key: cord-0253287-kn0jegpu
authors: Visci, G.; Di Felice, G.; Teglia, F.; Angelini, M.; Boffetta, P.
title: Effect of SARS-CoV-2 infection on outcome of cancer patients: A systematic review and meta-analysis of studies of unvaccinated patients
date: 2021-10-23
journal: nan
DOI: 10.1101/2021.10.20.21265284
sha: eefc78a1571ab47aea99cfbe0d75074a022ccb8e
doc_id: 253287
cord_uid: kn0jegpu

Background. Since the beginning of the SARS-Cov-2 pandemic, cancer patients affected by COVID-19 have been reported to experience poor prognosis; however, a detailed quantification of the effect of SARS-CoV-2 infection on outcome of unvaccinated cancer patients has not been performed.Methods. To carry out a systematic review of the studies on outcome of unvaccinated cancer patients infected by Sars-Cov-2, a search string was devised which was used to identify relevant publications in PubMed up to December 31, 2020. We selected three outcomes: mortality, access to ICU, and COVID-19 severity or hospitalization. We considered results for all cancers combined as well as for specific cancers. We conducted random-effects meta-analyses of the results, overall and after stratification by region. We also performed sensitivity analyses according to quality score and assessed publication bias.Results. For all cancer combined, the pooled odds ratio (OR) for mortality was 2.32 (95% confidence interval [CI] 1.82-2.94, I2 for heterogeneity 90.1%, 24 studies), that for ICU admission was 2.39 (95% CI 1.90-3.02, I2 0.0%, 5 studies), that for disease severity or hospitalization was 2.08 (95% CI 1.60-2.72, I2 92.1%, 15 studies). The pooled mortality OR for hematologic neoplasms was 2.14 (95% CI 1.87-2.44, I2 20.8%, 8 studies). Data were insufficient to perform a meta-analysis for other cancers. In the mortality meta-analysis for all cancers, the pooled OR was higher for studies conducted in Asia than studies conducted in Europe or North America. There was no evidence of publication bias.Conclusions. Our meta-analysis indicate a two-fold increased risk of adverse outcomes (mortality, ICU admission and severity of COVID-19) in unvaccinated cancer patients infected with SARS-CoV-2 compared to uninfected patients. These results should be compared with studies conducted in vaccinated patients; nonetheless, they argue for special effort to prevent SARS-CoV-2 infection in patients with cancer.Funding. No external funding was obtained.

Methods. To carry out a systematic review of the studies on outcome of unvaccinated cancer patients 23 infected by Sars-Cov-2, a search string was devised which was used to identify relevant publications in 24 PubMed up to December 31, 2020. We selected three outcomes: mortality, access to ICU, and COVID-19 25 severity or hospitalization. We considered results for all cancers combined as well as for specific cancers. 26 We conducted random-effects meta-analyses of the results, overall and after stratification by region. 27 We also performed sensitivity analyses according to quality score and assessed publication bias. 28 Results. For all cancer combined, the pooled odds ratio (OR) for mortality was 2.32 (95% confidence 29 interval [CI] 1.82-2.94, I 2 for heterogeneity 90.1%, 24 studies), that for ICU admission was 2.39 (95% CI 30 1.90-3.02, I 2 0.0%, 5 studies), that for disease severity or hospitalization was 2.08 (95% CI 1.60-2.72, I 2 31 92.1%, 15 studies). The pooled mortality OR for hematologic neoplasms was 2.14 (95% CI 1.87-2.44, I 2 32 20.8%, 8 studies). Data were insufficient to perform a meta-analysis for other cancers. In the mortality 33 meta-analysis for all cancers, the pooled OR was higher for studies conducted in Asia than studies 34 conducted in Europe or North America. There was no evidence of publication bias. 35 Conclusions. Our meta-analysis indicate a two-fold increased risk of adverse outcomes (mortality, ICU 36 admission and severity of in unvaccinated cancer patients infected with SARS-CoV-2 37 compared to uninfected patients. These results should be compared with studies conducted in 38 vaccinated patients; nonetheless, they argue for special effort to prevent SARS-CoV-2 infection in 39 patients with cancer. 40 Funding. No external funding was obtained. 41 Introduction 47 Since the emergence of SARS-CoV-2, many studies have been conducted on the outcomes of 48 in order to identify factors associated with a higher death rate and a more severe infection course. Some 49 groups of patients at increased risk of severe COVID-19, morbidity and mortality have been identified, 50 including elderly patients, and those with comorbidities, as hypertension, diabetes, chronic kidney 51 disease or COPD [1] . Cancer patients are also a high-risk group due to their compromised immune 52 systems and vulnerability to infection resulting from their disease and treatments [2] . 53 It is generally assumed that cancer patients are at higher risk for severe COVID-19 and death attributed 54 to . However, cancer encompasses a very heterogeneous group of diseases with a diverse 55 range of subtypes and stages. In addition, not all cancers are equal in terms of incidence, prognosis, and 56 treatment. This must be taken into account when the type of cancer is not specified [4] . For this reason, 57 although descriptions and analyses of risk factors, clinical courses, and mortality in cancer patients 58 infected with SARS-CoV-2 have been reported, a quantitative assessment of the effect of COVID-19 in 59 patients with cancer would be important to guide clinical decision-making. 60 We aimed at conducting a systematic review of the epidemiological features of the studies of in cancer patients conducted before the implementation of vaccination campaigns, and to provide a 62 quantitative estimate of the risk of cancer patients for severe infection course and COVID-19 mortality, 63 compared to uninfected cancer patients. We decided to restrict our review to studies of unvaccinated 64 patients because (i) they provide the clearest picture of the effect of SARS-CoV-2 infection on outcome

