Cardiovascular disease incidence and cancer risk in two large European prospective cohorts

Abstract

Our study investigated the association between incident cardiovascular disease (CVD) and subsequent cancer risk in two large European prospective cohorts. We included 568,926 adults from the European Prospective Investigation into Cancer and Nutrition and United Kingdom Biobank, all of whom were free of CVD, cancer and type 2 diabetes at baseline. Multivariable Cox proportional hazards regression models were used to estimate cancer hazard ratios (HRs) and 95% confidence intervals (CIs) in relation to incident CVD events. CVD was treated as a time‐varying exposure, and models accounted for time since CVD diagnosis and other lifestyle factors. Study‐specific estimates were pooled using meta‐analysis. Over a median follow‐up of 10.9 years, 51,559 participants developed cancer, including 2344 with a prior CVD diagnosis. In men and women combined, CVD was associated with a higher risk of cancer within the first year after diagnosis of CVD (hazard ration [HR] = 1.65, 95% CI: 1.46–1.86), but not between 1 and 5 years (HR = 1.05, 95% CI: 0.94–1.19) or after 5 years of CVD diagnosis (HR = 1.01, 95% CI: 0.93–1.09). Results were consistent across both cohorts. In sex‐specific analyses, CVD was associated with an increased cancer risk 1–5 years post‐diagnosis in women (HR = 1.13, 95% CI: 1.06–1.21), while in men, no such association was observed (HR = 1.02, 95% CI: 0.90–1.16). Men with a newly diagnosed CVD had a higher cancer risk within the first year of CVD, likely due to overdiagnosis, but no association was observed beyond 1 year. In women, a diagnosis of CVD was associated with an increased risk of cancer for up to 5 years post‐diagnosis, suggesting that cancer overdiagnosis is less likely. These findings should be interpreted given the lack of CVD medication data.

What's new?

Cardiovascular disease (CVD) and cancer share risk factors and biological mechanisms, raising questions about potential associations between the two, particularly regarding CVD duration and cancer onset. Whether CVD influences subsequent cancer risk, however, remains inconclusive. Here, using data from two European cohorts, the authors evaluated relationships between CVD diagnosis and cancer risk. Across cohorts, within one year of CVD diagnosis, cancer risk was increased. Sex‐specific analyses show that this was primarily the case in men. Women with CVD experienced an elevated cancer risk 1–5 years post‐diagnosis. Further study is needed to understand relationships between CVD surveillance and cancer.

Abbreviations

AF

atrial fibrillation

CHD

coronary heart disease

CI

confidence interval

CVD

cardiovascular diseases

EPIC

European Prospective Investigation into Cancer and nutrition

HF

heart failure

HR

hazard ratio

IARC

International Agency for Research on Cancer

T2D

type 2 diabetes

UKB

United Kingdom Biobank

1. INTRODUCTION

An aging population combined with improved medical care is leading to a rise in the number of people living with one or more non‐communicable diseases. ¹ Cardiovascular diseases (CVD) and cancer are the leading causes of morbidity and mortality worldwide. ² CVD and cancer share several risk factors such as smoking, obesity and lack of physical activity. ³ They also share common biological pathways including inflammation, oxidative stress and dysregulation of the immune system. ³ In addition, certain medications used for CVD management, such as aspirin, may also alter cancer risk. ⁴ These elements may explain the co‐occurrence of these diseases within an individual. ⁵ The emerging concept of ‘reverse cardio‐oncology’ highlights the growing interest in understanding the potential links between CVD and subsequent cancer.^{6, 7}

Epidemiological studies to date reported conflicting results on the relationship between CVD onset and the risk of developing cancer, with heterogeneous results depending on the type of CVD and whether duration of CVD was taken into account.^{8, 9, 10, 11, 12, 13, 14}

Leveraging data from two large cohorts, the European Prospective Investigation into Cancer and nutrition (EPIC) and United Kingdom Biobank (UKB), we assessed the association between the onset of CVD and the risk of cancer. This relationship was evaluated by time since CVD diagnosis accounting for common risk factors.

2. METHODS

2.1. Study population

The EPIC is an ongoing prospective cohort study conducted in 10 European countries (Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, The Netherlands and the United Kingdom). Between 1992 and 2000, over 520,000 participants were recruited from the general population, with a few exceptions. In France, Norway, Utrecht (Netherlands) and Naples (Italy), only women were recruited. Additionally, the study included women attending local population‐based breast cancer screening programmes in Utrecht (The Netherlands) and Florence (Italy). Moreover, some centres in Italy and Spain enrolled members of local blood donor associations. During the recruitment, data on lifestyle, diet, physical activity, health, anthropometric measurements and biological samples were collected from all participants, as detailed elsewhere. ¹⁵

UKB is an ongoing prospective cohort study conducted across England, Wales and Scotland. A total of 502,650 participants aged between 39 and 71 years were recruited from 2006 to 2010, as detailed elsewhere. ¹⁶ At recruitment, data on lifestyle, diet, physical activity, health, anthropometric measurements and biological samples were collected from all the participants.

In our analysis, EPIC participants from France, Greece, Sweden and Norway were excluded due to the unavailability or inconsistency of information on incident CVD events and diagnosis dates. We excluded individuals who had existing cancer, cardiovascular disease (CVD) or type 2 diabetes (T2D) at the time of recruitment to focus on incident events of CVD and to ensure validity of CVD ascertainment. Additionally, participants with missing data in any covariates were excluded. As a result, the present analyses included a total of 221,4548 participants from the EPIC study and 343,731 participants from UKB. More details regarding the exclusions can be found in Figures S1 and S2.

