Survival Among Patients With High-Risk Gastrointestinal Cancers During the COVID-19 Pandemic

Key Points Question How did incidence, stage, and survival among patients with high-risk gastrointestinal (HRGI) cancers change after the start of the COVID-19 pandemic? Findings In this cohort study with 156 937 patients, operative mortality and 1-year survival curves among patients with HRGI cancers were preserved during the pandemic. However, there was substantial underdiagnosis in 2020, with no proportional increase in newly diagnosed cases or stage migration throughout the remainder of the year. Meaning These findings suggest that cancer clinicians continued to deliver quality care during 2020 for patients able to be diagnosed; however, underdiagnosis and lack of health care access may have also contributed to lives lost.


Introduction
The COVID-19 pandemic presented challenges to the delivery of cancer care. 1 Resource diversion toward patients with COVID-19 led to a scarcity of personal protective equipment, health care personnel, and facility space that may have otherwise been used for patients with cancer. 2,3Thus, substantial decreases in cancer screening and diagnosis in 2020 have been described. 4,5rthermore, previous work has shown the increased risk of worse outcomes for patients with cancer and COVID-19, especially in 2020, prior to the widespread availability of vaccinations.Specifically, early reports during the pandemic demonstrated that active COVID-19 infection was associated with increased perioperative and long-term mortality among patients with cancer. 6,7though challenges in the delivery of care during the pandemic apply to all cancers, certain disease sites may have been particularly affected.For example, patients with high-risk gastrointestinal (HRGI) cancers, including esophageal, gastric, primary liver, and pancreatic cancers, already have higher rates of presenting at advanced stages as well as an increased risk of perioperative mortality and worse survival, [8][9][10][11][12] which may have been exacerbated by COVID-19 infection or pandemic-related stressors.Thus, the frequency of newly diagnosed cases, staging, and mortality among patients with HRGI cancers during the pandemic may have been significantly altered and serves as a relevant case study for assessing changes in cancer care and outcomes during this time.However, the extent to which these critical cancer datapoints were affected among this population due to barriers created by the pandemic remains unknown.
The National Cancer Database (NCDB) identifies 72% of patients newly diagnosed with cancer in the US annually. 13Previous work analyzing the data collection infrastructure of the NCDB during the pandemic demonstrated that case abstraction remained intact, validating the use of this comprehensive, national database in future cohorts. 14Although prior studies have evaluated national-level changes in cancer diagnosis 5 and treatment, [15][16][17][18] additional site-specific analyses are essential to understand what happened as a result of the pandemic and to anticipate changes in patient needs moving forward.We hypothesized that during the onset of the pandemic, there was a period of underdiagnosis as well as an increase in patients presenting with advanced-stage disease.
Additionally, we hypothesized that rates of 1-year and operative mortality increased during the pandemic, potentially due to pandemic-related stressors or COVID-19 infection.Thus, our objective was to evaluate the frequency of newly diagnosed cases, staging, and mortality, including overall 1-year survival and short-term operative mortality, among patients diagnosed with HRGI cancers during the COVID-19 pandemic.

Data Source
This retrospective cohort study queried the NCDB for patients diagnosed with HRGI cancer (esophageal, gastric, primary liver, or pancreatic) between January 1, 2018, and December 31, 2020, using data abstracted for the 2021 NCDB participant user file to ensure that all patients had a minimum of 12 months' follow-up.The NCDB represents one of the most comprehensive cancer registries in the world 13,19

Patient Population
Patients aged 18 years or older with newly diagnosed esophageal, gastric, primary liver, or pancreatic cancer as well as patients with only 1 primary cancer diagnosis in their lifetime were included.These 4 disease sites were chosen specifically based on their propensity for worse survival outcomes and higher rates of operative mortality [10][11][12] compared with other primary malignant neoplasms, representing a pertinent case study for detecting early changes to long-term outcomes during the pandemic.Data were defined using the International Classification of Diseases for Oncology, Third Edition, by primary site topography and histology codes. 20Patients with missing demographic or clinical staging data were excluded (eFigure 1 in Supplement 1).