This systematic review was conducted according to the PRISMA statement [5] . We submitted the 71 protocol (available as Supplementary File 1) to the PROSPERO Registry. To carry out the systematic 72 review of the scientific literature, the following string was used for the PubMed database: 73 (neoplas* [TIAB] OR tumor* [TIAB] OR cancer* [TIAB] OR malignancy [TIAB] ) AND (2019 novel 74 coronavirus [TIAB] OR OR COVID19 [TIAB] OR OR 2019-nCoV [TIAB] ). 75 In order restrict the review to studies populations on unvaccinated cancer patients, we included papers 76 published in peer-reviewed journals up to December 31, 2020. We excluded abstracts and non-peer-77 reviewed reports, articles in languages other than English, and studies including children. We also 78 excluded reviews, meta-analysis and case reports, and studies with less than 50 patients or less than 10 79 events. Finally, we excluded studies in which diagnosis of SARS-Cov-2 infection was not made by PCR 80 testing. 81 The articles were independently reviewed and abstracted by 2 pairs of reviewers [GDF and MA; GV and 82 FT], on the basis of title, abstract and full text; the disagreement between the authors of the reviews 83 (6.1% of all studies) and was resolved through discussion with a fifth reviewer [PB] . 84 We selected the following outcomes: mortality, ICU admission, severity of COVID-19 symptoms, and 85 hospitalization: we combined these latter two outcomes because the definition of severity was 86 heterogeneous across studies and the number of available studies was low. We excluded from the 87 review studies addressing the impact of SARS-Cov-2 infection on prevention, diagnosis, and treatment of 88 cancer patients, as well as studies on the oncogenic effect of the virus, e.g., analyses of cancer-related 89 alterations. In addition, we carried out a back-search by inspecting the lists of references of articles 90 selected for the review. 91 Figure 1 shows the flowchart for selection of the studies. Details on the studies retained in each step of 92 the process are available from the authors. 93 We abstracted the following parameters from the articles retained for the review: country, sample size, 94 number of person affected by cancer and by SARS-Cov-2 infection, cancer type and comparison group 95 (patients without cancer or patients with a different type of cancer), outcome, and risk estimate 96 (relative risk or odds ratio [OR]) with 95% confidence interval (CI). If the risk estimate or the CI were not 97 reported in the publication, we calculated them from the raw data, if possible. We also performed a 98 quality assessment (QA) based on a modified version of CASP score [6] , that included 10 criteria. 99 100 Statistical analysis 101 We conducted random-effects [7] meta-analyses of the risk estimates for the combinations of cancers 102 and outcomes with more than five independent results. We also conducted stratified meta-analyses 103 according to geographic region, to explore potential sources of heterogeneity, that we quantified using 104 the I 2 test [8] . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 To evaluated results stability, we performed sensitivity analyses by quality score and repeated the meta-106 analysis after excluding one study at a time. We also conducted secondary analyses excluding studies 107 with results calculated on the basis of raw data. Furthermore, we considered the funnel plot and 108 performed the Egger's regression asymmetry test to assess publication bias [9] . 109 Finally we conducted a cumulative meta-analysis, based on date of publication of subsequent studies. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021.

