Background
The COVID-19 pandemic, since its emergence in December 2019, has caused significant mortality and morbidity across the globe. In early 2021, more than 100 million people had been infected with more than 2 000 000 deaths. It has disrupted many aspects of healthcare delivery and overwhelmed many hospitals and their intensive care resources.
Patients with cancer are an at-risk population as they are generally older, immunocompromised and have multiple comorbidities. However, the outcomes of patients with cancer with COVID-19 remains conflicted despite multiple cohort studies and meta-analyses. A multicentre study of 105 patients with cancer in China suggested that patients with cancer with COVID-19 infection had higher risk of severe disease outcomes as compared with 536 age-matched non-cancer patients.1 Another study from the Chinese Centre for Disease Control and Prevention Group reported a case fatality rate (CFR) of 2.3% (1023/44672) among confirmed COVID-19 cases and a more than double CFR of 5.6% (6/107) among the patients with cancer with COVID-19.2 In contrast, a database report of 5688 COVID-19 positive patients from a tertiary hospital in New York found no significant difference in risk of death between cancer and non-cancer patients (RR 1.15, 95% CI 0.84 to 1.57) although there was an increased risk of intubation among patients with cancer (RR 1.89, 95% CI 1.37 to 2.61),3 and other cross-sectional studies also showed similar risk of morbidity and mortality as the general population4 5 though with a small number of patients. A few meta-analyses had recently addressed this as well and found that there was increased risk of severe COVID-19 infection among patients with cancer on active oncological treatment,6 especially chemotherapy.7
Hence, we aimed to review the current literature and to determine first the proportion of patients with cancer among the reported COVID-19 cases, second whether patients with cancer with COVID-19 had more severe outcomes and higher mortality compared with the general population and lastly, if there were any specific characteristics of patients with cancer that were associated with an increased risk of severe COVID-19 infection.
Methods
Search methods and study selection
A systematic search of the literature was conducted to identify studies that include patients with cancer with COVID-19 infection. The investigators searched biomedical databases including PubMed, Medline and Google Scholar using the terms COVID-19, coronavirus, nCoV, SARS-CoV-2, characteristics, comorbidity, mortality, death, risk, oncology, cancer, malignancy, neoplasm, tumor, tumour, and carcinoma (online supplemental appendix 1). The search was conducted on 29 January 2021 for a search period from 1 November 2019 to 31 December 2020. The references of relevant meta-analysis (MA) and published reviewers were searched for additional eligible studies. The articles included met the following criteria: (1) COVID-19 diagnosis should be confirmed by real time PCR, or based on 2019 Novel COVID-19 Diagnostic criteria, if applicable (online supplemental appendix 1), (2) studies should contain clinical data of at least ten COVID-19 patients with at least one patient with a diagnosis of cancer (including haematological malignancies), (3) severity outcomes (defined as hypoxaemia, admission to intensive care unit (ICU), and/or mechanical ventilation) and/or death outcomes should be presented and (4) studies should be available in English and full text. Exclusion criteria included studies that reported only specific patient populations, that is, intensive care setting, paediatric population or single tumour type cohorts. Studies published online ahead of print and selected preprint papers were included at the discretion of the authors. For the pooled analysis on severity and death outcomes of cancer and general COVID-19 patients, we mandated that the studies included at least 10 patients with cancer.
Supplemental material
The literature search and study selection were performed independently by three of the authors (SH, QZ and VSY), and discrepancies were resolved by consensus. The study followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (online supplemental appendix 1).
Data extraction
Two investigators (SH and VSY) independently graded the studies based on Newcastle-Ottawa Scale (NOS) for assessing the quality of non-randomised studies and discrepancies were resolved by consensus (online supplemental appendix 2). Data extraction was performed by SH, and the parameters extracted included the location (hospital and city) of the study, type of study, number of patients, demographics including median or mean age, gender, proportion of patient with comorbidities including cancer, and the clinical outcomes of severe disease and death (of patients with cancer and overall COVID-19 population) were collected for the available studies.
Statistical analysis
All statistical analyses were performed using the STATA V.15 software. MA using random-effects models was performed to analyse the pooled proportion of cancer among COVID-19 patients. The included studies were stratified into subgroups by region (within Hubei Province, China vs outside Hubei Province, including other countries), median/mean age (above or equal to vs below 60 years old) and gender (proportion of male more than or equal to 60% vs less than 60% in population). We further analysed the pooled relative risks (RR) of severe COVID-19 infection outcomes and related deaths for patients with cancer with COVID-19 versus the general population with COVID-19. The subgroup of lung cancer and stage IV patients were also evaluated for RR of severe COVID-19 infection outcomes. The results were presented as proportions (decimal) with 95% CI. Heterogeneity between studies was analysed by the I2 statistics, where I2 >50% was considered moderate heterogeneity and χ2 p<0.05 was considered statistically significant. χ2 tests were performed to look for associations between patient with cancer characteristics (age, gender, lung cancers, stage IV cancers, comorbidities and recent systemic therapy) and severe COVID-19 outcomes in the individual patient data (IPD).
