Survival of Patients Diagnosed With Cancer During the COVID-19 Pandemic
Todd Burus, Haluk Damgacioglu, Bin Huang, Thomas C. Tucker, Ashish A. Deshmukh, Krystle A. Lang Kuhs
TL;DR
Cancer patients diagnosed during the early years of the COVID-19 pandemic had worse survival rates compared to those diagnosed before the pandemic, with an estimated 13% more deaths within a year.
Contribution
This study is the first to show that pandemic-related disruptions in cancer care led to measurable declines in short-term survival rates for cancer patients in the US.
Findings
Patients diagnosed in 2020 and 2021 had significant reductions in 1-year survival rates compared to pre-pandemic years.
The survival decline was observed for both early- and late-stage cancer diagnoses, with late-stage cases showing larger drops.
The study estimates 17,390 more cancer-related deaths within a year of diagnosis during the pandemic's first two years.
Abstract
Were the survival rates of patients diagnosed with cancer in the US in 2020 and 2021 different than in the years before the COVID-19 pandemic? This population-based cohort study of more than 1 million US individuals found that patients diagnosed with cancer in 2020 and 2021 (compared with 2015 to 2019) had significant reductions in survival when diagnosed at both early and late stages, resulting in an estimated 17 390 more deaths (13.1%) within 1 year of diagnosis than expected. Individuals diagnosed with cancer in the US during 2020 and 2021 experienced worse survival during the first year after their diagnosis than those diagnosed before the COVID-19 pandemic. This cohort study investigates the association of the COVID-19 pandemic with short-term survival among patients diagnosed with cancer in 2020 and 2021 using US population-based cancer registry data. The effects of COVID-19…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2| Group | Patients, No. (%) | |||||
|---|---|---|---|---|---|---|
| 2015-2019 | 2020 | 2021 | ||||
| Cases | Deaths | Cases | Deaths | Cases | Deaths | |
| Overall No. | 2 471 372 | 366 968 | 473 781 | 69 779 | 534 231 | 73 798 |
| Stage at diagnosis | ||||||
| Early stage | 1 216 058 (49.2) | 42 978 (11.7) | 228 692 (48.3) | 8487 (12.2) | 266 540 (49.9) | 9195 (12.5) |
| Late stage | 1 255 314 (50.8) | 323 990 (88.3) | 245 089 (51.7) | 61 292 (87.8) | 267 691 (50.1) | 64 603 (87.5) |
| Age group, y | ||||||
| <65 | 1 252 085 (50.7) | 124 814 (34.0) | 232 146 (49.0) | 22 813 (32.7) | 256 885 (48.1) | 23 249 (31.5) |
| ≥65 | 1 219 287 (49.3) | 242 154 (66.0) | 241 635 (51.0) | 46 966 (67.3) | 277 346 (51.9) | 50 549 (68.5) |
| Race and ethnicity | ||||||
| Hispanic | 367 991 (14.9) | 52 329 (14.3) | 74 759 (15.8) | 10 747 (15.4) | 86 411 (16.2) | 11 969 (16.2) |
| Non-Hispanic Black | 280 365 (11.3) | 45 354 (12.4) | 54 320 (11.5) | 8655 (12.4) | 62 523 (11.7) | 9166 (12.4) |
| Non-Hispanic White | 1 623 286 (65.7) | 243 755 (66.4) | 303 836 (64.1) | 45 112 (64.6) | 335 769 (62.9) | 46 876 (63.5) |
| Other non-Hispanic | 177 959 (7.2) | 24 495 (6.7) | 35 271 (7.4) | 4982 (7.1) | 42 266 (7.9) | 5532 (7.5) |
| Unknown | 21 771 (0.9) | 1035 (0.3) | 5595 (1.2) | 283 (0.4) | 7262 (1.4) | 255 (0.3) |
| Urbanicity | ||||||
| Metropolitan | 2 174 826 (88.0) | 313 505 (85.4) | 416 680 (87.9) | 59 463 (85.2) | 471 631 (88.3) | 62 932 (85.3) |
| Nonmetropolitan | 294 426 (11.9) | 52 988 (14.4) | 56 691 (12.0) | 10 218 (14.6) | 62 136 (11.6) | 10 763 (14.6) |
| Unknown | 2120 (0.1) | 475 (0.1) | 410 (0.1) | 98 (0.1) | 464 (0.1) | 103 (0.1) |
| Group and stage at diagnosis | 2020 | 2021 | ||
|---|---|---|---|---|
| Observed CSS, % (95% CI) | Difference from expected, percentage points (95% CI) | Observed CSS, % (95% CI) | Difference from expected, percentage points (95% CI) | |
