Two-Decade Trends in Colorectal Cancer and Cardiovascular Disease-Related Mortality in the US Adult Population, 1999 to 2019.

METHODS
Deaths occurring within the United States related to CVD and CRC were extracted using the Centers for Disease Control and Prevention Wide‐Ranging Online Data for Epidemiologic Research (CDC WONDER).
21
The Multiple Cause‐of‐Death Public Use record death certificates were studied to identify records in which both CVD and CRC were mentioned as either contributing or underlying causes of death on nationwide death certificates. CRC was identified with International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM), codes C18‐C20, and CVD was identified with ICD‐10‐CM codes I00‐I99 (diseases of the circulatory system) in individuals aged ≥25 years. This cutoff was selected because age‐adjusted mortality rates (AAMRs) are only available in 10‐year increments for the adult population in CDC WONDER; furthermore, mortality estimates for younger adults (<25 years) were sparse.
The CRC+CVD‐related deaths, population sizes, demographics (including sex, race, ethnicity, and age) and regional information (ie, urban–rural, state and census regions) were extracted from 1999 to 2019. Race and ethnicities were defined as non‐Hispanic White, non‐Hispanic Black or African American (will be referred to as “Black” in this paper), Hispanic or Latino (will refer to as “Hispanic”), non‐Hispanic American Indian or Alaskan Native (will refer to as “Indigenous American”), and non‐Hispanic Asian or Pacific Islander patients. These race and ethnicity categories were previously used within analyses from the CDC WONDER and relied on reported data on death certificates. Age groups were defined as 25 to 34, 35 to 44, 45 to 54, 55 to 64, 65 to 74, 75 to 84, and ≥85 years. For urban–rural classifications, the 2013 National Center for Health Statistics Urban–Rural Classification Scheme was used to divide the counties into metropolitan (ie, large central metropolitan, large fringe metropolitan, medium metropolitan, and small metropolitan) and nonmetropolitan (ie, micropolitan and noncore) categories.
22
Regions were classified into Northeast, Midwest, South, and West according to the Census Bureau definitions.
Crude mortality rates and AAMRs per 100 000 population were determined. Crude mortality rates were determined by dividing the number of CRC+CVD‐related deaths by the corresponding US population of that year. As previously described, AAMRs were calculated by standardizing the CRC+CVD‐related deaths to the year 2000 US standard population.
23
Data were extracted from the CDC WONDER platform into Microsoft Excel spreadsheets. The extracted files were formatted to ensure consistency in age, sex, and race and ethnicity categorizations. The cleaned data sets were then exported and uploaded into the Joinpoint Regression Program for statistical analysis (Joinpoint version 4.9.0.0, National Cancer Institute). The Joinpoint Regression Program was used to identify statistically significant changes in annual mortality trends over time.
24

This method fits a series of connected straight‐line segments (“linear splines”) to the data, with “joinpoints” (or nodes) representing statistically significant inflection points in the trend. The optimal number and location of joinpoints were determined automatically by the software using Monte Carlo permutation testing, with an overall significance level of 0.05. The model selected up to 3 joinpoints (ie, 4 segments) as the best fit for the 1999 to 2019 (ie, 21 years of data) study period.
Joinpoint regression models mortality rates on the log scale (ie, log of the age‐adjusted rate). This transformation stabilizes variance, as mortality rates are often skewed, and allows multiplicative changes in rates to be modeled as linear on the log scale. The slope of each fitted segment on the log scale was exponentiated and expressed as the annual percentage change (APC), representing the estimated yearly percent change in mortality. The final segmented trends represent periods where the APC was statistically homogeneous, meaning each segment reflects a period where mortality rates changed at a consistent yearly rate until a statistically significant change in slope (ie, a joinpoint) was detected.
These breakpoints were identified through permutation testing rather than chosen arbitrarily. In practical terms, Joinpoint regression identifies the calendar years where mortality trends significantly changed, rather than assuming a constant rate of change during the entire study period.
To summarize the observed mortality trend throughout the full study period, the weighted mean of the segment‐specific APCs was computed and presented as the average APC, with corresponding 95% CIs. Both APCs and average APCs were reported to increase or decrease if the slope depicting the alteration in mortality deviated significantly from zero, determined by the 2‐tailed test.
In addition to model‐derived APC slopes estimated from Joinpoint regression, we calculated lag‐1 APC values to capture short‐term fluctuations in the combined CRC and CVD‐related mortality trends. Lag‐1 APC was defined as the observed APC in AAMRs between 2 consecutive years, using the following formula:where
AAMR
t
=age‐adjusted mortality rate in year t.

