Time-dependent bias when analyzing pulmonary aspergillosis associated with COVID-19 – Authors’ response

We thank Justin de Brabander and colleagues for their interest in our findings on COVID-19-associated pulmonary aspergillosis (CAPA) reported by the MYCOVID cohort. They raised a very interesting question about the time factor and competing risks in studying the relationship between CAPA and mortality, as evidenced by their own cohort.

Initially, we observed a statistically significant association between the occurrence of CAPA and mortality using a Cox model. Next, we decided to refine the Cox model analysis for mortality estimation with a benchmark analysis, considering the average day of proven or probable CAPA development (11 days in our cohort).
1
  • Gangnous JP
  • Dannaoui E
  • Fekkar A
  • et al.
Fungal infections in mechanically ventilated patients with COVID-19 during the first wave: the French multicentre MYCOVID study.

But, as suggested by de Brabander and colleagues, we performed an additional analysis, taking into account the time-dependent bias due to the occurrence of CAPA on the mortality estimate.

2
Practical methods for competing risk data: a review.

Austin and colleagues previously defined a competing risk as an event whose occurrence prevents the occurrence of the primary event of interest.

3
Introduction to the analysis of survival data in the presence of competing risks.

Here, the competing states were ‘time to death’ or ‘discharge’, both during intensive care unit (ICU) stay. No follow-up data was collected after ICU discharge, as mentioned in the manuscript.

2
Practical methods for competing risk data: a review.

Using this model, the cause-specific hazard ratio (CS-HR) at discharge due to CAPA was 0.57 (95% CI 0.38–0.84) and the CS-HR of death due to CAPA was 2.08 (1·47–2·93), compared to 1·45 (1·03–2·03) found with the initial log-rank analysis.

1
  • Gangnous JP
  • Dannaoui E
  • Fekkar A
  • et al.
Fungal infections in mechanically ventilated patients with COVID-19 during the first wave: the French multicentre MYCOVID study.

Overall, we provide evidence that the magnitude effect of CAPA on mortality when considering immortal bias and a cause-specific model increased the strength of the association between the development of CAPA and mortality in the MYCOVID cohort.

J-PG reports personal fees from Gilead and grants and personal fees from Pfizer, outside of submitted work. ED reports grants and non-financial support from Merck Sharp & Dohme and Gilead and non-financial support from Pfizer and Astellas, outside of submitted work. J-FT reports personal fees from Pfizer, Merck, Astellas, and Gilead, outside of submitted work. J-RZ reports grants and personal fees from Merck Sharp & Dohme and personal fees from Novartis and Pfizer, outside of submitted work. M-EB reports grants from Pfizer while conducting the study; grants and non-financial support from Gilead and Pfizer, and non-financial support from Merck Sharp & Dohme, outside of submitted work. JM, BL and SR declare no competing interests.

The references

  1. 1.
    • Gangnous JP
    • Dannaoui E
    • Fekkar A
    • et al.

    Fungal infections in mechanically ventilated patients with COVID-19 during the first wave: the French multicentre MYCOVID study.

    Respir Med lancet. 2021; ten: 180-190

  2. 2.

    Practical methods for competing risk data: a review.

    Statistical methods Res. med. 2012; 21: 257-272

  3. 3.

    Introduction to the analysis of survival data in the presence of competing risks.

    Traffic. 2016; 133: 601-609

Related Articles

Comments are closed.