key: cord-0849520-xy9zxsa2
authors: Salvatore, M.; Hu, M. M.; Beesley, L. J.; Mondul, A.; Pearce, C. L.; Friese, C. R.; Fritsche, L. G.; Mukherjee, B.
title: COVID-19 outcomes by cancer status, type, treatment, and vaccination
date: 2022-04-21
journal: nan
DOI: 10.1101/2022.04.19.22274047
sha: 4df872275cbd9b0e2d0c840b26c8cf53e6918d68
doc_id: 849520
cord_uid: xy9zxsa2

Background: Observational studies have identified patients with cancer as a potential subgroup of individuals at elevated risk of severe SARS-CoV-2 (COVID-19) disease and mortality. Early studies showed an increased risk of COVID-19 mortality for cancer patients, but it is not well understood how this association varies by cancer site, cancer treatment, and vaccination status. Methods: Using electronic health record data from an academic medical center, we identified 259,893 individuals who were tested for or diagnosed with COVID-19 from March 10, 2020, to February 2, 2022. Of these, 41,218 tested positive for COVID-19 of whom 10,266 had a past or current cancer diagnosis. We conducted Firth-corrected, covariate-adjusted logistic regression to assess the association of cancer status, cancer type, and cancer treatment with four COVID-19 outcomes: hospitalization, intensive care unit (ICU) admission, mortality, and a composite "severe COVID-19" outcome which is the union of the first three outcomes. We examine the effect of the timing of cancer diagnosis and treatment relative to COVID diagnosis, and the effect of vaccination. Results: Cancer status was associated with higher rates of severe COVID-19 infection [OR (95% CI): 1.18 (1.08, 1.29)], hospitalization [OR (95% CI): 1.18 (1.06, 1.28)], and mortality [OR (95% CI): 1.22 (1.00, 1.48)]. These associations were driven by patients whose most recent initial cancer diagnosis was within the past three years. Chemotherapy receipt was positively associated with all four COVID-19 outcomes (e.g., severe COVID [OR (95% CI): 1.96 (1.73, 2.22)], while receipt of either radiation or surgery alone were not associated with worse COVID-19 outcomes. Among cancer types, hematologic malignancies [OR (95% CI): 1.62 (1.39, 1.88)] and lung cancer [OR (95% CI): 1.81 (1.34, 2.43)] were significantly associated with higher odds of hospitalization. Hematologic malignancies were associated with ICU admission [OR (95% CI): 1.49 (1.11, 1.97)] and mortality [OR (95% CI): 1.57 (1.15, 2.11)], while melanoma and breast cancer were not associated with worse COVID-19 outcomes. Vaccinations were found to reduce the frequency of occurrence for the four COVID-19 outcomes across cancer status but those with cancer continued to have elevated risk of severe COVID [cancer OR (95% CI) among those fully vaccinated: 1.69 (1.10, 2.62)] relative to those without cancer even among vaccinated. Conclusion: Our study provides insight to the relationship between cancer diagnosis, treatment, cancer type, vaccination, and COVID-19 outcomes. Our results indicate that it is plausible that specific diagnoses (e.g., hematologic malignancies, lung cancer) and treatments (e.g., chemotherapy) are associated with worse COVID-19 outcomes. Vaccines significantly reduce the risk of severe COVID-19 outcomes in individuals with cancer and those without, but cancer patients are still at higher risk of breakthrough infections and more severe COVID outcomes even after vaccination. These findings provide actionable insights for risk identification and targeted treatment and prevention strategies.

The EHR data were collected from Michigan Medicine patients who were tested or treated for COVID-19 between the dates of March 10, 2020, through February 2, 2022.

There were 351,843 individuals in this initial cohort. To distinguish between pre-existing diagnoses and diagnoses that may be related to underlying COVID-19, we restricted patient data to at least 14 days before the first COVID-19 positive test for those who tested positive, and the first COVID-19 test for those who tested negative (the "index test").

Diagnoses that are sex-specified and were discordant with the individual's EHR-recorded sex were removed (n = 4,124). Individuals who did not have a diagnosis (cancer or otherwise) prior to the 14-day threshold were removed (n = 40,266). The analysis is further restricted to those who were adults (> 18) at the index test (n = 47,558). Finally, two individuals were removed because of their EHR-recorded age was zero or negative resulting in the analytic tested cohort of 259,893 individuals ( Figure S1) .

