased risk of developing breast cancer, but this risk may extend to their daughters and granddaughters as well.”
Since a preclinical model usually means animal, I went through the full study and learned they were referencing another study with rats: “Mammary tumorigenesis and the number of TEBs also were increased in daughters, granddaughters and great-granddaughters of dams exposed to ethinyl estradiol during pregnancy.” (A rat dam is pregnant or a mother.)
I looked up that preclinical study and found the researchers “examined whether maternal exposures to high-fat diet or ethinyl-oestradiol during pregnancy lead to multi- or transgenerational inheritance of mammary cancer in a carcinogen-induced rat model.” It doesn’t test DES in rat models at all! (It does test another estrogen, but it’s not DES.)
Meanwhile, the 100 referenced study says later, “The granddaughters are still too young to assess whether they might also be at an increased risk of developing breast cancer.” It also includes a theoretical model explaining how breast cancer risk might occur in DES granddaughters.
With this info, let’s return to the sentence quoted in the lengthy review: “It now appears that granddaughters of women prescribed DES during pregnancy are also experiencing an elevated incidence of breast cancer [100].”
That’s a flatly false statement. No studies show an increased incidence of breast cancer in DES granddaughters. There is zero evidence to back up the statement, yet it’s mentioned more than once in the review — and it’s one sentence on one page of a 40-page review with 848 references.
It’s possible there is another study with DES that tests breast cancer risk in rats (and I strongly suspect there is), but that’s not what was cited, and it would still be in rats, not humans. I can’t go down all the rabbit holes in a research review of this size, so with this example and others that seemed to overreach, I had to conclude the review was too unreliable to take anything in it at face value.
Evidence reviews are considered high on the levels of evidence pyramid, but they require just as much skepticism and critical examination as any other study.
