Comment Re:Correlation vs causation (Score 2) 17
From the study itself (pages 16 and 17):
Some methodological issues of this study need to be considered. First, the study was based on aggregate data at the municipality level and the interpretation of the results depends on the assumption of homogeneity of exposure within municipalities
95% of the population got their water from aqueducts, if the PFASs make it that far I think it's safe to assume a fairly uniform distribution within the affected municipalities.
Second, we could not adjust the association of PFAS exposure with the risk of cardiovascular disease
Of course they couldn't because they don't know the full mechanism. Suppose PFAS exposure causes hypertension and hypertension causes heart disease. Well you adjust for hypertension and the heart disease effect goes away and you falsely conclude that the PFAS didn't cause heart disease.
Third, our period of observation began approximately 15years after the onset of exposure because of limita- tions in availability of mortality data. As a consequence, we excluded from the exposed population those munici- palities with groundwater contamination since 1966
Again, if you don't know the pre-exposure rate of mortality then what exactly are you supposed to calculate?
And fourth, comparisons with other studies from the same area should be made with some caution because of the possibility that the municipalities considered to rep- resent the âcontaminated areaâ(TM) and the uncontaminated (or reference) area have been selected using data from different reports of regional authorities or have been defined according to criteria partly different from ours. The selection of the uncontaminated area is particularly prone to arbitrary assumptions and between-studies variation, with a risk of misclassification of exposure.
As for the last one: "we're unsure if our data is what we think it is".
They're saying the PFAS exposure doesn't perfectly follow the boundaries of the regions.... but it's good enough.
How the fuck did this even get peer-reviewed?
Because a natural experiment is never going to be completely clean and flawless data. But if you find a strong effect in a natural experiment, even with the limitations, it's still damn valuable data.