After exclusions, the tested cohort (n = 259,893) represents a non-probabilistic sample due to the testing protocol at Michigan Medicine, which focused initially on symptomatic and high-risk patients in the early stages of the pandemic. Of those who tested positive for COVID-19 (n = 41,218), 10,266 (24.9%) individuals had a cancer diagnosis recorded in their EHR and 4,846 (11.8%) had an initial cancer diagnosis within the past three years.

We considered four outcomes: COVID-19-related (1) hospitalization, (2) ICU admission, and (3) mortality in addition to (4) a composite "severe COVID" outcome which is the union of (1)- (3) . Those who were considered COVID-19 positive either had a positive test or were diagnosed with COVID-19. COVID-19-related hospitalization was defined as a hospitalization or discharge that occurred within 14 days prior to through 30 days after a COVID-19 diagnosis. COVID-19-related ICU admission was defined as being admitted to the ICU within 14 days prior to through 30 days after a COVID-19 diagnosis.

The mortality outcome was all-cause and was defined as a death (including nonhospitalization deaths) that occurred within 14 days prior to through 60 days after a COVID-19 diagnosis (i.e., includes post-mortem COVID-19 testing). The severe COVID-19 outcome captured anyone who was identified by any of the three previous outcomes.

For identifying individuals with cancer, we aggregated International Classification of Diseases, Ninth and Tenth Revision (ICD-9 or ICD-10) codes into broader phenotype descriptions, called phecodes (developed by Denny et al. 36 ). Cancer was defined as ever having at least one phecode for cancer recorded in the patient EHRs (Table S1) . We also define five cancer types: melanoma, hematologic malignancy, breast cancer, prostate cancer, and lung cancer (Table S1) . Of those with cancer in the test-positive cohort, 6.4% (n = 654) were ever diagnosed with at least two of these five cancer types. We also constructed a variable corresponding to the time since the most recent initial cancer diagnosis: within the last three years, three to ten years ago, and ten or more years ago.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 7 We considered three types of cancer treatment: chemotherapy, radiation therapy, and surgery. Chemotherapy was defined as having at least one of the chemotherapyrelated Current Procedural Terminology (CPT) codes or ICD-9 or ICD-10 codes recorded in the EHR as listed in Table S2 . Radiation therapy and surgery were defined similarly.

These treatments could have taken place at any time prior to 14 days before the first positive COVID-19 test.

For the statistical analysis, we defined "radiation only" and "surgery only" variables, as radiation therapy without corresponding chemotherapy or surgery codes, and surgery without corresponding radiation or chemotherapy codes. A surgery was characterized as cancer-related only if a cancer diagnosis code was recorded during the same encounter.

In addition, we constructed a variable corresponding to the time since most recent chemotherapy treatment: within last year, one to three years ago, and more than three years ago. Figure S2 shows the distribution of cancer treatment patterns among COVID-19 positive individuals with cancer.

We extracted self-reported age, sex, race/ethnicity, smoking status (never/past/current), alcohol consumption (never/past/current), and body mass index (BMI) from the EHR. Socioeconomic status measures were defined by US census tract for the year 2010 based on the patient's residential address in the EHR and 5-year (2013-8 with income in the last 12 months below the poverty level, and (4) the proportion of the population (age 16 and older) unemployed. In our data, the NDI is recorded as an ordinal quartile variable using the raw mean values.

For non-cancer comorbid conditions, we used ICD-9 and ICD-10 codes from the EHR (including only those diagnoses greater than 14 days prior to first COVID-19 test or diagnosis) to construct binary disease indicators for respiratory conditions, circulatory conditions, type 2 diabetes, kidney disease, liver disease, and autoimmune disease (qualifying phecodes for each comorbidity listed in Table S3 ). The comorbidity score, which excludes cancer, was calculated as a sum of these indicators and ranges from 0-6 33 .