2.2. CVD assessment

Non‐fatal CVD incident events were coded using the 10th Edition of the International Classification of Diseases (ICD‐10). The main coronary disease endpoints were defined as any coronary heart disease (CHD), comprised of myocardial infarction (I21, I22), angina (I20) or other CHD (I23–I25). Cerebrovascular events were defined as any haemorrhagic stroke (I60–I61), ischaemic stroke (I63), unclassified stroke (I64) and other acute cerebrovascular events (I62, I65–69, F01). For secondary analyses in UKB, we further identified atrial fibrillation (AF) (I48) and heart failure (HF) (I50).

In EPIC, non‐fatal coronary events were ascertained by different methods depending on the follow‐up procedures used by each centre, using active follow‐up through questionnaires or linkage with morbidity and hospital registries, or both. ¹⁷

In UKB, incident CVD events were continuously tracked using electronic linkage with hospital admission data. These data were gathered from Hospital Episode Statistics in England, Scottish Morbidity Record and Patient Episode Database for Wales, covering regions across England, Scotland and Wales.

2.3. Outcome assessment

Incident first primary cancer events in EPIC were ascertained by population cancer registries in Denmark, Italy, The Netherlands, Spain and the UK and through a combination of health insurance records, cancer pathology registries and active follow‐up in Germany. In UKB, these cancer cases were ascertained through cancer registries and data on cancer diagnoses were provided by NHS Digital and Public Health England for participants from England and Wales and by NHS Central Register (NHSCR) for participants residing in Scotland. Data on cancer incidence were coded according to the International Classification of Diseases for Oncology (ICD‐O‐3) and to the ICD‐10.

We examined all cancers combined (excluding non‐melanoma skin cancer), obesity‐related cancers ¹⁸ (namely, meningioma, multiple myeloma, adenocarcinoma of the oesophagus, thyroid, postmenopausal breast, gallbladder, stomach, liver, pancreas, kidney, ovaries, uterus and colorectum), alcohol‐related cancers ¹⁹ (namely, cancers of upper aerodigestive tract [lip and oral cavity, pharynx, larynx and oesophagus] and cancers of the colorectum, liver and female breast), and smoking‐related cancers ²⁰ (namely, cancers of upper aerodigestive tract, lung, nasal cavities and sinuses, liver and bile duct, renal pelvis, pancreas, uterine cervix, urinary bladder, kidney, urethra, stomach and myeloid leukaemia). Associations were also evaluated for lung, colorectal and female breast cancer as they are among the most common cancers.

2.4. Assessment of covariates

In EPIC, data on socio‐demographics, lifestyle, such as education level, smoking status, alcohol consumption, physical activity, menopausal status in women and use of hormone therapy in postmenopausal women were collected at recruitment. Information on participants' diet was assessed via country‐specific dietary questionnaires. ¹⁵ Height and weight were measured at recruitment using a standardised protocol, by trained medical staff; in the Oxford centre, height and weight were self‐reported. Body mass index (BMI) was computed as weight/height² (kg/m²). Incident T2D cases were ascertained through multiple sources across the different centres including self‐report, linkage to primary care registers, secondary care registers, medication use (drug registers), hospital admission and mortality data.

In UKB, a standardised touchscreen questionnaire was used to assess lifestyle factors such as educational level, Townsend deprivation index, smoking history, alcohol intake frequency, physical activity, menopausal status and use of hormone therapy for postmenopausal women. Information on participants' diet was assessed by using the UKB food frequency questionnaire. Height and weight were measured at recruitment using a standardised protocol by trained medical staff, and BMI was computed. Incident T2D cases were ascertained with electronic linkage to hospital admissions data.

2.5. Statistical analysis

Baseline characteristics of the study participants stratified by sex are presented as median and interquartile for continuous variables and as frequencies for categorical variables. To investigate the association between CVD and cancer (overall and lifestyle‐related cancer groups) risk, we fitted Cox proportional hazards models with age as the primary time between age at recruitment and either age at diagnosis of cancer, death or censoring (lost or end of follow‐up), whichever occurred first. Analyses were conducted separately in EPIC and UKB and results meta‐analysed. CVD occurrence was modelled with time‐varying indicators in models also including time since CVD onset (time‐varying accounting for non‐linearity through natural splines).

Models in EPIC and UKB were stratified by sex, recruitment centre and age at recruitment (5‐year categories). Additionally, models in EPIC were adjusted for BMI, alcohol intake, an indicator for alcohol consumption status, height, smoking history, physical activity, educational level, modified relative Mediterranean Diet score, ²¹ T2D status, duration of T2D, menopausal status and use of Hormone replacement therapy (HRT).

Models in UKB were adjusted for BMI, alcohol intake frequency, height, smoking history, physical activity, education level, Townsend deprivation index, healthy diet score, ²² T2D status, duration of T2D, menopausal status and use of HRT.

For both cohorts, the menopausal status variable was allowed to change from pre‐ to postmenopausal among women who turned 55 years during follow‐up. ²³ More details regarding covariates can be found in Supporting Information S1: Text 1.

Association between CVD and cancer risk was examined by categories of time since CVD diagnosis (<1 year; between 1 and 5 years; >5 years). Associations were evaluated for the overall onset of CVD, as well as separately for the onset of stroke and CHD, using a model that included two time‐dependent indicators.

