The fact-check article on deaths after COVID-19 vaccination by AFP is scientifically incorrect.

by Yasusi Suzumura, YSP Medical Information Laboratory
01/19/2024



I consider the fact-check article by AFP to be scientifically incorrect. This fact-check article is about my paper on deaths after COVID-19 vaccination. I examined why it is scientifically incorrect by comparing the fact-check article with my paper.

Kinoshita's comments (Cov-Navi) in the fact-check article

Kinoshita explained that the number of deaths used in the study could not be taken as a comprehensive source due to Japan having a passive reporting system where clinicians are only required to report adverse events only when they suspected a link to vaccination. "Clinicians are more likely to report sudden deaths occurring soon after vaccination - within 10 days," he told AFP on December 21, 2023. "Therefore, this study does not suggest that the occurrence of deaths after vaccination is higher in the early period compared to the later period."


Quoted from the paper.

The percentage of reported deaths during the risk period was 71% in Group 1 and 70% in Group 2 (Figures 1, 2), indicating that the mortality rate during the risk period may have been higher than that during the control period. However, it is not appropriate to conclude that there is an association between vaccination and deaths after vaccination. This is due to the existence of a reporting bias. In Japan, physicians have reported only cases in which they suspected that the vaccination was associated with death. Thus, in these cases, the greater the number of days from vaccination to death, the less likely the cases were to be reported by physicians. Consequently, this bias could have led to a higher mortality rate in the reported cases during the risk period compared with that in the control period.


Kinoshita's comments are way off the mark.


Quoted from the paper.

In contrast, the author believes that comparing sex ratios by period is less prone to reporting bias. This is because the decision to report by a physician is often not influenced by the sex of the patient. Therefore, this method may be useful for analyzing data with reporting bias.


I think Kinoshita should criticize the analysis of the sex ratios.


Raina MacIntyre's comments (professor of global biosecurity) in the fact-check article

"Most of the deaths [in the study] were related to cardiovascular disease, which is the leading cause of death and illness in the world and occurs at a higher rate in men than women, particularly under the age of 65," MacIntyre said on December 21, 2023. MacIntyre also noted that people who were prioritised for vaccination against Covid-19 in early 2021 -- the study period -- were older or had multiple chronic conditions, putting them at higher risk of dying regardless of their vaccination status.


A high or low value is meaningful only when compared to that of a control. Raina MacIntyre has not compared a high rate to that of a control. I don't believe Raina McIntyre commented following the basics of statistics.


Chunhuei Chi's comments (professor and director of the Center for Global Health) in the fact-check article

"The author acknowledged several limitations, including not considering the effects of vaccination after 11 days, and sex bias in reporting deaths, that limited the strengths of the evidence and conclusion," Chi said on December 21, 2023.


Limitations are described in all papers. Generally, you cannot deny a paper solely based on limitations.


William Schaffner's comments (professor of preventive medicine) in the fact-check article

"A rigorous study would have compared the vaccinated population with a comparable unvaccinated population similar in age and sex distribution with a similar distribution of underlying chronic illnesses," Schaffner told AFP on the same day. This methodological bias was typical of most studies claiming higher mortality after vaccines, said Siddharth Sridhar, clinical assistant professor at the department of microbiology at the University of Hong Kong (archived link). "Better quality cohort studies invariably find no increase in all-cause mortality after vaccination," he said on December 22, 2023.


A cohort study is a high-quality method. However, it has weaknesses. The paper describes its weaknesses.

Quoted from the paper.

In a cohort study in Japan published in 2022, there was no significant increase in all-cause mortality owing to vaccination [1]. This does not contradict the results of the present study. When the mortality rate of vaccination is exceptionally low, even if there is an association, no significant difference may be found in the cohort study. The results of a cohort study provide support for vaccine safety. However, this does not indicate that vaccine-related deaths are nonexistent; it only indicates that their number is not large enough to make a significant difference.


Furthermore, quoted from the paper.

Thus, when a cohort study does not detect a significant difference due to a low mortality rate, further statistical analysis is necessary to determine whether to provide financial support. Vaccines are administered to all people, including healthy individuals. Thus, vaccines require a higher level of safety than pharmaceuticals used for treatment and should have an exceptionally low vaccination mortality rate. Therefore, it is necessary to analyze vaccine safety with statistical methods that can detect significant differences even when the vaccination mortality rate is exceptionally low.


William Schaffner is suspected of commenting without properly reading the paper.


Kinoshita and MacIntyre's comments in the fact-check article

The World Health Organisation states that the Covid-19 vaccines are safe and effective, and estimates that in 2021 alone the jabs helped saved 14.4 million lives worldwide (archived link).


There is no guarantee that what WHO's statement is always correct. If there is a question, it is the professional thing to do to raise the issue. The main theme of the paper is, "Isn't there a problem with verifying vaccine safety using only cohort studies?"

Quoted from the paper.

An SCRI and a comparison of sex ratios by period may be useful in examining the association between vaccination and deaths after vaccination when a cohort study does not detect a significant difference due to a low mortality rate. The latter approach may be particularly useful for analyzing data with reporting bias. The author believes that this approach may not provide conclusive evidence, but it can offer valuable insights into assessing vaccine safety.


I am not claiming that this study may provide conclusive evidence of an association between vaccination and deaths. Since the subject of the paper is statistical methods, I believe statistics experts should fact-check it.


[Conclusion]
I propose that the AFP asks statistics experts to fact-check, and reconsider.