We constructed a vaccination status variable with four mutually exclusive categories: before vaccination, while partially vaccinated, while fully vaccinated (but not boosted), and after booster receipt. "Before vaccination" includes both those whose vaccination status is unvaccinated/unknown and those COVID-19 that were diagnosed prior to the availability of vaccines. Vaccination status was calculated based on the number of and manufacturer of COVID-19 vaccinations that took place at least 14 days prior to testing positive for COVID-19. Individuals who were identified as receiving a vaccine but for whom we did not have the time of the vaccination were categorized as missing. Additionally, individuals who received non-FDA-approved vaccines, including Astrazeneca, Novavax, Sinopharm, and Sinovac, were also categorized as missing.

Those with missing vaccination data were excluded from vaccination analyses. In our vaccination analyses, we adjusted for a 2020 indicator that was 1 when someone was . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 21, 2022.

9 COVID-19 positive in 2020 and 0 if they were COVID-19 positive after 2020. This was to account for the lack of access to vaccinations in 2020 and for potential differences in treatments and in effects of circulating variants.

For each of the four binary outcomes, we performed logistic regression with Firth bias correction to correct for separation issues 38 . We reported the unadjusted estimates for odds ratio (OR) as well as the estimates adjusting for three different sets of covariates (referred to as adjustment set 1, 2, and 3, respectively): where COVID is an indicator variable representing the following outcomes within the tested positive cohort comparing:

1. Those who experienced a severe COVID outcome (1) to those who did not (0) 2. Those who were hospitalized (1) to those who were not (0) 3. Those who were admitted to the ICU (1) to those who were not (0), and 4. Those who were deceased within 60 days of initial COVID-19 diagnosis (1) to those who were not (0), to assess the association between cancer status and COVID-19 outcomes.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 21, 2022. ; https://doi.org/10.1101/2022.04.19.22274047 doi: medRxiv preprint In the above model, cancer status refers to the cancer-related indicators of interest.

In addition to history of cancer, we also fit models for cancer diagnosis timing (most recent initial cancer diagnosis 0-3 years ago, 3-10 years ago, and 10 or more years ago), cancer treatment (where it represents chemotherapy, radiation only, or surgery only), chemotherapy treatment timing (most recent chemotherapy treatment less than 1 year ago, 1-3 years ago, and 3 or more years ago), and for cancer diagnosis (where is represents having a diagnosis of melanoma, a hematologic malignancy, breast cancer, prostate cancer, or lung cancer).

We also carried out interaction analyses by cancer status according to the model: 

We conducted two additional analyses: an analysis after adjusting for vaccination status and an analysis considering the recency of the cancer diagnosis. The goal of the vaccination analysis is to assess whether vaccination status is associated with lower odds of severe COVID-19 outcomes among individuals with and without a cancer diagnosis.

For this analysis, we fit a model of the form:

logit( ( = 1|vax status, cancer status, covariates)) = 0 + cancer status + vax status + cancer status * vax status + β 2020 (COVID positive test in 2020) + covariates where vaccination status (described in "Covariates" subsection above) is a categorical variable with the categories before vaccination (reference), partially vaccinated, fully vaccinated (but not boosted), and boosted. A schematic representing how individuals were classified with respect to vaccination status is presented in Figure 1B .

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 21, 2022. ;

In our analyses, we considered categorical variables regarding the timing of cancer diagnoses, types, and treatments. The construction of timing of cancer diagnosis (within past three years, three to ten years ago, and ten or more years ago) and chemotherapy (within past year, one to three years ago, and three or more years ago) are described above and their results reported in Table 2 and Table 3 , respectively. We constructed similar variables for cancer treatment considering the time of the cancer treatments relative to the positive COVID-19 test: within past three years and three or more years ago. We conducted the same analyses as presented in the "Statistical Analysis" subsection above using these variables as the independent variable (results presented in Section S2).

The analytic cohort of COVID-19 tested individuals (n = 259,893) was 41.7% (n = Table 1 .

In the test-positive cohort, 24.9% (n = 10,266) had a prior diagnosis of cancer.

Among all individuals who tested positive for COVID-19 in our data, patients with cancer . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted April 21, 2022. ; 13 comprised 40.7% (n = 1,205) of those with severe COVID, 39.5% (n = 1,047) of those who were hospitalized, 39.5% (n = 238) of those who were admitted to the ICU, and 51.8% (n = 270) of those who died. The number and proportion of individuals with and without cancer is presented in Figure 2 .

