Risk of Cardiovascular Events Among Patients With Head and Neck Cancer

Key Points

Question

What is the risk of cardiovascular (CV) events in patients with head and neck squamous cell carcinoma (HNSCC) and how are CV events associated with all-cause mortality?

Findings

In this cohort study of 35 897 US veterans with HNSCC, CV risk factors were highly prevalent (67% hypertension, 22% diabetes, 51% hyperlipidemia) and suboptimally controlled at time of cancer diagnosis. Ten-year cumulative incidence of stroke and myocardial infarction was 12.5% and 8.3%, respectively; incident stroke and myocardial infarction were associated with a 47% and 71% increased risk of all-cause mortality, respectively.

Meaning

The results of this cohort study suggest that patients with HNSCC are at a significant risk of CV events and CV risk mitigation is essential to survivorship care.

Abstract

Importance

Cardiovascular (CV) disease is a substantial cause of morbidity and mortality in cancer due to shared risk factors and exposure to potentially cardiotoxic cancer therapy. However, our understanding of CV risk in patients with head and neck squamous cell carcinoma (HNSCC) is limited.

Objective

To define CV risk profiles, incident stroke, myocardial infarction (MI), and mortality in patients with HNSCC.

Design, Setting, and Participants

This retrospective, population-based cohort study included 35 897 US veterans with newly diagnosed HNSCC from January 1, 2000, to December 31, 2020. Data were analyzed from May 2022 to January 2023.

Exposures

Demographic, cancer-specific, and treatment characteristics.

Main Outcomes

Prevalence of CV risk factors, medication use, and control at HNSCC diagnosis; cumulative incidence of stroke and MI; and all-cause death.

Results

Of 35 857 US veterans with HNSCC (median [IQR] age, 63 [58-69] years; 176 [0.5%] American Indian or Alaska Native, 57 [0.2%] Asian, 5321 [16.6%] Black, 207 [0.6%] Native Hawaiian or Other Pacific Islander, and 26 277 [82.0%] White individuals), there were high rates of former or current smoking (16 341 [83%]), hypertension (24 023 [67%]), diabetes (7988 [22%]), and hyperlipidemia (18 421 [51%]). Although most patients were taking risk-lowering medications, 15 941 (47%) had at least 1 uncontrolled CV risk factor. Black race was associated with increased risk of having uncontrolled CV risk factor(s) (relative risk, 1.06; 95% CI, 1.03-1.09), and patients with larynx cancer had higher rates of prevalent and uncontrolled risk factors compared with other cancer subsites. Considering death as a competing risk, the 10-year cumulative incidence of stroke and MI was 12.5% and 8.3%, respectively. In cause-specific hazards models, hypertension, diabetes, carotid artery stenosis, coronary artery disease, and presence of uncontrolled CV risk factor(s) were significantly associated with stroke and MI. In extended Cox models, incident stroke and MI were associated with a 47% (95% CI, 41%-54%) and 71% (95% CI, 63%-81%) increased risk of all-cause death, respectively.

Conclusion

The results of this cohort study suggest that in HNSCC, the burden of suboptimally controlled CV risk factors and incident risk of stroke and MI are substantial. Modifiable CV risk factors are associated with risk of adverse CV events, and these events are associated with a higher risk of death. These findings identify populations at risk and potentially underscore the importance of modifiable CV risk factor control and motivate strategies to reduce CV risk in HNSCC survivorship care.

This cohort study examines cardiovascular risk profiles, incident stroke, myocardial infarction, and mortality in US veterans with head and neck squamous cell carcinoma.

Introduction

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide and is diagnosed in 66 000 patients annually in the US.¹ With improved survival in HNSCC due to the increased incidence of more favorable human papillomavirus (HPV)–related cancers, as well as treatment advances, there has been increased attention to the risks of other chronic diseases.² These other illnesses account for more than half of the deaths after cancer treatment.^3,4 A Surveillance, Epidemiology, and End Results (SEER) study of more than 70 000 US patients with HNSCC showed that the risk of competing mortality increases over time and actually exceeds HNSCC-related mortality approximately 6 years after cancer diagnosis.⁵ In particular, multiple large population-based studies have shown that cardiovascular (CV) disease (CVD) is the leading noncancer cause of morbidity and mortality in patients with HNSCC.^3,4,5,6 Patients in the US who receive a diagnosis of HNSCC have a more than 3-fold higher incidence of developing CVD, including acute myocardial infarction (MI), during the first 2 years after diagnosis compared with matched individuals without cancer.⁷ A SEER-Medicare analysis of more than 6000 patients older than 65 years with HNSCC treated with radiotherapy from 1992 and 2002 reported up to a 34% 10-year incidence of cerebrovascular events.⁸ A large study from Korea found that the risks of MI and stroke were 38% and 48% higher, respectively, in patients with head and neck cancer compared with noncancer controls.⁹

