Volume 127, Issue 2 p. 266-274
Original Article
Free Access

Multivariate mortality analyses in COVID-19: Comparing patients with cancer and patients without cancer in Louisiana

Michael J. Lunski MD

Corresponding Author

Michael J. Lunski MD

Ochsner Cancer Institute, New Orleans, Louisiana

Corresponding Author: Michael J. Lunski, MD, Gayle and Tom Benson Cancer Center, Ochsner Cancer Institute, 1514 Jefferson Hwy, New Orleans, LA 70121 ([email protected]).

Search for more papers by this author
Jeffrey Burton PhD

Jeffrey Burton PhD

Ochsner Medical Center, New Orleans, Louisiana

Search for more papers by this author
Karine Tawagi MD

Karine Tawagi MD

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
Diana Maslov MD

Diana Maslov MD

Ochsner Medical Center, New Orleans, Louisiana

Search for more papers by this author
Victoria Simenson MD

Victoria Simenson MD

Ochsner Medical Center, New Orleans, Louisiana

Search for more papers by this author
Daniel Barr MS-III

Daniel Barr MS-III

University of Queensland School of Medicine, Brisbane, Queensland, Australia

Search for more papers by this author
Helen Yuan MD

Helen Yuan MD

Ochsner Medical Center, New Orleans, Louisiana

Search for more papers by this author
Daniel Johnson MD

Daniel Johnson MD

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
Marc Matrana MD, MS

Marc Matrana MD, MS

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
John Cole MD

John Cole MD

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
Zoe Larned MD

Zoe Larned MD

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
Brian Moore MD

Brian Moore MD

Ochsner Cancer Institute, New Orleans, Louisiana

Search for more papers by this author
First published: 28 October 2020
Citations: 40

Michael J. Lunski thanks the wonderful staff and his colleagues at the Ochsner Cancer Institute and the Ochsner Medical Center for their help and encouragement with this study. He also thanks his loving wife, who supported him throughout this research study.

Abstract

Background

This is the largest and only multivariate study evaluating the difference in mortality from coronavirus disease 2019 (COVID-19) between patients with cancer and patients without cancer in the United States. The objective was to assess COVID-19 mortality rates in patients with cancer versus patients without cancer and uncover possible statistically significant characteristics contributing to mortality.

Methods

This retrospective study analyzed patients with cancer and patients without cancer who tested positive for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) from March 1 through April 30, 2020. This was a multicenter study in the state of Louisiana throughout the Ochsner Health System in both tertiary and nontertiary centers. Patients older than 18 years were eligible. Three hundred twelve patients with cancer were compared with 4833 patients without cancer.

Results

Mortality was found to be higher in the cancer group. Patients of advanced age with cancer had a significant increase in mortality (odds ratio [OR], 5.96; P < .001). Other significant risk factors for increased mortality were male sex (OR, 2.15), a history of chronic kidney disease (OR, 3.84), and obesity (OR, 1.30). In hospitalized patients with cancer, adverse vital signs on admission, decreased absolute lymphocyte counts, thrombocytopenia, elevated creatinine, lactic acidosis, and elevated procalcitonin all seemed to increase the risk of death. Among patients with cancer, active or progressive disease (P < .001) and recent therapy (OR, 2.34; 95% confidence interval, 1.08-5.08) were shown to increase mortality.

Conclusions

Patients with cancer have increased mortality in the setting of infection with SARS-CoV-2 in comparison with patients without cancer. Patients with cancer who are 65 years of age or older and those with certain comorbidities have the greatest risk of death. Recent cancer-directed therapy and disease status also seem to play roles in mortality.

Lay Summary

  • This is the largest study of patients with cancer versus patients without cancer to date and is the first multivariate analysis study comparing these 2 patient populations.
  • This study confirms the hypothesis that patients with cancer are at increased risk for mortality and that there are multiple characteristics posing the potential to risk-stratify these patients in the setting of a future outbreak.

Introduction

As the world currently faces the global pandemic of coronavirus disease 2019 (COVID-19), which is caused by the novel coronavirus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the impact on patients with cancer is an important area of investigation. At the time of this publication, the state of Louisiana with a population of 4.6 million has had one of the highest rates of COVID-19 per capita in the United States. To date, more than 400,000 tests have been performed in Louisiana, and they have revealed 99,000 confirmed cases. More than 3600 COVID-19–related deaths have been reported in the state, with the highest impact in the greater New Orleans area.1 Reports to date have described the impact of age, sex, race, socioeconomic status, and comorbidities related to COVID-19 morbidity and mortality.2

Cancer is the second most common cause of death globally. As of January 1, 2019, there were more than 19 million people living with cancer in the United States, and it is projected that in 2030, there will be more than 22 million.3 Cancer rates are particularly high in Louisiana because of a combination of environmental, lifestyle, and socioeconomic issues. In the year 2020, it is projected that there will be 26,000 new cases of cancer in the state of Louisiana and 9000 deaths from cancer. The incident rate for cancer in Louisiana is estimated at 481 cases per 100,000 people.4 This incidence makes Louisiana a unique population to study during this global pandemic.