|
| ||||
| Early stage | 96.20 (96.11 to 96.28) | −0.44 (−0.54 to −0.34) | 96.43 (96.36 to 96.51) | −0.27 (−0.37 to −0.16) |
| Late stage | 74.18 (74.00 to 74.35) | −1.34 (−1.75 to −0.93) | 74.94 (74.77 to 75.10) | −1.20 (−1.69 to −0.71) |
|
| ||||
| <65 | ||||
| Early stage | 98.08 (98.00 to 98.16) | −0.25 (−0.37 to −0.13) | 98.24 (98.17 to 98.31) | −0.14 (−0.27 to −0.01) |
| Late stage | 82.01 (81.79 to 82.23) | −0.92 (−1.59 to −0.25) | 82.80 (82.58 to 83.01) | −0.64 (−1.46 to 0.17) |
| ≥65 | ||||
| Early stage | 94.29 (94.15 to 94.43) | −0.66 (−0.85 to −0.48) | 94.69 (94.57 to 94.81) | −0.41 (−0.61 to −0.21) |
| Late stage | 66.83 (66.57 to 67.10) | −1.77 (−2.14 to −1.40) | 67.86 (67.61 to 68.11) | −1.59 (−2.00 to −1.18) |
|
| ||||
| Hispanic | ||||
| Early stage | 95.81 (95.59 to 96.02) | −0.21 (−0.54 to 0.12) | 95.92 (95.72 to 96.12) | −0.13 (−0.50 to 0.24) |
| Late stage | 76.23 (75.81 to 76.65) | −1.42 (−2.24 to −0.61) | 76.45 (76.05 to 76.84) | −1.69 (−2.65 to −0.73) |
| Non-Hispanic Black | ||||
| Early stage | 96.34 (96.11 to 96.58) | −0.47 (−0.73 to −0.22) | 96.75 (96.55 to 96.95) | −0.12 (−0.36 to 0.12) |
| Late stage | 72.42 (71.90 to 72.94) | −1.18 (−1.90 to −0.45) | 73.31 (72.81 to 73.80) | −0.94 (−1.73 to −0.14) |
| Non-Hispanic White | ||||
| Early stage | 96.14 (96.05 to 96.24) | −0.50 (−0.62 to −0.39) | 96.38 (96.29 to 96.47) | −0.35 (−0.47 to −0.23) |
| Late stage | 73.53 (73.31 to 73.76) | −1.24 (−1.69 to −0.80) | 74.33 (74.12 to 74.55) | −1.05 (−1.58 to −0.52) |
| Other non-Hispanic | ||||
| Early stage | 96.61 (96.33 to 96.90) | −0.27 (−0.81 to 0.27) | 96.69 (96.43 to 96.94) | −0.19 (−0.83 to 0.45) |
| Late stage | 76.51 (75.91 to 77.12) | −2.09 (−2.82 to −1.36) | 77.31 (76.75 to 77.87) | −2.01 (−2.77 to −1.25) |
|
| ||||
| Urban | ||||
| Early stage | 96.34 (96.26 to 96.43) | −0.39 (−0.48 to −0.30) | 96.61 (96.54 to 96.69) | −0.19 (−0.27 to −0.10) |
| Late stage | 74.83 (74.64 to 75.01) | −1.30 (−1.79 to −0.82) | 75.61 (75.43 to 75.79) | −1.14 (−1.73 to −0.56) |
| Rural | ||||
| Early stage | 95.05 (94.79 to 95.32) | −0.79 (−1.18 to −0.40) | 95.05 (94.80 to 95.30) | −0.92 (−1.34 to −0.49) |
| Late stage | 69.67 (69.15 to 70.20) | −1.46 (−2.11 to −0.81) | 70.15 (69.65 to 70.66) | −1.59 (−2.28 to −0.89) |
| Site | Mortality difference, No. (95% CI) | |
|---|---|---|
| 2020 | 2021 | |
| All sites | 9162 (6958 to 11 365) | 8228 (5411 to 11 044) |
| Female breast | 88 (−273 to 448) | 60 (−415 to 536) |
| Prostate | 249 (38 to 461) | 346 (71 to 621) |
| Colon and rectum | 1150 (704 to 1595) | 645 (88 to 1202) |
| Corpus and uterus NOS | 189 (−79 to 457) | 130 (−217 to 477) |
| Melanoma | 81 (−34 to 196) | 15 (−107 to 137) |
| Pancreas | 333 (23 to 644) | 652 (298 to 1006) |
| Liver and IBD | 31 (−359 to 421) | 231 (−285 to 747) |
| Esophagus | 70 (−105 to 245) | −16 (−218 to 186) |
| Lung and bronchus | 1131 (−5 to 2268) | 1966 (538 to 3393) |
| Brain and ONS | 341 (148 to 533) | 197 (−4 to 397) |
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCOVID-19 and healthcare impacts · Lymphadenopathy Diagnosis and Analysis · Economic and Financial Impacts of Cancer
Introduction
Several studies have reported significant COVID-19 pandemic–related disruptions to cancer care in the US.^1,2,3,4,5,6,7,8,9,10^ Yet, with current cancer registry data covering only cases diagnosed through 2022, existing research has mostly been restricted to evaluating declines in cancer screening and diagnosis. An important and unstudied question is the association of these disruptions with increased cancer morbidity and mortality. While it will likely take several years to gain a complete picture of how the pandemic has impacted cancer outcomes, 1-year cause-specific survival (CSS) among patients diagnosed in 2020 and 2021 can be evaluated using available data. In this study, we analyzed 1-year CSS to determine whether—and to what extent—short-term cancer-related outcomes changed during the first 2 years of the COVID-19 pandemic.