The basis for elevated CV risk among patients with HNSCC is hypothesized to be secondary to (1) shared risk factors that predispose to a patient to cancer and CVD (including smoking, alcohol use, and older age^3,4,10) and (2) exposure to potentially cardiotoxic therapies as part of cancer treatment. US veterans are a particularly high-risk population, with disproportionately higher rates of alcohol and tobacco use compared with the general population,¹¹ both of which predispose patients to HNSCC in a dose-dependent fashion^12,13,14 through mechanisms, including DNA damage and field cancerization.¹⁵ Tobacco use has also been firmly established as a major contributor to CVD.⁴⁵

With the growing incidence of HPV-related oropharyngeal HNSCCs, which have a more favorable cancer prognosis, there has been a growing population of aging cancer survivors at risk for CVD. Although HPV-related cancers are traditionally associated with younger nonsmokers, there remain many patients with HPV-related cancer who have a substantial smoking history; these patients have worse outcomes than their nonsmoking counterparts.^16,17 In a US veteran–based study, 75% of patients with a diagnosis of HPV-positive oropharyngeal cancer from 2000 to 2012 had a more than 10 pack-year smoking history.¹⁸ Furthermore, HPV-related and unrelated cancers are treated with potentially cardiotoxic therapies, including radiation to the neck, cisplatin, and fluorouracil, all of which have been shown to exacerbate susceptibility to CV events, including stroke^{8,19,20,21,22} and MI.^23,24 In 1 study of patients with HNSCC postradiotherapy, HPV-positive status was associated with a more than 4 times increased risk of cerebrovascular events.²⁵ Thus, understanding and mitigating CV risk is essential to improving long-term survivorship care and outcomes in patients with HPV-positive and HPV-negative HNSCC.

Cancer treatment–related CV toxic effects and the competing risk of CVD in patients with HNSCC have been recognized for more than a decade, but baseline CV risk profiles and control, and the association of these factors with longitudinal CV events and mortality, have not been well characterized in the HNSCC population. Smaller studies have suggested that CV risk factors in patients with HNSCC are common and suboptimally managed. For example, a single-center study of 115 patients treated with radiotherapy found that 23% of patients had documented CVD before cancer diagnosis, of whom 59% were not receiving antiplatelet therapy and 30% were not receiving statin therapy.²⁶ In another single-center HNSCC cohort, older age and diabetes were associated with increased CVD risk.²⁷ Although prior seminal studies in cardio-oncology have shown that CV risk factors and cardiotoxic therapies, such as radiotherapy, are associated with major adverse CV events in patients with cancer,^28,29 some studies have suggested that traditional risk factors and risk prediction models may underperform in the HNSCC population.³⁰ As population-based studies of CV risk prevalence and control are lacking, evidence-based guidelines detailing the specific populations that may benefit most from CV screening, when and how to assess risk, and how best to mitigate CV risk do not exist for the HNSCC population.

To address this knowledge gap, we defined CV risk profiles and incident stroke and MI among a large population of US veterans with new diagnoses of HNSCC. We hypothesized that prevalent and suboptimally controlled CV risk factors, as well as exposure to cardiotoxic chemotherapy and radiotherapy, would be associated with stroke and MI risk. We further hypothesized that these incident CV events would be associated with an increased all-cause mortality risk.

Methods

Data Source and Study Population

The electronic health record data of US veterans with a new diagnosis of HNSCC within the Veterans Health Administration from 2000 to 2020 were obtained and accessed through the Corporate Data Warehouse (CDW). We included all complete analytic cases of HNSCC from 2000 to 2020 from the VA central cancer registry, which compiles and stores data on cancer diagnosis and treatment from VA medical centers across the US.³¹ Patients with distant metastatic disease at diagnosis were excluded. This research protocol was approved by the institutional review boards at the VA medical centers of Philadelphia, Pennsylvania, and Salt Lake City, Utah. Data were collected with a waiver of informed consent and Health Insurance Portability and Accountability Act authorization.

Covariates

Patient-level information, including age, sex, race, smoking status, Eastern Cooperative Oncology Group (ECOG) performance status, and baseline comorbidities, was obtained from tables within the CDW. Cancer-related variables, including diagnosis year and tumor (T) and nodal (N) stage, were obtained from cancer registry data. The primary site of cancer was classified using a combination of site, diagnosis, and procedure codes (eTable 1 in Supplement 1).^21,32 Human papillomavirus status was not available in the data set. Exposure to radiation and chemotherapy as part of initial cancer treatment was obtained from the VA central cancer registry, medication tables, and administrative billing codes within CDW using a 6-month postdiagnosis window (eTable 1 in Supplement 1).