A previous report out of China has shown a higher risk of severe events (a composite endpoint defined as the percentage of patients who were admitted to the intensive care unit and required invasive ventilation or who died) in patients with cancer infected with SARS-CoV-2 in comparison with patients without cancer (39% vs 8%; hazard ratio, 5.34; confidence interval [CI], 1.80-16.18; P = .0026); however, the sample size of patients with cancer was small, with only 18 patients having a reported history of cancer.5 Multiple studies have described prognostic factors in cohorts of patients with cancer only.6 More recently, a meta-analysis from the United States, Spain, and Canada again showed high 30-day all-cause mortality in patients with cancer, regardless of whether they had active cancer, were on anticancer treatment, or both however, they were not compared with a noncancer cohort of patients.7 To date, there has not been a large study in the United States reviewing patients with cancer who are infected with SARS-CoV-2 versus patients without cancer and their characteristics.

The objectives of this study were to 1) compare the mortality of patients with cancer and patients without cancer in the state of Louisiana who were cared for in a single health care system and were infected with SARS-CoV-2 and 2) to identify possible risk factors for mortality in hopes of stratifying at-risk patients with current concerns for future outbreaks.

Materials and Methods

Patient Selection and Data Retrieval

This was a 2-month retrospective, observational analysis of patients with cancer and patients without cancer who were infected with SARS-CoV-2 between March 1, 2020, and April 30, 2020, throughout the Ochsner Health System in the state of Louisiana. It included 36 different hospitals and clinics across the state, which ranged from tertiary care centers to primary care offices. Patients who were tested and had completed results for SARS-CoV-2 by a rapid nasopharyngeal polymerase chain reaction (PCR) test were included. Reported, unconfirmed, or suspected cases were not included. Antibody testing was not reviewed or included. International Classification of Diseases, Tenth Revision codes were used to identify patients with active cancer or a history of cancer. Throughout the Ochsner system, 2801 patients with cancer had been tested between the study dates. These 2801 charts were individually reviewed to confirm a cancer history. Among the tested patients, there were 327 positives. The 327 charts then underwent complete chart review; 312 patients had sufficient data documented to be included in the demographic study. To meet the inclusion criteria for the noncancer reference group, patients had to be adults with 1) a positive COVID-19 PCR test ordered between March 1 and April 30, 2020, 2) at least 1 encounter in the Ochsner system between January 1, 2019, and the date of their COVID-19 test, 3) no cancer diagnosis recorded between January 1, 2019, and the date of their COVID-19 test, and 4) a known mortality status (alive or deceased). All data for the noncancer group were abstracted from Ochsner's electronic health record system (Epic). The institutional review board of Ochsner Medical Center approved this study.

A predesigned form was used for data collection. The collected data included age, race, sex, presence of comorbidities (diabetes, hypertension and chronic obstructive pulmonary disease [COPD], chronic kidney disease, coronary artery disease, obesity, and cirrhosis), and data on hospitalization and death. Diabetes was defined as having an A1c concentration > 6.5% or being on antihyperglycemic medications. COPD was defined by an abnormal pulmonary functioning test or bronchodilator medication use. Chronic kidney disease was defined as a serum creatinine level > 1.5 mg/dL. Also, any of the aforementioned diagnoses were accounted for if the patient had documentation of the disease in his or her previous medical history. Active therapy was defined as any form of chemotherapy, immunotherapy (42 days), targeted therapy, or a combination within 30 days of a positive test. Recent therapy was defined as any therapy within 30 days of a positive test. The Eastern Cooperative Oncology Group (ECOG) status and tumor staging were taken from the most recent progress notes available.

Statistical Analysis

To meet the inclusion criteria for the noncancer reference group, patients had to be adults with 1) a positive COVID-19 PCR test ordered between March 1 and April 30, 2020, 2) at least 1 encounter in the Ochsner system between January 1, 2019, and the date of their COVID-19 test, 3) no cancer diagnosis recorded between January 1, 2019, and the date of their COVID-19 test, and 4) a known mortality status (alive or deceased). All data for the noncancer group were abstracted from Ochsner's electronic health record system (Epic). Patients in the cancer group were identified by the principal investigator, and data were abstracted directly from patient charts.

The presented data compare patient characteristics by cancer status. The characteristics of all patients who tested positive for COVID-19 are summarized in Table 1 (n = 5145). The characteristics of the subset of hospitalized patients are shown in Table 2 (n = 1617), with patients missing all admission laboratory tests or vitals excluded. All measures are presented as frequencies and percentages and were compared between cancer groups by chi-square tests for association. Threshold values used to dichotomize vital signs and laboratory tests shown in Table 2 were predetermined according to clinical guidelines.

