Cardiovascular Disease Risk Among Cancer Survivors: The Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Background:

Over 80% of adult patients diagnosed with cancer survive long-term. Long-term complications of cancer and its therapies may increase the risk of cardiovascular disease (CVD), but prospective studies utilizing adjudicated cancer and CVD events are lacking.

Objectives:

Assess the risk of CVD in cancer survivors in a prospective community-based study.

Methods:

We included 12,414 ARIC participants. Cancer diagnoses were ascertained via linkage with state registries supplemented with medical records. Incident CVD outcomes were coronary heart disease (CHD), heart failure (HF), stroke, and a composite of these endpoints. We used multivariable Poisson and Cox regression to estimate the association of cancer with incident CVD.

Results:

Mean age was 54, 55% were female, and 25% Black. 3,250 (25%) participants had incident cancer over a median 13.6 years of follow-up. Age-adjusted IR of CVD (per 1,000 person-years) were 27.0 (24.7, 29.1) for cancer survivors and 12.0 (11.5, 12.4) for non-cancer controls. After adjustment for cardiovascular risk factors, cancer survivors had significantly higher risks of CVD (HR 1.37, 95% CI 1.26, 1.50), HF (HR 1.52, 95% CI 1.38, 1.68), and stroke (HR 1.22, 95% CI 1.03, 1.44), but not CHD (HR 1.11, 95% CI 0.97, 1.28). Breast, lung, colorectal, and hematologic/lymphatic cancers, but not prostate cancer, were significantly associated with CVD risk.

Conclusions:

Compared to persons without cancer, adult cancer survivors have significantly higher risk of CVD, especially HF, independent of traditional cardiovascular risk factors. There is an unmet need to define strategies for CVD prevention in this high-risk population.

Condensed abstract:

There is a growing number of cancer survivors in the general population. We used data from the prospective ARIC cohort to estimate the excess burden of CVD and individual CVD subtypes in cancer survivors compared to persons without prior cancer, adjusting for baseline and time-varying shared risk factors. We found that cancer survivors had 37% greater risk of incident CVD. The excess CVD risk was independent of shared risk factors, varied by primary cancer, and was predominantly driven by HF, with less robust associations between cancer and stroke and CHD. Cancer survivors will benefit from enhanced CVD prevention strategies.

Introduction

Cancer survivors are a rapidly growing population with specific health needs (1). It is estimated that 17 million adults living in the US are cancer survivors, representing 5% of the adult population, and this number is projected to increase (2). The growing number of cancer survivors results in part from an increase in cancer incidence due to aging of the population from advances in cancer screening, improvements in early detection, and therapeutics leading to significant improvements in cancer prognosis and long-term survival (3). However, patients diagnosed with cancer often have a high burden of chronic health conditions related to late effects of cancer and its treatments and are now living long enough such that non-cancer deaths are surpassing the risk of cancer-related deaths (4, 5).

Cardiovascular disease (CVD) is highly prevalent and the leading cause of death among some cancer survivors (1). CVD and cancer share numerous risk factors and pathophysiological mechanisms that may predispose patients to both conditions (6). Additionally, several cancer treatments may cause cardiac toxicity contributing to a higher risk of CVD in cancer survivors (7). Despite increasing recognition of a close link between cancer and CVD, few prospective studies have rigorously assessed the excess risk of CVD in cancer survivors, particularly for cancers with onset in adulthood.

Until recently, the published literature was limited to randomized clinical trials of highly selected patient populations with limited follow-up and lack of generalizability (8, 9). More recently, large observational studies have demonstrated an excess risk of CVD in cancer survivors compared to controls (5, 10–12). However, these studies have important limitations including retrospective designs, use of prescription claims and/or billing codes for classification of CVD, and variable quality of information on risk factors that may confound the associations of cancer and CVD (5, 10, 11, 13). Understanding the true excess burden of CVD and its subtypes in cancer survivors, as well as the degree to which CVD risk in this population is explained by shared risk factors, can inform clinical and public health strategies for CVD prevention in this unique patient population.

We undertook a prospective cohort analysis of data from the community-based, Atherosclerosis Risk in Communities (ARIC) Study to estimate the associations of adult cancer survivorship with incident CVD and CVD subtypes (coronary heart disease (CHD), stroke, and heart failure (HF)). We evaluated whether the burden of CVD related to cancer was independent of traditional CVD risk factors and whether associations differed by type of primary cancer.