We identified a total of 3488 publications from the literature search, and excluded three because they 117 were duplicates. We screened the titles and abstracts of 3485 articles: we excluded 3145 of them 118 because not relevant (Figure 1) , and retained 340 articles as potentially eligible. 119 After reviewing the full-texts, we excluded 303 articles because these did not meet the inclusion criteria, 120 and included the remaining 37 studies in the review: we finally included 35 of them in the quantitative 121 synthesis. 122 Among the 35 studies, 30 reported results for all cancers combined, and 8 for hematologic neoplasms 123 (three of these reported both sets of results). Results for other specific cancers were sparse, and we 124 could not conduct meta-analyses for them. Out of the 35 studies, 13 were from Europe, 11 from North 125 America (all from USA), and 11 from Asia (9 from China and two from Iran). Fifteen studies were 126 considered good quality (CASP score >9.5), nineteen studies were of moderate quality (9.5 ≥ CASP score 127 > 6), whereas one was considered inadequate (CASP score ≤ 6). 128 Tables 1 and 2 show the details of the studies included in the analysis. In the analysis by geographic region (Figure 2 ), the association between SARS-CoV-2 infection and 136 mortality in cancer patients was stronger, and less heterogeneous, in studies from Asia (OR 2.92; 95% CI 137 2.13-4.01, I 2 37.8%) than in studies from either Europe (OR 2.37; 95% CI 1.65-3.40; I 2 67.9%) or North 138 America (OR 1.97; 95% CI 1.31-2.97; I 2 95.9%). Too few studies were available on the other outcomes to 139 justify a meta-analysis stratified by region of origin. 140 The cumulative meta-analysis, based on date of publication of subsequent studies of mortality (all type 141 of cancer), showed a stronger association in the studies published before July 2020 than in studies 142 published later. In the sensitivity analysis based on QA, the pooled OR of mortality results of studies with acceptable 148 quality was not different from that of results of good-quality studies: OR 2.25 (95% CI 1.73-2.94) vs. OR 149 2.50 (95% CI 1.47-4.26). When we repeated the analysis after excluding one study at a time, we did not 150 identify a major effect of any single study; in particular, the exclusion of the only study that suggested a 151 negative association between SARS-CoV-2 infection and mortality [18] yielded a pooled OR of 2.41 (95% 152 CI 1.95-2,99, I 2 85.5%). The association with mortality was less pronounced in studies whose results 153 . CC-BY 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021. ; https://doi.org/10. 1101 /2021 were reported by the authors (OR 2.11; 95% CI 1.55-2.87) compared to studies whose results were 154 calculated by us (OR 2.66; 95% CI 1.97-3.60%), although the difference was not statistically significant. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint cancer, by assessing whether the infection in these patients has a more severe course than in a control 176 group affected by the infection but without cancer. On the other hand, it is important to identify the 177 effects that the pandemic itself has determined in patients with cancer, including reduced access to 178 treatment, delay in diagnosis for postponed screening, increased time between follow-up visits, and 179 change in treatment organization. Acquiring more severe infection could be due to both components. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint We were not able to derive pooled results for other specific cancers. Results for patients with 215 hematologic and solid neoplasms were compared in some individual studies. In particular, Desai et al. 216 [53] reported a higher mortality in the former group, but the comparison was not adjusted for age and 217 type of therapy.

Although our study provides the most precise measure to date of the effect of COVID-19 in cancer 220 patients, it suffers from some limitations. Many studies included in our analysis did not provide results 221 adjusted for important determinants such as sex, age, comorbidities, and therapy. As mentioned above, 222 we were not able to analyse specific cancers other than hematologic neoplasms, because results were 223 too sparse.

In conclusion our meta-analysis confirms, by giving a more precise and accurate estimation, evidence to 226 the hypothesis of an association between all type of cancer (and more specific hematologic neoplasm) 227 and a worst outcome on Mortality, ICU admission and Severity of Future studies will be able to better analyse this association for the different subtypes of cancer too. 229 Furthermore, they will eventually be able to evaluate whether the difference among vaccinated 230 population is reduced. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Availability of data and material: all the primary data are available from the first Author. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021. ; Articles excluded from quantitative analysis because of lack of risk measurement (n=2)

. CC-BY 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint (2020) Westblade et al. (2020) Dai et al. (2020) AUTHORS Wang et al. (2020) Shoumariyeh et al. (2020) Ganatra et al. (2020) Sun et al. (2020) Tian et al. (2020) Zhang et al. (2020) Atalla et al. (2020) Liang et al. (2020) Song et al. (2020) Bauer et al. (2020) Ganatra et al. (2020) Brar et al. (2020) 2 Overall (I-squared = 87.5%, p = 0.000) AUTHORS Gottlieb et al. (2020) Ygenoglou et al. (2020) Ygenoglou et al. (2020) Dai et al. (2020) Gottlieb et al. (2020) Ygenoglou et al. (2020) Dai et al. (2020) 2 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted October 23, 2021. ; https://doi.org/10. 1101 /2021 

FigS2 Quality Assestment by CASP_all type of cancer Outcome 1 (Mortality): Acceptable studies (CASP>6-9.5) vs Good studies (CASP>9

Overall (I-squared = 90.1%, p = 0.000)

Subtotal (I-squared = 84.0%, p = 0.000)

Subtotal (I-squared = 89