Patient and public involvement
The research question was formed based on the urgent need to understand the effect of COVID-19 on patients with cancer and to address the clinical decisions that follows this understanding. No patient was directly involved in this systematic review and MA. We would like to thank the authors and also the patients that were involved in the publications that we had included in our MA.
Results
Eligible studies and characteristics
We identified 3083 relevant publications after literature search and review. Additional studies were identified from the references of published meta-analyses. After screening and eligibility assessment, we included in our MA a total of 57 case series involving 63 413 patients (figure 1; online supplemental appendix 2; table 1). Studies that did not meet the MA criteria were summarised in online supplemental appendix 2; table 2.
Flow chart of study identification and exclusion for data analysis.
Among the 57 studies, 37 studies were from public hospitals in China (24 from Hubei province) with the rest from worldwide (USA, Canada and Spain in CCC19 study) (1), Europe (1), Korea (1), Iran (1), France (1), Spain (1), Italy (2), UK (3) and USA (9) (table 1). Most studies were retrospective cohort studies and a few observational cross-sectional studies. The quality of all the included studies ranged from 6 to 8 (out of 9) using the NOS (online supplemental table 3). Among the 57 case series, 13 contained detailed descriptions of COVID-19 patients with cancer (table 2). Only five studies provided detailed IPD (online supplemental appendix 3; table 1). Forty-seven studies were used for proportion analyses for cancer diagnoses among COVID-19 patients. For the pooled analysis on RR of severe outcomes among cancer COVID-19 patients, we identified 10 case series that contained severity data, and 7 studies that contained mortality data for analysis (studies with at least 10 patients with cancer). Five studies contained severity outcomes of lung cancer versus other patients with cancer with COVID-19 infection, and four studies contained severity outcomes of patients with stage IV vs non-stage IV cancer.
Baseline characteristics, comorbidities and severity outcomes of 57 COVID-19 case series
Baseline characteristics of patients with cancer with COVID-19 infections from 13 case series
Overall, there were more males than females affected, with a median age above 60 in all case series. This median age was notably higher than the non-cancer populations reported in previous studies.8 Severe COVID-19 disease among patients with cancer ranged from 9% to 54% in 13 eligible studies (table 2). Overall, 230 patients from five case series were identified with oncological history, stage, recent treatment history and clinical outcomes (online supplemental appendix 3; table 1). However, among these patients, there was inconsistent reporting in those on active systemic chemotherapy and some clinical data on stage of cancer were missing, limiting the usefulness of the dataset in predicting the outcomes of COVID-19 in patients with active cancer.
Mortality outcomes of patients with cancer with COVID-19
Overall, the mortality rates of patients with cancer with COVID-19 infection ranged widely from 5.9% to 50% across 20 studies (table 3). The pooled RR of death was significantly higher in cancer patients with COVID-19 infection than in the general population with COVID-19 infection based on seven studies as shown in table 3 (RR 1.41; 95% CI 1.15 to 1.73), I2=26.2%, χ2 p=0.229 (figure 5A). The pooled mortality rate was estimated at 0.22 (95% CI 0.21 to 0.33) among patients with cancer with COVID-19, which was significantly higher than that of the overall COVID-19 patients, at 0.09 (95% CI 0.07 to 0.11; online supplemental appendix 3; figure 2A–C).
Mortality rate among patients with cancer diagnosed with COVID-19 infection*
(A) Random effects pooled relative risk (RR) of death for cancer COVID-19 patients versus total population of COVID-19 patients. (B) Random effects pooled RR for severe outcome of cancer COVID-19 patients versus total population of COVID-19 patients.