The heterogeneity of associations was evaluated by stratifying the analysis by sex. In UKB, the types of CVD, specifically stroke and CHD, were considered as the exposures in the analysis. To address residual confounding by smoking, we also investigated associations by baseline smoking. In UKB, we also examined other types of CVD (HF and AF) and their association with cancer or lifestyle‐related cancer risk. We also examined the difference in risk of colorectal and lung cancer by CVD types. To explore potential detection bias following CVD diagnosis, we grouped cancer types known to be prone to overdiagnosis (neuroblastoma, prostate cancer, thyroid cancer, lung cancer, melanoma, breast cancer), ²⁴ as they are often detected through screening as slow‐growing lesions that may never cause symptoms or affect mortality. We then assessed whether their association with CVD differed during the first year after diagnosis. Additionally, we assessed the association between CVD and the risk of colorectal and breast cancer, both screenable cancers, while further adjusting for screening history specific to each cancer type. Due to case numbers and data availability, these analyses were limited to these two cancers in UKB. We additionally adjusted for statin use in UKB (results not shown), but this did not change the estimates. It should be noted that information on prescription dates, duration of use and some other medications was not available. We additionally conducted age‐stratified analyses to explore potential variation in the association across age groups at CVD diagnosis. We also estimated crude cancer incidence rates per 1000 person‐years across categories of CVD status.

3. RESULTS

3.1. Study population

In the EPIC study, the median age at enrolment was 52.9 years (Interquartile range (IQR): 46.8–58.7). At baseline, 57% were overweight and/or obese and 28% were current smokers (Table 1). In UKB, the median age at recruitment was 57 years (IQR: 49–62), with 65% considered as overweight and/or obese and 7.9% being current smokers (Table 2).

TABLE 1.

Baseline characteristics of the study population in European Prospective Investigation into Cancer and nutrition (recruitment between 1992 and 2000), by sex.

	Female (N = 137,123)	Male (N = 84,331)	Overall (N = 221,454)
Age at assessment (years)
Median [Q1, Q3]	52.6 [46.1, 58.5]	53.3 [47.5, 58.7]	52.9 [46.7, 58.6]
Follow‐up (years)
Median [Q1, Q3]	11.1 [9.67, 12.6]	10.7 [9.53, 12.3]	10.9 [9.62, 12.5]
Height (cm)
Median [Q1, Q3]	162 [157, 166]	174 [170, 179]	166 [160, 173]
BMI categories, n (%) a
Underweight/normal weight	67,522 (49)	27,396 (32)	94,918 (43)
Overweight	48,081 (35)	43,551 (52)	91,632 (41)
Obesity	21,520 (16)	13,384 (16)	34,904 (16)
Education, n (%)
Longer education	22,194 (16)	22,578 (27)	44,772 (20)
Secondary school	20,532 (15)	9563 (11)	30,095 (14)
Technical/professional school	38,445 (28)	21,657 (26)	60,102 (27)
Primary school completed	44,963 (33)	25,990 (31)	70,953 (32)
None	8601 (6.3)	3512 (4.2)	12,113 (5.5)
Not specified	2388 (1.7)	1031 (1.2)	3419 (1.5)
Smoking intensity, n (%)
Never	70,739 (52)	24,556 (29)	95,295 (43)
Current, 1–15 cig/day	19,977 (15)	8575 (10)	28,552 (13)
Current, 16–25 cig/day	9241 (6.7)	7563 (9.0)	16,804 (7.6)
Current, 26+ cig/day	1768 (1.3)	2872 (3.4)	4640 (2.1)
Former, quit ≤10 years	11,635 (8.5)	11,265 (13)	22,900 (10)
Former, quit 11–20 years	10,369 (7.6)	10,419 (13)	20,788 (9.4)
Former, quit 20+ years	9641 (7.0)	10,272 (12)	19,913 (9.0)
Current, pipe/cigar/occas	3753 (2.7)	8809 (11)	12,562 (5.7)
Alcohol consumption (g/day)
Median [Q1, Q3]	4.00 [0.434, 12.2]	16.7 [6.30, 36.2]	7.65 [1.22, 20.8]
Alcohol consumers, n (%)
Never	13,206 (9.6)	1086 (1.3)	14,292 (6.5)
Mediterranean diet score, n (%) b
5 to ≤6 (healthier)	1217 (0.9)	591 (0.7)	1808 (0.8)
4 to <5	14,763 (11)	7703 (9.1)	22,466 (10)
3 to <4	43,129 (31)	21,784 (26)	64,913 (29)
2 to <3	48,188 (35)	28,981 (34)	77,169 (35)
1 to <2	25,540 (19)	20,552 (24)	46,092 (21)
0 to <1 (unhealthier)	4286 (3.1)	4720 (5.6)	9006 (4.1)
Physical activity, n (%)
Active	26,560 (19)	23,519 (28)	50,079 (23)
Moderately active	28,810 (21)	20,905 (25)	49,715 (22)
Moderately inactive	47,816 (35)	26,135 (31)	73,951 (33)
Inactive	33,937 (25)	13,772 (16)	47,709 (22)
Ever use HT, c n (%)
No	114,897 (84)	0 (0)	114,897 (84)
Yes	22,226 (16)	0 (0)	22,226 (16)
Menopause status, c n (%)
Premenopausal	44,778 (33)	0 (0)	44,778 (33)
Postmenopausal	66,798 (49)	0 (0)	66,798 (49)
Perimenopausal	20,183 (15)	0 (0)	20,183 (15)
Surgical menopause	5364 (3.9)	0 (0)	5364 (3.9)
Country of recruitment, n (%)
Denmark	26,651 (19)	24,196 (29)	50,847 (23)
Germany	24,285 (18)	18,623 (22)	42,908 (19)
Italy	29,485 (21)	13,146 (16)	42,631 (19)
Spain	21,350 (16)	13,305 (16)	34,655 (16)
The Netherlands	19,075 (14)	4686 (5.6)	23,761 (11)
United Kingdom	16,277 (12)	10,375 (12)	26,652 (12)
Age at CVD diagnosis (years) d
Median [Q1, Q3]	65.4 [59.8, 71.0]	63.0 [57.6, 68.4]	64.0 [58.28, 69.74]

Abbreviations: BMI, body mass index; CVD, cardiovascular diseases; HT, hormone therapy.