Methods

Study Population

The ARIC Study is a prospective community-based cohort initiated in 1987 with the intent of studying the risk factors and natural history of CVD. A total of 15,792 participants from one of 4 US communities (Jackson, Mississippi; Washington County, Maryland; suburbs of Minneapolis, Minnesota; and Forsyth County, North Carolina) were enrolled at the initial study visit (1987–1989). Participants were aged 45–64 years at enrollment and predominantly Black or White adults. Participants were followed prospectively with continuous surveillance for incident CVD and serial study examinations that occurred every 3 years following baseline (1987–1989; 1990–1992; 1993–1995; and 1996–1998). ARIC visits 5–8 occurred in 2011–2013, 2016–2017, 2018–2019, and 2020. Additional study details have been previously published (14). All participants provided informed consent and the institutional review boards associated with all ARIC Study centers approved the study protocol.

Within ARIC, 15,641 participants consented to cancer research and were linked to cancer registries. From these, we excluded 139 participants with missing data on cancer; 910 participants with self-reported prevalent cancer at baseline (visit 1, 1987–1989), as information about these cases was limited; 1,430 participants with prevalent CVD at baseline; 88 participants who did not self-identify as Black or White race, as well as Black participants from the Minneapolis suburbs and Washington county centers (due to small numbers); and 804 participants missing information on the covariates of interest, leaving a total of 12,421 participants included in the main analyses. Of those, 3,250 (25%) participants developed cancer during ARIC follow up.

Ascertainment of Cancer Cases

Data on cancer cases occurring between 1987 and December 31^st, 2015 were obtained via linkage of the ARIC study with state cancer registries from Minnesota, North Carolina, Maryland, and Mississippi. These data were supplemented with information obtained directly from ARIC participants or their family members, hospital discharge summary codes, and review of medical records (15). Cancer survivors included in this analysis were those participants who had a diagnosis of a first primary invasive cancer (excluding those with non-melanoma skin cancer) during ARIC follow-up and who were free of CVD at the time of cancer diagnosis.

An expert panel adjudicated all cases of bladder, breast, colorectal, liver, lung, pancreatic, and prostate cancer (15). For adjudicated cases, when possible, stage at diagnosis was determined from the cancer registry or medical records using the pathologic TNM stage (tumor extent, lymph node involvement, presence of metastasis). When this was not available, staging was determined from the cancer registry or clinical TNM stage from cancer registry or medical records according to Surveillance, Epidemiology, and End Results (SEER) summary stage.

Ascertainment of CVD Events

Incident CVD was the main outcome of interest and was defined as a composite of incident CHD, stroke, or HF from baseline through December 31^st, 2015, as information on cancer cases is not yet available beyond this date. As secondary outcomes, we also considered incident CHD, stroke, and HF as individual endpoints. Participants were followed continuously for any possible CVD event via annual telephone calls, community-wide hospital surveillance, and linkage to state and national death indexes. Additional details of CVD surveillance in ARIC have been previously published (16, 17). An expert panel adjudicated all cases of CHD and stroke. Incident CHD was defined as a definite or probable non-fatal myocardial infarction, or definite fatal CHD. Stroke was defined as definite or probable ischemic or hemorrhagic stroke. HF was defined as the first hospitalization or death related to HF, using ICD-9 code 428 or ICD-10 code I-50 in the main analyses. We performed sensitivity analyses considering HF events based on ICD codes between baseline and December, 2004 and HF events adjudicated by an expert panel from January 2005 (start of ARIC HF adjudication) onwards (18).

Covariates of Interest

Information about participants’ demographics and alcohol drinking status (former, current, never) was obtained via questionnaire at visit 1. Information on all other covariates of interest was obtained at visit 1 and at each subsequent ARIC visit and included as time-varying covariates in regression models. Smoking status was categorized as former, current, or never, and additional information was obtained on number of cigarette-pack years consumed. All medications used in the prior 2 weeks were brought in by participants and recorded at each study visit. Body mass index (BMI) was calculated based on measured height and weight in kg/m². High-density lipoprotein (HDL) cholesterol and triglycerides were measured in serum using enzymatic assays, and low-density lipoprotein (LDL) cholesterol was calculated using the Friedewald equation (19). Hypertension was defined as a measured blood pressure ≥140/90 or use of antihypertensive medication. Diabetes mellitus was defined as a fasting blood glucose ≥ 126 mg/dl, non-fasting blood glucose ≥ 200 mg/dl, history of physician diagnosis, or current use of hypoglycemic medication. Estimated glomerular filtration rate (eGFR) was calculated based on measured creatinine using the CKD-EPI equation (20).