AAMR
t−1=age‐adjusted mortality rate in the preceding year.

This metric provided short‐term year‐to‐year variability in mortality rates, complementing Joinpoint‐derived APC estimates that reflected smoothed long‐term trends. Statistical significance was set at a P value of ≤0.05. This analysis was performed using Python (version 3.10) for data processing and visualization.
Institutional review board approval was not applicable because this study used publicly available, aggregate, deidentified data from CDC WONDER
21
and did not involve individual human subjects. Analytic code (ie, Joinpoint project files) and nonidentifiable derived data that support the findings are available from the corresponding author on reasonable request. This study was reported per Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
25

RESULTS
Between 1999 and 2019, a total of 1, 303, 016 deaths related to CRC occurred in people aged ≥25 years in the United States. The AAMRs of CRC decreased from 38.2 in 1999 to 25.7 in 2013 (APC −3.0 [95% CI, −3.1 to −2.8]), followed by a slower but significant decrease to 22.8 by 2019 (APC −1.8 [95% CI, −2.3 to −1.0]) (Figure 1).
Of those with a CRC‐related death, a total of 394, 871 (31.8%) patients also had CVD (Table S1). The overall AAMR per 10000 of the entire cohort was 8.8 (95% CI, 8.7–8.8) (Table). The CRC+CVD‐related AAMR per 100 000 was 12.1 in 1999 and 6.7 in 2019. The CRC+CVD‐related AAMR was stable between 1999 and 2002 (APC of −2.0 [95% CI −3.3 to 0.4]), decreased from 2002 to 2015 (APC of −3.8 [95% CI, −5.1 to −3.6]), and then was stable from 2015 to 2019 (APC −0.8 [95% CI, −2.0 to 1.7]) (Figure 1; Figures S1 and S2). Trends in average APC stratified by sex, age, race and ethnicity, region, and metropolitan status are provided in Table S2.

Demographic Trends

Sex
The overall AAMR per 100 000 was higher in men (11.1 [95% CI, 11.0–11.1]) compared with women (7.2 [95% CI, 7.1–7.2]) (Figure 2; Table S3). The AAMR in men was 15.5 in 1999 and 8.6 in 2019. From 1999 to 2015, the AAMR decreased, but then stabilized from 2015 to 2019 (APC −0.5 [95% CI, −1.6 to 1.3]) (Table S3).

Race and Ethnicity
The overall AAMRs were highest for non‐Hispanic Black or African American individuals (12.2 [95% CI, 12.1–12.3]) followed by non‐Hispanic Whites (8.6 [95% CI, 8.6–8.7]), non‐Hispanic Indigenous Americans (7.6 [95% CI, 7.2–8.0]), and Hispanic or Latino (6.9 [95% CI, 6.8–7.0]), and the lowest for non‐Hispanic Asian or Pacific Island people (6.2 [95% CI, 6.1–6.3]) (Figure 3; Table S3).

Age
The overall CRC+CVD‐related crude mortality rates were highest among the elderly group aged ≥85 years (112.3 [95% CI, 111.7–112.9]), while those aged 25 to 34 years (0.1 [95% CI, 0.07–0.12]) had the lowest rates. For the 35‐ to 44‐year age group, the crude mortality rates increased annually by 1.8% from 1999 to 2019 (Figure 4; Table S4).

Geographical Patterns

Metropolitan Versus Nonmetropolitan
Overall, the AAMR related to CRC+CVD was higher in nonmetropolitan (9.8 [95% CI, 9.7–9.8]) counties than in metropolitan (8.6 [95% CI, 8.5–8.6]) counties. AAMR for metropolitan (12.1) versus nonmetropolitan (12.2) was similar in 1999, but this gap widened by 2019, with an AAMR of 6.3 in metropolitan areas versus 8.5 in nonmetropolitan areas (Figure 5; Table S5).

Statewide
Among the states, AAMR varied widely from 4.5 in Utah to 13.8 in New York (Figure 6; Table S6). States in the >90th percentile of mortality had AAMR that was >2‐fold higher than those in the bottom 10th percentile. States with AAMR >90th percentile were New York, Nebraska, Mississippi, West Virginia, and California, whereas states with AAMR <10th percentile were Idaho, New Mexico, Arizona, Utah, and Montana (Figure 6; Table S6).

