Can the impact of unmeasured confounders be excluded according to Bernstein’s study?
As Bernstein’s study was observational, it is impossible to exclude the impact of unmeasured confounders. Lack of randomization in this cohort was a flaw. All participants were American, and the National Health and Nutrition Examination Survey (NHANES) revealed that Americans consume PM at more than three times the global average. Participants in Bernstein’s cohort may have an exaggerated exposure to PMs that represent one nationality, which is not a true representation of the global population, these results, therefore, lack generalisability. Streppel et al. highlight that food frequency questionnaires (FFQs) lack reproducibility. A limitation of this study was that dietary data were obtained from FFQ’s. FFQs’ with blank items at baseline for 70 men and ten women were included in the study. Implausibly high energy intakes were also included, and this could have impacted validity.
Abnet assessed the effect of cooking temperatures on the carcinogenicity of meats. This study found that cooking meats at high temperatures directly over a flame generates the formation of carcinogens, namely heterocyclic amines (HCAs) and PAHs. Abnet concluded that consumption of these carcinogens increases cancer incidence. This is a well-established confounder which Bernstein failed to adjust for in his study, highlighting a significant limitation. It is possible that if this had been adjusted for, as in previous studies (John et al., 2011), the association between PM consumption and CRC incidence might have reported different results. Hodgson and Rustgi highlight the importance of familial malignancy history. However, hereditary susceptibility to cancer is not always adjusted for when researching PM and CRC. This is surprising as Sulz et al. (2014) report that those with a first- degree relative with CRC, have a 2-4-fold higher risk of developing the malignancy. While the exclusion criteria for many of the included studies required omission of those who had a cancer diagnosis, it was discovered that 14.6% of males and 17.7% of females in Bernstein’s study had a first-degree relative with CRC. With 32.3% of the cohort predisposed to an elevated risk of CRC, this could have significantly impacted results. Unmeasured confounders are a shortcoming observed across multiple cohorts included by the IARC.