Baseline CV Risk Factors, Treatment, and Control

For all baseline measures, the index date was the date of first HNSCC diagnosis. Presence of baseline CV risk factors and prevalent stroke or MI were defined using International Classification of Diseases, Ninth Revision (ICD-9) and ICD-10 coding definitions (≥2 outpatient or ≥1 inpatient diagnosis codes any time before 6 months after HNSCC diagnosis), which were adapted from prior VA-based studies (eTable 1 in Supplement 1).^33,34,35 Laboratory, vital sign, and medication data were obtained from VA data tables. Cardiovascular risk–reducing medication use was defined as 1 or more recorded administrations from 1 year before 6 months after HNSCC diagnosis. Uncontrolled hypertension was defined as a systolic blood pressure of 140 mm Hg or more or diastolic blood pressure of 90 mm Hg or more using the median of values recorded during the year prior to diagnosis date. Uncontrolled cholesterol was measured as a low-density lipoprotein cholesterol level of more than 130 mg/dL (to convert to mmol/L, multiply by 0.0259) or total cholesterol level of more than 240 mg/dL using the closest value before and within 3 years of the diagnosis date. Uncontrolled blood glucose levels were measured as a hemoglobin A_1C level of 7% or greater (to convert to the proportion of total hemoglobin, multiply by 0.01) using the closest value before and within 3 years of diagnosis date (eTable 1 in Supplement 1). Uncontrolled CV risk factors were defined as having either uncontrolled blood pressure or cholesterol and/or glucose levels. These thresholds were adapted from prior work and chosen to reflect relatively lenient thresholds that may be appropriate targets for an older cancer population (for example, as opposed to a hemoglobin A_1C target of 6.5% or systolic blood pressure less than 130 mmHg).³⁵

Outcomes: Incident Risk of Stroke and MI and Overall Survival

Stroke-free and MI-free survival were defined as the time from HNSCC diagnosis to first occurrence of stroke or MI, respectively. Incident stroke and MI were defined as 2 or more outpatient diagnosis codes or 1 or more inpatient diagnosis codes, with the first date of diagnosis used (eTable 1 in Supplement 1).^36,37 Death was ascertained from vital status data and other tables within the CDW and considered as a competing risk in the incident stroke/MI analysis. Patients without documented stroke/MI or death were censored at the time of loss to follow-up. Patients with prevalent stroke and MI before their cancer diagnosis date were excluded from respective incident event analyses. Overall survival was defined as the time from cancer diagnosis to all-cause death, which was censored at time of loss to follow up.

Statistical Analysis

We calculated the prevalence of CV risk factors, prior MI/stroke, and rates of medication use and CV risk factor control in the entire cohort and by primary cancer site. We performed a log binomial regression to examine associations, as defined by relative risk (RR), between patient factors (age < or ≥65, sex, race, year of diagnosis, primary tumor site, T/N stage, preexisting coronary artery disease) and the probability of having uncontrolled CV risk factor(s) at baseline.

To quantify the risk of incident stroke and MI, we estimated (1) the cumulative incidence of stroke with death as a competing event and (2) the cumulative incidence of MI with death as a competing event. We calculated 5-year and 10-year cumulative incidence estimates with 95% CIs.

Next, we assessed the association between baseline covariates and risk of incident stroke and MI using cause-specific hazard models, as the primary interest was in the etiologic associations between risk factors and CV events. These were summarized using cause-specific hazard ratios (csHRs). The following covariates were included in the multivariable model: age, sex, race, year of diagnosis, smoking status, primary cancer site, ECOG performance status, and T and N stage; baseline hypertension, diabetes, hyperlipidemia, carotid artery stenosis, coronary artery disease, kidney disease, and atrial fibrillation; receipt of radiotherapy; receipt of chemotherapy; and having 1 or more uncontrolled CV risk factors at baseline. These covariates were selected a priori based on clinical judgment as relevant to CVD. Multiple imputation using chained imputations was performed for variables with missing data to minimize bias within at-random missingness conditions.³⁸ Results were combined across imputations using Rubin rules. A sensitivity analysis was performed that was restricted to patients with complete data across all covariates. An additional sensitivity analysis using a Fine and Gray competing risks approach and subdistribution hazard ratio was also explored.