TABLE 1. Comparison of the Characteristics of COVID-19+ Patients by Cancer Status (n = 5145)
Variable Patients With Cancer (n = 312) Patients Without Cancer (n = 4833) P
Age ≥ 65 y, No. (%) 167 (53.5) 1169 (24.2) <.001
Sex, No. (%) .017
Female 170 (54.5) 2962 (61.3)
Male 142 (45.5) 1871 (38.7)
Race, No. (%) .001
Black/AA 197 (63.1) 2956 (61.2)
White 108 (34.6) 1499 (31.0)
Hispanic 7 (2.2) 142 (2.9)
Other/unknown 0 (0.0) 236 (4.9)
Comorbid conditions, No. (%)
Cirrhosis 10 (3.2) 20 (0.4) <.001
Chronic kidney disease 80 (25.6) 902 (18.7) .002
COPD 35 (11.2) 334 (6.9) .004
Coronary artery disease 34 (10.9) 192 (4.0) <.001
Diabetes 116 (37.2) 872 (18.0) <.001
Hypertension 246 (78.9) 1709 (35.4) <.001
Obesity 143 (45.8) 2339 (48.4) .390
Clinical outcomes, No. (%)
Hospitalized 166 (53.2) 1488 (30.8) <.001
Died 66 (21.2) 418 (8.7) <.001
  • Abbreviations: AA, African American; COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019.
TABLE 2. Characteristics of Hospitalized COVID-19+ Patients by Cancer Status (n = 1617)
Variable Patients With Cancer (n = 157) Patients Without Cancer (n = 1460) P
Age ≥ 65 y, No. (%) 107 (68.2) 675 (46.2) <.001
Sex, No. (%) .478
Female 75 (47.8) 741 (50.7)
Male 82 (52.2) 719 (49.3)
Race, No. (%) .003
Black/AA 100 (63.7) 987 (67.6)
White 55 (35.0) 368 (25.2)
Hispanic 2 (1.3) 39 (2.7)
Other/unknown 0 (0.0) 68 (4.5)
Admission vitals, No. (%)
Systolic blood pressure < 100 mm Hg 77/150 (51.3) 726 (49.7) .708
Heart rate > 120 bpm 40/150 (26.7) 416 (28.5) .636
Temperature ≥ 100.4 °F/38 °C 98/150 (65.3) 915 (62.7) .520
Spo2 < 90% 87/150 (58.0) 920/1457 (63.1) .215
Admission laboratory tests, No. (%)
White blood cell count < 4 K/µL 42 (26.8) 255 (17.5) .004
Absolute neutrophil count < 1.5 K/µL 15/147 (10.2) 34/1440 (2.4) <.001
Absolute lymphocyte count < 1 K/µL 93/145 (64.1) 747/1422 (52.5) .008
Hemoglobin < 10 g/dL 49 (31.2) 317 (21.7) .007
Platelet count < 150 K/µL 54 (34.4) 393 (27.3) .058
Creatinine > 1.5 mg/dL 66 (42.0) 571 (39.1) .476
Venous lactate > 2.2 mmol/L 24/144 (16.7) 215/1230 (17.5) .808
Troponin I ≥ 0.06 ng/mL 35/135 (25.9) 306/1195 (25.6) .936
Brain-type natriuretic peptide > 100 pg/mL 36/112 (32.1) 290/1036 (28.0) .355
Procalcitonin > 0.25 ng/mL 57/134 (42.5) 450/1193 (37.7) .277
CRP > 8.2 ng/mL 126/138 (91.3) 1134/1241 (91.4) .977
Ferritin > 300 ng/mL 82/110 (74.6) 763/994 (76.8) .603
d-Dimer > 3 mg/L 24/74 (32.4) 195/619 (31.5) .871
Clinical outcomes, No. (%)
Intubated 39 (24.8) 404 (27.7) .450
Died 56 (35.7) 372 (25.5) .006
  • Abbreviation: AA, African American; COVID-19, coronavirus disease 2019; CRP, C-reactive protein; Spo2, oxygen saturation as measured by pulse oximetry.
  • Some vitals and laboratory tests were not recorded or were not available for all patients; these measures are shown in the table with the appropriate denominator.