Among those who tested positive for COVID-19, having a cancer diagnosis was significantly associated with higher rates of severe COVID [OR (95% . Melanoma and breast cancer diagnoses did not show a significant difference in any of the outcomes in the fully adjusted model, though the results suggest potential for melanoma (e.g., hospitalization OR (95% CI): 1.18 (0.99, 1.41)).

Prostate cancer was negatively associated with hospitalization and ICU admission.

Results from our analyses of COVID-19 outcomes by cancer type are shown in Table 4 . is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ;

Supplementary Section 2. 60-day all-cause mortality was not considered as an outcome for interaction analyses due to sample size.

The vaccination results demonstrate that increased vaccination coveragefrom partially vaccinated ( Table 4 . However, across every stratum of vaccination status, individuals with cancer are at elevated risk for severe COVID (e.g., cancer OR (95% CI) for severe COVID among fully vaccinated individuals: 1.69 (1.10, 2.62); Figure 3 ).

In this study, we examined a cohort of 41,218 individuals who tested positive for is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ; https://doi.org/10.1101/2022.04.19.22274047 doi: medRxiv preprint individuals whose most recent initial cancer diagnosis was within the past three years since there were no associations found in those whose most recent initial cancer diagnosis was more than three years ago.

We also examined differences in risk by cancer treatment and by cancer type. Our study provides evidence that chemotherapy receipt is associated with worse COVID-19 outcomes, including hospitalization [OR (95% CI): 1.96 (1.72, 2.24)]. Moreover, this association exhibited attenuation as time since most recent chemotherapy treatment increased and remained significant when three or more years ago. While there are some smaller studies found no difference in outcomes for chemotherapy patients 16 , a recent paper by Chavez-MacGregor et al. 42 found that recent chemotherapy treatment is associated with mortality in COVID-19 patients. This finding is consistent with both chemotherapy's potential for immunosuppression 43 and our finding that hematologic malignancy diagnoses were associated with higher odds of hospitalization, for which myelosuppressive chemotherapy is a primary form of treatment 44 . We also found that chemotherapy receipt continued to be positively associated with higher rates of severe (Table S4 ). While we found these associations plausible, it is likely that they are overstated since chemotherapy could be a proxy for patients with high-stage cancer. Radiation-only was not found to be significantly associated with COVID-19 ICU admission or mortality, though our sensitivity analysis suggests recent radiation-only treatment may be associated with severe COVID, COVID-19 hospitalization, and ICU admission (Table S5) . Given this information, extra COVID-. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ;

17 19 protection measures, increased surveillance, and early and aggressive COVID-19 treatment may be warranted for patients who receive chemotherapy.

We identified that hematologic malignancies and lung cancer diagnoses were associated with higher odds of severe COVID and COVID-19 hospitalization.

Hematologic malignancies were additionally associated with higher likelihood of ICU admission and mortality (a result reported by Fu et al. 45 ). However, melanoma, breast cancer, and prostate cancer were not. In fact, we found that prostate cancer was associated with lower odds of hospitalization and ICU admission. We believe this finding is attributable to high rates of prostate cancer diagnosis among individuals with indolent disease who are otherwise healthy. Our cancer type analysis suggests that only hematologic malignancies are associated with COVID-19 mortality ( individuals with a cancer diagnosis, 47% (n = 4,846) had an initial cancer diagnosis in the three years prior to testing positive. We also conducted additional analyses by timing of most recent cancer diagnosis (within 3 years, 3-10 years, 10 or more years). Fourth, the data comes from a single site -specifically a large, academic healthcare system in . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. Michiganand may not be representative of the state or US population, limiting the generalizability of the results. Fifth, the outcomes are defined using time periods around the index test, which means the outcomes could be the result of something unrelated to COVID-19 (e.g., death due to end-stage cancer rather than COVID-19). We calculated the absolute rates of hospitalization, ICU admission, and fatality in COVID-19 test-positive and in matched and unmatched test-negative patients ( Table 6) . We see that the absolute rates of ICU admission and fatality were higher in the COVID-19 positive cohort than in the COVID-19 negative cohorts. Additionally, the cancer-no cancer outcome rate ratios are stable across COVID-19 positive and negative status, suggesting that there is not a synergistic effect between COVID-19 and cancer for these outcomes.

