Impact of COVID-19 on 2021 cancer incidence rates and potential rebound from 2020 decline

Abstract

The COVID-19 pandemic led to substantial declines in cancer incidence rates in 2020, likely because of disruptions in screening and diagnostic services. This study aimed to assess the impact of the pandemic on cancer incidence rates in the United States using 2021 incidence data from the Surveillance, Epidemiology, and End Results program. The analysis compared observed 2021 cancer incidence rates with expected prepandemic trends, evaluating changes by individual cancer site and stage. Although incidence overall and in many cancer sites the rates were close to prepandemic levels, they did not exhibit a recovery that incorporated the delayed diagnoses from 2020. There were exceptions, however, such as metastatic breast cancer, which showed significantly higher observed rates than expected (rate ratio = 1.09, 95% confidence interval = 1.04 to 1.13). Ongoing monitoring and targeted interventions are needed to address the long-term consequences of the COVID-19 pandemic on cancer care and outcomes.

The COVID-19 pandemic has substantially affected cancer diagnosis in the United States (1-6). A substantial decline in cancer incidence rates across multiple cancer types in 2020 was noted, coinciding with the start of the COVID-19 pandemic. This decline was attributed to disruptions in cancer screening, diagnostic services, and routine medical visits during the pandemic, not because of any issues with data reporting or registry operations (5,7). This distinction is important in understanding the true impact of the pandemic on cancer care and cancer incidence trends. The missed cancer diagnoses during the pandemic are likely to result in delayed diagnosis and more cases diagnosed with advanced disease in the coming years. Therefore, it is crucial to assess whether cancer rates have returned to prepandemic levels in the subsequent years. The aim of the study was to assess the impact of the COVID-19 pandemic on observed cancer incidence rates in the Surveillance, Epidemiology, and End Results (SEER) areas using the newly released 2021 incidence data (8), the second year of the pandemic. In this study, we assessed the impact of the COVID-19 pandemic on 2021 cancer incidence rates and potential rebound from the 2020 decline. The definition of a rebound in cancer incidence cases implies a recovery in cases following the observed decline in 2020 that surpasses prepandemic levels by including cases that were missed or diagnosed later because of disruptions to health-care access. In this article, we loosely define a rebound as a statistically significant increase in rates in 2021 compared with what would have been expected or estimated in the absence of the COVID-19 pandemic.

Data were obtained from 22 cancer registries in the SEER Program, representing about 48% of the US population (8). For this analysis, we selected all caners combined and 5 individual cancer sites: cancers with screening recommendations (female breast, prostate), cancer types detected mainly by symptoms (lung and bronchus [although lung cancer has screening guidelines, most lung cancers are still detected based on symptoms], pancreas), and cancers with frequent incidental detection (thyroid). Cancer stage was categorized as localized, regional, or distant using combined Summary Stage (9). Cancer incidence rates were calculated after accounting for reporting delay (10), but stage-specific reporting delay was not available for lung, pancreatic, or thyroid cancers. Incidence rates were age standardized to the 2000 US general population using SEER*Stat, version 8.4.3, software (11).

Next, we used Joinpoint Trend Analysis Software, version 5.1.0, (12) to estimate prepandemic trends by cancer site and stage. The Joinpoint model estimated the number and location of joinpoints and the log-linear trend between joinpoints. The Joinpoint models were fitted to 2000-2019 incidence rates to estimate prepandemic trends by cancer site (all stages) using a maximum of 3 joinpoints and to 2004-2019 rates to estimate trend by cancer site and stage (consistent staging is not available before 2004), allowing a maximum of 2 joinpoints. The last segment allows at least 3 data points (including the last joinpoint). The resulting trend across each calendar interval is described by the slope of the line segment. Using the model’s parameters (ie, estimated intercept and slope from the last segment), we projected the expected trends to 2020 and 2021 to represent respective expected prepandemic incidence rates. To determine whether rates have returned to prepandemic levels, we compared the 2021 observed and expected rates by computing the rate ratio for 2021 (2021 observed rate divided by 2021 expected rate) (see Table 1). The 95% confidence intervals (CIs) for rate ratios were presented, and Supplementary Material (available online) presents the formula for obtaining the 95% confidence interval for the rate ratio in 2021.