Mortality was assessed among COVID-19+ patients and among the hospitalized subgroup. Unadjusted and covariate-adjusted multivariate logistic models were used to investigate the raw, unadjusted, and covariate-adjusted associations between cancer (yes vs no) and death. Only non-Hispanic Black and non-Hispanic White patients were included because of the small sample sizes of other race groups. All covariates included in the multivariate models were chosen for their clinical relevance and potential association with mortality. For the body mass index and all admission laboratory tests and vitals with missingness less than 25%, missing data were imputed via multiple imputation by fully conditional specification. One hundred full data sets were imputed. The multivariate mortality model for all COVID-19+ patients incorporated a main effect for cancer status and covariates for demographics (age, sex, and race) and comorbid conditions (chronic kidney disease, COPD, coronary artery disease, diabetes, hypertension, and obesity). Cirrhosis was excluded because of its low observed prevalence.

The multivariate mortality model for hospitalized COVID-19+ patients incorporated a main effect for cancer status and covariates for demographics (age, sex, and race), admission vitals (blood pressure, heart rate, and oxygen saturation as measured by pulse oximetry), and admission laboratory tests (absolute neutrophil count, absolute lymphocyte count, hemoglobin, platelet count, creatinine, venous lactate, troponin I, and procalcitonin). Admission laboratory tests and vital signs were not considered if 1) more than 25% of the data were missing or 2) no association with mortality was observed in bivariate analyses (results not shown). Mortality results are presented in Tables 3 and 4 as odds ratios (ORs). The ORs for cancer versus noncancer are presented as unadjusted and covariate-adjusted estimates. Adjusted ORs for all covariates included in the multivariate model are also presented to show the associations of each with mortality. All estimates are accompanied by 95% CIs.

TABLE 3. Unadjusted and Covariate-Adjusted Odds Ratios for Death Among COVID-19+ Patients (n = 4760)
Variable Odds Ratio (95% CI)
Unadjusted Adjusted
Cancer (yes vs no) 2.87 (2.14-3.85) 2.03 (1.44-2.87)
Age ≥ 65 y (yes vs no) 5.96 (4.71-7.54)
Sex (male vs female) 2.15 (1.73-2.67)
Race (Black vs White) 0.97 (0.77-1.23)
Chronic kidney disease (yes vs no) 3.84 (2.98-4.93)
COPD (yes vs no) 1.34 (0.96-1.88)
Coronary artery disease (yes vs no) 1.43 (0.99-2.08)
Diabetes (yes vs no) 0.92 (0.70-1.21)
Hypertension (yes vs no) 0.60 (0.47-0.78)
Obesity (yes vs no) 1.30 (1.03-1.63)
  • Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019.
  • The analysis does not include Hispanic or other/unknown patients (385 of 5145).
TABLE 4. Unadjusted and Covariate-Adjusted Odds Ratios for Death Among Hospitalized COVID-19+ Patients (n = 1510)
Variable Odds Ratio (95% CI)
Unadjusted Adjusted
Cancer (yes vs no) 1.62 (1.14-2.31) 1.36 (0.89-2.08)
Age ≥ 65 y (yes vs no) 2.85 (2.14-3.79)
Sex (male vs female) 1.24 (0.93-1.65)
Race (Black vs White) 0.72 (0.53-0.98)
Systolic blood pressure < 100 mm Hg 1.72 (1.29-2.29)
Heart rate > 120 bpm 1.65 (1.22-2.22)
Spo2 < 90% (yes vs no) 3.16 (2.28-4.38)
Absolute neutrophil count < 1.5 K/µL 0.98 (0.41-2.29)
Absolute lymphocyte count < 1 K/µL 1.71 (1.28-2.29)
Hemoglobin < 10 g/dL 1.12 (0.81-1.54)
Platelet count < 150 K/µL 1.70 (1.28-2.27)
Creatinine > 1.5 mg/dL 1.80 (1.33-2.45)
Venous lactate > 2.2 mmol/L 1.62 (1.13-2.30)
Troponin I ≥ 0.06 ng/mL 1.31 (0.95-1.82)
Procalcitonin > 0.25 ng/mL 1.59 (1.14-2.20)
  • Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; Spo2, oxygen saturation as measured by pulse oximetry.

Secondary analyses were performed to investigate associations between mortality and patient and clinical characteristics of COVID-19+ patients with cancer. The characteristics of all patients with cancer are presented in Table 5 (n = 312). Cancer stage is presented only for patients with solid tumor (n = 255). Finally, the prevalence of cancer types is summarized in the supporting information (Supporting Tables 1 and 2). Chi-square tests and Fisher exact tests were used to evaluate associations between patient death and categorical measures. Two-sample t tests and Wilcoxon rank sum tests were used to assess continuous measures. To further investigate the association between mortality and active therapy, a multivariate logistic regression model was constructed. Only non-Hispanic Black and non-Hispanic White patients with cancer were included because of the small sample sizes of patients of other races (n = 7). The model included an indicator of active therapy status along with covariates for demographics (age ≥ 65 years, race, and sex), medical history (smoking status, chronic kidney disease, COPD, diabetes, hypertension, and obesity), and the type of diagnosis. Both unadjusted and covariate-adjusted ORs for mortality are presented in Table 6 along with 95% CIs. All analyses were performed with SAS/STAT software from the SAS System for Windows (version 9.4; SAS Institute, Inc, Cary, North Carolina).