Methods
This study was deemed exempt from review by the University of Kentucky Institutional Review Board, and informed consent was waived because the data were deidentified and publicly available. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Data Sources
We obtained 1-year CSS data for patients diagnosed with a first primary malignant cancer in the US between January 1, 2020, and December 31, 2021, using data from the Surveillance, Epidemiology, and End Results 21 Registries (SEER-21) database.^11^ SEER-21 covered approximately 41.9% of the US population during the study period. Cases in which the individual was diagnosed by death certificate and/or autopsy only, or recorded as alive with no survival time available, were excluded, as were in situ cases or cases with unknown staging. We collected data for the overall population and stratified by sex, age, race and ethnicity, urbanicity, and stage at diagnosis. Age was stratified at age 65 years or older to account for the beginning of Medicare eligibility and the fact that advanced age was a leading risk factor for COVID-19 mortality. Race and ethnicity were defined as Hispanic, non-Hispanic Black, non-Hispanic White, and other non-Hispanic races according to the SEER-recommended race and origin variable.^12^ Other non-Hispanic races included individuals who identified as American Indian and Alaska Native and Asian or Pacific Islander, which were combined due to low case counts. Urbanicity was defined by the 2013 Rural Urban Continuum Codes (RUCC) for county of residence, with RUCC 1-3 defined as metropolitan and RUCC 4-9 as nonmetropolitan.^13^ Stage at diagnosis was defined using SEER’s combined summary stage variable with expanded regional codes, with early stage defined as localized disease only, and late stage including regional and distant stage diagnoses.^11^
We also collected separate CSS data for individual cancer sites according to 2 mutually exclusive categories of interest, which we labeled low-survival cancers and high-incidence/high-survival cancers. Low-survival cancers are susceptible to survival outcome changes with even minor disruptions in care delivery, whereas small reductions in CSS for high-incidence/high-survival cancers could entail large increases in total cancer deaths due to the large number of total diagnoses. We defined low-survival cancers as common cancer sites with current 5-year relative survival rates of approximately one-third or lower.^14^ These sites included the pancreas, liver and intrahepatic bile duct, esophagus, lung and bronchus, and the brain and other nervous system sites. We defined high-incidence/high-survival cancers as common cancer sites with a current age-adjusted incidence greater than or equal to 20.0 per 100 000 people and a 5-year relative survival of approximately two-thirds or higher.^14^ These sites included the female breast, prostate, colon and rectum, corpus and uterus (not otherwise specified), and melanoma. Cancer sites were identified according to the International Classification of Diseases for Oncology, Third Revision Site Recode/World Health Organization 2008 definition.^15^
Statistical Analysis
