slothman92, on 10 June 2012 - 12:31 AM, said:
What do you mean by good papers? What are your qualifications? The only real way to decide is to read the paper and see if the methodology is thourough and the conlusions draw from the results are appropriate. The journals are all rated by JCR if that helps you.
If you don't want to believe they are *good* because they go against the majority of other papers then believe that. Their results are their results, i don't see any reason for them to fabricate or falsify their results, we saw how that turned out for Andrwe Wakefield.
The highest quality evidence takes into account the results of many studies and weights them according to their own quality. This is represented by a
systematic review, and the Cochrane review I posted is the gold standard for that. Almost as good are
meta-analyses which similarly combine the results of many studies to arrive at an overall conclusion. Next comes a
randomized controlled trial where variables are actually controlled not just "adjusted for" after the fact. Continuing down the evidence hierarchy we have
cohort studies and
case-control studies, where prospective studies are to be preferred to retrospective ones.
Finally, we have the paper you linked, which is a
cross-sectional observational study, so there is actually no experimental control of variables. An actual trial would assign neonates to either a "vaccine" or "no vaccine" group and then follow them for a period to determine whether there was a differing risk of autism diagnosis (or whatever other outcome you want to use). If a trial is not possible, you would follow babies who had been vaccinated or not from birth, and follow-up later for that outcome (a
prospective cohort study).
This study has several problems. First, the outcome variable was "a dichotomous (yes/no) variable created in response to the following survey question and presentation of a card with a choice of diagnoses: “Looking at this list, has a doctor or other professional ever told you that [sample child’s name] had any of these conditions . . . ( i.e., autism)?” Refusals to answer, responses of “don’t know,” and missing values were counted as missing data."
This is a retrospective subjective outcome that is subject to recall bias and probably some element of selection bias. They also treated the "don't knows" or "refusals" as missing, which means they were excluded from the analysis, but that introduces non-response bias. The authors appear to make no attempt to account for this.
Stunningly, the actual numbers suggest that this study was not adequately powered. The number of girls with autism (n = 9) is so small that the analysis was conducted on boys only. Except even in the autism group, there were only 31 observations, 9 who had been vaccinated and 22 who had not. That's against the non-autism group, with 1258 who had been vaccinated and 6092 who hadn't. Once again, they exclude data on the vaccination status where there is a "don't know" or "refusal" or incomplete response, without accounting for this non-response bias. However, with such small numbers, the bias could actually be considerable, and the data is still self-reported and subject to recall bias.
The authors do admit that, as a cross-sectional study, no causal interpretations can be made.
Anyway, as to my "qualifications", here they are:
M.Sc. (Applied Mathematics), MMath (Biostatistics), MD Candidate (May 2013)