Steve Kirsch

I know what the analysis was. It does NOT test whether vaccination increases the risk of death or causes harm. Instead, it conditions on death having occurred, compares two arbitrarily defined time periods, and adjusts for no confounders. What it actually tests is: “Among children who survived through day 90, had a matched immunization record so that they received at least one immunization before death, and eventually died before age three, were those vaccinated during days 60–90 more likely to die during days 90–120 rather than sometime after day 120, compared with children who received none of their vaccines during days 60–90?” An OR greater than 1 does not show that “vaccines are unsafe.” The study does not even include the population denominators needed to determine whether vaccinated or unvaccinated infants had a higher risk of death. Its “unvaccinated” group is not truly unvaccinated; it consists of children who were not recorded as receiving the specified vaccines during days 60–90. -- every single child received at least one vaccine in this study, it is just a question of when. The paper does not report how many were vaccinated before day 60, during days 91–120, or after day 120. Likewise, the children classified as “alive” were not alive—they simply died after day 120. The analysis therefore shows only that, among children who all eventually died, the timing of death differed according to whether vaccination was recorded during the selected 60–90-day window as opposed to some other time period. There is also a potentially massive guaranteed-survival, or immortal-time-type, bias arising from the requirement that every included child have at least one immunization record. For a child classified as “unvaccinated” during days 60–90 whose qualifying vaccination occurred later, inclusion in that group required the child to survive until that later vaccination. A child vaccinated after day 120 was, by definition, incapable of being counted as a death during days 90–120. Even a child vaccinated during days 91–120 had to survive at least until the vaccination date. Thus, future survival may have helped determine membership in the reference group, structurally lowering the proportion of deaths during days 90–120 among the supposedly “unvaccinated.” Unless the authors report the full vaccination-date distribution in that group, this bias cannot be quantified, but it could readily manufacture an OR greater than 1, and in fact for any children who received vaccines after day 60, it is guaranteed to induce bias making OR>1 On top of that, by restricting the entire analysis to children who eventually died, the authors hard-coded outcome-dependent selection and collider bias into the design. This compounds the already serious uncontrolled confounding between children vaccinated on schedule and those not vaccinated during that window. Prematurity, underlying illness, hospitalization, healthcare access, missed well-child visits, medical deferral, and many other factors affect both vaccination timing and mortality timing. These differences would produce an OR greater than 1 by themselves, even if vaccination had no harmful effect whatsoever. To be clear, I am not arguing that the preprint should have been removed. I am simply pointing out that the analysis does not test the hypothesis the authors claim it tests. It would have been straightforward to directly evaluate whether the risk of death was elevated after vaccination using the underlying birth-cohort denominators and exact vaccination and death dates. Why the authors instead chose this awkward and convoluted analysis is unclear. Perhaps they tried more direct analyses and did not obtain the result they wanted, although there is no evidence available to establish that. What is clear is that this analysis cannot support the paper’s stated conclusions and would never survive proper peer review without a fundamental and complete redesign.

This study was a proportional mortality analysis—a recognized and frequently utilized methodology. Any concerns regarding potential biases or over-interpretation of the data could have been addressed through a simple revision, an author’s note, or an updated version of the paper. There is one primary bias in the study: selection bias. It operates to move OR <1 under the null hypothesis that vaccines are safe. But the study found OR >1. So if it wasn't the vaccine, how did selection bias work in REVERSE?? Explain that one.

i reviewed the judges decision. The judge, who normally thinks everyone is guilty, essentially wrote this guy is the most innocent guy I've ever seen. DOJ wants a second shot at trying to put this doctor in jail for life when the doctor did NOTHING wrong. Not even a coding error.