Data were analyzed from May 13 to May 27, 2025. We calculated 1-year CSS rates overall and by stage at diagnosis with 95% CIs for patients diagnosed with malignant cancer in 2020 and 2021 using SEER*Stat, version 8.4.5 (National Cancer Institute).^16^ Annual 1-year CSS rates were also calculated for 2015 to 2019 for comparison. CSS rates were calculated among the total eligible population using the actuarial method (eMethods in Supplement 1). One-year CSS estimates the probability of not having a cancer-related cause of death within 1 year of diagnosis and is calculated as the number of persons still living 1 year after a cancer diagnosis divided by the total number of persons diagnosed with cancer. CSS is a measure of net cancer survival in which noncancer causes of death are censored. We chose CSS for this analysis instead of relative survival due to concerns over having appropriate life tables for computing the expected survival of individuals in the absence of a cancer diagnosis during the COVID-19 pandemic.^17,18,19^
Previous research has indicated substantial differences in the number of cancer cases diagnosed between 2020 and 2021.^2^ The increase in 2021 may be attributed, in part, to the diagnosis of cases that went undiagnosed in 2020. Therefore, to account for potential changes in stage distribution during the COVID-19 pandemic, we analyzed survival trends separately by stage at diagnosis. Stage-specific 1-year CSS trends for 2015 to 2019 were fit with proportional hazards models using the JPSurv web application.^20^ Fitted models estimated trends as the average absolute change in survival (AACS) with 95% CIs. AACS represents the absolute year-to-year percentage point difference in 1-year CSS between 2015 and 2019. The fitted models for 2015 to 2019 were then used to estimate expected stage-specific 1-year CSS rates for 2020 and 2021 and to calculate the absolute survival difference between the observed and expected 1-year CSS rates. Additionally, we calculated the 1-year mortality difference as the difference in the number of observed and expected deaths within 1 year of diagnosis. Stage-specific 1-year mortality differences were obtained by multiplying the absolute survival difference by the appropriate observed cancer incidence rate and the standard US population estimate of 274 633 642 (or 137 316 821 for sex-specific cancers). Total 1-year mortality differences in 2020 and 2021 were calculated as the sum of stage-specific 1-year mortality differences. Because CSS is not an additive measure, the 1-year mortality difference for all cancer sites will not necessarily equal the sum of 1-year mortality differences among complementary subgroups.
We defined pointwise significant changes in CSS to have occurred in 2020 or 2021 if the 95% CI for the absolute difference between observed and expected 1-year CSS rates did not contain 0. All analyses were performed and visualizations created using the R statistical programming language, version 4.4.1 (R Foundation).
Results
One-Year Cause-Specific Survival for All Cancer Sites
During the first 2 years of the COVID-19 pandemic, a total of 1 008 012 individuals were diagnosed with cancer, with 473 781 cases (47.0%) reported in 2020 (50.1% female and 49.9% male; 51.0% diagnosed at ≥65 years; 11.5% Black, 15.8% Hispanic, 64.1% White, and 7.4% other race, including Alaska Native, American Indian, Asian, and Pacific Islander) and 534 231 cases (53.0%) in 2021 (50.3% female and 49.7% male; 51.9% diagnosed at ≥65 years; 11.7% Black, 16.2% Hispanic, 62.9% White, and 7.9% other race, including Alaska Native, American Indian, Asian, and Pacific Islander) (Table 1). Among these, 143 577 individuals experienced a cancer-related death within 1 year of diagnosis (69 779 in 2020 and 73 798 in 2021), and 63 967 were lost to follow-up (26 784 in 2020 and 37 183 in 2021). The resulting 1-year CSS rates were 84.84% in 2020 (95% CI, 84.74%-84.95%) and 85.69% in 2021 (95% CI, 85.59%-85.78%). Total case numbers by site and stage at diagnosis are included in eTables 1 and 2 in Supplement 1.