Independent Variables
The following variables were evaluated across the entire cohort: year of diagnosis, age at diagnosis (years), sex, race and ethnicity, facility type, insurance status, and Charlson-Deyo score (comorbidities). 21Importantly, month of diagnosis is not publicly available within the NCDB; however, because this analysis was performed by the American College of Surgeons Cancer Department staff, month of diagnosis was available and included.Race and ethnicity were selfreported and categorized as Asian including Hawaiian or Pacific Islander (hereinafter, Asian), Hispanic, non-Hispanic Black (hereinafter, Black), and non-Hispanic White (hereinafter, White).Racial and ethnic disparities in cancer care have been widely described before; thus, we adjusted for race and ethnicity in the current study.Among patients diagnosed with HRGI cancers in 2020, the proportion of patients diagnosed with COVID-19 was evaluated and stratified by perioperative infections, including 30 days before or after the date of definitive resection defined by prior studies. 6

Primary Outcomes
The primary outcomes were trends in newly diagnosed cases, stage at diagnosis, and mortality.
Clinical stage was determined with the eighth edition of the American Joint Committee on Cancer staging system. 22Mortality included an evaluation of 1-year survival and operative mortality.The NCDB defines operative mortality as death within 30 or 90 days of the most definitive primary site surgery. 21Thus, evaluation of operative mortality was limited to patients who underwent curativeintent resection.

Statistical Analysis
Patient demographic characteristics, facility type, disease site, and stage were evaluated across study years and compared using χ 2 tests.To measure variance in the monthly reporting of HRGI cases, the number of newly diagnosed cases were plotted by month and compared across years using a repeated-measures analysis of variance (ANOVA) to assess for temporal changes.Next, the proportion of patients presenting at each stage was similarly plotted by month and compared across years with a repeated-measures ANOVA to measure changes in stage at diagnosis.
We then sought to evaluate potential changes in mortality during the pandemic, including 1-year overall survival and operative mortality rates.Kaplan-Meier methods in conjunction with log rank and Wilcoxon rank sum tests were used to assess 1-year survival between years.A multivariable Cox proportional hazards regression was used to assess the association of the year of diagnosis with survival, adjusting for patient demographic characteristics, facility type, disease site, and clinical stage and clustered by facility.To assess operative mortality, monthly trends in rates of both 30-day and 90-day mortality were compared across years using a repeated-measures ANOVA to identify potential increases in surgical risk during the pandemic among patients who underwent resection.
Differences in operative mortality were additionally assessed between years and compared using χ 2 tests.Multivariable logistic regression models, adjusting for patient demographics, facility type, disease site, and clinical stage, were used to evaluate the association of the diagnosis year with 30-day and 90-day operative mortality and similarly clustered by facility.

JAMA Network Open | Oncology
Survival Among Patients With High-Risk Gastrointestinal Cancers During the COVID-19 Pandemic Recognizing that different primary disease sites may have been affected by pandemic-related stressors in different ways, sensitivity analyses were performed examining newly diagnosed cases, stage, and mortality across all 4 disease sites and studied as separate cohorts.
All statistical tests were 2-sided, with significance determined using a threshold of α < .05.All statistical analyses were conducted using SAS, version 9.4 (SAS Institute Inc).Data were analyzed between August 23 and September 4, 2023.

Frequency of Newly Diagnosed HRGI Cancers
In general, there was a significant decrease in newly diagnosed HRGI cancers in March through May 2020 compared with prior years, representing at least 3000 fewer cases during this time and greater than 1500 fewer new cases in April 2020 alone (P = .002;Figure 1).However, the frequency of newly diagnosed cases per month returned to prepandemic levels by July 2020, with no evidence of a proportional increase, or rebound throughout the remainder of 2020.

Stage at Diagnosis
There was a slight increase in the proportion of patients presenting with stage IV disease in 2020 (43.9%) compared with 2018 (41.9%) and 2019 (42.2%) (P < .001;Table ).When trends throughout the year were evaluated, there was a noticeable decrease in patients diagnosed with stage I (−3.9%) and stage II (−2.3%) disease as well as a dramatic increase in patients presenting with stage IV disease (7.1%), primarily in March through May 2020 (P < .001; Figure 2).However, throughout the remainder of 2020, the proportion of patients presenting at each stage mirrored the trends in stage observed in prior years.

