Association of cancer and outcomes of patients hospitalized for COVID-19 between 2020 and 2023

Study design and setting

This was a prospective cohort study that utilised secondary data from the COVID-19 clinical database hosted by the Infectious Diseases Data Observatory (IDDO). The database contains individual patient data from more than 800,000 hospitalised patients in more than 1,200 institutions from 54 countries across 6 continents. The data were collected using the ISARIC-WHO case report form as a part of the ISARIC-WHO Clinical Characterisation Protocol.^15,16

Study population

We included hospitalised patients of any age with clinically or laboratory-diagnosed SARS-CoV-2 infection. Patients were enrolled between 30^th January 2020 and 10^th January 2023. Patients with unknown cancer status were excluded. Patients admitted for complications due to COVID-19 were followed from the time of hospital admission to discharge or death.

Study variables

We compared the differences in demographic characteristics, comorbidities, treatment with intensive interventions, length of hospitalisation, death (defined as 30-day in-hospital mortality), and hospital outcomes to characterise hospitalised COVID-19 patients with and without cancer.

Severe disease was defined as treatment with higher-level care, including one or more of the following events: admission to an ICU, treatment with invasive mechanical ventilation (IMV), non-invasive ventilation (NIV), high-flow nasal cannula (HFNC), inotropes and/or vasopressors. Length of hospital stay was censored at 100 days.

The presence of cancer was self-reported by patients or relatives and recorded as a binary variable classified as malignant neoplasm in the ISARIC-WHO case report form. Cancer was defined as having a current solid organ or haematological malignancy. Malignancies that had been declared ‘cured’ ≥5 years with no evidence of ongoing disease, non-melanoma skin cancer and benign growths or dysplasia were not included in this definition. Those with unknown cancer status were excluded.

Data collection and validation

We used prospectively collected, international observational data on demographics, clinical features and outcomes of patients hospitalized with COVID-19 with or without cancer (coded as ‘malignant neoplasm’). Data were collected using the ISARIC-WHO Clinical Characterisation Protocol and contributed to a central repository at the University of Oxford, England. Participating sites used the ISARIC-WHO case report form to enter data onto a Research Electronic Data Capture (REDCap, https://www.project-redcap.org/ version 8.11.11, Vanderbilt University, Nashville, TN) database or used local databases before uploading to the central data repository.¹⁷ Open Data Kit is a suitable open access alternative (https://getodk.org). Centrally collated data were wrangled and mapped to the structure and controlled terminologies of the Study Data Tabulation Model (https://www.cdisc.org/standards/foundational/sdtm, version 1.7, Clinical Data Interchange Standards Consortium, Austin, TX) using Trifacta^® software version 9.7.1 (http://trifacta.com). OpenRefine is a suitable open access alternative (https://openrefine.org/) to using Trifacta^®. The data collection, aggregation, curation, and harmonisation process has been previously described.¹⁶ Though more than 50% of the data were collected from low- and middle-income countries, most data on patients with cancer were collected from patients in higher income countries, per World Bank classification. Our statistical analysis plan was designed to explore differences between patient outcomes between these two economic regions as a proxy for the quality of the healthcare setting in a country.

Analysis and statistical method

Continuous variables such as age and length of hospital stay were summarised as means with standard deviations or medians with interquartile ranges depending upon the distribution of data. Categorical variables (sex, presence of cancer, hospital exit outcomes, etc) were summarised as frequencies and percentages.

Categorical variables such as death and treatment with intensive interventions between patients with cancer and those without cancer were compared using the chi-square test. Continuous variables such as length of hospital stay were compared between the two groups using the unpaired t-test or Mann Whitney U test depending on the distribution of data. A Kaplan-Meier curve was plotted to show the cumulative incidence of mortality during hospitalization. To assess the independent effect of cancer on mortality in hospitalized COVID-19 patients, a survival analysis model was fitted to the data. The model was adjusted for the following confounders: age, sex, and country income-level with no explicit adjustment made for further co-morbidities. Unadjusted and adjusted hazard ratios with 95% confidence intervals were reported as measures of association. In addition, we undertook two further sensitivity analyses using different adjustment sets. Denominators on individual analyses differ due to availability of data on different variables across the dataset. A P-value of <0.05 was considered statistically significant.

Information on country income level was obtained from the World Bank (https://datacatalog.worldbank.org/search/dataset/0038543).

All analyses were performed using R version 4.2.2 (IDE PBC, Boston, MA, USA), an open access software. (R: The R Project for Statistical Computing (https://www.r-project.org/).

Ethics considerations

Execution of the ISARIC-WHO Clinical Characterisation Protocol was approved by the WHO Ethics Review Committee (RPC571 and RPC572, 25 April 2013) and by local or national ethics committees for participating sites. Approvals (dates unknown) include the South Central—Oxford C Research Ethics Committee for England (Ref. 13/SC/0149), the Scotland A Research Ethics Committee (Ref. 20/SS/0028) for Scotland, and the Human Research Ethics Committee (Medical) at the University of the Witwatersrand in South Africa as part of a national surveillance programme (M160667), which collectively represent most of the data. Written patient consent for data to be collected and used in research was obtained or waived according to local norms determined by the responsible Ethics Committee. The data were collected using the ISARIC-WHO case COVID-19 report form, locally-tailored versions of the form, or independently designed forms. Arrangements surrounding the pooling, storage, curation and sharing of these data are covered by the IDDO Governance processes.¹⁸

All data were deidentified and ensured of low risk for identification of individuals by a statistical disclosure process prior to sharing. Data were shared under a Data Access Agreement following approval from the IDDO Data Access Committee.¹⁹ Execution of this secondary analysis was approved by the Union Ethics Advisory Group of the International Union against Tuberculosis and Lung Disease, Paris, France (EAG number 18/23, dated 8^th September 2023).