^a Following World Health Organization (WHO) categorisation of body mass index (BMI), underweight (BMI <18.5 kg/m²), normal (BMI ≥18.5 and <25 kg/m²), overweight (BMI ≥25 and <30 kg/m²), obesity (BMI ≥30 kg/m²).

^b The score is from the highest to lowest adherence (6 to 0).

^c In women.

^d Information during follow‐up.

TABLE 2.

Baseline characteristics of the study population in United Kingdom Biobank (recruitment between 2006 and 2010), by body mass index categories.

	Female (N = 181,072)	Male (N = 162,659)	Overall (N = 343,731)
Age at assessment (years)
Median [Q1, Q3]	56.0 [49.0, 62.0]	57.0 [49.0, 63.0]	57.0 [49.0, 62.0]
Follow‐up (years)
Median [Q1, Q3]	10.9 [10.1, 11.7]	10.9 [10.0, 11.6]	10.9 [10.1, 11.7]
Height (cm)
Median [Q1, Q3]	163 [159, 167]	176 [171, 180]	169 [162, 176]
BMI categories, a n (%)
Underweight and normal	76,007 (42)	43,048 (27)	119,055 (35)
Overweight	66,358 (37)	81,566 (50)	147,924 (43)
Obesity	38,707 (21)	38,045 (23)	76,752 (22)
Qualifications, n (%)
College or University degree	63,534 (35)	61,282 (38)	124,816 (36)
A levels/AS levels or equivalent	23,064 (13)	18,003 (11)	41,067 (12)
O levels/GCSEs or equivalent	42,698 (24)	30,774 (19)	73,472 (21)
CSEs or equivalent	9560 (5.3)	8767 (5.4)	18,327 (5.3)
NVQ or HND or HNC or equivalent	7631 (4.2)	14,521 (8.9)	22,152 (6.4)
Other professional qualifications nursing, teaching	10,140 (5.6)	6989 (4.3)	17,129 (5.0)
None of the above	24,445 (13)	22,323 (14)	46,768 (14)
Deprivation index
Median [Q1, Q3]	−2.22 [−3.68, 0.268]	−2.27 [−3.72, 0.317]	−2.25 [−3.70, 0.292]
Smoking intensity, n (%)
Never	109,491 (61)	83,239 (51)	192,730 (56)
Former, light	22,619 (13)	21,148 (13)	43,767 (13)
Former, intermediate	9016 (5.0)	11,795 (7.3)	20,811 (6.1)
Former, heavy	2426 (1.3)	5563 (3.4)	7989 (2.3)
Former, no information	22,360 (12)	21,357 (13)	43,717 (13)
Current light	5789 (3.2)	5561 (3.4)	11,350 (3.3)
Current intermediate	5054 (2.8)	5730 (3.5)	10,784 (3.1)
Current heavy	1687 (0.9)	3485 (2.1)	5172 (1.5)
Current, no information	2630 (1.5)	4781 (2.9)	7411 (2.2)
Alcohol frequency, n (%)
Never	15,114 (8.3)	8787 (5.4)	23,901 (7.0)
Special occasions only	24,928 (14)	10,631 (6.5)	35,559 (10)
One to three times a month	23,469 (13)	14,253 (8.8)	37,722 (11)
Once or twice a week	47,248 (26)	41,833 (26)	89,081 (26)
Three or four times a week	39,504 (22)	44,219 (27)	83,723 (24)
Daily or almost daily	30,809 (17)	42,936 (26)	73,745 (22)
Diet score, b n (%)
6, healthier	6032 (3.3)	3369 (2.1)	9401 (2.7)
5	41,840 (23)	26,251 (16)	68,091 (20)
4	83,846 (46)	61,072 (37)	14,4918 (42)
3	39,487 (22)	50,069 (31)	89,556 (26)
2	8460 (4.7)	17,468 (12)	25,928 (7.5)
1, unhealthier	1407 (0.8)	4430 (2.7)	5837 (1.7)
Physical activity, n (%)
High	70,190 (39)	70,340 (43)	140,530 (41)
Moderate	78,054 (43)	62,198 (38)	140,252 (41)
Low	32,828 (18)	30,121 (19)	62,949 (18)
Menopausal status, c n (%)
Premenopausal	48,717 (27)	0 (0)	48,717 (27)
Postmenopausal	106,050 (59)	0 (0)	106,050 (59)
Not sure	26,305 (15)	0 (0)	26,305 (15)
Ever use HT, c n (%)
No	115,895 (64)	0 (0)	115,895 (64)
Yes	65,177 (36)	0 (0)	65,177 (36)
Age at CVD diagnosis (years) d
Median [Q1, Q3]	65.0 [58.0, 70.0]	67.0 [62.0, 71.0]	66.0 [60.0, 70.0]

Abbreviations: AS, Advanced Subsidiary; BMI, body mass index; CSEs, Certificate of Secondary Education; CVD, cardiovascular diseases; GCSEs, General Certificate of Secondary Education; HT, hormone therapy; HND, Higher National Diploma; HNC, Higher National Certificate; NVQ, National Vocational Qualifications.

^a Following WHO categorisation of body mass index (BMI), underweight (BMI <18.5 kg/m²), normal (BMI ≥18.5 and <25 kg/m²), overweight (BMI ≥25 and <30 kg/m²), obesity (BMI ≥30 kg/m²).

^b The score is from the highest to lowest adherence (6 to 0).

^c In women.