In 2020, there were 228 692 early-stage cancer diagnoses, with a corresponding 1-year CSS rate of 96.20% (95% CI, 96.11%-96.28%) (Table 2; Figure 1). This number increased to 266 540 early-stage diagnoses in 2021, with a 1-year CSS of 96.43% (95% CI, 96.36%-96.51%). Compared with survival trends from 2015 to 2019, the 1-year CSS for early-stage diagnoses was 0.44 percentage points lower than expected in 2020 (95% CI, −0.54 to −0.34 percentage points) and 0.27 percentage points lower in 2021 (95% CI, −0.37 to −0.16 percentage points) (Figure 2).
Changes in 1-Year Cause-Specific Survival (CSS) Rates Among Patients Diagnosed With Cancer in 2020 and 2021One-year CSS rates with 95% CIs among patients diagnosed with early-stage cancer (A)and late-stage cancer (B) between 2015 and 2021. Shaded areas represent 95% CIs. The overall number of patients at risk was 1 008 012.
Changes in 1-Year Cause-Specific Survival (CSS) Rates Among Patients Diagnosed With Cancer in 2020 and 2021 by Cancer SiteAbsolute differences in 1-year CSS rates among patients with early-stage cancer diagnoses in 2020 (A), early-stage cancer diagnoses in 2021 (B), late-stage cancer diagnoses in 2020 (C), and late-stage cancer diagnoses in 2021 (D). The overall number of patients are risk was 473 781 in 2020 and 534 231 in 2021. IBD indicates intrahepatic bile duct; NOS, not otherwise specified; ONS, other nervous system.aSignificant difference from expected.
For late-stage diagnoses, 245 089 cases were identified in 2020 with a 1-year CSS of 74.78% (95% CI, 74.00%-74.35%), while 267 691 late-stage cases were recorded in 2021 with a 1-year CSS of 74.94% (95% CI, 74.77%-75.10%). In both years, 1-year CSS for late-stage cases was significantly lower than expected based on prepandemic survival trends, with 1.34 percentage points lower in 2020 (95% CI, −1.75 to −0.93 percentage points) and 1.20 percentage points lower in 2021 (95% CI, −1.69 to −0.71 percentage points).
Reductions in survival among both early-stage and late-stage cancers contributed to an estimated mortality difference of 9162 (95% CI, 6958-11 365) cancer-related deaths within 1 year of diagnosis in 2020 and 8228 (95% CI, 5411-11 044) in 2021 (Table 3). Overall, the first 2 years of the pandemic were associated with 17 390 more cancer-related deaths (13.1%) than expected within 1 year of diagnosis.
All population subgroups examined experienced 1-year CSS reductions for early-stage diagnoses in both years, except for non-Hispanic Black individuals in 2021 and individuals of Hispanic ethnicity and other non-Hispanic race in 2020 and 2021 (Table 2). Late-stage survival reductions occurred in all population subgroups in 2020 and in all except individuals younger than 65 years in 2021. Absolute survival reductions of greater than 1.00 percentage points occurred in both years for late-stage diagnoses among individuals of other non-Hispanic race (−2.09 [95% CI, −2.82 to −1.36] percentage points and −2.01 [95% CI, −2.77 to −1.25] percentage points, respectively) and individuals aged 65 years and older (−1.77 [95% CI, −2.14 to −1.40] percentage points and −1.59 [95% CI, −2.00 to −1.18], respectively).