Finally, to quantify the association between incident stroke/MI and risk of subsequent death, we used a Cox regression with overall survival as the outcome and stroke or MI as a time-varying covariate. This model was adjusted for age, sex, race, diagnosis year, smoking status, primary site, ECOG performance status, T and N stage, hypertension, diabetes, hyperlipidemia, carotid artery stenosis, coronary artery disease, kidney disease, and atrial fibrillation. The hazard ratio for death associated with the stroke/MI time-varying covariate reflected the association between the occurrence of stroke/MI (at any point from the cancer diagnosis date onward) and overall risk of death. It was a comparison in risk of death between patients who had stroke or MI (1) vs patients who did not (0) at any time in follow-up. Stata, version 17 (StataCorp) was used to conduct all analyses. Statistical significance was defined as 2-sided P < .05.

Results

Baseline Characteristics and CV Risk

Population

The study cohort comprised 35 857 patients with nonmetastatic HNSCC. The median (IQR) age was 63 (58-69) years, 35 477 (98.9%) were male, and 5321 (16.6%) self-identified as Black or African American (Table 1). The most common primary cancer sites were the oropharynx, larynx, and oral cavity; 16 341 patients with recorded smoking status (83.2%) had a history of smoking. As part of their initial cancer treatment (within 6 months of diagnosis), 24 039 patients (67.0%) received radiotherapy, and 12 116 (33.8%) received chemotherapy (58.2% cisplatin, 25.5% carboplatin, and 16.3% cetuximab).

Table 1. Baseline Characteristics.

Variable (n = 35 857)	No. (%)a
Age, median (IQR), y	63 (58-69)
Male	35 477 (98.9)
Race
American Indian or Alaska Native	176 (0.5)
Asian	57 (0.2)
Black or African American	5321 (16.6)
Native Hawaiian or Other Pacific Islander	207 (0.6)
White	26 277 (82.0)
Missing	3819
Year of diagnosis
2000-2005	9548 (26.6)
2006-2010	10 403 (29.0)
2011-2015	9991 (27.9)
2016-2020	5915
Smoking history
Current smoker	13 486 (68.7)
Former smoker	2855 (14.5)
Never smoker	3300 (16.8)
Missing	16 216
Primary site
Oropharynx	12 970 (36.2)
Larynx	12 724 (35.5)
Oral cavity	7023 (19.6)
Hypopharynx	1896 (5.3)
Nasal cavity/sinus	754 (2.1)
Nasopharynx	490 (1.4)
T stage
0	3171 (9.7)
1	8293 (25.2)
2	9131 (27.8)
3	5822 (17.7)
4	6432 (19.6)
Missing	3008
N stage
0	16 893 (51.4)
1	3768 (11.5)
2	11 132 (33.9)
3	1081 (3.3)
Missing	2983
ECOG performance status
0	5945 (46.5)
1	5340 (41.7)
≥2	1509 (11.8)
Missing	23 063
Charlson Comorbidity Index score
0	10 676 (29.8)
1	7593 (21.2)
≥2	17 588 (49.1)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; N, node; T, tumor.

^a No. (percentage of nonmissing data) for categorical variables; median (IQR) for continuous variables.

Baseline CV Risk

Within the cohort, there were substantial rates of prevalent hypertension (24 023 [67.0%]), diabetes (7988 [22.3%]), and hyperlipidemia (18 421 [51.4%]) (Table 2). Patients with larynx cancer had slightly higher rates of CV risk factors than those with cancers of other primary sites. Black patients had higher rates of hypertension (73.5% vs 66.8%) and diabetes (24.1% vs 22.4%) but not other CV risk factors compared with patients of other racial and ethnic groups. Overall, 2430 patients (6.8%) had prevalent MI, and 4042 (11.3%) had prevalent stroke (occurring before their HNSCC diagnosis); these patients were excluded from analyses of incident MI and stroke, respectively.

Table 2. Proportion of Baseline Risk Factors, and Risk Factor Medication Use and Control, Overall and by Primary Site.