TABLE 5. Characteristics of Patients With Cancer by Patient Status: Alive or Deceased (n = 312)
Variable Deceased P a
No (n = 246) Yes (n = 66)
Age ≥ 65 y, No. (%) 116 (47.2) 51 (77.3) <.001
Sex, No. (%) .097
Female 140 (56.9) 30 (45.5)
Male 106 (43.1) 36 (54.6)
Race, No. (%) .616
African American 158 (64.2) 39 (59.1)
White 82 (33.3) 26 (39.4)
Hispanic 6 (2.4) 1 (1.5)
History of smoking, No. (%) <.001
No 154 (62.6) 19 (28.8)
Yes 92 (37.4) 47 (71.2)
Diagnosis type, No. (%) .033
Hematologic 42 (17.1) 19 (28.8)
Oncologic 204 (82.9) 47 (71.2)
Disease status, No. (%) <.001
Active/progressive disease 30 (12.2) 25 (37.9)
Posttreatment/maintenance 175 (71.1) 31 (46.9)
>5 y from therapy/NED 41 (16.7) 10 (15.2)
Time since diagnosis, No. (%) .079
0-1 y 43 (17.5) 21 (31.8)
1-2 y 44 (17.9) 16 (24.2)
2-3 y 31 (12.6) 5 (7.6)
3-4 y 26 (10.6) 3 (4.55)
4-5 y 21 (8.5) 4 (6.1)
5-10 y 59 (24) 14 (21.2)
10-15 y 22 (8.9) 3 (4.55)
Active treatment, No. (%)b .006
No 210 (85.4) 46 (69.7)
Yes 36 (14.6) 20 (30.3)
Recent therapy, No. (%) .033
None 159 (64.6) 39 (59.1)
Surgery 5 (2) 0 (0)
Radiation 2 (0.8) 0 (0)
Hormonal 44 (17.9) 7 (10.6)
Chemotherapy 12 (4.9) 7 (10.6)
Immunotherapy 4 (1.6) 0 (0)
Targeted 9 (3.7) 4 (6.1)
Combination 11 (4.5) 9 (13.6)
ECOG, No. (%) <.001
0 54 (22.0) 2 (3.0)
1 94 (38.2) 14 (21.2)
2 57 (23.2) 21 (31.8)
3 32 (13.0) 19 (28.8)
4 9 (3.7) 10 (15.2)
Stage, No. (%) .054
S1 95 (45.9) 15 (31.3)
S2 53 (25.6) 10 (20.8)
S3 26 (12.6) 8 (16.7)
S4 33 (15.9) 15 (31.3)
Admitted to hospital, No. (%) 105 (42.7) 61 (92.4) <.001
Hospital LOS, median (IQR)c 6 (3-15) 7 (4-12) .925
Intubated, No. (%) 13/105 (12.4) 30/61 (49.2) <.001
Admitted to ICU, No. (%) 17/105 (16.2) 31/61 (50.8) <.001
Comorbidities, No. (%)
Diabetes mellitus 85 (34.6) 31 (47.0) .064
Hypertension 189 (76.8) 57 (86.4) .092
COPD 23 (9.3) 12 (18.2) .044
Cirrhosis 6 (2.4) 4 (6.1) .228
CKD 54 (22.0) 26 (39.4) .004
Obesity (BMI ≥ 30 kg/m2) 116 (47.2) 27 (40.9) .366
  • Abbreviations: BMI, body mass index; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; ECOG, Eastern Cooperative Oncology Group; ICU, intensive care unit; IQR, interquartile range; LOS, length of stay; NED, no evidence of disease.
  • a Categorical measures were compared via chi-square tests for association or Fisher exact tests where appropriate; continuous measures were compared via t tests (normally distributed) or Wilcoxon rank sum tests (nonnormally distributed).
  • b Does not include hormonal therapy, recent surgery, or radiation.
  • c LOS was compared only for patients admitted to the hospital (105 alive and 61 deceased) and was measured in days.
TABLE 6. Unadjusted and Covariate-Adjusted Odds Ratios for Death Among COVID-19+ Patients With Cancer (n = 305)
Odds Ratio (95% CI)
Model 1 Model 2 Model 3 Model 4 Model 5
Active therapy (yes vs no) 2.27 (1.19-4.31) 2.31 (1.16-4.61) 3.01 (1.45-6.26) 1.80 (0.86-3.75) 2.34 (1.08-5.08)
Race (White vs Black) 1.29 (0.69-2.43) 1.12 (0.59-2.16) 1.29 (0.68-2.44) 1.12 (0.58-2.16)
Sex (female vs male) 0.95 (0.51-1.76) 0.97 (0.52-1.83) 1.01 (0.54-1.89) 1.04 (0.55-1.97)
Smoking history (yes vs no) 3.75 (1.94-7.22) 3.56 (1.82-6.95) 4.05 (2.07-7.89) 3.86 (1.95-7.65)
CKD (yes vs no) 1.91 (0.99-3.67) 1.44 (0.73-2.84) 1.88 (0.97-3.65) 1.42 (0.71-2.82)
COPD (yes vs no) 1.08 (0.46-2.52) 0.95 (0.40-2.25) 1.14 (0.49-2.67) 1.00 (0.42-2.38)
Diabetes (yes vs no) 1.36 (0.73-2.51) 1.45 (0.77-2.72) 1.42 (0.76-2.65) 1.52 (0.80-2.88)
Hypertension (yes vs no) 1.24 (0.52-2.97) 1.13 (0.46-2.76) 1.25 (0.52-3.04) 1.13 (0.46-2.78)
Obesity (yes vs no) 0.97 (0.52-1.79) 1.14 (0.61-2.15) 0.93 (0.50-1.73) 1.09 (0.57-2.06)
Age ≥ 65 y (yes vs no) 3.27 (1.58-6.75) 3.31 (1.59-6.88)
Diagnosis type (hematologic vs oncologic) 2.17 (1.03-4.60) 2.25 (1.03-4.88)
  • Abbreviations: CI, confidence interval; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019.
  • Hispanic patients (n = 7) were not included in the analysis.