Census Regions
Overall, the AAMR related to CRC+CVD was highest in the Northeast (10.6 [95% CI, 10.5–10.6]), followed by the West (9.0 [95% CI, 9.0–9.1]), Midwest (8.4 [95% CI, 8.4–8.5]), and South (7.9 [95% CI, 7.8–7.9]) (Figure 7; Table S7).

DISCUSSION
In this comprehensive nationwide analysis, we identified >1.3 million deaths attributable to CRC that occurred during the 2‐decade period from 1999 to 2019 in the United States. Among these fatalities, more than one‐quarter (31.8%) were also associated with CVD. Despite observing a decline in the overall AAMR related to the co‐occurrence of CRC and CVD until 2015, a stagnation in this trend was noted during the period from 2016 to 2019. Considerable disparities related to sex, race or ethnicity, and geographical location were observed in the combined CRC+CVD mortality, with higher AMMRs among men, non‐Hispanic Black or African American individuals, the elderly population aged ≥85 years, and in nonmetropolitan areas. Notably, a significantly rising trend in the combined mortality was documented among those aged 35 to 44 years.
To our knowledge, this is the first nationwide analysis to assess trends in CRC+CVD‐related mortality spanning >2 decades in the United States. This time period represents an evolution in the screening, treatment, and management of CRC, which has resulted in significantly reduced mortality rates. However, this improvement in survival was expected to increase the burden of noncancer‐related deaths. Our study provided valuable insights into the correlation between CRC+CVD‐related mortality. These findings warrant careful evaluation of risk factors and subsequent management of CVD in patients diagnosed with CRC.
The strong association of CRC and CVD mortality has been demonstrated in prior research. Several prior studies identified CVD as a common cause of noncancer‐related deaths in patients with CRC.
5
,
6
Zhang et al analyzed a total of 197, 699 CRC cases from the Surveillance, Epidemiology, and End Results (SEER) database and found that CVD accounted for at least 4 of 10 noncancer‐related mortalities.
26
Similarly, a cohort study of 563, 298 patients with CRC reported CVD as the leading noncancer cause of death.
11
The relationship between CRC and CVD has been postulated to result from a combination of several mechanisms. Chronic inflammation of the gut appears to contribute to the development of CRC.
27
Proinflammatory cytokines such as tumor necrosis factor α and interleukin 1β have been shown to play a crucial role in the progression of CRC.
28
Similarly, inflammatory processes are strongly implicated in CVD.
29
,
30
,
31

Some studies have suggested that exposure to chemotherapy for the treatment of CRC may contribute to increased risk of CVD, although the data are inconsistent. Fluorouracil and capecitabine are among the most commonly used chemotherapeutic agents for the treatment of CRC. These drugs have been demonstrated to exert cardiotoxic effects, although the severity of complications varies depending on factors such as dosage and preexisting CVD.
26
The occurrence of fluorouracil‐induced cardiotoxicity is reported to range from 1.2% to 18%.
32
Coronary vasospasm is considered the most noteworthy underlying mechanism for this side effect.
33
,
34
Other mechanisms mainly include endothelial injury that promotes vasoconstriction, accumulation of cardiotoxic by‐products leading to direct myocardial injury, and a procoagulant state.
35
Kenzik et al evaluated the incidence of new‐onset CVD in 72, 408 older adults with CRC and identified radiation therapy as an independent predictor of CVD (hazard ratio [HR], 1.18 [95% CI, 1.11–1.25]).
14
Zhang et al examined 197 699 patients with CRC and found chemotherapy to be an independent risk factor for the development of CVD among patients with CRC in their multivariate analysis.
26
In contrast, Koo et al analyzed a total of 412 patients in a retrospective cohort analysis and reported a lower risk of major adverse cardiovascular events in patients with CRC receiving chemotherapy compared with those not undergoing chemotherapy (HR, 0.37 [95% CI, 0.19–0.75]).
36