Characteristic	No. (%)a
	Full population (n = 35 857)	By primary cancer site
		Oropharynx (n = 12 970)	Larynx (n = 12 724)	Oral cavity (n = 7023)	Other (n = 3140)b
Hypertension	24 023 (67.0)	8369 (64.5)	8783 (69.0)	4762 (67.8)	2109 (67.2)
Receiving antihypertensive medicationc	20 344/24 023 (84.7)	6960/8369 (83.2)	7616/8783 (86.7)	3991/4762 (83.8)	1777/2109 (84.3)
Uncontrolled (SBP >140)d	10 424/32 830 (31.8)	3524/11 811 (29.8)	3922/11 965 (32.8)	2029/6169 (32.9)	949/2885 (32.9)
Diabetes	7988 (22.3)	2950 (22.7)	2960 (23.3)	1448 (20.6)	630 (20.1)
Receiving glucose-lowering medicationc	6161/7988 (77.1)	2283/2950 (77.4)	2313/2960 (78.1)	1111/1448 (76.7)	454/630 (72.1)
Uncontrolled (HbA1C >7)d	2904/19 011 (15.3)	1100/7152 (15.4)	1085/6815 (15.9)	498/3492 (14.3)	221/1552 (14.2)
Hyperlipidemia	18 421 (51.4)	6891 (53.1)	6741 (53.0)	3282 (46.7)	1507 (48.0)
Receiving lipid-lowering medicationc	13 492/18 421 (73.2)	4905/6891 (71.2)	5154/6741 (76.5)	2338/3282 (71.2)	1095/1507 (72.7)
Uncontrolled (LDL >130)d	5821/29 298 (19.9)	2269/10 688 (21.2)	2035/10 599 (19.2)	1027/5510 (18.6)	490/2501 (19.6)
≥1 CVRF uncontrolledc	15 941/33 351 (47.8)	5721/12 009 (47.6)	5836/12 078 (48.3)	2983/6337 (47.1)	1401/2927 (47.9)
Current or former smokere	16 341/19 641 (83.2)	5386/6990 (77.1)	6338/7205 (88.0)	3208/3760 (85.3)	1409/1686 (83.6)
Prevalent MI	2430 (6.8)	884 (6.8)	931 (7.3)	416 (5.9)	199 (6.3)
Prevalent stroke	4042 (11.3)	1301 (10.0)	1572 (12.4)	806 (11.5)	363 (11.6)
Carotid artery stenosis	1833 (5.1)	570 (4.4)	718 (5.6)	388 (5.5)	157 (5.0)
Coronary artery disease	9191 (25.6)	3041 (23.4)	3629 (28.5)	1750 (24.9)	771 (24.6)
Kidney disease	2045 (5.7)	705 (5.4)	771 (6.1)	395 (5.6)	174 (5.5)
Atrial fibrillation	7,64 (2.1)	294 (2.3)	237 (1.9)	165 (2.3)	68 (2.2)
Received radiotherapy	24 039 (67.0)	9923 (76.5)	8811 (69.2)	3062 (43.6)	2243 (71.4)
Received chemotherapy	12 116 (33.8)	6580 (50.7)	2859 (22.5)	1317 (18.8)	1360 (43.3)

Abbreviations: CVRF, cardiovascular risk factors; HbA_1c, hemoglobin A_1c; LDL, low-density lipoprotein cholesterol; MI, myocardial infarction; SBP, systolic blood pressure.

^a Percentages are out of the population number in the header unless another denominator was noted.

^b Hypopharynx, nasal cavity/sinus, nasopharynx.

^c Denominator includes patients with prevalent risk factor (diabetes, hypertension, and hyperlipidemia).

^d Denominator includes all patients with at least 1 recorded blood pressure in the year (hypertension) or HbA_1c (diabetes) or LDL (hyperlipidemia) level during the 3 years before cancer diagnosis. Detailed definition criteria provided in the Methods and eTable 1 in Supplement 1.

^e Denominator includes all patients with recorded smoking status.

Among those with prevalent CV risk factors, most were prescribed risk-lowering medications: 20 344 patients (84.7%) with hypertension were taking antihypertensive medication, 13 492 (73.2%) with hyperlipidemia were taking lipid-lowering medication, and 6161 (77.1%) with diabetes were taking antihyperglycemic medication. However, rates of suboptimal risk factor control were high: of patients with recorded data, 10 424 (31.8%) had uncontrolled blood pressure, 5821 (19.9%) had uncontrolled lipids, and 2904 (15.3%) had uncontrolled glucose levels. Overall, 15 941 patients (47.8%) with recorded data had at least 1 uncontrolled CV risk factor (Table 2).

On multivariable log binomial regression, being 65 years or older at diagnosis (RR, 1.07; 95% CI, 1.04-1.09) and Black race (RR, 1.06; 95% CI, 1.03-1.09) were each associated with a higher likelihood of having uncontrolled baseline CV risk factor(s). Higher T and N stage was modestly associated with a lower likelihood of uncontrolled CV risk factor(s), while the presence of coronary artery disease was not associated with likelihood of CV control (eTable 2 in Supplement 1).

Stroke/MI Incidence and Associations With Risk Factors and Death

Among patients without a history of stroke, the cumulative incidence of stroke was 8.0% (95% CI, 7.7%-8.4%) at 5 years and 12.5% (95% CI, 12.1%-12.9%) at 10 years when considering death as the competing risk (Figure 1). Among patients without a history of MI, the cumulative incidence of MI was 5.4% (95% CI, 5.1%-5.6%) at 5 years and 8.3% (95% CI, 8.0%-8.6%) at 10 years (Figure 2).

