Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19

12 major claims checked against the paper's own evidence: all adequately supported.

Found 3 reported tests, but none could be recomputed (missing degrees of freedom or sample size).

6 integrity concerns flagged (1 high).

4 copyedit issues flagged: mostly clarity, consistency, punctuation.

This published paper has significant robustness gaps. An informed reader should be cautious: the extremely large protective effect for ACE inhibitors (OR 0.33) is inconsistent with subsequent RCTs and likely driven by uncorrected confounding or data issues; the missing data/code availability prevents independent verification; and the ethics documentation is incomplete. These issues together justify a call for an independent re-analysis or a correction detailing the sensitivity of results to confounding and missing data assumptions.

1. 1.
   HIGHdata code

   Add a data availability statement specifying how deidentified registry data can be accessed (e.g., through a managed-access process from Surgisphere) and deposit a deidentified aggregate dataset in a public repository.

   The absence of any data access mechanism prevents independent verification and is a fundamental transparency failure.
2. 2.
   HIGHethics

   State the name and protocol number of the IRB or ethics committee that granted the exemption from ethics review, or explain why no review was required by that body.

   The current statement is vague and does not meet the reporting standard for human subjects research using non-public data.
3. 3.
   HIGHethics

   Include a statement on whether informed consent was obtained or a waiver granted, with justification (e.g., 'waiver of consent was granted because the study used deidentified data').

   Informed consent documentation is required for human subjects research; its absence is a reporting gap.
4. 4.
   HIGHrigor

   Provide the generic names (e.g., lisinopril, losartan) of the ACE inhibitors, ARBs, and other cardiovascular drugs analyzed, rather than only drug classes.

   Replication requires knowing which specific agents were studied, as effect sizes may differ across drugs within a class.
5. 5.
   HIGHstatistics

   Report exact p-values alongside odds ratios and 95% confidence intervals for all primary and secondary analyses.

   Exact p-values allow readers to assess statistical significance precisely and are standard in clinical research reporting.
6. 6.
   HIGHstatistics

   Conduct and report sensitivity analyses treating missing data as missing rather than absent, or explicitly justify the 'absent if not coded' assumption with evidence.

   The assumption that missing EHR fields mean the condition is absent is well known to introduce information bias; sensitivity analysis is needed to assess the impact.
7. 7.
   HIGHrigor

   Report a formal power analysis or sample-size justification, or explain why the available sample size is adequate for the planned analyses.

   Without a power analysis, readers cannot assess whether the study was adequately powered to detect the reported effect sizes, especially for subgroup or interaction analyses.
8. 8.
   HIGHreporting

   Add a reference to the STROBE reporting guideline for observational studies in the Methods section.

   Referencing STROBE demonstrates adherence to established reporting standards and improves transparency.
9. 9.
   HIGHrigor

   Describe the approach to outlier handling (e.g., none were excluded, or identify the method used) in the Statistical Analysis section.

   Outlying observations can distort regression results; readers need to know how they were managed.
10. 10.
    MEDIUMreporting

    Clarify whether the funder (Surgisphere) had any role in study design, data analysis, or manuscript preparation, beyond funding the registry's maintenance.

    Full disclosure of funder involvement is essential for assessing potential conflicts of interest.
11. 11.
    MEDIUMreporting

    State whether the analysis plan was pre-registered, and if not, explicitly note that no analysis plan was pre-registered.

    Pre-registration signals that analyses were pre-specified; its absence should be acknowledged so readers can adjust for potential reporting bias.
12. 12.
    MEDIUMstatistics

    Verify and report that all logistic regression assumptions (e.g., linearity of logit, absence of multicollinearity, influential observations) were checked and met.

    Only linearity of the logit was confirmed; other assumptions are critical for the validity of the model.
13. 13.
    MEDIUMrigor

    Provide the full list of participating hospitals and countries in a supplementary table to improve transparency of the registry.

    Geographic and institutional details help readers assess generalizability and potential selection biases.
14. 14.
    LOWcopyedit

    Rephrase the parenthetical definition of the tipping-point analysis in Statistical Analysis for better readability (e.g., 'A tipping-point analysis was performed to determine the effect size and prevalence an unmeasured confounder would need to shift the upper boundary of the confidence interval toward null').

    Minor clarity improvement suggested by the copyedit pass.
15. 15.
    LOWcopyedit

    Reword the missing-data assumption statement to 'it was assumed that the characteristic was absent' to improve clarity.

    Minor redundancy in phrasing flagged by the copyedit pass.