Skip to main content
SearchLoginLogin or Signup

Review 1: "The Safety of Antivirals and Neutralising Monoclonal Antibodies Used in Prehospital Treatment of Covid-19"

Reviewers emphasize addressing potential confounding factors and providing more comprehensive methodological details for stronger conclusions.

Published onJun 13, 2024
Review 1: "The Safety of Antivirals and Neutralising Monoclonal Antibodies Used in Prehospital Treatment of Covid-19"
1 of 2
key-enterThis Pub is a Review of
The safety of antivirals and neutralising monoclonal antibodies used in prehospital treatment of Covid-19
The safety of antivirals and neutralising monoclonal antibodies used in prehospital treatment of Covid-19

Abstract Objective This proof-of-principle pharmacovigilance study used Electronic Health Record (EHR) data to examine the safety of sotrovimab, paxlovid and molnupiravir in prehospital treatment of Covid-19.Method With NHS England approval, we conducted an observational cohort study using OpenSAFELY-TPP, a secure software-platform which executes analyses across EHRs for 24 million people in England. High-risk individuals with Covid-19 eligible for prehospital treatment were included. Adverse events (AEs) were categorised into events in the drug’s Summary of Product Characteristics (SmPC), drug-reactions and immune-mediated. Cox models compared risk across treatments. A pre-pandemic record analysis was performed for comparative purposes.Results Between 2021-2023, 37,449 patients received sotrovimab, paxlovid or molnupiravir whilst 109,647 patients made up an eligible-but-untreated population. The 29-day rates of AEs were low: SmPC 0.34 per 1000 patient-years (95%CI 0.32-0.36); drug-reactions 0.01(95% CI0.01-0.02) and immune-mediated 0.03(95%CI 0.03-0.04), and similar or lower than the pre-pandemic period. Compared with the eligible but untreated population, sotrovimab and paxlovid associated with a risk of SmPC AE [adjHR 1.36(95%CI 1.15-1.62) and 1.28(95%CI 1.05-1.55), respectively], whilst sotrovimab associated with a risk of drug-reactions [adjHR 2.95(95%CI 1.56-5.55)] and immune-mediated events [adjHR 3.22(95%CI 1.86-5.57)].Conclusion Sotrovimab, paxlovid and molnupiravir demonstrate acceptable safety profiles. Although the risk of AEs was greatest with sotrovimab, event rates were lower than comparative pre-pandemic period.

RR:C19 Evidence Scale rating by reviewer:

  • Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.


Review: The authors addressed an important safety question on the antiviral medications used for COVID-19. Overall, the manuscript was well-written, and the authors conducted thorough analyses, but please consider the following points to improve the study and the overall manuscript.

Overall, the authors compared COVID-19 patients treated with antiviral medications to those not treated with any of these medications. It is unclear why the authors reported an "acceptable safety profile" when the authors observed increased risks of all adverse events that the authors examined. Although some sensitivity analyses led to statistically insignificant findings, the direction of the estimates was consistent across most analyses. Given this, the authors should acknowledge an increased risk of adverse events observed rather than interpreting the findings as acceptable. 

Relatedly, the analysis of historical comparison was unclear and did not account for the time-varying risks within individuals, e.g., treated patients 4 years ago may not have the same underlying risk for the adverse outcomes as untreated patients 4 years ago as patients' health wax and wane over time and that is difficult to control for. Instead of comparing the same patients 4 years ago, please consider conducting the self-controlled designs, which is an appropriate choice given the short risk period. It also controlled for all confounding and thus was much more interpretable than the current analyses. 

Given the increased potential for bias of the untreated comparison, please consider comparing the risk of adverse outcomes among different anti-viral medications themselves, which is of interest to clinicians choosing between different available treatment options. 

  1. While the untreated population was restricted to patients who underwent testing, it is reasonable to expect that patients who were treated might have had more regular contact with the healthcare system (or their primary care providers) and more frequent opportunities for surveillance than the untreated population. This opportunity for bias could have explained the increased risk of adverse events that the authors observed. This could be reduced by controlling for healthcare utilization during baseline, e.g., rates of hospitalization, emergency room visits, screening, annual wellness visits/medical checkups, etc. Please also analyze one or more of these events during the follow-up as the control outcome to assess the potential for differential surveillance between the treated and untreated populations during the follow-up.