^d Information during follow‐up.

After a median follow‐up period of 10.9 years in both studies, a total of 33,390 participants developed a CVD (with 8161 cases from EPIC and 25,134 cases from UKB). In the EPIC study, after a median follow‐up of 3.3 years after a CVD diagnosis, 631 participants developed cancer. Similarly, in the UKB study, 1713 participants were diagnosed with cancer after a median follow‐up of 2.3 years (Figure S3).

3.2. Association of CVD with cancer

A strong association between CVD and cancer was observed within the first months following a CVD diagnosis, as shown in Figure 1. This association declined sharply over time. Figure 2 confirmed these findings, with an HR equal to 1.65 (95% CI: 1.46–1.86) during the first year after CVD onset. Between 1 and 5 years after CVD onset, the HR decreased to 1.05 (95% CI: 0.94–1.19) and 5 years after CVD onset HR was 1.01 (95% CI: 0.93–1.09).

FIGURE 1.

Association between cardiovascular diseases (CVD) and cancer risk by time since CVD (in years). (A) EPIC study, (B) UK Biobank study.

FIGURE 2.

Association between cardiovascular diseases (CVD) and cancer risk by categories of time since CVD, in years. Model in European Prospective Investigation into Cancer and nutrition (EPIC): adjusted for time educational level, smoking intensity, alcohol consumption, alcohol consumption status, type 2 diabetes status, body mass index, height, use of hormonal therapy, menopausal status, physical activity, Mediterranean diet score, stratified by centre, sex and age at recruitment (5‐year categories). Model in United Kingdom Biobank (UKB): adjusted for qualifications, deprivation index, smoking intensity, alcohol consumption, type 2 diabetes status, body mass index, height, use of hormonal therapy, menopausal status, physical activity, healthy diet score, stratified by centre, sex and age at recruitment (5‐year categories). I ²: index which quantifies the dispersion of effect sizes in a meta‐analysis. CI, confidence interval; HR, hazard ratio.

When stratified by sex (Figure 3), the association between CVD and cancer risk tended to be higher in women than in men within the first year after CVD diagnosis (HR_women: 1.78, 95% CI: 1.54–2.05; HR_men: 1.57, 95% CI: 1.32–1.86). Furthermore, a persistent association between CVD and cancer risk was observed in women from 1 year up to 5 years following CVD diagnosis with HR_women equal to 1.13 (95% CI: 1.06–1.21), but not in men (HR_men equal to 1.02, 95% CI: 0.90–1.16).

FIGURE 3.

Association between cardiovascular diseases (CVD) and cancer risk by categories of time since CVD (in years), by sex. Model in European Prospective Investigation into Cancer and nutrition (EPIC) among women: adjusted for time since CVD, educational level, smoking intensity, alcohol consumption, alcohol consumption status, type 2 diabetes status, body mass index, height, use of hormone therapy, menopausal status, physical activity, Mediterranean diet score, stratified by centre, and age at recruitment (5‐year categories). Model in United Kingdom Biobank (UKB) among women: adjusted for time since CVD, qualifications, deprivation index, smoking intensity, alcohol consumption, type 2 diabetes status, body mass index, height, use of hormone therapy, menopausal status, physical activity, healthy diet score, stratified by centre and age at recruitment (5‐year categories). Model in EPIC among men: adjusted for time since CVD, educational level, smoking intensity, alcohol consumption, alcohol consumers status, type 2 diabetes status, body mass index, height, physical activity, Mediterranean diet score, stratified by centre, and age at recruitment (5‐year categories). Model in UKB among women: adjusted for time since CVD, qualifications, deprivation index, smoking intensity, alcohol consumption, type 2 diabetes status, body mass index, height, physical activity, healthy diet score, stratified by centre, and age at recruitment (5‐year categories). I ²: Index which quantifies the dispersion of effect sizes in a meta‐analysis. CI, confidence interval; HR, hazard ratio.

3.3. Crude incidence rate of cancer by CVD categories

As shown in Table S1, in both cohorts, crude incidence rates of cancer were markedly higher among participants with CVD, particularly during the first year following diagnosis of CVD. In EPIC, the incidence rate of cancer among participants without prior CVD was 7.63 per 1000 person‐years (95% CI 7.52–7.75), compared to 18.00 (15.07–21.33) in the first year after CVD, and 14.58 (12.99–16.30) and 16.77 (14.48–19.33) for 1–5 and >5 years after diagnosis of CVD, respectively. In UKB, similar patterns were observed: 9.01 (8.91–9.11) per 1000 person‐years in participants without CVD, 23.78 (21.65–26.07) within the first year after CVD, 14.74 (13.74–15.80) between 1 and 5 years and 16.05 (14.60–17.60) beyond 5 years.

3.4. Lifestyle‐related cancers

Our analysis revealed an elevated risk of obesity‐related cancers within the first year after a CVD diagnosis, with a HR of 1.76 (95% CI: 1.26–2.45). However, this risk decreased beyond the first year, with the confidence interval (CI) including the null (Figure S4). For alcohol‐related cancer, we also observed an association within the first year following the CVD diagnosis. In the EPIC study, an elevated risk persisted between 1 and 5 years after the CVD diagnosis (HR: 1.27, 95% CI: 1.04–1.54), while the association in UKB included the null. An association was observed for smoking‐related cancer risk within the first year after a CVD diagnosis (HR = 2.08, 95% CI: 1.65–2.61).