Statistical Analyses

We compared the Visit 1 (pre-cancer) characteristics of those individuals who did and did not develop incident cancer during follow-up, using the t-test for continuous variables and the chi square test for categorical variables.

We used survival analysis methods to estimate the association of the cancer with subsequent risk of CVD. In our main analyses, incident cancer was modeled as a time varying exposure. Participants without prior cancer contributed person time to the non-cancer group from Visit 1 until the development of CVD, incident cancer, censoring or death. Those with incident cancer contributed person time to the cancer group starting at the date of cancer diagnosis until development of CVD, censoring or death.

We used Poisson regression to calculate adjusted incidence rates of CVD and CVD subtypes for cancer survivors and for persons who did not develop cancer during follow up. We used Cox regression models to assess the association of cancer with incident CVD, CHD, stroke, and HF, with two levels of adjustment. Model 1 included demographics: age, sex, race-center and educational level. We used a second adjustment model with robust adjustment for covariates thought to be potential confounders of the associations of cancer with CVD. Model 2 included baseline age, sex, race-center, educational level, drinking status and non-steroidal anti-inflammatory drug use (including aspirin), as well as baseline and time-varying smoking status, smoking pack-years, BMI, LDL-cholesterol, HDL-cholesterol, triglycerides, use of lipid lowering medications, prevalent hypertension, prevalent diabetes mellitus, and eGFR. By including time-varying covariates in this model, we ensured CVD risk factors assessed at multiple ARIC visits were included in the adjustment model. These analyses were performed overall and stratified by race and sex, with tests for interaction with cancer survivorship status. For analyses stratified by sex, we created a category of non-sex-related cancers including all cancers that could have occurred in both men and women, and excluding breast, cervical, endometrial, ovarian, and prostate cancers.

We performed similar exploratory analyses considering the associations of the most common cancers with incident CVD and CVD subtypes. These included post-menopausal female breast cancer, prostate, lung, colorectal, and hematopoietic and lymphatic cancers.

We conducted multiple sensitivity analyses to confirm our findings. First, we repeated the main analyses excluding participants who died within one year of cancer diagnosis since these deaths most likely occurred as a result of the cancer itself or its treatments. Second, we created a nested matched cohort sample, where each participant who developed cancer was matched on sex, race, and age at the time of cancer diagnosis (±5 years) to two participants without a history of cancer. Participants were followed from time of matching until development of CVD, censoring, or death. If a control developed cancer after matching, follow up in the non-cancer group would be censored and from that point on they would become a cancer case. In these matched analyses, we used adjusted Cox regression to assess the association of cancer survivorship with incident CVD, CHD, stroke, or HF. This study design also allowed us to construct analyses stratified by date of matching (prior to 1995, 1995–2000, 2000–2005, 2005 onwards).

All covariates were selected a priori based on previously published literature. Analyses were performed using Stata SE 15.

Results

At visit 1, the mean age of the study population was 54 years, 55% were female, and 25% were Black adults. Participants who developed cancer during ARIC follow-up were older, more likely to be male, and less likely to be Black compared to participants who did not develop cancer (Table 1). They were also more likely to be current or former smokers and to have higher pack-years of smoking and lower HDL-cholesterol, but less likely to have diabetes, compared to persons who did not develop cancer. There were no significant differences in BMI, LDL-cholesterol, triglycerides, use of cholesterol lowering medication, prevalence of hypertension, or NSAID use between the two groups.

Table1.