One-Year Cause-Specific Survival for Low-Survival Cancers
The 5 low-survival cancer sites considered accounted for 17.8% of all cases included in 2020 and 2021 and 52.3% of all cancer deaths within 1 year of diagnosis. Significant reductions in 1-year CSS for early-stage diagnoses occurred in 2020 for esophageal cancer (−3.89 percentage points; 95% CI, −6.78 to −0.99 percentage points) and brain cancer (−2.94 percentage points; 95% CI, −4.29 to −1.60 percentage points), and in 2021 for liver cancer (−2.80 percentage points; 95% CI, −5.14 to −0.45 percentage points), esophageal cancer (−3.67 percentage points; 95% CI, −6.33 to −1.01 percentage points), lung cancer (−0.89 percentage points; 95% CI, −1.59 to −0.19 percentage points), and brain cancer (−1.77 percentage points; 95% CI, −3.17 to −0.37 percentage points) (Figure 2; eTable 3 in Supplement 1). Survival reductions were also observed in 2021 for late-stage diagnoses of pancreatic cancer (−2.13 percentage points; 95% CI, −3.16 to −1.09 percentage points) and lung cancer (−2.01 percentage points; 95% CI, −3.74 to −0.30 percentage points). Overall, significant 1-year mortality differences were observed for brain cancer in 2020 (341 more deaths than expected; 95% CI, 98-148), lung cancer in 2021 (1966 more deaths than expected; 95% CI, 538-3393), and pancreatic cancer in both years (333 and 652 more deaths than expected; 95% CIs, 23-644 and 298-1006).
One-Year Cause-Specific Survival for High-Incidence/High-Survival Cancers
The 5 high-incidence/high-survival cancer sites considered accounted for 44.0% of all cases reported in 2020 and 2021 and 14.0% of all cancer deaths within 1 year of diagnosis. In 2020, 1-year CSS rates were significantly lower than expected for early-stage colorectal cancer (−1.08 percentage points; 95% CI, −1.55 to −0.61 percentage points), uterine cancer (−0.42 percentage points; 95% CI, −0.74 to −0.10 percentage points) and melanoma (−0.12 percentage points; 95% CI, −0.24 to −0.01 percentage points) and for late-stage prostate cancer (−0.64 percentage points; 95% CI, −1.25 to −0.03 percentage points) and colorectal cancer (−1.48 percentage points; 95% CI, −2.24 to −0.72 percentage points). One-year CSS rates were significantly lower than expected in 2021 for early-stage female breast cancer (−0.09 percentage points; 95% CI, −0.17 to −0.04 percentage points), early-stage colorectal cancer (−0.78 percentage points; 95% CI, −1.29 to −0.26 percentage points) and late-stage prostate cancer (−0.77 percentage points; 95% CI, −1.45 to −0.08 percentage points). Individuals diagnosed with prostate cancer and colorectal cancer experienced significantly more deaths than expected within 1 year of diagnosis in both 2020 and 2021, resulting in an estimated 25.4% more prostate cancer deaths and 16.3% more colorectal cancer deaths than would have occurred without survival reductions.
Discussion
To our knowledge, this population-based cohort study is the first to examine short-term survival outcomes among US patients with newly diagnosed cancers during the first 2 years of the COVID-19 pandemic. We found significant reductions in 1-year CSS rates among patients diagnosed with both early-stage and late-stage cancer in 2020 and 2021 compared with trends from 2015 to 2019. Overall, survival reductions in 2020 and 2021 resulted in an estimated 17 390 more cancer deaths (13.1%) than expected within the first 12 months after diagnosis. Reductions occurred among all considered population subgroups, with the greatest impact among individuals of other non-Hispanic race and ethnicity and individuals aged 65 years or older. We also observed site-specific decreases among a variety of low-survival and high-incidence/high-survival cancer sites. We found the results for colorectal, prostate, and pancreatic cancers most concerning because all experienced significant differences in 1-year mortality in 2020 and 2021 after accounting for differences in stage distribution. Overall, our findings indicate that individuals diagnosed with cancer in the US during 2020 and 2021 experienced poorer cancer-related outcomes in the first year after diagnosis than those diagnosed before the COVID-19 pandemic.
