Association between COVID-19 Primary Vaccination and Severe Disease Caused by SARS-CoV-2 Delta Variant among Hospitalized Patients: A Belgian Retrospective Cohort Study

Round 1

Reviewer 1 Report

 

General comment

The paper “Acute otitis media pneumococcal disease burden and

nasopharyngeal colonization in children due to serotypes

included and not included in current and new pneumococcal

conjugate vaccines” is an interesting and well written article that investigate vaccine effectiveness against progression to severe COVID-19 20 (acute respiratory distress syndrome (ARDS), intensive care unit (ICU) admission and/or death) and 21 in-hospital death in a cohort of hospitalized COVID-19 patients.

 

Major comments

The authors should better explain some aspects of Methods and Results. The number of participating hospitals in the abstract (73) does not match the number that appears in the methods (103). Explain whether deaths in the hospital is a category of the severity variable. In this case, the number of ARDS, ICU and Death that appear in table 1 and 2 do not match the number of sever cases. Please explain better.

Specific comments

1)      Introduction: not match the dates of 3^rd wave with figure 1. Explain better

2)      In Methods explain how may general population coverage the 73 or 103 hospitals.

3)      There are an error in one OR of table 4-5 (OR=8!).

4)      Discuss the role of previous infection in the vaccination effect.

Author Response

Dear reviewer,

In this cohort of hospitalized COVID-19 patients, there were 736 severe COVID-19 cases and 385 in-hospital deaths. 

Major comments

1. "The number of participating hospitals in the abstract (73) does not match the number that appears in the methods (103)."

The non-exhaustive Clinical Hospital Survey allows surveillance of 103 acute care hospitals across Belgium. For the period chosen to restrict on Delta infections, as well as after applying in- and exclusion criteria for the cohort in this study, data was selected from 'only' 73 hospitals in this period. These 73 hospitals are still representative for the situation in hospitalized patients in Belgium though.

I added the following sentence to the section "Study population" to better explain this. "The Clinical Hospital Survey allows surveillance of 103 acute care hospitals across Belgium, but for the period chosen, as well as after applying in- and exclusion criteria for the cohort in this study, data was selected from 73 hospitals in this period."

2. Explain whether deaths in the hospital is a category of the severity variable. In this case, the number of ARDS, ICU and Death that appear in table 1 and 2 do not match the number of severe cases. Please explain better.

This is actually explained in section 2.3.1 Outcomes as "In this study, ICU admission, diagnosis of ARDS and in-hospital death were operationalized as dichotomous (yes/no) variables. Patients with ‘yes’ for one or more of these three variables were classified as severe COVID-19 cases." The reason why the numbers do not "add up" in table 1 and 2 is because a patient often has more than one condition fulfilled, so a patient that died is likely to have been in the ICU and/or has had ARDS as well, but is only counted as 1 patient in the "severe cases" category. 

Specific comments

1)      Introduction: not match the dates of 3^rd wave with figure 1. Explain better

I do not exactly understand this comment (as the dates from the introduction for the 3rd wave defined by Sciensano 15/02/21 - 27/06/21 are clearly mentioned at the bottom of figure 1). Do you mean that the alignment of the dates of the other subfigures in figure 1 do not perfectly align? I did this by hand, so I can try to tweak it, but it will never be fully on scale. The main message of this figure is actually to visualize the period of data inclusion and to roughly scetch the context of that time (which was not during the actual third wave, but during the time after when Delta VOC was circulating dominantly AND most of the Belgian population had received an opportunity to receive the primary vaccination schedule).

I hope this clears up the intention of this figure. Let me know if you would still like me to tweak/improve the figure a bit so the dates align even better. I would need a bit more time to do this. 

2)      In Methods explain how may general population coverage the 73 or 103 hospitals.

As mentioned above, I added the following sentence to the section "Study population" in M&M to better explain this. "The Clinical Hospital Survey allows surveillance of 103 acute care hospitals across Belgium, but for the period chosen, as well as after applying in- and exclusion criteria for the cohort in this study, data was selected from 73 hospitals in this period."