Baseline (Visit 1) Characteristics of the Study Population by Incident Cancer During ARIC Follow-up^a

Variable	No Cancer (n=9,171)	Cancer (n=3,250)	P value
Age (yrs)	53.6 (5.7)	54.5 (5.7)	<0.001
Sex (%)
Female (%)	58.3	47.1	<0.001
Male (%)	41.7	52.9
Race (%)			0.005
Black	26.1	23.4
White	73.9	76.6
Education Level (%)			0.66
Less than high school	22.2	21.6
High school or vocational school	41.1	41.9
Some college education of higher	36.7	36.4
Cigarette smoking status (%)			<0.001
Current	24.0	28.8
Former	30.7	32.2
Never	45.3	39.0
Pack-years of smoking	276 (400)	354 (443)	<0.001
Drinking status (%)			<0.001
Current	56.0	60.5
Former	18.1	17.2
Never	26.0	22.3
Body mass index (kg/m2)	27.5 (5.3)	27.4 (5.0)	0.39
Lipids (mg/dl)
LDL-cholesterol	137.3 (39.7)	136.6 (36.9)	0.42
HDL-cholesterol	52.7 (17.0)	51.7 (16.8)	0.003
Triglycerides	122.2 (63.7	123.9 (62.4)	0.20
Cholesterol lowering medication (%)	2.2	2.5	0.48
Hypertension (%)	31.4	30.1	0.18
Diabetes mellitus (%)	10.6	7.8	<0.001
eGFR, mL/min/1.73 m2	103.2 (15.4)	102.2 (14.3)	0.002
NSAID use (%)	54.8	53.8	0.34

^a)Values are means (SD) or percentages.

A total of 3,250 (25%) participants free of CVD were diagnosed with a first primary cancer after visit 1, with a median time to cancer diagnosis of 13.6 years. Post-menopausal breast cancer was the most common cancer among women (35%), whereas prostate cancer was the most common cancer among men (40%). Lung, colorectal, and hematopoietic and lymphatic were the most common primary non-sex-related cancers (Table 2).

Table 2.

Distribution of First Primary Cancer in Women and Men

Cancer	Women (n=1,532)	Men (n=1,718)
Breast	529 (34.5)	-
Local	193 (12.6)	-
Regional	130 (8.5)	-
Distant	16 (1.0)	-
Stage unknown	190 (12.4)	-
Cervical and endometrial	121(7.9)	-
Ovarian	52 (3.4)	-
Prostate	-	681 (39.6)
Pathological or clinical TNM 1	-	34 (2.0)
Pathological or clinical TNM in (2A, 2B, 2C0)	-	414 (24.1)
Pathological or clinical TNM 3	-	59 (3.4)
Pathological TNM 4	-	31 (1.8)
Stage unknown	-	143 (8.3)
Lung	184 (12.0)	236 (13.7)
Local	33 (2.2)	39 (2.3)
Regional	43 (2.8)	58 (3.4)
Distant	54 (3.5)	67 (3.9)
Stage unknown	54 (3.5)	72 (4.2)
Colorectal	156 (10.2)	158 (9.2)
Local	43 (2.8)	42 (2.4)
Regional	44 (2.9)	47 (2.7)
Distant	21 (1.4)	15 (0.9)
Stage unknown	48 (3.1)	54 (3.1)
Hematopoietic and lymphatic	123 (8.0)	133 (7.7)
Renal	48 (3.1)	58 (3.4)
Bladder	38 (2.5)	116 (6.8)
Melanoma	38 (2.5)	58 (3.4)
Pancreatic	35 (2.3)	49 (2.9)
Stomach	18 (1.2)	26 (1.5)
Other digestive	28 (1.8)	41 (2.4)
Central nervous system	38 (2.5)	30 (1.8)
Thyroid	16 (1.0)	7 (0.4)
Other	91 (5.9)	268 (15.6)

The median follow-up time to CVD was 14 years (from visit 1 date) among those who never developed cancer and 5.2 years (from date of cancer diagnosis) among those who developed cancer. During the follow-up period there were 3,723 incident CVD events; 1,824 CHD, 1,162 strokes, and 2,665 HF events. Median times from cancer diagnosis to any CVD event were: 6.2 years for breast, 6.3 years for prostate, 1.3 years for lung, 5.1 years for colorectal, and 3.1 years for hematopoietic and lymphatic cancers.

Age-adjusted incidence rates (IR) of CVD per 1,000 person-years were 12.0 (95% CI 11.5, 12.4) for participants who did not develop cancer and 23.1 (95% CI 21.4, 25.1) person-years for all cancer survivors. After robust adjustment for potential confounders (Model 2), IR remained higher among those who developed cancer (17.4 (95% CI 14.8, 20.5) person-years) than among those without cancer (11.0 (95% CI 9.6, 12.7) person-years), resulting in an IR difference of 6.4 (95% CI 5.2, 7.8) person-years.