  2. The authors acknowledged the potential for milder or asymptomatic disease in the untreated population. However, the reverse could be true in that the treated population had mild or moderate disease (which is the indication) and the untreated had more severe disease (but no hospitalization), which precluded them from treatment. How would the authors minimize the potential for such bias? Please also explicitly acknowledge this possibility in the discussion.

  3. In the study population section, the authors stated the criteria for eligibility for anti-viral treatment. It is unclear what inclusion criteria the authors used for the treated population. Why did the authors impose different inclusion criteria for treated and untreated populations? High-risk conditions were not required for the treated population but were required for the untreated population. The analysis restricted to high-risk conditions for both treated and untreated was done only as the sensitivity analysis. Please consider switching it to the primary analysis. 

  4. The description of the data source was reasonably clear. However, were the authors able to obtain COVID-19 testing data, hospitalization, and treatment records for all eligible study populations? What are the percentages of successful linkage of TPP with additional data sources? Did the authors restrict to the study population with the successful linkage of all three additional data sources? 

  5. Relatedly, how well does the TPP database (and its linked data sources) capture the patients over time? How commonly could patients switch to receiving care at different hospitals or institutions not captured by the database? Please report the distributions of patients censored for different reasons and the median follow-up time during this 29-day risk window.

  6. More information and citations are needed for the target trial emulation. Please provide explicit details on what hypothetical target trial the authors emulated. This could be a few sentences in addition to the table/figure. For example, the authors emulated a hypothetical 4-arm trial of patients tested for positive, mild or moderate COVID-19 infection receiving the following drugs vs the reference group (please clarify) with the following inclusion criteria, etc. Please provide citations for per-protocol analyses and immortal time bias. 

  7. It is understandable that the authors identified only the first-ever code for the condition. However, how did the authors define 'first ever'? What lookback period (before the index date) was used to define the 'first ever'? Did the authors exclude patients with these conditions or did they keep the patients with prevalent codes in the study population? Please clarify these points and consider excluding these patients with prevalent conditions noted during the baseline period from the study population (at least as a sensitivity analysis).

  8. How did the authors identify nausea and vomiting outcomes without access to free text notes (which the authors mentioned)?

  9. The authors mentioned that covariates were assessed during the baseline period but did not clarify what baseline period was used. What was the baseline period? How many months before what date? 

  10. What is the extent of missing data on ethnic group, IMD, and BMI, all of which are important covariates to control for? The authors mentioned that they used the missing data category. 

  11. More information is needed on how the authors adjusted for covariates. What weighting method did the authors use (IPTW, SMR weighting, or overlap weighting, etc.)? What is the target population? The entire study population, treated or untreated? Please also provide appropriate citations for PS (Rosenbaum and Rubin, 1983) and weighting methods. 

  12. The authors mentioned the potential bias due to looking into the future after the COVID-19 test. Did the authors look into the future after testing until the end of available person-times to ensure they're never treated? On the other hand, as long as the authors defined patients who were not treated on the 2nd day after testing as untreated and censored patients upon being treated afterward, then it should not be a problem. Please clarify what the authors did. Please also consider re-defining the untreated as 5 days after testing (to make it similar to the treated) instead of 2 days. Then the authors won't need to look into the future, and there will be no such problems as the authors mentioned.

  13. Given that this safety surveillance study relies heavily on the coding definitions of the outcomes, please report any previous validation studies that have been done on this data source regarding the specific outcome events reported in this study.

Minor comments:

  1. While the authors presented the code lists online, it is difficult to go through the online link to look up the codes used to define the inclusion criteria and outcomes. Please present them in the appendix. It was mentioned in the outcomes section that details of definitions and code lists were presented in the supplementary appendix 6, but this reviewer could not find them.

  2. The authors sometimes referred to the risk period as 29 days and sometimes as 28 days. Please be consistent throughout. What is the reason for the choice of 29 days as the risk window?

  3. There is presumably a typing error in the exposure section. What is the "pre-protocol approach?" I assume the authors were referring to a per-protocol analysis, is that right? Or what else were the authors referring to?

  4. Please provide citations for sequential trial designs and cloning methods.

No comments here
Why not start the discussion?