Demographics and comorbidities

Of the 560,547 individuals analysed, 27,243 (4.9%) had cancer. Furthermore, 219,922 (39.2%) individuals that met the criteria for analysis were hospitalised in high-income countries. There were differences in age, sex, country income level, and other comorbidities in the group of patients with cancer versus those without cancer. Those with cancer were older (84.4% versus 46.3% aged ≥60 years), were more likely to be male (58.1% versus 49.1%) and were more likely to come from a high-income country (90.6% versus 36.6%). Of the 10 comorbidities most common in the whole population, all except obesity were more prevalent in the group of patients with cancer ( Table 1).

Mortality, severity, and length of hospitalization

Patients with cancer had higher 30-day in-hospital mortality (29.1% vs 18.0%) and longer duration of hospitalization (median of 12 days (IQR 6.0-22.0) vs 8 days (IQR 4.0-14.0)) compared with those without cancer ( Table 2 and Figures 1 and 2).

Figure 1. Boxplot showing length of hospitalisation among COVID-19 patients with and without cancer hospitalised between 2020-2023 and enrolled to the ISARIC-WHO Clinical Characterisation Protocol.

Figure 2. Kaplan-Meier plot of COVID-19 patients with and without cancer hospitalised between 2020-2023 and enrolled to the ISARIC-WHO Clinical Characterisation Protocol.

However, patients with cancer were reported to have received higher-level care slightly less often than those without cancer (28.9% vs 29.8%) including lower rates of ICU admission (12.6% vs 17.1%) and invasive mechanical ventilation (4.5% vs 7.6%). There were similar levels of treatment with high-flow nasal cannulas (17.5% vs 16.1%), extracorporeal membrane oxygenation (0.1% and 0.5%), non-invasive ventilation (11.6% vs 11.7%), and treatment with inotropes or vasopressors (3.5% vs 4.5%) across both groups ( Table 2).

The effect of cancer and other comorbidities on 30-day in-hospital mortality among COVID-19 patients is reported in Table 3. Hospitalised COVID-19 patients with cancer had a higher risk of 30-day in-hospital mortality compared to those without cancer. The hazard ratio of dying from cancer, adjusted for age, sex and country income level was 1.18 (1.15-1.2).

Adjusted hazard ratios were higher for age and gender compared with those for cancer. Adjusted for sex and country income level, individuals aged ≥ 60 years had the highest hazard ratio 2.43 (2.39-2.46). Adjusted for age and country income level, male sex had a hazard ratio of 1.19 (1.18-1.21).

Among all comorbidities, only diabetes mellitus (HR: 1.32, 95%CI: 1.31-1.34) and chronic pulmonary disease (HR: 1.30, 95%CI: 1.28-1.33) were more strongly associated with an increased risk of death compared with cancer, after adjusting for age, sex and country income level. Two sensitivity analyses were conducted and the results are presented in Table 4 and Table 5. The findings of the sensitivity analyses indicate that the quantified hazards ratio for cancer remained unchanged when adjusted for different comorbidities (Tabel 4). In addition, the quantified association between any of the predictors and outcome remained relatively stable with some/minor differences in the estimated hazards ratio, apart from chronic neurological disorder ( Table 5). However, it has to be cautioned that such a multivariable model with all the predictors included is subject to large missingness.

Strengths and limitations

One key strength of our study was the use of a large sample size, orders of magnitude larger than most previous studies. Therefore, our estimates should be more generalisable and should have a higher power to demonstrate significant associations than previously published studies. We adhered to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines for reporting study findings.

This analysis has several limitations. Though recruitment of patients included in the database used for this analysis targeted those admitted for complications due to COVID-19, the reason for other admissions is not recorded in the database and therefore we cannot verify other reasons for admission but analysed the hospital outcomes of patient with cancer and COVID -19. Any inclusion of patients admitted for other reasons may have had an impact on the subsequent treatment pathway. This study encompassed all patients with current cancer, but we couldn't distinguish between various cancer types as detailed information on each patients’ cancer diagnosis, staging and treatment modalities were not available. However, other studies have highlighted lung cancer and haematological cancers as being those most closely linked to mortality in COVID-19 patients.^13,24,25

Differences in reporting of the type and details of cancer diagnosis across the available literature make it challenging to make comparisons. Studies that analysed data through the use of electronic health records may have included patients in remission.

Our study includes patients from January 30, 2020, to January 10, 2023. During this period, COVID-19 underwent significant changes in genomics, treatment, and epidemiology, with vaccines introduced at varying times across countries. However, our dataset lacks genotyping and reliable vaccination information, which are crucial for analysing temporal changes accurately. Without data on these key factors, especially vaccination status, we cannot provide a robust analysis of changes over time. The impact of evolving vaccination rates on outcomes is likely substantial but impossible to calculate with our current data.

We acknowledge this limitation, and it has informed changes to ISARIC's case report forms for future outbreaks to address these data gaps. The majority of data on patients with cancer (90.6%) were collected from patients in high income countries. So, no inferences could be drawn from patient outcomes linked with World Bank income classifications.