In analyses adjusted for demographics (Model 1), cancer survivors had 42% higher risk of developing incident CVD compared to those who did not develop cancer (HR 1.42, 95% CI 1.30, 1.56; Supplemental Table 1). Results were minimally attenuated and remained significant after robust adjustment for shared risk factors between cancer and CVD (Model 2: HR 1.37, 95% CI 1.26, 1.50; Figure 1 and Supplemental Table 1). When considering specific subtypes of CVD, overall cancer survivorship was significantly associated with incident HF (HR 1.52, 95% CI 1.38, 1.68) and stroke (HR 1.22, 95% CI 1.03, 1.44), but not with CHD (HR 1.11, 95% CI 0.97, 1.28) (Model 2; Table 3). Results were unchanged in sensitivity analyses considering adjudicated cases of HF (HR for the association of cancer survivorship with incident HF 1.58, 95% CI 1.43, 1.75).

Figure 1.

Association of Cancer Survivorship with Cardiovascular Disease, by Cancer Type Caption: Forest plot demonstrating the association of cancer and cancer subtypes with incident CVD, after adjustment for demographic variables and shared risk factors between cancer and CVD (Model 2).

a) Model 2: Adjusted for baseline age, sex, race-center, educational level, drinking status, and time-varying smoking status and cigarette pack-years, NSAID use (including aspirin), and time-varying, body mass index, LDL-cholesterol, HDL-cholesterol, triglycerides, use of lipid lowering medications, hypertension, diabetes mellitus and eGFR.

b) N = total number of participants included in the analyses, including cancer survivors and persons without cancer.

Table 3.

Associations (HR, 95% CI) of Cancer with Cardiovascular Disease Subtypes

		Coronary Heart Disease	Stroke	Heart Failure
Any cancer	Model 1	1.15 (1.01, 1.32)	1.24 (1.05, 1.46)	1.59 (1.44, 1.75)
	Model 2	1.11 (0.97, 1.28)	1.22 (1.03, 1.44)	1.52 (1.38, 1.68)
Breast cancer	Model 1	1.23 (0.90, 1.68)	0.82 (0.53, 1.25)	1.64 (1.33, 2.03)
	Model 2	1.21 (0.88, 1.65)	0.80 (0.52, 1.23)	1.58 (1.28, 1.95)
Prostate cancer	Model 1	0.96 (0.76, 1.22)	1.10 (0.82, 1.48)	1.05 (0.87, 1.28)
	Model 2	0.99 (0.78, 1.26)	1.12 (0.84, 1.51)	1.08 (0.89, 1.31)
Lung cancer	Model 1	2.08 (1.35, 3.20)	3.11 (1.99, 4.86)	3.78 (2.92, 4.90)
	Model 2	1.41 (0.91, 2.18)	2.40 (1.53, 3.78)	2.73 (2.10, 3.55)
Colorectal cancer	Model 1	1.17 (0.82, 1.69)	1.29 (0.83, 1.99)	1.32 (1.00, 1.75)
	Model 2	1.19 (0.83, 1.72)	1.28 (0.83, 1.98)	1.32 (1.00, 1.75)
Hematopoietic and lymphatic cancer	Model 1	1.74 (1.14, 2.65)	1.58 (0.93, 2.68)	3.05 (2.37, 3.93)
	Model 2	1.76 (1.15, 2.69)	1.60 (0.94, 2.71)	3.22 (2.51, 4.18)

^a)Model 1: Adjusted for baseline age, sex, race-center and educational level.

^b)Model 2: Adjusted for baseline age, sex, race-center, educational level, drinking status and NSAID use (including aspirin), and baseline and time-varying smoking status, cigarette pack-years, body mass index, LDL-cholesterol, HDL-cholesterol, triglycerides, use of lipid lowering medications, hypertension, diabetes mellitus and eGFR

Note: Analyses of breast cancer also adjust for hormone use in Model 2.

We did not find significant differences in the association of cancer and incident CVD by race (p-for-interaction = 0.76) (Table 4). There was a stronger association of survivorship from non-sex-related cancers (i.e., excluding breast, cervical, endometrial, ovarian, and prostate cancers) with incident CVD among women (HR 1.96, 95% CI 1.66, 2.31, Model 2) versus men (HR 1.57, 95% CI 1.35, 1.83, Model 2; p-for-interaction <0.01; Table 4).

Table 4.

