Authors’ reply

We would like to thank Professor P. Zagożdżon and Dr. A. Jarynowski for their important comments on our paper published recently in Polish Archives of Internal Medicine.1 In their letters, they rightly highlighted several methodologic aspects of our analysis that are crucial for correct and comprehensive understanding of the article message. The registry-based studies are inherently prone to several types of bias, and unfortunately force some simplifications of the analysis, as compared with cohort studies with precise patient-level data. Therefore, in our paper, we aimed to clearly inform the readers about its limitations.

We agree with Dr. A. Jarynowski that our paper is not the first one to describe the vaccination program in Poland, however, the methodology of our study differs significantly from the report issued by the National Institute of Public Health-National Institute of Hygiene2 by using vaccine effectiveness (VE) as the main outcome measure, and contrary to the report by Meslé et al,3 it covers the entire population of Poland, thus painting a more complete picture.

We also agree with Dr. A. Jarynowski that inclusion of adverse events is important in such an analysis, to better inform the readers about the net benefit of an intervention. Unfortunately, the cited report does not contain the date of the vaccination, and therefore prevents us from using it in the context of the studied time period. Moreover, please note that none of the deaths mentioned in the report had a confirmed causal relationship with the administration of the vaccine. Finally, as Dr. A. Jarynowski correctly points out, the number of deaths (n = 10) bearing temporal (but unconfirmed causal) relationship with the vaccine administration would not have significantly altered interpretation of the results of the paper in which a total number of deaths exceeded 70 000.

We agree with Professor P. Zagożdżon and Dr. A. Jarynowski that immortal time bias is an important issue, and could potentially affect interpretation of the results. As mentioned multiple times across the manuscript, all analyses were performed for separate large registries, and we did not have access to individual patients’ data. Therefore, we had limited tools to cope with the immortal time bias. Nevertheless, the number of deaths occurring in the time period between the vaccine administration and achieving the fully vaccinated status was low and very unlikely to alter the message of the manuscript.

The mentioned gradual decrease of the VE is important and may be observed in Figure 4 of our manuscript. We thank Dr. A. Jarynowski for highlighting this crucial issue, however, we would like to emphasize that the choice of the analyzed time period was determined by the data availability and not by our intent to exaggerate the effectiveness of the vaccine.

We are aware that simplification of the vaccination status in our study affects its interpretation. Therefore, we clearly stated this fact in the methodology section, and included this aspect as the first and most important limitation of our manuscript.

The definition of COVID-19–associated death was not simply death when the presence of SARS-CoV-2 infection was confirmed using the real-time polymerase chain reaction assay, but when the processes leading to death were initiated by COVID-19. Only when a treating physician reported COVID-19 as an underlying cause of death, it would later be counted as COVID-19–associated death. Such an approach was suggested by the World Health Organization.4 We disagree with the opinion that the vaccination status affected the perception of the cause of death and, similarly to Professor P. Zagożdżon, we are unable to find a reference to support this statement. We do not have any as to information whether COVID-19–associated death occurred in or out of a hospital, and hence we are unable to provide the analysis suggested by Professor P. Zagożdżon.

We find some statements by Professor P. Zagożdżon to be rather personal opinions than evidence-based statements (association between the vaccination status and chances of patient’s death being attributed as related to COVID-19; reasons to omit the second dose of the vaccine). However, our most important disagreement with Professor P. Zagożdżon pertains to available evidence on the VE based on randomized controlled trials (RCTs). Available literature provides compelling data from meta-analyses of RCTs on the VE in preventing both hospitalization and COVID-19–associated death.5

Summing up, we believe that our study adds important data to the current literature and confirms the effectiveness of the BNT162b2 vaccine in the Polish population. We agree with Professor P. Zagożdżon and Dr. A. Jarynowski that our study has several important limitations, and we are grateful for their highlighting the aspects not mentioned in the manuscript. We are convinced that these letters to the Editor allow the readers to better and more consciously interpret the results of our study.