3.5. Specific cancer sites

The meta‐analysis for specific cancers showed an increased risk for colorectal cancer within the first year after CVD diagnosis (HR: 1.43, 95% CI: 1.08–1.90). However, no association was found between 1 and 5 years (HR: 1.00, 95% CI: 0.76–1.32) or beyond 5 years (HR: 0.88, 95% CI: 0.69–1.12) (Figure S5). For breast cancer risk in women, no associations with CVD were found. CVD and lung cancer risk were strongly associated within the first year following CVD diagnosis, with a HR equal to 2.68 (95% CI: 2.13, 3.36), and between 1 and 5 years after CVD diagnosis, with a HR equal to 1.25 (95% CI: 1.05, 1.50) (Figure S5) but not beyond a period of 5 years following CVD diagnosis (HR: 1.19, 95% CI: 0.94, 1.51).

Among current smokers in the meta‐analysis, CVD was associated with lung cancer both within the first year and beyond 5 years after CVD diagnosis (HR: 1.63, 95% CI: 1.23–2.16). For former and never smokers, the association was observed within the first year and between 1 and 5 years post‐diagnosis (HR: 1.63, 95% CI: 1.27–2.07) (Figure S6).

3.6. Specific CVD types

Whether for the association between CHD and cancer risk or between stroke and cancer risk, we only observed an association during the first year of diagnosis of the CVD studied (respectively, HR = 1.55, 95% CI: 1.41–1.71 and HR = 1.84, 95% CI: 1.40–2.44).

In UKB, for both AF and HF, their association with cancer was significant only during the first year of diagnosis of the CVD studied (respectively, HR_UKB = 2.21, 95% CI: 1.85, 2.49 and HR_UKB = 2.56, 95% CI: 1.87, 3.52). These associations were higher than for CHD and stroke (respectively, HR_UKB = 1.59, 95% CI: 1.42, 1.77 and HR_UKB = 2.08, 95% CI: 1.76, 2.48) (Figure S7). For obesity‐ and smoking‐related cancer risk, the associations with AF in the first year after diagnosis of AF were strong and persisted between 1 and 5 years after this period (respectively, HR_UKB = 1.20, 95% CI: 1.01, 1.42 and HR_UKB = 1.24, 95% CI: 1.01, 1.32) (Figure S8). For colorectal cancer, HR estimates were overall similar for different CVD types. In contrast, the association between AF and lung cancer was strongest within the first year after CVD diagnosis (HR: 5.09, 95% CI: 3.79–6.84) compared with other CVD types, and remained positive thereafter, both 1 and 5 years and more than 5 years following the CVD diagnosis (Table S2).

3.7. Evaluation of overdiagnosis

First, the association between CVD and the risk of cancers known for overdiagnosis showed a similar pattern as for overall cancer risk—a strong positive association in the first year after CVD, which attenuated to no association after longer follow‐up (Figure S9). Second, adjusting for screening history did not significantly alter the observed associations between CVD and colorectal or breast cancer risk (Figure S10).

3.8. Stratification by age at CVD diagnosis

The association between CVD and cancer risk within the first year after CVD diagnosis differed by age at recruitment. In EPIC, the association was weaker among participants recruited before age 60 (HR = 1.24; 95% CI: 0.95–1.63) than among those recruited at age 60 or older (HR = 1.75; 95% CI: 1.40–2.18). In contrast, the opposite pattern was observed in UKB, where the association was stronger among participants recruited before age 60 (HR = 1.95; 95% CI: 1.64–2.32) than among those recruited at older ages (HR = 1.65; 95% CI: 1.48–1.85). When cancer occurred between 1 and 5 years after CVD diagnosis, a modest association was observed in the meta‐analysis among participants recruited before age 60 (HR = 1.10; 95% CI: 1.01–1.10), but not among those recruited at older ages (Figure S11).

4. DISCUSSION

In this study, data from two large European cohorts were leveraged to investigate the association between incident CVD events and the risk of overall cancer and lifestyle‐related cancers in models accounting for the time since CVD diagnosis. A strong positive association was identified between CVD and cancer risk within the first year following a CVD event. However, no significant association was observed for cancers diagnosed more than 1 year after CVD onset. Between 1 and 5 years after a CVD diagnosis, a weaker positive association was observed in EPIC but not in UKB, and in women but not in men.

The strong drop in cancer risk over time following a CVD diagnosis and the results of our sensitivity analysis in groups of cancers that are known for overdiagnosis suggest increased surveillance as a main driver of elevated cancer risk related to CVD diagnosis. Nevertheless, there remains suggestive evidence for a potential biological link between CVD and certain types of cancers such as lung cancer or cancers in women, where an elevated cancer risk persisted up to 5 years after a CVD event. The relationship between CVD and subsequent cancer risk may differ by both cardiovascular and cancer subtype, reflecting variation in shared pathophysiology, risk exposures and surveillance. Nevertheless, our study was not sufficiently powered to investigate these subtype‐specific associations in greater detail. Future studies should investigate subclinical CVD (e.g., using early markers of heart disease), where surveillance bias is less likely to affect cancer diagnosis.

In a cohort study using anonymised healthcare reimbursement claims data, the health status of approximately 27 million individuals was monitored over time. Individuals who experienced a CVD event had a 13% higher hazard of developing cancer than those who had not. While the study accounted for self‐reported risk factors, which may have confounded the observed associations, it did not consider the time since CVD diagnosis. ²⁵ Similarly, a recent study in UKB showed that individuals with prior CVD had a 14% higher risk of cancer compared to those without CVD but without accounting for the time since CVD diagnosis, which may overestimate cancer risk. ²⁶

Several studies that investigated the association between types of CVD and cancer risk also modelled time since CVD onset.^{11, 27} In a Korean national cohort study, an association between HF and cancer risk was observed, which persisted after excluding participants who developed cancer within 2 years of HF diagnosis. ²⁷ Similarly, in a large population‐based study in Denmark, ¹¹ cancer incidence among patients with HF was higher compared to the incidence of the general population, beyond the first year after HF diagnosis, although the evaluation did not account for the role of smoking history. ¹¹ In UKB, our findings were inconclusive for an association with cancer risk beyond the first year after diagnosis of HF (Figure S7).

