Cancer and COVID-19 - Authors' reply.

We thank Dimitrios Moris and col­leagues and Alexandre Malek and colleagues for their insightful commentary about the CCC19 study findings. We value the opportunity to further characterise mortality outcomes beyond our initial report.

With a median of 30 days (IQR 21–90) follow-up, as of Aug 21, 2020, 30-day all-cause mortality increased to 20% (154 of 754 patients who either died within 30 days or had at least 30 days of follow-up). Planned time-to-event analyses will refine these estimates. 121 (79%) deaths were attributed to respiratory failure (appendix). In our cohort, the category of respiratory failure encompasses deaths from any respiratory failure syndrome. Although respiratory failure caused by cancer, its therapies, or other comorbidities could confound the cause of death attribution, this is unavoidable. Diagnostic procedures are challenging with COVID-19, autopsies are rare, and Vital Statistics Reporting Guidance specifically directs medical certifiers to list COVID-19 as the underlying cause of death, with the most immediate cause of death (eg, respiratory failure) listed first. Because this method might overestimate COVID-19-related deaths, we reported all-cause mortality.

We share Moris and colleagues' concerns regarding the potential increase of cancer mortality caused by delays in cancer screening, diag­nosis, and care delivery because of severe acute respiratory syndrome coronavirus 2—a concern already borne out in some early analyses. The prospect of widening existing racial and socioeconomic disparities in cancer outcomes is of real concern and central to forthcoming analyses of our cohort.

Malek and colleagues discuss import­ant limitations in applying the results to patients with haematological malignancies. The small number of patients with specific haematological malignancies in the CCC19 cohort required the broad categorisation with non-exclusive categories, resulting in apparent numerical discrepancies. Only 53 (26%) patients with haema­tological malig­nancies were in remission, limiting the conclusions in this subgroup (appendix). Subsequent studies have shown a high risk of severe COVID-19 outcomes for patients with haema­tological malignancies.4, 5 Despite a larger sample size, these analyses still do not have the power to identify, at the granular level, associations between the clinical status of the haematological malignancy, therapeutic modalities, and outcomes. The CCC19 cohort now includes more than 900 patients with haematological malignancies, and further analyses are underway.

Regarding laboratory data, 449 (48%) patients presented with mild COVID-19, most of whom had no available base­line laboratory data. Additionally, our sample size only allowed the interroga­tion of the reported clinical variables (which were established a priori) in multivariable modelling. We agree that examining the independent prognostic value of laboratory parameters is vital; we will soon present a larger analysis addressing this. Although we firmly support robust methods and highlight limitations, it is imperative to deliver timely and valuable information to the community.