Severity outcomes of patients with cancer with COVID-19
Further analyses were conducted for the 10 case series with available severity outcomes for both cancer and general COVID-19 patient cohorts. We found that patients with cancer had higher risk of severe COVID-19 outcomes compared with the general COVID-19 population, with a pooled RR of 1.49 (95% CI 1.18 to 1.87) and substantial heterogeneity, I2=66.7%, χ2 p=0.001 (figure 5B). Also, we noted that the pooled proportion of severe outcomes in patients with cancer was 0.42 (95% CI 0.30 to 0.54), which was higher than the proportion of severe cases of all COVID-19 patients at 0.27 (95% CI 0.21 to 0.33; online supplemental appendix 3; figure 3A–C). There was no statistically significant difference in RR of severe outcomes between those patients (available data from five studies identified) with lung cancers vs non lung cancers (RR at 1.46; 95% CI 1.84 to 2.52, I2=48.1%, χ2 p=0.7; figure 6A). There was no statistically significant difference in RR of severe outcomes between stage IV vs non stage IV cancers from four studies identified (RR 1.48; 95% CI 0.89 to 2.47; figure 6B, online supplemental figure 4).
(A) Random effects pooled relative risk (RR) for severe outcome of lung cancer versus other patients with cancer infected with COVID-19. (B) Random effects pooled RR for severe outcome of patients with stage IV cancer vs non-stage IV cancer infected with COVID-19.
The severe outcomes of COVID-19 in patients with cancer was also analysed from the available 230 IPD and summarised in online supplemental table 1. Only patient age and gender appeared to be significantly associated with severity outcomes by χ2 test (online supplemental table 2).
Discussion
We performed a systematic review and MA to elucidate the proportion of cancer diagnoses among COVID-19 patients, and whether patients with cancer had more severe outcomes compared with the general population. We also looked at the patient characteristics that may be associated with severe clinical outcomes among these patients. To our knowledge, this is the largest MA that directly addresses the clinical outcomes of COVID-19 patients with cancer, including the RR of severity and death as compared with general COVID-19 patients.
First, we found that the pooled proportion of patients with cancer among COVID-19 cases was 4%. This result was similar to the proportion rate from an earlier MA9 of 11 studies with a proportion of 2%, but higher than the 0.2% described by the Chinese Centre for Disease Control and Prevention Group.2 We found that the cancer proportion was higher among older patients and in male patients with COVID-19. In the subsequent IPD analysis, it was also advanced age and male gender that were significantly associated with severity of COVID-19 infection. We also noted that the cancer proportion among the COVID-19 case series from Hubei province was 3% (online supplemental appendix 3; figure 1), which was higher than the estimated cancer prevalence of 0.45% in the region.10 However, cancer proportion can vary across geographical locations and specialist medical centres.
We have shown that COVID-19 patients with cancer had significantly more severe outcomes compared with the general COVID-19 patients, with a pooled RR of 1.49 (95% CI 1.18 to 1.87, p=0.001). The pooled estimated proportion of severe COVID-19 outcomes was higher among patients with cancer in our MA compared with the largest cancer cohort data (CCC19).11 Overall, the proportion of patients with cancer with severe COVID-19 outcomes varied across studies but remained high, highlighting the vulnerability of patients with cancer to SARS-CoV-2 infection.
We also observed that lung cancer was the most common cancer among those affected by COVID-19 in Chinese studies (table 2) and this may be due to the fact that lung cancer is the most common cancer in China, rather than a direct relation to the risk of COVID-19 infection.12 A few case series noted that patients with lung cancers had significantly worse outcomes compared with non-lung cancer cohorts,8 13 which was also supported by data from TERAVOLT,14 an international database of thoracic cancers (not included in our MA as it is a single tumour registry). TERAVOLT demonstrated 33% mortality and high morbidity among patients with thoracic malignancies, but with low ICU admission rates, presumably partly due to resource allocations and patient prognosis. However, in our MA, we did not find the RR of severity to be significant for lung cancers (vs other tumour types) or stage IV cancers (vs stage I–III) across four studies. This was also echoed by our IPD analysis, where lung cancer and the stage of cancer were not significantly associated with severe outcomes. This could reflect that lung cancer or stage IV cancer did not necessarily confer a higher risk of severe COVID-19 infection, but given the small number of included studies for this part of the analysis, this conclusion should be cautiously interpreted.
Separately, haematological cancers also showed severe outcomes with high mortality rates in multiple cohort studies.8 15 The available individual data in our study were not large enough to analyse the influence of modality of systemic treatment (targeted therapy, immunotherapy or chemotherapy), surgery or radiotherapy on the outcome of COVID-19 infections. Recent studies did not consistently find active systemic therapy to be associated with risk of severe COVID-19 infection16 though some studies suggest that patients on recent chemotherapy10 or immunotherapy8 were at higher risk. These questions remain unanswered and require further collaborative efforts to identify risk factors of severe COVID-19 infection among patients with cancer. One suggestion is to have standardised reporting format for cancer COVID-19 cohorts including specific tumour type, stage, date of last chemotherapy or systemic treatment, comorbidities, active versus past cancer diagnosis and also the severity and mortality data for each subgroup.
