Cancer is associated with the severity and mortality of patients with COVID-19: a systematic review and meta-analysis

2.1. Eligibility criteria

Studies that met the following criteria were included: (1) patients have a laboratory-confirmed diagnosis of COVID-19; (2) described the prevalence of cancer in infected patients, or provided data of cancer between patients with severe or non-severe illness or between non-survivors and survivors; (3) published in Chinese and English.

We excluded following studies: (1) studies with a sample size of fewer than 20 patients; (2) studies focused on only suspected cases or suspected cases and confirmed cases; (3) studies did report data related to cancer patients; (4) review articles, protocols, guidelines, consensus, comments, abstracts, letters, and editorials.

2.2. Literature search

We conducted comprehensive searches in PubMed, EMBASE.com, the Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science, Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure (CNKI), and Wanfang Database up to April 12, 2020. Search terms included the following words: “COVID-19”, “coronavirus disease-19”, “new coronavirus”, “2019-nCoV”, “novel corona virus”, “novel coronavirus”, “nCoV-2019”, “novel coronavirus pneumonia”, “2019 novel coronavirus”, “coronavirus disease 2019”, “SARS-CoV-2”, “severe acute respiratory syndrome coronavirus 2”, “neoplasms”, “neoplasia”, “tumor”, “tumour”, “cancer”, “malignancy”, “clinical characteristic” “clinical feature”, “risk factor”, and “comorbidities”. The search strategy of PubMed is presented in Appendix Word 1. Reference lists of eligible studies and relevant systematic reviews were manually searched for potentially eligible studies.

2.3. Study selection process

The retrieved records were imported into EndNote X8 (Thomson Reuters (Scientific) LLC Philadelphia, PA, US) software for management. Two authors independently (YG and ML) screened the titles and abstracts of the records to determine if they met the inclusion criteria. Then, the same two authors retrieved the full text of all potentially eligible studies and assessed the eligibility of each study according to the inclusion criteria. Regarding multiple studies from the same teams or studies with samples from the same settings, we evaluated the time frame and detailed data of the study. For studies with overlapping data, we included the study with a larger sample size. Conflicts were resolved by discussions with a third reviewer (JHT).

2.4. Data extraction and quality assessment

A standardized data extraction form was developed using Microsoft Excel 2016 (Microsoft Corp, Redmond, WA, www.microsoft.com) through discussions with the review team and was revised after piloting on a random of five studies. The data extracted included: (1) study characteristics (first author, year of publication, country of the corresponding author, journal name, publication language, study setting, recruitment time frame), population characteristics (age, sex, sample size), and outcomes of interest (number of cancer patients, sever cases, non-severe cases, deaths, and survivors). The severe illness was defined in this study as patients experiencing acute respiratory distress syndrome (ARDS), requiring mechanical ventilation, requiring vital life support, or requiring intensive care unit admission (ICU) support [12, 13]. We used the Newcastle-Ottawa quality assessment scale (NOS) to assess the quality of included studies [14]. Studies that obtained more than 7 stars were considered as high quality, 5-7 stars were considered as moderate quality, and lower than 5 stars were considered as low quality [15]. The data extraction and quality assessment were performed by one reviewer (YG, ML, SZS, or YMC) and checked by a second reviewer (YS and JC). Discrepancies were resolved by consensus or by the discussion with a third reviewer (JHT).

2.5. Statistical analysis

We conducted a rate meta-analysis to estimate the pooled prevalence and its 95% confidence interval (CI) for cancer among patients with COVID-19. Pairwise meta-analyses were conducted to compute the odds ratio (OR) and 95% CI of cancer prevalence in COVID-19 patients with or without severe illness, and non-survivors or survivors. The meta-analyses were performed using the inverse variance method with the random-effects model. The statistical heterogeneity was assessed with the I² statistic, and value of < 25%, 26-50%, and > 50% was considered as low, moderate, and high level of heterogeneity, respectively [16].

We planned to conduct subgroup analyses of the outcomes between different countries. Sensitivity analyses were conducted by excluding studies with a sample size of less than 100. We also performed univariate meta-regression analyses to assess if either the outcomes or the heterogeneity was associated with the publication languages and number of centers of the study conducted. The funnel plot and Eggers test were adopted to detect the publication bias. All analyses were conducted using Stata (13.0; Stata Corporation, College Station, Texas, USA Stata) and the statistical level of significance was set at P < 0.05.

3.1. Screening results

We identified 1592 records through the literature search, among which 838 were from English databases, 751 were from Chinese databases, and 3 were from other sources. After removing duplicates and reviewing the titles and abstracts, 1466 records were excluded. Through full-text evaluation of the remaining 126 records, 92 records were further excluded. Finally, 34 studies [17–50] were included for our meta-analyses. The flowchart of the screening process is presented in Figure 1.