Mortality
When mortality was evaluated, 1-year survival rates in 2020 were 47.4% compared with 50.7% in both 2018 and 2019 (P < .001;eTable 1 in Supplement 1).Despite this variation, 1-year survival curves for 2020 reflected those of 2018 and 2019 (log rank P = .30,Wilcoxon P = .20;Figure 3).These findings were similarly demonstrated in a multivariable Cox regression, in which patients diagnosed in 2020 were not more likely to experience mortality at 1 year (hazard ratio, 0.99; 95% CI, 0.97-1.01)compared with prior years after adjusting for potential confounders.Importantly, after repeating this analysis with only stages I through III disease, 1-year survival rates in 2020 were 64.1% compared with 67.9% in 2018 and 67.7% in 2019 (P < .001).Despite a similar variation as within the overall cohort, our findings remained such that patients diagnosed in 2020 were not more likely to experience 1-year mortality on a multivariable Cox regression (hazard ratio, 0.98; 95% CI, 0.96-1.03).
For operative mortality, 39 412 patients underwent definitive resection.Of these patients, there was a slight decrease in the proportion of patients undergoing surgery in 2020 (25.7% in 2018, 25.3% in 2019, and 24.3% in 2020; P < .001;eTable 2 in Supplement 1).Similarly, changes in the

JAMA Network Open | Oncology
Survival Among Patients With High-Risk Gastrointestinal Cancers During the COVID-19 Pandemic proportion of patients treated with chemotherapy and radiation were small between years.Importantly, although there was slight variance in the monthly rate of 30-day mortality events (2.1% in 2018, 2.0% in 2019, and 2.1% in 2020; P = .04;Figure 4A), the plotted trends did not reflect worse outcomes during the early months of the pandemic, and no variance in 90-day mortality was observed (4.3% in 2018, 4.4% in 2019, and 4.6% in 2020; P = .10;Figure 4B).Although there was a small increase in the proportion of unknown operative mortality data in 2020 (from 0.6% to 2.1%; P < .001),changes in the rate of operative mortality between years were minimal (eTable 1 in Supplement 1).These results remained consistent on multivariable analysis as well; after adjusting for potential confounders, patients were not more likely to experience 30-day (odds ratio, 0.96; 95%

Sensitivity Analysis
Recognizing the potential for site-specific differences, trends in newly diagnosed cases, staging, and mortality were repeated across the 4 cancer sites included in this analysis.Specifically, similar trends  in case counts were seen across all 4 diseases, with the most substantial decreases exhibited for primary liver and pancreatic cancers (eFigure 2 in Supplement 1).We similarly identified a slight increase in the proportion of patients presenting with stage IV disease, most noticeably for gastric cancer (eTable 3 in Supplement 1).In addition, although there was a slight decrease in 1-year survival in 2020, rates of operative mortality reflected those of the larger cohort across years (eTable 4 in Supplement 1), with 1-year survival curves mirroring those of prior years as well (eFigure 3 in Supplement 1).

Discussion
The delivery of cancer care was undoubtedly affected during the COVID-19 pandemic. 4,5,23Early reports revealing increased postoperative mortality due to COVID-19 infection 6 influenced the creation of consensus statements regarding the triage of elective procedures, often including cancer operations. 24,25These guidelines may have particularly affected certain populations, such as those with HRGI cancers, in which resection is the only chance for cure and delays in management may lead to worse outcomes.The findings of this national study of patients with newly diagnosed HRGI cancers suggest substantial underdiagnosis in the first year of the pandemic.Additionally, patients more frequently presented with advanced-stage disease; however, at a time when the pandemic was disrupting health care, it is a tribute to the efforts of cancer clinicians that 1-year survival curves and operative mortality remained unchanged, as our findings suggest.
There was a considerable decrease in newly diagnosed HRGI cancers at the onset of the pandemic. 5However, as 2020 progressed, the frequency of newly diagnosed cases returned to prepandemic levels, which is consistent with previous reports. 5Obviously, there was not a decrease in the number of patients who developed cancer as a result of the pandemic; rather, notably fewer patients were diagnosed, likely due to pandemic-related stressors.To the credit of the cancer programs, the findings of this study suggest that the frequency of newly diagnosed HRGI cancers quickly recovered within the second half of 2020.Because more than 3000 fewer cases were reported at the beginning of the pandemic, it is likely that the early lack of available health care, including scarce resources and fears of contracting COVID-19, led to a reluctance to seek medical care 2,3,26,27 and may have resulted in substantial loss of life for patients with HRGI cancers who would have otherwise had access to diagnosis and treatment.Specifically, these data call attention to a critical health care systems failure in our ability to respond to national crises.Although the COVID-19 public health emergency has been lifted, these results are broadly applicable to future pandemics,