Previously, Mani et al^17^ reported that persons living with cancer were at a greater risk of death from COVID-19 infection in 2020 than those without a cancer diagnosis, even at less than 1 year from diagnosis. Using cancer-specific mortality for our study allowed us to minimize concerns about COVID-19–related mortality and focus attention on experiences along the cancer care continuum. Given that available treatments should have been no worse than before the pandemic, decreases in 1-year CSS for patients diagnosed in 2020 and 2021 suggest issues with timely diagnosis and treatment.^21,22,23,24^ Our findings of significant 1-year survival rate reductions overall and for many of the cancer sites examined support this hypothesis and align with prior research into potential increased mortality due to treatment delays during the COVID-19 pandemic.^25^
Considering specific sites as either low-survival or high-incidence/high-survival allowed us to examine CSS in 2020 and 2021 from different perspectives. It was not surprising to find that several cancer sites with traditionally low survival rates experienced greater-than-expected 1-year mortality associated with pandemic-related disruptions during 2020 and 2021. It was somewhat surprising, however, that with just 1 year of follow-up, we already detected significantly lower CSS for several high-incidence/high-survival cancer sites as well. While the magnitude of reductions for these sites was generally less than observed among low-survival cancer sites, their higher incidence rates translated to a substantial increase in estimated lives lost vs expected. Underdetection of less advanced or asymptomatic cases during 2020 and 2021 may help explain some of the reductions observed in short-term survival for high-incidence/high-survival cancer sites. However, accounting for stage at diagnosis in our final assessment mitigated this concern to some extent. Moreover, even if some of these results are due to underdetection, this is not very reassuring because undetected cases could have progressed further before being discovered—something that may already be reflected in the survival rates from 2021. Identifying policy missteps and failures in health system preparedness that may have contributed to underdetection of cancer—as well as the previously mentioned delays in time to treatment initiation and treatment disruptions—are fundamental issues remaining to be addressed in the wake of the COVID-19 pandemic. It is important to investigate these questions further and determine whether other explanations exist as to why high-survival cancer sites were vulnerable to poorer outcomes during the COVID-19 pandemic.
Reductions in cancer screening leading to a change in the composition of cancers diagnosed in 2020 and 2021 are also important to consider when discussing survival decreases. Though some studies reported reduced female breast cancer screening during 2020, Doan et al^26^ found that screening disruptions among Medicare enrollees were largely relegated to the initial pandemic months of March through May. Additionally, prior studies from our group found that breast cancer incidence returned to expected levels by June 2020.^1,2^ These observations may help explain the limited 1-year survival impact we witnessed for female breast cancer. In contrast, prolonged screening disruptions and corresponding estimates of underdiagnosis for all stages of colorectal and lung cancers are consistent with our findings of reduced stage-specific survival for these sites.^1,2,3,4,5,6,27,28^ Disruptions to lung cancer screening have been attributed to COVID-19 infection concerns and strained health care capacity.^5,21,26,29,30^ Similar concerns were also raised regarding a lower volume of colonoscopies performed in 2020.^6,31^ As such, rebuilding capacity and reminding patients about their eligible cancer screenings will be critical to limiting the long-term negative impacts of the COVID-19 pandemic on early cancer detection. Rebuilding capacity and taking other steps to instill health system resilience would also be effective measures to help prevent delays in time to treatment initiation and treatment disruptions for patients with cancer in general.^24,32,33,34^
Strengths and Limitations