Figure 1. Cumulative Incidence of Stroke.

Accounting for competing risk of death, over time from cancer diagnosis.

Figure 2. Cumulative Incidence of Myocardial Infarction.

Accounting for competing risk of death, over time from cancer diagnosis.

After adjusting for patient-specific (age, sex, race, and ECOG performance status) and cancer-specific (primary site, year of diagnosis, and T and N stage) covariates, factors significantly associated with risk of incident stroke included current smoking (csHR, 1.16; 95% CI, 1.04-1.30), hypertension (csHR, 1.39; 95% CI, 1.29-1.50), diabetes (csHR, 1.15; 95% CI, 1.06-1.25), carotid artery stenosis (csHR, 22.18; 95% CI, 19.10-25.75), coronary artery disease (csHR, 1.24; 95% CI, 1.14-1.34), and kidney disease (csHR, 1.31; 95% CI, 1.12-1.53). Receipt of radiotherapy and chemotherapy as part of HNSCC treatment was associated with a 1.16-fold (95% CI, 1.07-1.26) and 1.18-fold (95% CI, 1.08-1.28) adjusted risk of stroke, respectively. Having at least 1 uncontrolled CV risk factor at HNSCC diagnosis was associated with 1.09-fold (95% CI, 1.02-1.17) adjusted risk of stroke (Table 3).

Table 3. Factors Associated With Incident Stroke and MI.

Variablea	csHR (95% CI)
	Stroke (n = 31 815)b	MI (n = 33 427)b
Age	1.02 (1.01-1.02)	1.01 (1.00-1.01)
Male (reference: female)	0.93 (0.68-1.29)	2.50 (1.34-4.67)
Black race	0.93 (0.85-1.03)	0.87 (0.78-0.98)
Year of diagnosis	0.96 (0.95-0.96)	1.03 (1.02-1.04)
Smoking status (reference: never smoker)
Current	1.16 (1.04-1.30)	1.11 (0.98-1.26)
Former	1.12 (0.97-1.30)	0.96 (0.80-1.15)
Primary site (reference: oropharynx)
Hypopharynx	1.13 (0.96-1.34)	1.30 (1.07-1.58)
Larynx	1.07 (0.98-1.17)	1.23 (1.11-1.37)
Nasal cavity/sinus	1.14 (0.91-1.44)	1.04 (0.78-1.38)
Nasopharynx	1.39 (1.08-1.81)	1.07 (0.75-1.54)
Oral cavity	1.08 (0.97-1.20)	1.21 (1.07-1.37)
ECOG performance status	1.04 (0.97-1.12)	1.03 (0.95-1.12)
Clinical T stage (reference: T1)
T2	1.04 (0.96-1.13)	1.07 (0.97-1.18)
T3	1.08 (0.98-1.20)	1.11 (0.98-1.25)
T4	1.01 (0.90-1.13)	1.31 (1.14-1.49)
Clinical N stage (reference: N0)
N1	0.97 (0.86-1.09)	0.91 (0.79-1.04)
N2	1.08 (0.98-1.19)	0.93 (0.83-1.04)
N3	1.18 (0.95-1.47)	1.02 (0.77-1.36)
Hypertension	1.39 (1.29-1.50)	1.29 (1.17-1.42)
Diabetes	1.15 (1.06-1.25)	1.27 (1.16-1.38)
Hyperlipidemia	1.00 (0.93-1.07)	0.98 (0.90-1.07)
Carotid artery stenosis	22.18 (19.10-25.75)	1.31 (1.14-1.52)
Coronary artery disease	1.24 (1.14-1.34)	4.04 (3.72-4.39)
Kidney disease	1.31 (1.12-1.53)	1.50 (1.30-1.74)
Atrial fibrillation	1.03 (0.78-1.36)	1.51 (1.19-1.93)
Received radiation therapy	1.16 (1.07-1.26)	1.02 (0.93-1.11)
Received chemotherapy	1.18 (1.08-1.28)	1.10 (1.00-1.22)
Uncontrolled CVRF(s)	1.09 (1.02-1.17)	1.19 (1.11-1.29)

Abbreviations: csHR, cause-specific hazard ratio; CVRF, cardiovascular risk factor; ECOG, Eastern Cooperative Oncology group; MI, myocardial infarction; N, node; T, tumor.

^a Variables listed include all covariates included in the multivariable model on multiply-imputed data.

^b Includes patients without prevalent stroke or MI, respectively.