Table 1.

Cancer incidence rates in 2021, comparing observed age-adjusted rates with expected rates^a

Cancer site	2021 Expected age-adjusted rate	2021 Observed age-adjusted rate	Rate ratio (observed/expected)	Rate ratio 95% CI
Overall and by stage
All sites	459.06	458.33	1.00	0.97 to 1.03
Breast	140.49	137.42	0.98	0.87 to 1.09
Prostate	135.06	127.97	0.95	0.80 to 1.10
Lung and bronchus	49.40	46.93	0.95	0.94 to 0.96
Pancreas	14.25	13.92	0.98	0.95 to 0.99
Thyroid	14.57	13.62	0.93	0.85 to 1.02
By stage at diagnosis
Breast, localized	87.16	89.75	1.03	1.00 to 1.06
Breast, regional	36.31	36.30	1.00	0.86 to 1.13
Breast, distant	7.76	8.44	1.09	1.04 to 1.13
Prostate, localized	89.55	88.40	0.99	0.92 to 1.05
Prostate, regional	16.82	16.40	0.98	0.91 to 1.04
Prostate, distant	13.00	12.90	0.99	0.96 to 1.03
Lung and bronchus, localized	14.31	13.19	0.92	0.77 to 1.07
Lung and bronchus, regional	9.26	8.94	0.96	0.89 to 1.04
Lung and bronchus, distant	21.73	20.80	0.96	0.87 to 1.04
Pancreas, localized	2.81	2.41	0.86	0.78 to 0.94
Pancreas, regional	3.61	3.49	0.97	0.92 to 1.02
Pancreas, distant	6.50	6.54	1.01	0.99 to 1.03
Thyroid, localized	9.51	8.55	0.90	0.77 to 1.02
Thyroid, regional	4.21	3.99	0.95	0.87 to 1.03
Thyroid, distant	0.30	0.38	1.25	0.99 to 1.50

^aSee Supplemental Material (available online) for details on confidence interval (CI) calculations. See site recode definition for details on cancer site codes: https://seer.cancer.gov/siterecode/icdo3_dwhoheme/index.html.

Table 1 presents cancer incidence rates for all cancers combined and 5 individual cancer sites in 2021, comparing observed age-adjusted rates with expected rates, along with rate ratios and 95% confidence intervals. For overall and 5 individual cancer sites, the observed rate was slightly lower than the expected rate, but the difference was not statistically significant except for lung cancer (rate ratio = 0.95, 95% CI = 0.94 to 0.96) and pancreatic cancer (rate ratio = 0.98, 95% CI = 0.95 to 0.99) (also see Figure 1, A-F). When considering cancer sites by stage, localized breast cancer showed a slightly higher observed rate than the expected rate, but the rate was not significant (see Table 1 and Figure 1, G); regional stage was lower than expected as well, but not significantly (see Table 1 and Figure 1, H). Distant breast cancer showed a significantly higher observed rate than the expected rate in 2021 (rate ratio = 1.09, 95% CI = 1.04 to 1.13) (see Table 1 and Figure 1, I). Similarly, distant-stage pancreatic cancer showed an uptick in 2021 rates, although it was not statistically significant. For thyroid cancer, which is often detected incidentally during routine medical procedures, incidence rates experienced the most significant decline during the COVID-19 pandemic (1). When analyzing 2021 thyroid cancer rates by stage, localized and regional stage rates were still below the expected levels (see Table 1 and Figure 1, J and K). The incidence of distant-stage thyroid cancers, however, had the largest rate ratio but showed borderline statistical significance (see Table 1 and Figure 1, L). Stage-specific trends for prostate, lung, and pancreatic cancer are shown in Supplementary Figure 1 (available online).

Figure 1.

Observed and modeled cancer incidence trends during the prepandemic period (2000-2019), projected incidence trends in 2020 and 2021 in the absence of the pandemic, and pandemic-affected 2020 and 2021 observed incidence rates by selected cancer types and stage at diagnosis. SEER-22, 2000-2021. SEER = Surveillance, Epidemiology, and End Results.