• Download figure
• Open in new tab

Figure 1.

The flowchart of the screening process

3.2. General characteristics and quality of studies

All included studies [17–50] were published in 2020, enrolled patients between December 11, 2019 and March 21, 2020. 31 studies [17–20, 23–25, 27–50] were performed in China and 3 studies [21, 22, 26] were conducted in Italy, France, and Korea. 23 studies [17–39] published in English and 11 studies [40–50] published in Chinese. The sample size per study ranged from 28 to 1,591 and the total sample size was 8,080 (4,867 males, 3,213 females). Seven studies [18, 20, 21, 23, 29, 33, 44] were rated as high quality and 27 studies [17, 19, 22, 24–28, 30–32, 34–43, 45–50] were rated as moderate quality according to the NOS scale. The detailed characteristics and quality of the included studies are summarized in Table 1.

3.3. Prevalence of cancer in COVID-19 patients

All included studies [17–50] reported the number cancer patients among the COVID-19 patients. The meta-analysis showed that the pooled prevalence of cancer in patients with COVID-19 was 2.0% (95% CI: 2.0% to 3.0%), with significant heterogeneity among the studies (I² = 68.8%, Figure 2). The Subgroup analysis indicated the prevalence was 5.0% (95% CI: 4.0% to 6.0%; 1 study [21], 1,591 patients) in Italy, 6.0% (95% CI: 3.0% to 14.0%; 1 study [22], 80 patients) in France, 4.0% (95% CI: 1.0% to 18.0%; 1 study [26], 28 patients) in Korea, and 2.0% (95% CI: 2.0% to 3.0%; 31 studies [17–20, 23–25, 27–50], 6,381 patients) in China (Figure 2).

Sensitivity analyses showed that the prevalence did not change after excluding studies with a sample size of less than 100 (Appendix Figure 1).

3.4. Association between cancer and the severity of COVID-19

Thirteen studies [20, 23, 24, 29, 32, 37, 40, 43–45, 47, 49, 50] involving 3,450 patients provided cancer data with comparison between COVID-19 patients with severe and non-severe illness. The meta-analysis revealed that cancer was associated with a 2.84-fold significantly enhanced risk of severe COVID-19 disease (OR = 2.84, 95%CI: 1.75 to 4.62, P < 0.001; I² = 7.0%) (Figure 3). After excluding studies with a sample size of less than 100, we observed a stronger association (OR = 3.83, 95%CI: 2.21 to 6.63) between cancer and COVID-19 severity (Appendix Figure 2).

3.5. Association between cancer and the mortality of COVID-19

Six studies [18, 19, 23, 27, 36, 38], totaling 2,671 samples, reported cancer data between dead and surviving COVID-19 patients. Cancer was observed to be associated with a significantly enhanced risk of death (OR = 2.60, 95%CI: 1.28 to 5.26, P = 0.008; I² = 6.2%), Figure 4. Sensitivity analysis by excluding studies with a sample size of less than 100 showed similar results (OR = 2.63, 95%CI: 1.14 to 6.06) (Appendix Figure 3).

3.6. Meta-regression analyses

Univariate meta-regression analyses revealed that publication languages and the number of centers of studies conducted were not the sources of heterogeneity or the factors affecting the cancer prevalence (Appendix Figure 4-5), the association between cancer and COVID-19 severity (Appendix Figure 6-7), and the association between cancer and the mortality of COVID-19 (Appendix Figure 8).

3.7. Publication bias

We found that there was a possibility of publication bias for cancer prevalence (P = 0.001) (Appendix Figure 9). Eggers tests indicated there was no statistically significant publication bias for the association between cancer and COVID-19 severity (P = 0.865) (Appendix Figure 10), and the association between cancer and COVID-19 mortality (P = 0.439) (Appendix Figure 11).

4.1. Principal findings

This study included 34 studies from China, Italy, France, and Korea, identified from a comprehensive search of seven electronic databases. We systematically evaluated the prevalence of cancer among COVID-19 patients and the association between cancer and the severity and mortality of patients with COVID-19. Our meta-analyses indicated that the pooled prevalence of cancer in patients with COVID-19 was 2.0% and the prevalence in Italy, France, and Korea were higher than that in China. Cancer was associated with a 2.84-fold significantly enhanced risk of severe COVID-19 disease and a 2.60-fold significantly enhanced risk of death in patients with COVID-19. Sensitivity analyses showed that the results did not change substantially after excluding studies with a sample size of less than 100.