For MI, statistically significant risk factors identified in cause-specific hazard models were hypertension (csHR, 1.29; 95% CI, 1.17-1.42), diabetes (csHR, 1.27; 95% CI, 1.16-1.38), carotid artery stenosis (csHR, 1.31; 95% CI, 1.14-1.52), coronary artery disease (csHR, 4.04; 95% CI, 3.72-4.39), kidney disease (csHR, 1.50; 95% CI, 1.30-1.74), and atrial fibrillation (csHR, 1.51; 95% CI, 1.19-1.93). Smoking and receipt of radiotherapy and chemotherapy were each not independently associated with risk of MI. Having 1 or more uncontrolled CV risk factors was associated with a 1.19-fold (95% CI, 1.10-1.29) risk of MI (Table 3). Compared with oropharynx cancer, larynx, oral cavity, and hypopharynx primary site were all associated with a higher risk of MI.

A sensitivity analysis using nonimputed data (complete case analysis) yielded similar associations. Diabetes was associated with stroke and MI; hypertension, carotid artery stenosis, and smoking were more strongly associated with stroke, whereas coronary artery disease, atrial fibrillation, and kidney disease were more strongly associated with MI. Uncontrolled CV risk factor(s) remained a significant risk factor for stroke and MI in sensitivity analyses (eTable 3 in Supplement 1). A sensitivity analysis using a Fine and Gray approach also yielded similar results (eTable 3 in Supplement 1).

Modeling incident stroke as a time-varying covariate in an extended Cox regression model for overall survival, incident stroke was associated with a 47% (95% CI, 41%-54%) increase in the risk of death in a multivariable-adjusted analysis. Incident MI was associated with a 71% (95% CI, 63%-81%) increase in risk of death in an adjusted analysis (eTable 4 in Supplement 1). In these adjusted models, baseline diabetes and hypertension were associated with 6% to 14% increased risk of all-cause mortality, whereas hyperlipidemia was associated with 25% decreased mortality risk (eTable 4 in Supplement 1).

Discussion

In this cohort study of CVD risk in more than 35 000 US veterans with a diagnosis of nonmetastatic HNSCC from 2000 to 2020, there were several findings. First, there was a substantial prevalence of modifiable CV risk factors, including hypertension, diabetes, hyperlipidemia, and smoking, at the time of cancer diagnosis. Second, although most patients were taking risk-reducing medications, rates of suboptimal risk factor control were still high, with almost half of patients having at least 1 uncontrolled CV risk factor, the most common being hypertension. Third, although patients with known CVD should be receiving intensified CV risk factor management for secondary prevention,³⁹ established CVD disease was not associated with improved CV risk factor control in this cohort. Fourth, uncontrolled risk factors were associated with an increased risk of stroke and MI; patients experienced an estimated 12.5% incidence of stroke and 8.3% incidence of MI at 10 years. Fifth, incident stroke and MI were associated with a 47% and 71% increased risk of all-cause mortality, respectively. These findings highlight the clinical effect of CV risk factors and CVD in the HNSCC population.

These results align with prior studies that showed a significant risk of cardiovascular events in patients with head and neck cancer. For instance, a SEER analysis showed a 25% to 34% 10-year incidence of cerebrovascular events in older (older than 65 years) patients with head and neck cancer.⁸ However, the present study further extends these findings by identifying several HNSCC subpopulations at high CV risk, for whom particular attention to CV risk factor mitigation should be paid. Black patients had higher baseline rates of hypertension and diabetes than patients of other racial and ethnic groups and were also more likely to have uncontrolled CV risk factor(s) in adjusted analyses. A recent single-center study also showed a higher prevalence of diabetes, but lower prevalence of dyslipidemia, in Black patients with HNSCC.⁴⁰ Patients with larynx cancer had the highest rate of prevalent CV risk factors, including smoking at diagnosis, which is unsurprising given the strong association between tobacco use and larynx cancer.⁴¹ Compared with oropharynx cancers, larynx, oral cavity, and hypopharynx primary site (smoking-associated, HPV-negative cancers) were each associated with a higher risk of MI. Traditional modifiable risk factors, including hypertension and diabetes, were associated with stroke and MI; baseline carotid artery stenosis was associated with incident stroke, whereas coronary artery disease and atrial fibrillation were associated with incident MI. Current smoking was also associated with stroke, but not MI; given the strong association between smoking and HPV-negative cancer subsites, it is possible that some of the effect of smoking was captured by adjustment for the primary site. Finally, consistent with prior studies, exposure to radiotherapy and chemotherapy as part of treatment for HNSCC was associated with risk of stroke.^{8,19,20,21,23,24} Overall, these findings support the role that traditional CV risk factors play within the HNSCC population²⁷ and identify specific cancer-related subgroups that are at high risk of incident CV events.