3)      There are an error in one OR of table 4-5 (OR=8!).

Oh yes, thank you for making me aware! It was supposed to be OR=0.80, I corrected it!

4)      Discuss the role of previous infection in the vaccination effect.

This study was not set up to study the effect of previous infections in the vaccination effect. To keep the focus on the vaccination effect we even explicitly decided in the study design phase to exclude patients with with definite previous infections to keep the internal validity of the study high. Additionally, even if we would include data on previous infections, validity of the analysis would be limited, as some people may have experienced a previous (possibly asymptomatic) infection without confirmation of a PCR/antigen test. It could be done, but in our opinion this would not add much extra, valuable information and with rather low validity. 

 

Thank you kindly for taking the time to review this work!

Reviewer 2 Report

In my opinion the presented manuscript has a very high quality as well as interest. The only changes to be considered are related to the tables. In my opinion the authors should indicate the p value of the comparation of the sample characteristics in tables 1 and two. For a clearer reading I also suggest that tables 4 to 7 should be simplified in one or two tables. In the present manuscript the distribution of information in the tables 4 to 7 is confusing. 

Definitely, I think that we are in front of a very interesting and high quality work. A high quality methodology as well as discussion including limitations (In fact the autrhors have been too critical with they work).

Author Response

Dear reviewer

1) including p values to show differences in baseline characteristics: We argue against this suggestion for the reason that it is not needed to perform statistical tests of descriptive "baseline" characteristics. There were no prior hypotheses about comparisons between the groups described in table 1 and 2 (and tableA1 and A2) and therefore do not warrant p values.

2) make table 4-7 into 2 tables: We agree and did this. We also did it for the supplementary tables that had this format.

Thank you kindly for taking the time to review this work!

Reviewer 3 Report

Dear Authors,

Authors investigated the association between COVID-19 primary vaccination and severe disease caused by SARS-CoV-2 Delta variant among hospitalized patients in Belgium. Thus, they investigated vaccine effectiveness against progression to severe COVID-19 (acute respiratory distress syndrome, intensive care unit admission and/or death) and in-hospital death in a cohort of hospitalized COVID-19 patients.

 

This study provided more real-world evidence on brand comparison information that is currently lacking in the context of COVID-19 progression to severe outcomes. 

 

Study analyzed data of high quality through linkage of extensive person-level databases which made correction for confounding and specific vaccine brands possible. Also, Authors used data from the nationwide Belgian surveillance system.

 

Authors performed a well-designed study with high quality data.

I would like to see a Cox regression with death as the dependent variable since this regression model takes into consideration the time until death. Also, a Kaplan-Meier survival curve between vaccinated and unvaccinated individuals would add valuable information. 

 

The conclusions are consistent with the evidence.

References are appropriate.

Tables and figures are well-written and self-explanatory.

Author Response

Dear reviewer

Thank you for the kind, positive feedback!

"I would like to see a Cox regression with death as the dependent variable since this regression model takes into consideration the time until death. Also, a Kaplan-Meier survival curve between vaccinated and unvaccinated individuals would add valuable information."

As for this suggestion, it was something we thought of and discussed during the designing and protocol writing phase of this study. One reason to put this type of survival analysis aside is that it does not add to the (choice of the) existing logistic regression analyses, because there is no need to account for censoring. The follow-up duration in our design was fixed (so no dynamic cohort). Another argument against a survival analysis is that the relevance of survival time within such a short period as we chose for this study is very limited and even ambiguous. From the point of view of the hospital bed capacity: if there was a shortage, then a patients who dies within 2 days puts lower burden on hospital beds than someone who dies 14 days after admission, though quality of life of those days can be very different and is very difficult to assess or compare. The primary aim is to leave the hospital alive; so that survival duration (which as mentioned is not a good proxy for severity in this case) is of less interest than total survival.

 

Thank you kindly for taking the time to review this work!