4.2. Comparison with other studies

A previous meta-analysis revealed that the pooled prevalence of malignancy among hospitalized COVID-19 patients was estimated to be 0.92% (95% CI: 0.56% to 1.34%) [51]. Another meta-analysis showed that the overall prevalence of cancer in patients with COVID-19 was 2.0% (95% CI: 2.0% to 3.0%), and the prevalence in studies with a sample size < 100 was slightly higher than that in studies with a sample size > 100 [52]. In the current study, we found the prevalence of cancer in COVID-19 patients was 2.0% (95% CI: 2.0% to 3.0%) and the prevalence did not change after excluding studies with a sample size of less than 100. Compared to these two studies, our study has several advantages that make it more conclusive. First, the present meta-analysis included 34 studies involving a total of 8,080 COVID-19 patients compared to no more than 3,561 patients in previous meta-analyses. Thus, our study had enlarged sample sizes and added statistical power of nearly 4500 cases. Second, studies included in previous meta-analyses were all from China, so the results may not apply to other countries and data selection bias may exist. However, our meta-analysis included studies from four countries, although only three studies from Italy, France, and Korea. Third, in addition to conducting subgroup analyses to evaluate the difference of prevalence between different countries, we also performed sensitivity analyses and meta-regression analyses and these analyses indicated that the results of our study were stable.

A previous meta-analysis found that there was no correlation between cancer and the severity of patients with COVID-19 [7]. However, our meta-analysis indicated that cancer was significantly associated with an increased risk of severe COVID-19 disease, which was inconsistent with the previous meta-analysis [7]. The previous reviewers conducted a meta-analysis based on only four studies with a sample of 1,356 patients [7]. We performed a meta-analysis of 3,450 cases from thirteen studies and sensitivity analysis by excluding studies with a sample size of less than 100 showed a stronger association. Therefore, the result of our study is more convincing.

4.3. Implications for research and practice

According to the current meta-analysis, the pooled prevalence of cancer in patients with COVID-19 was 2.0%. Combined with previously published results [9, 51, 52], we can conclude that patients with cancer have an increased risk of COVID-19. The development of cancer is usually related to a blunted immune status [53, 54], and anti-cancer treatments (such as chemotherapy and surgery) can also put cancer patients in an immunosuppressive state [8]. Therefore, immunodeficiency may be the main reason for cancer patients susceptible to COVID-19. Our subgroup analyses found that the prevalence of cancer among COVID-19 patients in Italy, France, and Korea were higher than that in China, although the result was limited by the sample size. These results suggest that cancer patients should be provided with special precautions and advised to use stronger personal protection.

Our meta-analysis found that cancer was associated with a 2.84-fold significantly increased risk of severe illness in patients with COVID-19, as well as with a 2.60-fold increased risk of death. Although our data are potentially limited by the sources of studies, these findings highlight the need for oncology professionals to be vigilant to the increased risk of serious illness and death associated with COVID-19 infection in patients with cancer or cancer survivors. A previous study showed that anti-tumor therapy within 14 days of COVID-19 diagnosis increased the risk of developing serious events and recommended that cancer patients with COVID-19 infection should avoid treatments causing immunosuppression [10]. However, there is currently no recommendation regarding the treatment strategies of immunotherapy, radiotherapy, chemotherapy, or delaying adjuvant therapy for cancer patients with COVID-19 [52]. The results of our meta-analysis also provide the latest references for the development of new guidelines. High-quality evidence-based guidelines clarify protection measures for cancer patients, and care and treatment strategies for cancer patients with COVID-19 are urgently needed.

4.4. Strengths and limitations

To the best of our knowledge, this is the first meta-analysis systematically evaluated the prevalence of cancer among COVID-19 patients, and the association between cancer and the severity and mortality of patients with COVID-19. Besides, we also conducted sensitivity analyses and meta-regression analyses to evaluate factors that may affect the results. However, our study also has some limitations. First, most of the studies included were from China, so the current findings may not fully reflect the global situation and should be interpreted with caution. Second, although this meta-analysis included 34 studies, there are few data available for subgroup analysis. Third, we performed sensitivity and meta-regression analyses to explore heterogeneity, but some factors were not evaluated due to limited data. Fourth, the patient overlap is still possible between a few studies, although we have ruled out many studies with overlap samples during the study selection and data extraction processes. Finally, we did not evaluate which types of cancer patients are more susceptible to COVID-19 or more associated with severe illness and mortality. As data from more countries become available, it is necessary to update this study and performed more comprehensive analyses to answer questions to guide clinical practice.