Results

Mortality Is Significantly Higher in Patients With Cancer and COVID-19

A direct comparison of patients with cancer and patients without cancer who were infected with SARS-CoV-2 showed that mortality was significantly higher among patients with cancer (P < .001). The overall unadjusted mortality among patients with cancer was found to have an OR of 2.87 (95% CI, 2.14-3.85). This was also confirmed in the multivariate analysis (OR, 2.03; 95% CI, 1.44-2.87).

Age, sex, and race seemed to affect mortality in a direct comparison of the 2 population groups. After multivariate analysis, we found that mortality in patients with cancer who are 65 years of age or older had an OR of 5.96 (95% CI, 4.71-7.54). Males with cancer (OR, 2.15; 95% CI, 1.73-2.67) also had significantly increased mortality. Race (OR, 0.97; 95% CI, 0.77-1.23) did not appear to play a role.

Comorbid conditions, including cirrhosis, chronic kidney disease, COPD, coronary artery disease, diabetes, and hypertension, all appeared to be risk factors during a direct comparison. However, in the multivariate analysis, chronic kidney disease (OR, 3.84; 95% CI, 2.98-4.93) and obesity (OR, 1.3; 95% CI, 1.03-1.63) seemed to be significant traits increasing mortality. Having a history of hypertension seemed to decrease the risk of mortality (OR, 0.60; 95% CI, 0.47-0.78; Table 3).

Hospitalized Patients With Cancer Have Similar Mortality Risk Once Admitted

During a direct comparison, patients with cancer had a statistically significant increase in mortality, and this was also seen in the unadjusted OR. However, upon further analysis with the aforementioned factors (Table 4), this risk did not seem to ultimately hold true.

Once again, age and race appeared to be significant risk factors in a head-to-head comparison between patients with cancer and patients without cancer. Multivariate studies showed that advanced age was associated with increased mortality. Non-Hispanic Blacks appeared to have a decreased mortality risk among hospitalized patients.

Vital signs upon admission did not seem to play a role, but after the multivariate analysis, abnormal vital signs during the initial presentation to the hospital, including hypotension, tachycardia, and hypoxia, all seemed to contribute to mortality among patients with cancer.

Leukopenia, neutropenia, lymphopenia, and anemia on the initial presentation to the hospital appeared to be risk factors in a direct comparison of patients with cancer and patients without cancer. When the aforementioned values were accounted for (Table 4), lymphopenia, thrombocytopenia, elevated creatinine, lactic acidosis, and elevated procalcitonin were all found to be risk factors for mortality in patients with cancer who were admitted.

Active Treatment Increases Mortality Risk in Patients With Cancer

Cancer-directed therapy was shown to increase the mortality of patients with cancer (P = .006) during the initial comparison. When corrections were made for confounders in the multivariate analysis, active therapy still appeared to be a significant risk factor for mortality (OR, 2.34; 95% CI, 1.08-5.08; Table 6). Other characteristics that had a negative impact on survival included a history of smoking, advanced age, and a diagnosis of a hematologic malignancy.