Results for age-adjusted incidence rates are shown for all cancers combined (panel A) and 5 cancer sites: female breast cancer (panel B), prostate cancer (panel C), lung and bronchus cancer (panel D), pancreatic cancer (panel E), and thyroid cancer (panel E). We also show stage-specific, age-adjusted rates for breast cancer and thyroid cancer: female breast cancer, localized (panel G); female breast cancer, regional (panel H); female breast cancer, distant (panel I); thyroid cancer, localized (panel J); thyroid cancer, regional (panel K); and thyroid cancer, distant (panel L). Note that stage-specific rates are available from 2004 on, whereas overall rates for the 5 cancer sites are available from 2000 on. For each panel, we show age-adjusted incidence rates in blue (observed rates as squares, modeled rates as lines). The line segments of each curve were selected using the Joinpoint program, and a new joinpoint is detected when the line segments change. The dashed red lines are projected trends in 2020 and 2021 in the absence of the pandemic. Using the Joinpoint model’s parameters (ie, estimated intercept and slope from the last segment), we projected the expected trends in 2020 and 2021 in the absence of the pandemic.

Some limitations of the study are noted. Our 2021 incidence rate projection assumes that the trend estimated for the last segment of the joinpoint model before 2019 would have continued in the absence of the pandemic, which may not accurately reflect real-world complexities and potential changes in underlying factors affecting cancer incidence over time. Extrapolating trends beyond the observed data range increases uncertainty, as shown by the lack of statistical significance for 2021 rate ratios. Finally, when a joinpoint occurs toward the end of the data series (eg, joinpoint in 2017 for female breast cancer overall trend), caution should be used when interpreting these projected trends for 2020 and 2021, as they may be less reliable because of the limited data points.

In summary, our analysis indicates that overall incidence rates for all cancer sites and the 5 specific cancer sites examined in this study returned to prepandemic levels. They did not, however, display a rebound to account for potential delayed diagnoses in 2020. Disruptions caused by the pandemic to the health-care system leading to delays in cancer screening and diagnoses (1,2) resulted in a decline in reported cancer incidence rates in 2020 (1), which seem not to have fully recovered by 2021. Some exceptions were observed; for example, the increased incidence of distant-stage breast cancers in 2021 compared with expected rates indicates that the suspension of breast cancer screening during the pandemic and a potential rebound effect may have contributed to the higher observed rate of distant breast cancer in 2021 (13,14). Similarly, there was an uptick in distant-stage pancreatic cancer, although the change was not statistically significant. Although 2021 did not mark the peak of the pandemic, it remained a transition year affected by new variants and new waves of COVID-19 cases, which continued to affect states’ decisions about whether to maintain or reinstate regulations throughout the year. Consequently, it remains uncertain whether 2021 witnessed a complete return to pre-2020 levels of medical care. Continuous monitoring is necessary to address the long-term consequences of the COVID-19 pandemic on cancer incidence and outcomes.

Data availability

The authors cannot share the data or code used in this project directly. The data and statistical software used are publicly available through the National Cancer Institute’s SEER Program. For data, visit https://seer.cancer.gov/data/; for the SEER*Stat statistical software, visit https://seer.cancer.gov/seerstat/; and for the Joinpoint software, visit https://surveillance.cancer.gov/joinpoint/.

Author contributions

Nadia Howlader, MS (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Supervision; Writing—original draft; Writing—review & editing); Huann-Sheng Chen, PhD (Methodology; Writing—review & editing); Anne-Michelle Noone, PhD (Writing—review & editing); Daniel Miller, BA (Software); Jeffry Byrne, BS (Software); Serban Negoita, DrPH (Writing—review & editing); Kathy Cronin, PhD (Conceptualization; Writing—review & editing); Angela Mariotto, PhD (Conceptualization; Writing—review & editing).

Funding

This work was funded by the Surveillance Research Program, Division of Cancer Control and Population Sciences, in the National Cancer Institute.

Conflicts of interest

The authors declare no potential conflicts of interest.