Overall, the pooled RR of death among the patients with cancer was 1.41 (95% CI 1.15 to 1.73), compared with general COVID-19 patients. However, pooled mortality risk of patients with cancer with COVID-19 should be interpreted with caution due to the limited sample size. Some case series also had patients who remained hospitalised or with incomplete patient outcome at the time of reporting. CFR were lower at 5.6% among cancer COVID-19 patients in a large Chinese report2 and 11.1% from a tertiary centre in New York, USA.3 Notably, among selected COVID-19 mortality reports from Italy and USA, around 20%–30% of deaths had a history of cancer.17 18 We would encourage mortality data to be taken from database reports instead of small case series from single centres.
In light of the severity of COVID-19 in patients with cancerpatients with cancer, efforts should be undertaken to reduce nosocomial exposure in patients with cancer, as nosocomial infections may contribute to incidence of COVID-19 in oncology patients.19 The decision to attend clinics and continue non-urgent systemic therapy should be made judiciously, especially in regions with high burden of disease transmission. This has been echoed by various oncology bodies, including American Society of Clinical Oncology and European Society of Medical Oncology.20 Given the higher risk of severe outcomes and mortality among patients with cancer, it is important that patients with cancer with COVID-19 infection should be monitored closely and decisions for ICU support discussed early. ICU outcomes should also be an important question for future retrospective studies in view of the ethical and clinical considerations of putting patients with advanced cancer on mechanical ventilation, especially in countries with significant outbreak and limited ICU resources. In light of the recent advances in COVID-19 vaccine efforts, another area of active discussion is that of the risk and benefit of vaccination for patients with cancer, especially those who are deemed to be significantly immunocompromised. Our data shows that there is risk of severe COVID-19 infection and risk of death among patients with cancer compared with general population. Hence, this factor should be strongly considered in the individualised decision for vaccination.
The limitations of this MA include the heterogeneity and retrospective nature of the case studies, including reporting and selection biases. A few studies were reported from the same hospitals and hence there may be overlapping patients although efforts were taken to reduce this. During screening of papers by the authors, we have ensured that the analysed studies were not reported from the same hospital during the same period of admission (online supplemental appendix 2; table 1). The small numbers and incomplete death outcomes at the time of reporting also restricted our mortality analysis. Due to the limited clinical data, we could not adjust for co-morbidities and other characteristics that may contribute to severe disease. It should also be noted that severity outcomes were defined inconsistently across some studies but efforts were taken to indicate the different definitions in table 1. Of note, some earlier studies also included clinically diagnosed COVID-19 cases.
While we acknowledge other risk factors such as age,21–23 gender,24 cardiovascular risk factors23 25 26 including diabetes mellitus27 28 and hypertension29 remain important in COVID-19 infection, this is the largest MA to show that malignancy is also a significant risk factor for severe disease and mortality. However, we did not identify other tumour or treatment-related factors associated with severe outcomes among patients with cancer with COVID-19. We also did not stratify for gender, age and comorbidities in our analysis due to limited available data for pooled analysis on severity and mortality. As more data becomes available in the literature, further analyses may be performed to discern the clinical outcomes of patients with cancer with COVID-19 infection, with specific attention to underlying lung cancers, haematological cancers and the effects of various systemic therapies or other treatment modalities. CCC19, UKCCMP (Coronavirus Cancer Monitoring Project UK) and TERAVOLT are some examples of collaborative efforts on the outcomes of patients with cancer with COVID-19 infection, though a limitation is that there is a lack of comparison against general population. Information on COVID-19 patients with cancer should also be collected in more detail and with consistency across studies, to allow further meta-analyses to ascertain the outcomes of patients with cancer with COVID-19 infection.
Conclusion
Given that COVID-19 may persist in the foreseeable future, it is imperative to continue active research and data collection on the effect of SARS-CoV-2 on oncology patients. Due to the clinical burden of their disease, oncology patients will continue to require systemic chemotherapy and urgent care at medical centres.30 As such, treating physicians need to be able to make more discerning decisions on the choice of systemic treatment and escalation of care.
Overall, we found through MA that the severity and mortality rates of COVID-19 in patients with cancer were higher than that in the general population. However, the dataset remains small and should be interpreted with caution. We hope that with collaborative efforts, comprehensive patient outcomes of COVID-19 patients, and specifically, of patients with cancer, can be collated to better understand the disease trajectory and outcomes of COVID-19.