Associations (HR, 95% CI) of Cancer with Cardiovascular Disease by Race and Sex

Cancer		Black	White	p Interaction
Any cancer	Model 1	1.44 (1.22, 1.71)	1.43 (1.29, 1.59)	0.26
	Model 2	1.40 (1.18, 1.66)	1.36 (1.23, 1.51)	0.76
Cancer		Women	Men	p Interaction
Non-sex-related cancer	Model 1	2.12 (1.80, 2.51)	1.69 (1.45, 1.97)	<0.01
	Model 2	1.96 (1.66, 2.31)	1.57 (1.35, 1.83)	<0.01

^a)Model 1: Adjusted for baseline age, sex, race-center and educational level.

^b)Model 2: Adjusted for baseline age, sex, race-center, educational level, drinking status and NSAID use (including aspirin), and baseline and time-varying smoking status, cigarette pack-years, body mass index, LDL-cholesterol, HDL-cholesterol, triglycerides, use of lipid lowering medications, hypertension, diabetes mellitus and eGFR

Age-adjusted incidence rates of CVD per 1,000 person years for survivors of specific cancers were: 16.6 (95% CI 13.7, 20.0) person-years for breast, 21.0 (95% CI 18.0, 24.6) person-years for prostate, 50.0 (95% CI 39.0, 64.2) person-years for lung, 25.4 (95% CI 20.1, 32.0) person-years for colorectal, and 41.0 (95% CI 32.1, 52.4) person-years for survivors of hematopoietic and lymphatic cancers. In fully adjusted analyses, we found that survivorship from breast, lung, colorectal, and hematopoietic and lymphatic cancers were each independently associated with incident CVD (Figure 1 and Supplemental Table 1). There was no significant association between prostate cancer and incident CVD.

We observed some differences in the associations of specific cancers with CVD subtype. In the fully adjusted analyses, increased HF risk was seen among survivors of breast (HR 1.58, 95% CI 1.28, 1.95), lung (HR 2.73, 95% CI 2.10, 3.55), colorectal (HR 1.32, 95% CI 1.00, 1.75), and hematopoietic and lymphatic (HR 3.22, 95% CI 2.49, 4.15), but not of prostate cancer (HR 1.08, 95% CI 0.89, 1.31), compared to persons without prior cancer (Table 3). Survivors of lung cancer had a more than two-fold increased risk of stroke than persons without cancer (HR 2.40, 95% CI 1.53–3.78; Table 3), but other cancers were not significantly associated with incident stroke. Only survivors of hematopoietic and lymphatic cancers were at significantly higher risk for CHD compared to persons without cancer (HR 1.76, 95%CI 1.15–2.69; Table 3).

Our results were not appreciably different in sensitivity analyses excluding participants who died within 1 year of cancer diagnosis, with a HR for incident CVD comparing cancer survivors to persons without cancer of 1.21 (95% CI 1.11, 1.33). We also found similar results for the associations of cancer survivorship with incident CVD, CHD, stroke, and HF when using a nested study population with matching of cancer survivors and individuals without cancer (Supplemental Table 2). However, in these analyses, we did not observe significant associations between breast or colorectal cancer survivorship and incident CVD (Supplemental Table 3). In analyses stratified by date of cancer diagnosis at the time of matching, we found that the association of cancer with incident CVD was similar for different time intervals of cancer diagnosis except for those with remote cancer diagnosis, prior to 1995, who had null associations (Table 5).

Table 5.

Associations of Cancer with Cardiovascular Disease in Nested Case-Control Sample, by Date of Cancer Diagnosis.

Date of Matching	HR (95% CI)
	Model 1	Model 2
<1995 (n=1,525)	1.13 (0.92, 1.39)	1.09 (0.89, 1.35)
1995-<2000 (n=1,565)	1.40 (1.14, 1.72)	1.33 (1.08, 1.63)
2000-<2005 (n=1,869)	1.39 (1.13, 1.69)	1.28 (1.04, 1.56)
≥2005 (n=2,726)	1.53 (1.23, 1.92)	1.44 (1.14, 1.80)

^a)Model 1: Adjusted for baseline educational level.