JAMA Network Open | Oncology
Survival Among Patients With High-Risk Gastrointestinal Cancers During the COVID-19 Pandemic national emergencies, and natural disasters, demonstrating the need for improved safeholds to ensure future lives are not lost unnecessarily.
We also observed a substantial increase in the diagnosis of advanced-stage disease at the onset of the pandemic.These findings are consistent with prior reports across numerous other cancer sites 5 and, thus, are more likely representative of the time during the pandemic when the patient sought care rather than true evidence of stage migration.Specifically, during the early months of the pandemic, patients with milder symptoms may have delayed seeking medical advice, whereas those with severe symptoms due to advanced disease had to seek care by necessity.This represents an opportunity to improve patient education by advising them to seek medical attention when any symptoms arise rather than just severe ones.In addition, given that screening mechanisms are absent for detecting HRGI cancers, timely diagnosis is dependent on primary care physician referrals, usually after the development of early-onset symptoms, followed by imaging, laboratory tests, or endoscopy.However, delays in routine primary care health checks, along with office closures and cancellations of elective procedures such as endoscopy, may have contributed to the proportional decrease in the diagnosis of early-stage disease. 28,29Furthermore, to evaluate the association of the pandemic with other aspects of cancer care such as screening, future work should evaluate changes in the presentation of screenable cancers as well as their long-term outcomes.
Despite the observed shifts in stage, we identified no difference in 1-year survival curves for patients diagnosed with HRGI cancers in 2020.These findings are contradictory to the hypotheses described in numerous prior studies, 5,30 illustrating diagnostic and treatment delays that were projected to substantially affect survivorship. 31,32However, prior work using the NCDB to evaluate alterations in cancer treatment during the pandemic demonstrated a substantial decrease in time to treatment initiation 18 which may contribute to the consistency in survival curves in the current study.
Although this cohort of patients will be followed over time, it is unlikely that mortality rates at further time points will differ substantially, given the present results.These data highlight a key point for cancer researchers across the US, that inclusion of this cohort in future studies is unlikely to confound survival analyses moving forward.
In this study, operative mortality rates remained stable compared with prior years, without evidence of deviation throughout the first year of the pandemic.These data highlight the tremendous efforts of cancer clinicians at the time, who came together on behalf of patients with cancer to ensure that they continued receiving quality care despite being faced with one of the deadliest pandemics in history. 33,34This is consistent with prior studies that showed success in continuing to perform necessary cancer operations to prevent lapses in treatment.For example, a COVID-minimal surgical pathway was created to standardize best practices in preventing COVID-19 and is jointly maintained by the American College of Surgeons Commission on Cancer and the American Cancer Society, abstracting data regarding the diagnosis and treatment of cancer at Commission on Cancer-accredited hospitals.All data within the NCDB are deidentified and Health Insurance Portability and Accountability Act compliant and thus deemed exempt from the American College of Surgeons institutional review board review.The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Figure 2 .D
Figure 2. Monthly Incidence of High-Risk Gastrointestinal (HRGI) Cancers by Stage in 2018 to 2020

Figure 3 .
Figure 3. One-Year Survival Among Patients With High-Risk Gastrointestinal (HRGI) Cancers Diagnosed in 2018 to 2020