Several studies have also examined the association between AF and cancer risk. A study using data from two large case–control studies conducted in Israel showed that the association between AF and cancer risk vanished after 3 months from AF onset. ²⁸ In contrast, in the Danish Diet, Cancer, and Health study, an association between incident AF events and cancer risk was consistently reported within 3 months, within and beyond 1 year from diagnosis. ¹⁴ A population‐based study of AF patients in Denmark showed a strong association between AF onset and overall cancer risk within the first 3 months from AF diagnosis, while weaker, yet statistically significant, associations were observed between 4 and 6 months (standardised incidence ratio SIR: 1.38, 95% CI: 1.32–1.45), and after 6 months (7–12 months SIR: 1.15, 95% CI: 1.11–1.20, 13–24 months SIR: 1.14, 95% CI: 1.11–1.18). ²⁹ Similar association patterns were documented between AF and smoking‐, alcohol‐ and obesity‐related cancers. ²⁹ In our study, results from UKB only indicated that AF was related to cancer risk within 1 year after AF diagnosis. These results were consistent for overall and for alcohol‐related cancer, while for obesity‐ and smoking‐related cancer associations were observed within 5 years from an AF diagnosis.

We also examined associations between CVD and the risk of specific cancers, including colorectal, female breast cancer and lung cancer, as well as the association with lung cancer by smoking status. Among former and never smokers combined, the association with lung cancer risk persisted during the first 5 years after CVD diagnosis, possibly due to residual confounding by smoking. Among current smokers, an inconsistent pattern was observed, with an association during the first year and beyond the fifth year after CVD diagnosis. Our findings were in line with two previous studies showing that CVD was associated with lung cancer risk.^{30, 31} A Mendelian randomisation analysis carried out in UKB did not support a direct causal relationship between CVD and lung cancer, suggesting that CVD may be a marker of shared risk factors rather than a direct cause. ³² For breast cancer risk, no association was observed with CVD, suggesting that the positive association, in women, between CVD (both within the first year and between 1 and 5 years) and overall cancer risk was not driven by breast cancer.

In our study, several potential confounding factors, which were assumed to be associated with CVD onset and cancer risk, were accounted for in statistical models. It is established that CVD and cancer share numerous risk factors, both non‐modifiable, such as age and genetics, and modifiable, such as obesity and smoking, which may partly explain their frequent co‐occurrence.^{3, 33} Shared risk factors may also trigger candidate biological mechanisms, including chronic inflammation, oxidative stress and metabolic dysregulation.^{34, 35}

Tobacco smoking and obesity were shown to stimulate inflammation and cause the release of inflammatory markers, such as cytokines, which might promote the initiation of CVD, ³⁶ as well as carcinogenesis and tumour progression ³⁷ In the CANTOS trial, patients after a myocardial infarction were randomised to receive either a placebo or canakinumab, a potent interleukin‐1β inhibitor that lowered C‐reactive protein (CRP) inflammatory levels. The treatment not only reduced CVD recurrence but also decreased lung cancer incidence and mortality.^{38, 39}

Mechanistic studies using mice models have identified several candidate mechanisms by type of CVD,^{7, 40, 41} such as HF, myocardial infarction or cardiac remodelling, that could accelerate cancer, including the secretion of proteins like serpin A3 and A1, fibronectin, paraoxonase‐1. Specifically, serpin A3 was shown to directly induce the growth of human colon cancer (HT‐29) cells. ⁴⁰ Additionally, it has been suggested that acute pathological stress from cardiac remodelling may enhance tumour proliferation and metastasis by releasing protumorigenic factors ⁴¹

Despite this evidence, the associations pattern between CVD onset and risk of cancer that was observed in the present and other studies^{11, 14, 27, 29} could be partly explained by surveillance bias, as patients with CVD are more exposed to health monitoring compared to the general population. As a result, careful and much needed surveillance of persons with CVD may lead to the detection of occult tumours ²⁹ or tumours at early stages ⁴² It has also been argued that overlapping symptoms of cancer and certain CVD may be initially attributed to heart diseases, thus leading to a delay in cancer diagnosis ¹⁴ Last, CVD management, for example, via the administration of anticoagulants to treat AF, may contribute to the earlier detection of cancers due to bleeding ⁴³

Our study has several strengths. Firstly, we leveraged individual‐level data from two large prospective cohorts of adult participants from six European countries. These cohorts had validated ascertainment of cancer and CVD occurrence^{16, 17} Secondly, to the best of our knowledge, this is the largest study to date that uses prospective data to investigate the association between incident CVD and cancer risk. Third, statistical analyses are stratified with regards to time since CVD and for the role of potential confounders assessed prior CVD and cancer. Lastly, we conducted a large range of sensitivity analyses to investigate potential biases.