Some studies report an increased incidence of CVD‐related mortality in patients with cancer following radiotherapy,
37
,
38
while others demonstrate no association between the two.
39
,
40
It is noteworthy that radiotherapy is often reserved for advanced tumor stages. These patients are more likely to succumb to poor oncological outcomes rather than CVD‐related complications. Furthermore, both CRC and CVD share common risk factors that could potentially explain the correlation between the 2 conditions. Among these risk factors, smoking, obesity, and diabetes are correlated with both CVD and CRC.
41
,
42
,
43
,
44
,
45
,
46
Other factors, such as poor diet and a sedentary lifestyle, also contribute to the disease burden of both conditions.
47
These mechanisms have been illustrated in Figure 8.
We demonstrated a downward trend in overall mortality related to CRC+CVD from 2002 to 2015, followed by stagnation until 2019. A concerning contributing factor was the increasing burden in younger populations, with increasing mortality in adults aged 35 to 44 years throughout our study period and in adults aged 45 to 54 years from 2004 onwards. This was likely driven by individual increases in CRC and CVD incidence and mortality. Dharwadkar et al showed that, since the early 1990s, despite decreases in CRC incidence and mortality rates in patients >50 years, incidence rates in patients <50 years had nearly doubled.
48
Earlier work by Siegel et al reported that CRC mortality in patients aged 20 to 54 years had declined from 1970 to 2004, but then increased annually by 1% until 2014.
49
Din et al also demonstrated an increase in CRC rates among those aged <50 years and residing in North America likely driven by rising rates of rectal cancer.
50
The rising prevalence in younger populations may be a result of dietary patterns, the increasing rates of obesity and metabolic syndrome, sedentary behavior, alcohol and tobacco use, maternal obesity and in‐utero antibiotic exposure, and intestinal dysbiosis (ie, characterized as an imbalance of intestinal flora increasing the likelihood of potential infections and inflammation).
48
,
51
,
52
,
53
,
54

Furthermore, the younger population may be more likely to ignore subtle signs of CRC, which can lead to delayed diagnosis at a higher stage, and may be less likely to pursue routine medical care, leading to inadequate screening and treatment for CRC and CVD.
39
In light of the rising incidence of CRC in younger adults, the US Preventive Services Task Force updated its 2021 guidelines to reduce the recommended age for initiating screening colonoscopy from 50 to 45 years.
55
Further efforts to increase awareness among young adults, control modifiable risk factors, and research efforts to better understand the causes of the rising CRC+CVD burden in young adults can aid in reducing incidence and mortality.
Our study demonstrated that the highest mortality rates related to CRC+CVD were seen in non‐Hispanic Black individuals compared with all other races. The mortality rates were almost 2‐fold greater than those in Asian or Pacific islander populations, who exhibited the lowest mortality rates. White populations had the second‐highest mortality rates from CRC+CVD. The higher mortality from CVD in the Black population with CRC is consistent with prior literature, and has been previously attributed in large part to socioeconomic inequalities, dietary differences, tumor stage, and higher comorbidities.
28
,
56
,
57
,
58
Encouragingly, a notable reduction in the mortality gap between Black and White populations was observed in this study. This may be a result of the Affordable Care Act Providing Coverage for preventive screening,
59
advancements in medical technology and therapies,
60
,
61
and national efforts to promote awareness of the risk factors affecting CRC and CVD health.
62
Despite this progress, there remained significant disparities in the absolute rates of CRC+CVD‐related mortality between Black and White populations.
The Indigenous American population demonstrated the greatest year‐to‐year variability in CRC‐ and CVD‐related mortality rates compared with other ethnic groups. This variability is partly attributable to the relatively small number of annual deaths within this population, which can produce unstable rate estimates. Beyond these statistical considerations, Indigenous American communities face well‐documented disparities in access to cardiovascular care
63
and cancer screening.
64
Thus, these fluctuations should be interpreted cautiously, and additional research is warranted to better clarify the relationship between CRC‐ and CVD‐related mortality in this underserved group.
During the study period, mortality rates related to CRC+CVD decreased in both urban and rural areas; however, there were significantly higher mortality rates in nonmetropolitan regions. These findings were consistent with epidemiological data that showed higher mortality rates of both CVD and CRC individually in rural areas.
65
,
66
,
67
It is well established that disparities in access to CRC screening in rural populations attributable to socioeconomic inequities, distance to care, and inadequate health care coverage have led to higher CRC mortality in these areas.
68
,
69
,
70
Rural residents are also known to possess higher rates of shared CVD risk factors such as cigarette smoking, hypertension, and obesity.
71
Mortality rates related to CRC and CVC decreased among all geographic regions of the United States, but there was a greater decrease in the Northeast compared with other regions in more recent years. Specific states with the highest mortality rates for CRC and CVD mortality included New York, Nebraska, Mississippi, West Virginia, and California. The exact causes for these geographical differences remain unknown.