Diagnosis and Disease Status Play Roles

Patients with a diagnosis of a hematologic malignancy were shown to have increased mortality in comparison with patients with an oncologic malignancy (31.1% vs 18.7%; P < .001). The most common cancer types among survivors were breast and genitourinary cancers (Supporting Table 1); the most common cancer types among deceased patients were hematologic malignancies and genitourinary cancers (Supporting Table 2).

The patient's disease stage and cancer status played significant roles. A total of 20 of 55 patients (36.4%) died with active or progressive disease at the time of infection, whereas 31 of 175 (17.7%) died during posttreatment follow-up or on maintenance therapy. In addition, 10 of 51 patients (19.6%) who were further than 5 years from treatment or did not have any evidence of disease died.

When we compared the aforementioned groups (Table 5), the time since diagnosis did not appear to be a statistically significant attribute. However, when we evaluated patients with a recent diagnosis (<1 year ago) and patients with a distant history of cancer (10-15 years ago), those with a recent diagnosis were more likely to die than those with a distant history of cancer (32.8% vs 12%).

Characteristics of the Cancer Patient Population Matter

As documented previously and shown again in our cancer patient population, advanced age was shown to increase mortality. Males and females were evenly accounted for, and sex did not seem to play a role in mortality. Although the majority of the patients studied in our cancer population were non-Hispanic Black (63.1%), mortality was not statistically affected by race.

A history of smoking was a significant negative factor: 33.8% of the patients with a history of smoking died, whereas 10.9% of the patients without a history of smoking died.

The ECOG status was associated with mortality. Patients with an ECOG status of 0 had a 3.6% death rate, those with an ECOG status of 1 had a 13% death rate, those with an ECOG status of 2 had a 27.6% death rate, those with an ECOG status of 3 had a 37.3% death rate, and those with an ECOG status of 4 had a 52.6% death rate.

Discussion

The mortality rate of COVID-19 is currently estimated at 6.3% worldwide, 5.9% in the United States, and 7.1% in Louisiana.8 These numbers will continue to be adjusted because there is likely a component of false-negatives and asymptomatic patients not being tested. These will likely drive the mortality rate down at some point in the future. For this study, only test-confirmed positives were included. In the noncancer group, the mortality rate was 8.7% throughout the Ochsner Health System. This appears to be slightly elevated in comparison with the government data reported previously. In our cohort of patients with cancer infected with SARS-CoV-2, we found the mortality rate to be 21.2%. This also appears to echo data from China, although that study had a much smaller sample size of patients with cancer.5 When mortality was adjusted in the multivariate analysis, the OR for mortality in patients with cancer and a COVID-19 infection was 2.03 (95% CI, 1.44-2.87; P = .05). Although overall mortality seemed to be significantly increased, once admitted, patients with cancer did not seem to have a significantly increased mortality risk (OR, 1.36; 95% CI, 0.89-2.08). This finding is most likely secondary to multiple factors. Our hypothesis would be that once patients require admission, the disease course becomes similar in these 2 patient populations. Indicators such as vital signs and laboratory results would potentially be more important. As reported many times in this article, age has a direct correlation with mortality, and this is likely more important than just a history of cancer when a patient is admitted with a COVID-19 infection. Clearly, a combination of cancer and advanced age certainly shows a marked increase in mortality, and this group appears to be the most at-risk population during this pandemic. Although we have seen previous reports that patients with cancer, among themselves, have a higher rate of mortality,9 this is the first large study to date to show increased mortality in patients with cancer in comparison with patients without cancer.

In our study, there were more COVID-19+ females than males among both patients with cancer and patients without cancer (54.5% and 61.3%, respectively), although males seemed to be more at risk for overall mortality with an OR of 2.15 (95% CI, 1.73-2.67). This is consistent with previous studies showing that men with a COVID-19 infection are at higher risk for worse outcomes and death independently of age10; however, males that were admitted to the hospital did not appear to be more at risk for mortality (OR, 1.24; 95% CI, 0.93-1.65). These data are hypothesis generating, and a further look at sex disparities with COVID-19 mortality needs further dedicated studies.

Black patients represented the largest racial demographic in our study. This is likely due to the population demographics in the state of Louisiana. The increased percentage of this population testing positive did not correlate with a significant risk of mortality overall, and among hospitalized patients, it appeared to confer a decreased risk of mortality in comparison with non-Hispanic Whites (OR, 0.72; 95% CI, 0.53-0.98). These data are congruent with previous data from our health care system, where among 3626 patients who tested positive for COVID-19, 70.4% were Black.2 Among our patients with cancer, race did not seem to be an underlying risk factor for mortality; this is similar to previous reports around the world.7

Our large, single health system spans the entire state of Louisiana. There are clinics and hospitals in rural, suburban, and urban areas. This provided our study with a myriad of different patient backgrounds, including the insured and uninsured, and allowed us to study a diverse patient population. Although there has been much research published regarding characteristics among patients with cancer at risk, our patient population here in Louisiana has unique demographics. Previous research failed to correlate a negative relationship between recent cancer-directed therapy and an increase in mortality.7 This is the first study to show that there should be concern about this at-risk patient population. After adjusting for comorbidities and age, we found that the OR for mortality in patients receiving active treatment increased. When we accounted for all comorbidities, including age and type of diagnosis, those undergoing active treatment continued to have a small but significant increase in mortality. There was not a difference in the type of recent treatment, although none of the patients who recently received immunotherapy died. There were only 4 such patients, and not much can be taken from this sample, although with the inflammatory effects of this infection, further analysis is warranted to evaluate whether this could be protective.

