| (Disclaimer: I’m not trying to make an argument that there is a link between vaccines and autism, I’m trying to understand a research paper and its methodology and conclusions) I just glanced through the study itself and not the article https://www.acpjournals.org/doi/10.7326/ANNALS-25-00997 and have questions for anyone that is familiar with this sort of thing: They did not do a non-vaccinated vs vaccinated people comparison. They looked at how many vaccinations each subject had and tabulated the total amount of aluminum they received (they have good records) in their vaccinations prior to age 2, and then looked for a correlation between higher amounts of aluminum adjutants and higher instances of all the conditions they reported on (the study does not focus on autism but many things, see figure 3 of the study) There wasn’t a sizable aluminum free cohort because most children in Denmark got vaccinated for the data set they have to work with (figure 2). Wouldn’t you need a sizable cohort of non vaccinated children? (I don’t know where in the world you would find that cohort except in countries that don’t have good healthcare systems which implies they don’t have solid tracking of health outcomes in general, or maybe the USA in the last 5-10 years.) The researchers discarded from their cohort 34,547 children for receiving too many vaccinations/too much aluminum (figure 1, right column) before age 2. Wouldn’t that data be relevant to look at? So to my laymen’s mind it doesn’t seem like they in any way ruled out a link between vaccines and autism. At the very best, they saw no relationship with the amount of aluminum a child received before 2 and the rate of chronic disease (figure 1). The numbers correspond to “adjusted hazard ratio (95% CI) per 1mg increase in aluminum received……” and in the chart Asperger’s is listed at 1.13 (.89 - 1.44) so they do potentially see an increase for Asperger’s, but with the data they have the confidence interval is not small enough to be sure one way or the other. But it does not seem to me that they proved what the article and submission title states. I would appreciate someone that is familiar with biomedical research that can elaborate on whether my conclusion is sound or faulty. Also, as I understand it aluminum is one of about 5 or so adjuvants used in vaccines at the present time, so what about the other adjuvants? |
### They did not do a non-vaccinated vs vaccinated people comparison
These studies are observational, and in that sense no manipulation (such as random treatment assignment) was made. So they need to rely on observational studies techniques.
From what I could gather, one of this observational techniques in inverse treatment probability weighting when they say
> "we adjusted for all baseline covariates using stabilized inverse probability of treatment weights"
The IPW technique basically tries to solve the problem that some groups of children might have a higher prevalence of the treatment by weighting them in a manner as to mimic random assignment. This is as close as we can get to comparing treated and untreated in observational settings.
Also, the division of the doses of vaccine is also a better characterization of "was vaccinated". If you think getting vaccinated causes something, then getting more vaccinated should increase the incidence of that something.
### Wouldn’t you need a sizable cohort of non vaccinated children?
Yes, but as you pointed out, that might not exist. So by creating comparable groups (via IPW) + treatment intensity/dose we can still arrive at some conclusions.
### The researchers discarded from their cohort 34,547 children for receiving too many vaccinations/too much aluminum (...) Wouldn’t that data be relevant to look at?
They probably did. A few things are important here. The data comes from administrative data sources, so mistakes can happen. While you need to trust the sources, there could be imputation typos, or just weird cases. So the researchers probably went after a notion of "clinically relevant vaccine dosage for this study" to know up to which point to consider data points, because from that point onward, it either is not interesting because its a rare treatment incidence, or just seems like a mistake.
### “adjusted hazard ratio (95% CI) per 1mg increase in aluminum received……” and in the chart Asperger’s is listed at 1.13 (.89 - 1.44) so they do potentially see an increase for Asperger’s
Important to say that with hazard ratios, a HR of 1 means that there is no change because we are in the multiplicative scale (as opposed to 0 meaning "treatment does nothing" in the additive scale).
So in this setting, after adjusting for the treatment probabilities, a HR CI of (.89 - 1.44) just translates to 'no effect'. Nothing out of the ordinary in terms of the interpretation.