Our study contributes important information to the area of cancer and treatment in the time of COVID-19. Specifically, cancer status alone appears to be associated with higher rates of COVID-19-related hospitalization, ICU admission, and mortality. This association is driven be people with recent cancer diagnoses. The existence and strength of an association is different based on cancer diagnosis (e.g., hematologic malignancies were associated with worse COVID-19 outcomes while breast cancer was not) and treatment (e.g., chemotherapy was associated with worse COVID-19 outcomes while surgery only was not). Additionally, chemotherapy appeared to be associated with worse COVID-19 outcomes even after the exclusion of cancer patients with hematologic malignancies. Finally, we provide evidence that vaccination is effective in reducing severe COVID in cancer patients. Future research should consider post-acute sequalae of COVID-19 (PASC, or "long COVID") as an outcome and look more closely at the role cancer types, treatments, and COVID-19 vaccination play in COVID-19 outcomes.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ; 

Two weeks Three to ten years Ten or more years . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ; https://doi.org/10.1101/2022.04.19.22274047 doi: medRxiv preprint Figure 2 . COVID-19 outcomes by cancer status. The bars present raw proportion of individuals with the outcome ( = 1) overall (green), among those with cancer (orange), and among those without cancer (purple). The first panel ("Severe COVID") represents the proportion of individuals who were hospitalized, admitted to the ICU, or died because of COVID-19. The error bars represent the 95% confidence intervals. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. ; Figure 3 . Odds ratio for severe COVID-19 corresponding to cancer status by COVID-19 vaccination strata. "Before vaccination" includes those whose vaccination status is unvaccinated/unknown as well as those who were diagnosed with COVID-19 prior to the availability of COVID-19 vaccines. Additionally, an indicator variable indicating whether the COVID-19 diagnosis was made in 2020 (1) or not (0) was added as a covariate in the model to account for differences in strains. OR for severe COVID corresponding to cancer status stratified b y vaccination . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 21, 2022. 36,036 * Analysis was limited to females and excluded sex as a covariate ** Analysis was limited to males and excluded sex as a covariate Adjustment 1: Age, race/ethnicity, sex Adjustment 2: Adjustment 1 + Neighborhood Disadvantage Index (quartile) Adjustment 3: Adjustment 2 + comorbidity score Bolded point estimates represents statistical significance at the 95% confidence level. 35,940 Models were adjusted for age, race/ethnicity, sex, Disadvantage Index (quartile) and comorbidity score, and an indicator variable for whether COVID-19 was diagnosed in 2020 (1) or not (0). Fully vaccinated does not include individuals who received a booster. Bolded point estimates represents statistical significance at the 95% confidence level.

Epidemiology of COVID-19

Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19). World Health Organization; 2020. Accessed

Chief JBLS. 1st known case of coronavirus traced back to November in China. livescience.com. Accessed

WHO Director-General's opening remarks at the media briefing on COVID-19 -11

Microsoft Bing COVID-19 Tracker

Patients with Cancer Appear More Vulnerable to SARS-CoV-2: A Multicenter Study during the COVID-19 Outbreak

Cancer management during the COVID-19 pandemic: Choosing between the devil and the deep blue sea

Integrated Survival Estimates for Cancer Treatment Delay Among Adults With Cancer During the COVID-19 Pandemic

Clinical characteristics, outcomes, and risk factors for mortality in patients with cancer and COVID-19 in Hubei, China: a multicentre, retrospective, cohort study

Cancer and Risk of COVID-19 Through a General Community Survey

Chemotherapy and COVID-19 Outcomes in Patients With Cancer

COVID-19 outcomes of patients with gynecologic cancer in New York City

Determinants of COVID-19 disease severity in patients with cancer

Rate of hospital stays, critical care stays, and deaths in COVID-19 positives, unmatched test-negatives, and age-and sex-matched test-negatives. n (%) Overall No Hospital Stay Hospital Inpatient Care*,** Critical Care** Death Time window relative to index test (days) (-14

COVID-19 positive

COVID-19 negative (unmatched) (n total) 257,122 231,632 (90.1)

5) 4,945 (2.1) 1,923 (0.8) Cancer 51,508 45,074 (87.5) 6,434 (12.5) 1,350 (2.6) 866 (1.7) No cancer 185