^b)Model 2: Adjusted for baseline educational level, drinking status and NSAID use (including aspirin), and baseline and time-varying smoking status, cigarette pack-years, body mass index, LDL-cholesterol, HDL-cholesterol, triglycerides, use of lipid lowering medications, hypertension, diabetes mellitus and eGFR

Discussion

In this prospective analysis of the community-based ARIC Study, we found strong independent associations of adult cancer survivorship with incident CVD and, in particular, with HF. After accounting for shared risk factors between cancer and CVD, cancer diagnosis was associated with an excess of 6.4 CVD cases per 1,000 person-years. Compared to persons without prior cancer, cancer survivors had a 37% higher risk of incident CVD and 52% higher risk of HF, and less strong associations with stroke or CHD. The risk of incident CVD and CVD subtypes varied by primary cancer type, with significant risk associations with for breast, lung, colorectal, and hematopoietic and lymphatic cancers, but no significant associations for prostate cancer. Importantly, the associations of cancer survivorship with CVD were largely unchanged between minimally adjusted analyses accounting for demographics and those robustly adjusted for traditional CVD risk factors, suggesting additional cancer-specific mechanisms likely contribute to the excess burden and risk of CVD in this population.

Few population-based studies have examined the risk of CVD in cancer survivors compared to controls, with important limitations. In a prior retrospective matched cohort study, Schoormans et al. found that only survivors of prostate and lung and trachea cancers had an increased risk of CVD compared to non-cancer controls (11). Despite inclusion of a large sample size, the study was limited by the use of drug dispensing information for the definition of CVD and confounding comorbidities. Two other large retrospective cohort studies have demonstrated variable associations between survivorship from different cancers and incident CVD (10, 13). Inferences from these studies are limited by retrospective design, variable information on shared risk factors between cancer and CVD, lack of CVD or cancer event adjudication, and variable follow up times. Additionally, most studies to date have had limited information on smoking, an important confounder of the associations of cancer with CVD. Our study adds to the literature by demonstrating strong associations between adult cancer survivorship and incident CVD in a diverse, prospective cohort, with continuously adjudicated cancer and CVD events, as well as detailed information on shared risk factors at multiple time points, for the first time. Our results are in line with a growing body of literature indicating that cancer survivors are a population at increased risk for CVD who may benefit from enhanced preventive measures.

While the mechanisms underlying the associations of cancer CVD are uncertain, several groups have speculated that shared cancer and CVD risk factors may be important contributors (6). Prior studies have found variable differences in the burden of CVD risk factors between cancer survivors and non-cancer patients. Using data from electronic health records, Armenian et al. found that cancer survivors were more likely to have hypertension, diabetes, dyslipidemia, excess weight, and a history of smoking than persons without cancer (10). Conversely, another large study using multiple linked electronic health records from the United Kingdom found only smoking, hypertension, and CKD were marginally more prevalent in cancer survivors (13). In the present study, aside from smoking, we did not find that persons who developed cancer had a significantly higher burden of pre-existing CVD risk factors. Differences in our findings may be explained by the timing of assessment of comorbidities which was done prior to as opposed to following cancer diagnosis as it is well established that the burden of CVD risk factors may increase following cancer diagnosis and treatment (21, 22). More importantly, our study benefited from direct assessment of these comorbidities rather than relying on data from electronic health records or drug dispensing information.

In our study, the associations of cancer with incident CVD were largely independent of traditional CVD risk factors. While CVD risk factors may mediate some of the observed associations observed, our findings suggest alternative mechanisms may contribute to the link between cancer and CVD (Central Illustration). Potential pathways involved in the cancer and CVD association include shared disease mechanisms such as systemic inflammation and oxidative stress (23, 24), a pro-inflammatory and prothrombotic state promoted by cancer itself (25), as well as cancer therapies. Variation in CVD risk across primary cancers suggest that the malignancy itself or cardiotoxicity from specific cancer treatments are likely central to CVD risk in this population (26). This is further supported by our findings of variable associations of cancer with CVD subtypes, with stronger associations with incident HF. For example, breast and hematopoietic and lymphatic cancers are typically managed with a combination of chemotherapy, often anthracycline-based, as well as chest radiation, both with well-established cardiotoxic potential (27). Similarly, chest radiation may be at least partly responsible for the increased risk of CHD in patients with hematopoietic and lymphatic cancers, that was not observed among other cancer survivors (28). Conversely, prostate cancer may be managed with active surveillance or local therapies without the risk of cardiotoxicity, which may explain why we did not observe an excess CVD risk in this subgroup (29). Prior epidemiologic studies evaluating the contributions of cancer therapies to CVD risk have been limited by factors such as retrospective design and limitations in assessments of cancer therapies and confounders. Additional studies are needed to elucidate the contribution of cancer therapies to the development of CVD in cancer survivors.