Our study also had limitations. The statistical model relied on lifestyle factors collected at recruitment in EPIC and UKB but did not account for changes in modifiable habits during follow‐up. Furthermore, participants free of CVD and cancer were included in this study, which may introduce selection bias in the study population. In addition, the UKB showed a low recruitment rate, approximately 5%.^{44, 45} They also tended to reside in less socioeconomically deprived areas and had a lower prevalence of long‐term conditions. ⁴⁶ Despite the use of large cohorts, the number of CVD patients who developed cancer is limited. Consequently, our ability to conduct more detailed and stratified analyses by cancer sites was limited. Data on medications was not available in EPIC and was limited also in UKB as we did not have information on the date of prescription or duration of drug use. However, some commonly prescribed CVD medications can influence cancer outcomes. For example, aspirin is known to prevent early neoplastic transformation and to have an anti‐metastatic effects, ⁴⁷ while angiotensin converting enzyme inhibitors and angiotensin receptor blockers are inversely associated with the incidence of some cancers, including colorectal cancer. ⁴⁸ Future studies with longitudinal data on CVD medications might better elucidate their role in cancer detection and prevention, and clarify how cancer incidence changes over time after a CVD diagnosis. In addition to breast and colorectal cancer screening, other organised screening programmes (e.g., cervical and, more recently, targeted lung cancer screening) may also affect cancer detection patterns and should be considered in future studies examining cancer incidence after CVD.

In this study, participants with newly diagnosed CVD showed an increased risk of cancer within the first year after CVD diagnosis, with markedly higher incidence rates compared with those without CVD. However, this association was not observed when cancer was diagnosed beyond this first year period, except for lung cancer and cancers in women. These associations were consistently observed for cancers related to obesity, alcohol consumption and smoking in both EPIC and UKB cohorts. Our results align with current cancer screening guidelines and do not support additional cancer surveillance among patients with CVD.

AUTHOR CONTRIBUTIONS

Emma Fontvieille: Writing – original draft; investigation; writing – review and editing; formal analysis. Vivian Viallon: Methodology; writing – review and editing; supervision. Laia Peruchet‐Noray: Writing – review and editing; visualization. Quan Gan: Writing – review and editing. N. Charlotte Onland‐Moret: Writing – review and editing. Yvonne Koop: Writing – review and editing. Anne Tjønneland: Writing – review and editing. Nicola Patricia Bondonno: Writing – review and editing. Verena Katzke: Writing – review and editing. Rudolf Kaaks: Writing – review and editing. Matthias B. Schulze: Writing – review and editing. Catarina Schiborn: Writing – review and editing. Calogero Saieva: Writing – review and editing. Vittorio Simeon: Writing – review and editing. Claudia Agnoli: Writing – review and editing. Rosario Tumino: Writing – review and editing. Fulvio Ricceri: Writing – review and editing. Leila Luján‐Barroso: Writing – review and editing. Maria‐José Sánchez: Writing – review and editing. Conchi Moreno‐Iribas: Writing – review and editing. Konstantinos K. Tsilidis: Writing – review and editing. Marc J. Gunter: Writing – review and editing. Adam Butterworth: Writing – review and editing. Elio Riboli: Writing – review and editing. Heinz Freisling: Writing – review and editing; validation; methodology; conceptualization; supervision; funding acquisition. Pietro Ferrari: Conceptualization; funding acquisition; supervision; project administration; writing – review and editing; validation; methodology.

FUNDING INFORMATION

This work was supported by the French National Cancer Institute (l'Institut National du Cancer) [INCa_18624, ‘ILIAD’]. The coordination of EPIC‐Europe is financially supported by the International Agency for Research on Cancer (IARC) and also by the Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, which has additional infrastructure support provided by the NIHR Imperial Biomedical Research Centre (BRC). The national cohorts are supported by: Danish Cancer Society (Denmark); Ligue Nationale Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l'Education Nationale (MGEN), Institut National de la Santé et de la Recherche Médicale (INSERM), French National Research Agency (ANR, reference ANR‐10‐COHO‐0006), French Ministry for Higher Education (subsidy 2102918823, 2103236497 and 2103586016) (France); German Cancer Aid, German Cancer Research Center (DKFZ), German Institute of Human Nutrition Potsdam‐Rehbruecke (DIfE), Federal Ministry of Education and Research (BMBF) (Germany); Associazione Italiana per la Ricerca sul Cancro‐AIRC‐Italy, Italian Ministry of Health, Italian Ministry of University and Research (MUR), Compagnia di San Paolo (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), the Netherlands Organisation for Health Research and Development (ZonMW), World Cancer Research Fund (WCRF) (The Netherlands); UiT The Arctic University of Norway; Health Research Fund (FIS)—Instituto de Salud Carlos III (ISCIII), Regional Governments of Andalucía, Asturias, Basque Country, Murcia and Navarra and the Catalan Institute of Oncology—ICO (Spain); Swedish Cancer Society, Swedish Research Council and County Councils of Skåne and Västerbotten (Sweden); Cancer Research UK (C864/A14136 to EPIC‐Norfolk; C8221/A29017 to EPIC‐Oxford), Medical Research Council (MR/N003284/1, MC‐UU_12015/1 and MC_UU_00006/1 to EPIC‐Norfolk; MR/Y013662/1 to EPIC‐Oxford) (United Kingdom).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENT

Ethical approval for the European Prospective Investigation into Cancer and nutrition (EPIC) study was granted by the Ethical Review Boards of the International Agency for Research on Cancer (IARC) and the Institutional Review Board of each participating EPIC centre. All participants from the UK Biobank cohort provided written informed consent, and the study itself received ethical approval from the Northwest Multi‐Centre Research Ethics Committee under the Research Tissue Bank (RTB) framework. The current approval reference number is: 21/NW/0157.

Supporting information

Data S1. Supporting Information.

Fontvieille E, Viallon V, Peruchet‐Noray L, et al. Cardiovascular disease incidence and cancer risk in two large European prospective cohorts. Int J Cancer. 2026;159(5):1139‐1152. doi: 10.1002/ijc.70458

DATA AVAILABILITY STATEMENT

Study is based on European Prospective Investigation into Cancer and nutrition and United Kingdom Biobank data that are available from their respective websites after approval. Further information is available from the corresponding author upon request.