A key strength of this study is its use of data from high-quality, population-based central cancer registries. However, we acknowledge certain limitations within this study. First, using relative survival instead of CSS may be preferred for measuring excess mortality at the specific sites we examined,^35,36^ but restricting to only 1 year of follow-up limits the differences that can emerge. Moreover, previous studies have shown that CSS is preferable if standard life tables are unavailable or fail to capture risk factors specific to patients with cancer, such as increased risk of COVID-19 mortality.^17,18,19,37^ Second, differences in the severity of cases diagnosed may exist and affect survival between the pandemic and prepandemic periods. Stratifying survival rates by stage at diagnosis helps mitigate this concern, though meaningful within-stage variations may exist beyond the scope of this study to assess. Further research is needed to clarify this difference and better prepare for future disruptions. Third, CSS estimates may change due to delays in reporting deaths for patients diagnosed in 2020 and 2021, though this effect is likely minimal given that the dataset used was not assembled until 22 months after the latest follow-up date. Fourth, patient history of nonfatal COVID-19 infection or other type of infection was not available, limiting our ability to consider how an earlier recovered SARS-CoV-2 infection could have led to worse overall frailty in a way that it would not have done in a patient without cancer. Fifth, although we restricted consideration to cancer-specific causes of death, the accuracy of cause-of-death information during the COVID-19 pandemic may be reduced due to misclassification of COVID-19–related mortality. However, in an additional analysis, we found larger reductions in relative survival (which does not depend on cause of death) than those found for CSS (eTable 4 in Supplement 1). This pattern indicates misclassification is unlikely to be a primary driver of reductions in CSS, although some misclassification may still be present. Prior literature also supports using SEER cause-of-death fields for estimating cancer-specific mortality despite some misclassification.^19,37,38,39^
Conclusions
Disruptions related to the COVID-19 pandemic in the US have had a substantial impact across the cancer care continuum. This cohort study’s findings of significantly decreased rates of short-term cancer survival during the pandemic’s first 2 years are in line with previously documented interruptions to screening, diagnosis, and treatment. Survival reductions were seen in a broad range of cancer sites and stages of diagnosis, regardless of available screenings or typical prognosis. Swift action should be taken to increase cancer screening, rebuild health care capacity, and improve patient communication to combat longer-term consequences and prepare for potential future disruptions. Continued surveillance is needed to assess whether additional changes in survival outcomes extended further into and after the pandemic.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Burus T, Lei F, Huang B, . Undiagnosed cancer cases in the US during the first 10 months of the COVID-19 pandemic. JAMA Oncol. 2024;10(4):500-507. doi:10.1001/jamaoncol.2023.6969 38386344 PMC 10884945 · doi ↗ · pubmed ↗
- 2Burus T, Lei F, Huang B, . COVID-19 and rates of cancer diagnosis in the US. JAMA Netw Open. 2024;7(9):e 2432288. doi:10.1001/jamanetworkopen.2024.32288 39240562 PMC 11380103 · doi ↗ · pubmed ↗
- 3Negoita S, Chen HS, Sanchez PV, . Annual Report to the Nation on the Status of Cancer, part 2: Early assessment of the COVID-19 pandemic’s impact on cancer diagnosis. Cancer. 2024;130(1):117-127. doi:10.1002/cncr.35026 37755665 PMC 10841454 · doi ↗ · pubmed ↗
- 4Schafer EJ, Islami F, Han X, . Changes in cancer incidence rates by stage during the COVID-19 pandemic in the US. Int J Cancer. 2024;154(5):786-792. doi:10.1002/ijc.34758 37971377 · doi ↗ · pubmed ↗
- 5Zhang X, Elsaid MI, De Graffinreid C, Champion VL, Paskett ED; Impact of COVID-19 on Behaviors across the Cancer Control Continuum in Ohio group. Impact of the COVID-19 pandemic on cancer screening delays. J Clin Oncol. 2023;41(17):3194-3202. doi:10.1200/JCO.22.01704 36735899 PMC 10256430 · doi ↗ · pubmed ↗
- 6Fedewa SA, Star J, Bandi P, . Changes in cancer screening in the US during the COVID-19 pandemic. JAMA Netw Open. 2022;5(6):e 2215490. doi:10.1001/jamanetworkopen.2022.15490 35657622 PMC 9166223 · doi ↗ · pubmed ↗
- 7Fedeli U, Barbiellini Amidei C, Han X, Jemal A. Changes in cancer-related mortality during the COVID-19 pandemic in the United States. J Natl Cancer Inst. 2024;116(1):167-169. doi:10.1093/jnci/djad 191 37688577 · doi ↗ · pubmed ↗
- 8Janczewski LM, Cotler J, Merkow RP, . Alterations in cancer treatment during the first year of the COVID-19 pandemic in the US. JAMA Netw Open. 2023;6(10):e 2340148. doi:10.1001/jamanetworkopen.2023.40148 37902756 PMC 10616721 · doi ↗ · pubmed ↗