Central Illustration. Cardiovascular Disease Risk in Cancer Survivors and Proposed Pathways Linking Cancer and Cardiovascular Disease.

Cancer related pathways underlying the associations of cancer with CVD include but are not limited to common biologic predisposition; inflammation, oxidative stress, a pro-thrombotic state promoted by the cancer itself; and cardiotoxic effects of various cancer therapies. Non-cancer related pathways include lifestyle factors such as diet and physical activity, and other shared risk factors between cancer and CVD such as smoking, obesity and diabetes.

Our findings have important clinical and public health implications. CVD screening and prevention practices among cancer survivors are highly variable and often neglected due to limited evidence guiding its practice as well as misconceptions regarding competing risks of cancer mortality (30). In the present study, close to half of cancer survivors developed CVD following cancer diagnosis, indicating that this population would likely benefit from aggressive screening and preventive interventions. However, we also demonstrate that the links between cancer and CVD go above and beyond traditional risk factors. Therefore, while attention to shared risk factors between cancer and CVD is needed, our data suggests traditional risk assessment tools are likely to underestimate the risks in this population and risk factor modification alone is likely insufficient to fully address CVD risk in this population. Furthermore, it is important to consider the variable associations of specific cancer subtypes with CVD and CVD subtypes, with some subsets of adult cancer survivors having particularly high risk. Further studies are needed to inform screening and preventive strategies specific to this unique patient population.

Our study has important limitations to consider. First, its observational nature means that we cannot eliminate the possibility of residual confounding. Second, while CHD and stroke outcomes were adjudicated by an expert panel, HF events were defined based on ICD codes with the possibility of misclassification. However, a prior study has showed high specificity of this definition (18). Furthermore, our sensitivity analyses considering adjudicated cases from 2005 onwards demonstrated similar results. Third, even with over 12,000 middle-aged adults at baseline, our power to detect small to moderate associations, especially for specific cancer subtypes and demographic groups, was likely limited. Lack of power may also explain the lack of association between breast and colorectal cancer and CVD in our matched analyses. While we had limited information on cancer staging, which may influence cancer treatments, we did not have sufficient power for stratified analyses. Similarly, we did not have information on cancer treatments, which could be directly contributing to the observed variability in CVD risk across cancer subtypes. Lastly, we cannot exclude the possibility of increased medical care and CVD surveillance among cancer survivors increasing the likelihood of CVD detection.

Despite such limitations, our study is unique as a large, diverse, community-based cohort study of men and women, with long-term follow-up. Our study also benefited from comprehensive cardiovascular surveillance, adjudication of both cancer cases and cardiovascular events, the ability to distinguish CVD subtypes, and rigorous measurement of cardiovascular risk factors at multiple time points, allowing us to conduct time-varying adjustment. Furthermore, we were able to conduct time to event analyses due to validation of cancer diagnosis with exact timing.

In summary, we found cancer survivors have an increased risk of CVD, particularly HF, compared to persons without cancer, and that this excess risk is not explained by traditional CVD risk factors. Elucidating the mechanisms underlying the excess risk of CVD among adult cancer survivors, from treatment toxicities to shared biological pathways, is needed in order to define novel strategies for predicting and preventing CVD in this population.

Clinical perspectives:

Clinical competencies:

Overall cancer survivors have an increased risk of CVD compared to persons without prior cancer, independently of shared cancer and CVD risk factors. This excess risk is predominantly driven by HF, with less robust associations between cancer and stroke, and in particular, CHD. Some cancer survivors are at greater CVD risk than others. Cancer survivors should be considered a high-risk population and may benefit from enhanced CVD preventative interventions. While CVD risk factor modification should be emphasized, it may not be sufficient to address this excess risk.

Translational outlook:

Cancer survivors have an increased risk of CVD, particularly HF, compared to persons without cancer, independently of shared risk factors. Future studies are needed to elucidate the mechanisms underlying the excess risk of CVD among adult cancer survivors, from treatment toxicities to shared biological pathways.

Abbreviation list:

ARIC

Atherosclerosis Risk in Communities

BMI

Body mass index

CHD

coronary heart disease

CVD

cardiovascular disease

eGFR

estimated glomerular filtration rate

HDL

high-density lipoprotein

HF

heart failure

LDL

Low-density lipoprotein