We believe that these findings have several potentially important clinically actionable implications. First, specific comorbidities (eg, hypertension, diabetes, coronary artery disease, carotid artery stenosis, and kidney disease), when present in patients with newly diagnosed HNSCC, should prompt clinicians to augment CV risk factor monitoring and mitigation. Second, we believe patients undergoing radiotherapy-based treatment for HNSCC, such as chemoradiotherapy for oropharynx cancer, should be carefully counseled about the risk of cerebrovascular events,²¹ and risk-reducing strategies should be pursued. Third, because having at least 1 uncontrolled CV risk factor at the time of cancer diagnosis was associated with a 9% to 19% higher hazard of stroke and MI, optimization of CV risk factors at the time of cancer diagnosis is clinically important and should not be disregarded while planning definitive-intent cancer therapy.

While stroke and MI are significant sources of morbidity, we found that incident stroke and MI were also associated with significantly increased risk of all-cause death. In this population of patients with cancer, occurrence of major CV events had a significant discernible association with the risk of all-cause mortality. Additionally, baseline diabetes and hypertension were associated with a higher risk of all-cause mortality; the apparent protective association of hyperlipidemia with survival may reflect higher statin use in this population, or a true association.⁴² These findings underscore the importance of efforts to mitigate CV risk and disease in patients with HNSCC, including lifestyle modification, prescription of risk-reducing CV medications to control hypertension and diabetes, and tobacco cessation. Despite a justifiable emphasis on cancer-related toxic effects and recurrence in this patient population, this study’s findings suggest that concurrent and competing CV risks, which tend to grow in clinical effect compared with cancer-related concerns as survivors age and recurrence risk declines, are important. Especially in patients with favorable cancer prognoses, including patients who have been cured of HNSCC, optimization of CV risk factors is a low-cost intervention that has the potential to improve not only CV event–free outcomes but also overall survival, possibly on a scale of novel cancer therapeutics.

Strengths and Limitations

This retrospective study had limitations. US veterans represent a particularly at-risk population with disproportionately high rates of alcohol and tobacco use¹¹; findings from this study may not be fully generalizable to the global HNSCC population. In particular, this population was almost entirely male, and the proportion of HPV-associated cancer may differ somewhat from the US population given the higher burden of tobacco use in the veteran population, although several studies have shown a preponderance of p16-positive oropharynx cancers in veterans as well.^43,44 We did not have HPV status available in the data set; oropharynx site is a useful but imperfect surrogate for this. Veterans with HNSCC diagnosed or treated outside of the VA system, who may differ in their CV and/or cancer profiles, were not included in the data set. We also did not have information on cause of death and were only able to analyze all-cause mortality, as opposed to cause-specific mortality. The data set only contained information on CV risk factor medication use and control at time of baseline cancer diagnosis; longitudinal CV risk factor treatment and control in the survivorship period is also an important question that was beyond the scope of this article.

These findings potentially have public health implications for US veterans and patients with HNSCC as a whole. As HNSCC cancer-specific outcomes and cure rates continue to improve due to the growing incidence of improved-prognosis HPV-related cancers compared with smoking-related HPV-negative cancers, as well as advances in treatment and supportive care, competing risks and long-term health events, such as MI and stroke, become increasingly important. Integration of multidisciplinary care with cardio-oncology clinicians can help to ensure appropriate emphasis on CV risk assessment and mitigation during and after cancer treatment to optimize survivorship in this high-risk population.

Conclusions

This cohort study presented a large population-based analysis of CV risk in patients with HNSCC and found a significant burden of baseline CV risk factors, many of which are suboptimally controlled, as well as a high rate of incident CV events after diagnosis. The presence and suboptimal control of modifiable risk factors, including hypertension and diabetes, particularly noted among Black individuals, were associated with increased risk of incident CV events; and patients who experienced these events were at higher risk for subsequent all-cause mortality.

Owing to epidemiologic shifts and therapeutic advances, cancer prognosis for patients with HNSCC has improved during the past few decades. An improved understanding and focus on modifying CV risk is important in improving survivorship care and long-term outcomes for these patients. This study defines the population-level effect of baseline and longitudinal CV comorbidities in a contemporary cohort of patients with HNSCC, highlights subgroups at high risk of CV events, and generates new data to inform the critical importance of personalized cardioprotection strategies in this population.

Supplement 1.

eTable 1. Coding definitions

eTable 2. Log binomial regression for having at least 1 uncontrolled cardiovascular risk factor at baseline

eTable 3. Sensitivity complete case analyses: cause-specific hazard ratios (csHR) and subdistribution hazard ratios (subHR) for stroke and myocardial infarction

eTable 4. Extended overall survival Cox regression with all-cause mortality as outcome and stroke/MI as time-varying covariates

Supplement 2.

Data sharing statement