Disease status also appeared to play a role in mortality. Patients with active or progressive disease had higher mortality than patients with a history of distant disease (>5 years from therapy/NED). This is not surprising because most patients with active or progressive disease are on active therapy and/or have a poor baseline functional status. The time since diagnosis did not appear to be a statistically significant variable, although there was a higher percentage of mortality associated with a diagnosis within 2 years of the infection. A direct comparison of a cancer diagnosis within the 2 years before the infection and a cancer diagnosis more than 2 years before the infection was significant for increased mortality (87 alive/37 dead vs 159 alive/29 dead; P = .004). Further multivariate analyses may be warranted to confirm these findings.

One of the goals of this study was to identify patient characteristics that could risk-stratify patients on admission to the hospital. Admission vitals, including hypotension, tachycardia, and hypoxia, are all significant risk factors for mortality as expected. We also confirmed that previously known prognostic laboratory markers associated with higher mortality, including lymphopenia, thrombocytopenia, elevated serum creatinine, lactic acidosis, and elevated procalcitonin, all seemed to be significant risk factors for hospitalized patients with cancer and COVID-19 in the multivariate analysis.11-13 A more recent report from New York showed in multivariate analyses some of the risks that may be associated with mortality in patients with cancer, although there was not a comparison group.14 This is the first time that they are shown and confirmed in comparison with a noncancer patient population.

In conclusion, this study is the first large comparison of patients with cancer and patients without cancer via statistical analysis in a US-only population. Although patient-derived data from this global pandemic are ever changing, it appears to be clear that the cancer patient population is at increased risk in comparison with the noncancer population; this is especially true for elderly patients with cancer and those with chronic comorbidities. This is also the first multivariate analysis of patients with cancer that exhibits an increased risk of mortality for patients receiving recent cancer-directed therapy during infection. Once admitted to the hospital, patients with cancer and patients without cancer appear to have similar risks of death, and this is likely attributable to the overall severity of this infection. Our data suggest that certain admission vital signs, laboratory values, and patient characteristics could help to determine patients with cancer at the highest risk for fatal outcomes from COVID-19. This study highlights that not all patient populations are the same and sheds light on the challenges of caring for patients with cancer and COVID-19.

Patients are negatively affected by delaying cancer treatments and, unfortunately, do not have the flexibility to postpone treatment to potentially decrease the risk of a COVID-19 infection. Because we now have data showing that patients on active therapy are at increased risk here in the United States, extra efforts to protect these patients must be made. Guidelines should be established by individual cancer centers for patients undergoing cancer-directed therapy as well as patients with certain nonmodifiable risk factors to ensure patient safety while still providing optimal cancer care. This will be a challenge for us going forward, and hopefully, some of these poor prognostic factors that have been uncovered will allow for risk stratification and improvements in patient care in the setting of possible future outbreaks.

Funding Support

No specific funding was disclosed.

Conflict of Interest Disclosures

Daniel Johnson reports compensation from Bristol-Myers Squibb for work with a speakers' bureau and meeting honoraria from BrightPath Pharmaceutical Company. Marc Matrana reports personal fees from Strata Oncology, AstraZeneca, DisperSol, Merck, Bristol-Myers Squibb, Astellas, Eisai, Janssen, and Seattle Genetics outside the submitted work. The other authors made no disclosures.

Author Contributions

Michael J. Lunski: Research, statistical analysis, and writing of the manuscript. Jeffrey Burton: Statistical analysis and compilation of the noncancer patient cohort. Karine Tawagi: Chart review and some revisions of the manuscript. Diana Maslov: Chart review and some revisions of the manuscript. Victoria Simenson: Chart review and some revisions of the manuscript. Daniel Barr: Chart review and some revisions of the manuscript. Helen Yuan: Chart review and some revisions of the manuscript. Daniel Johnson: Review of final recommended changes to the manuscript and professional input. Marc Matrana: Review of final recommended changes to the manuscript and professional input. John Cole: Review of final recommended changes to the manuscript and professional input. Zoe Larned: Mentoring. Brian Moore: Review of final recommended changes to the manuscript and professional input.