Future Directions
Our analysis reveals that the highest mortality burden exists among elderly individuals, with a concerning increase observed among younger populations, as well as among men, Black individuals, and residents of nonmetropolitan areas. This underscores the importance of targeted preventative strategies to address the combined burden of CRC and CVD in these populations. These disparities signal overlapping exposures to modifiable risk factors such as obesity, smoking, hypertension, and diabetes, which are known to contribute to both diseases. Of note, the persistence of higher mortality rates in rural and some minority communities suggests a lack of existing preventative strategies in these groups. Our findings also highlight the need for earlier and more aggressive cardiovascular risk assessment in patients with CRC, particularly among high‐risk subgroups. Integrating a cardio‐oncology framework at the time of CRC diagnosis may represent a promising approach. At the population level, initiatives to expand access to cancer screening and cardiovascular care in underserved groups, including Indigenous American communities, are warranted. Last, the interplay between systemic inflammation, risk factors, treatment‐related cardiotoxicity, and preexisting disease warrants individual‐level investigations conducted in diverse and high‐risk populations to assess variations among racial, geographical, and socioeconomic groups.

Limitations
While our study provides important insights into the CRC+CVD mortality trends during a period of 2 decades, it is important to address certain limitations in this analysis. First, the CDC WONDER primarily relied on death certificates as a data source, which may have resulted in inaccuracies in mortality data. Second, the use of ICD‐10‐CM codes may add to the misclassification bias. Third, our study did not account for the specific types of CVD‐related events, which could provide further clarity on the extent of the correlation between the 2 diseases. Fourth, while cancer therapy–related cardiotoxicity is a plausible contributor to the observed association, treatment‐specific information is not available in CDC WONDER. Thus, we were unable to directly evaluate the impact of chemotherapy or radiotherapy on CVD risk in this analysis. Last, since this is an ecological trend analysis based on aggregated mortality data, causal inferences between CRC and CVD cannot be established. These findings should therefore be interpreted with caution and warrant confirmation through individual‐level studies.

CONCLUSIONS
In conclusion, almost one‐third of all deaths related to CRC were also related to CVD in this study. Although advances in prevention and treatment have led to decreases in CRC+CVD‐related mortality, the trend has stagnated in recent years, with a greater disease burden seen in men, non‐Hispanic Black or African American individuals, and those living in nonmetropolitan areas. These findings suggest that following a CRC diagnosis, patients may benefit from increased efforts towards CVD risk assessment and prevention. Furthermore, emphasis on preventive efforts targeting shared risk factors such as smoking, obesity, and diet, especially in the younger population, may reduce the incidence of both conditions.

Sources of Funding
None.

Disclosures
Salim S. Virani has received research support from the National Institutes of Health, the National Institute for Health and Care Research (UK), USA, Department of Veterans Affairs USA, and Tahir and Jooma Family and Asharia Family. Martha Gulati is supported by contracts from the National Heart, Lung, and Blood Institutes numbers N01‐HV‐068161, N01‐HV‐068162, N01‐HV‐068163, N01‐HV‐068164, grants U01 HL064829, U01 HL649141, U01 HL649241, K23 HL105787, K23 HL125941, K23 HL127262, K23HL151867, T32 HL069751, R01 HL090957, R03 AG032631, R01 HL146158, R01 HL146158‐04S1, R01 HL124649, R01 HL153500, U54 AG065141, General Clinical Research Center grant MO1‐RR00425 from the National Center for Research Resources, the National Center for Advancing Translational Sciences Grant UL1TR000124, Department of Defense grant PR161603 (CDMRP‐DoD), and grants from the Gustavus and Louis Pfeiffer Research Foundation, Danville, NJ, The Women’s Guild of Cedars‐Sinai Medical Center, Los Angeles, CA, The Ladies Hospital Aid Society of Western Pennsylvania, Pittsburgh, PA, and QMED, Inc., Laurence Harbor, NJ, the Edythe L. Broad and the Constance Austin Women’s Heart Research Fellowships, Cedars‐Sinai Medical Center, Los Angeles, CA, the Barbra Streisand Women’s Cardiovascular Research and Education Program, Cedars‐Sinai Medical Center, Los Angeles, CA, the Society for Women’s Health Research, Washington, DC, the Linda Joy Pollin Women’s Heart Health Program, the Erika Glazer Women’s Heart Health Project, the Adelson Family Foundation, Cedars‐Sinai Medical Center, Los Angeles, CA, Robert NA. Winn Diversity in Clinical Trials Career Development Award (Winn CDA), and the Anita Dann Friedman Endowment in Women’s Cardiovascular Medicine & Research. Consultant fees/honoraria: Novartis, New Amsterdam and Medtronic Inc, unrelated to this work. The remaining authors have no disclosures to report. This work is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute, the National Institutes of Health, or the US Department of Health and Human Services.

Supporting information
Tables S1–S7
Figures S1–S2