Venous thromboembolism and 20-year cancer risk: a Danish population-based cohort study

Abstract

Background

Venous thromboembolism (VTE) is associated with an elevated short-term cancer risk, but data on the long-term risk are conflicting. The mechanisms linking patients with VTE to elevated long-term cancer risk remain largely unknown.

Objectives

To investigate long-term cancer risk after VTE and whether this risk pertained predominantly to cancers sharing risk factors with VTE.

Methods

Using nationwide Danish registries, we followed patients with first-time VTE (1996-2022) until cancer diagnosis, death, emigration, or December 31, 2022, whichever occurred first. We used the Aalen–Johansen estimator to compute cumulative cancer incidence, treating death as a competing event. Standardized incidence ratios (SIRs) compared observed events with expected events based on general population rates. We analyzed cancer incidence overall and by site-specific and etiology groups for smoking-, obesity-, alcohol-, hormone-, and infection-related cancers, and “all other” cancers.

Results

We followed 138,049 patients for a median of 5.5 years. The 1-year cancer risk was 4.0%. The 1-year SIR was 3.45 (95% CI, 3.36-3.55), then ranged from 1.09 to 1.16 over 20 years of follow-up, reaching 1.00 (95% CI, 0.90-1.11) thereafter. SIR was elevated in all cancer groups within 1 year and afterward, but was driven by site-specific cancers often related to smoking, obesity, and alcohol (lung, pharynx, larynx, esophagus, stomach, liver, pancreas, and colon), as well as hematological cancers.

Conclusion

Compared with the general population, patients with VTE had an increased cancer risk for up to 20 years. Shared risk factors, such as lifestyle-related factors and genetic predisposition, may contribute to this increased risk.

Essentials

• •VTE may be an early sign of undiagnosed cancer.
• •We studied 138,049 Danish patients with a first-time VTE from 1996 to 2022.
• •Cancer risk was more than 3-fold higher in the first year and remained elevated for 20 years.
• •Shared lifestyle factors may partly explain the long-term link between VTE and cancer.

1. Introduction

The association between cancer and venous thromboembolism (VTE) has been recognized since the 1800s, when Bouillaud and Trousseau reported cases of episodic migratory thrombophlebitis in cancer patients [1,2]. It is now well established that VTE is a common complication in patients with cancer [3]. Additionally, it has been demonstrated repeatedly over the past 40 years that VTE also may be a marker of an as-yet undetected (occult) cancer [2,4,5], evidenced by a 2- to 4-fold increased risk of a cancer diagnosis within the following year compared with the general population [[6], [7], [8], [9], [10]].

Whether an elevated cancer risk persists beyond the initial years following a VTE diagnosis is debated. Some studies report a continued 1.1- to 1.3-fold increase in relative risk [6,7,9], whereas others report no association [11,12]. Still, patients with VTE have an excess mortality risk compared with the general population for up to 30 years after a VTE diagnosis, with cancer representing a common cause of death [13]. The mechanism underlying a potentially increased long-term cancer risk remains largely unclear, but shared risk factors linking VTE to cancer have been suggested as potential contributing factors [6,10]. If so, patients with VTE may be at a particularly high risk of certain cancers that share risk factors with VTE (eg, smoking and obesity [14]).

We carried out a population-based cohort study in Denmark from 1996 to 2022, aiming to investigate 1) whether the increased cancer risk in patients with VTE persisted in the long term and 2) whether this risk pertained predominantly to certain cancers that share risk factors with VTE.

2. Methods

2.1. Setting and data sources

We conducted this population-based cohort study in Denmark. The Danish healthcare system provides tax-supported healthcare to all Danish residents, ensuring universal access to hospitals and private practitioners, as well as partial reimbursement for prescribed medicines [15]. At birth or upon immigration, all Danish residents are assigned a unique 10-digit Civil Registration Number (CPR number). The CPR number enables unambiguous individual-level data linkage across Danish registries [16].

We obtained information on diagnoses from the Danish National Patient Registry (DNPR) [17]. The DNPR was established in 1977 and contains detailed nationwide information on every nonpsychiatric inpatient hospital contact since 1978 [17]. Information on nonpsychiatric outpatient clinic contacts, emergency department contacts, and psychiatric contacts was added in 1995. The DNPR records a range of information, including date of contact, date of discharge, a primary diagnosis, multiple secondary diagnoses (optional), and surgical procedures. The primary diagnosis indicates the main reason for hospitalization, whereas secondary diagnoses supplement the primary diagnosis, eg, by specifying the chronic conditions relevant to the hospital contact. Diagnoses were recorded according to the International Classification of Diseases (ICD), Eighth Revision until 1994 and the 10th Revision thereafter. Surgical procedures have been coded according to the Nordic Medico-Statistical Committee Classification of Surgical Procedures since 1996.

We obtained information on selected redeemed prescriptions from the Danish National Prescription Registry, which has recorded all medical prescriptions dispensed by Danish community pharmacies since 1995 [18]. Redeemed prescriptions are coded according to the Anatomical Therapeutic Chemical Classification System.

We compiled information on cancer diagnoses from the Danish Cancer Registry. This registry has maintained records of all incident primary cancer cases in Denmark since 1943 [19].

From the Danish Civil Registration System [16], which issues CPR numbers and maintains information on migration and vital status, with daily electronic updates, we extracted data on the Danish general population.

2.2. VTE

This study included all patients with a first-time inpatient or outpatient clinic diagnosis of VTE (ie, deep venous thrombosis [DVT], superficial venous thrombosis [SVT], or pulmonary embolism [PE]) recorded in the DNPR from January 1, 1996, to December 31, 2022. Patients with a cancer diagnosis in the Danish Cancer Registry, other than nonmelanoma skin cancer, prior to their VTE were excluded. The date of the first hospital contact for VTE was used as the index date. Relevant ICD codes are provided in Supplementary Table S1.

2.3. Cancer

The primary outcome was invasive cancer (excluding in situ cancers and nonmelanoma skin cancers). To investigate whether patients with VTE were at particularly high risk of cancers of a given etiology, cancers were categorized into groups based on etiology. Although the multifactorial etiology of cancers makes the grouping of cancers difficult, we categorized cancer cases into the following groups: 1) smoking-related; 2) obesity-related; 3) alcohol-related; 4) hormone-related; 5) infection-related; and 6) all other cancers [20]. Some cancers were allowed to occur in multiple groups; e.g., oropharyngeal cancers were considered smoking-related, alcohol-related, and infection-related. ICD codes are provided in Supplementary Table S1.

2.4. Covariates

We obtained information on preadmission comorbidities from the DNPR using all available hospital history. We included all individual diseases in the Charlson Comorbidity Index (CCI; Table 1), excluding solid tumors, leukemia, lymphoma, and metastatic solid tumors [21], because we required study participants to have no history of cancer prior to the index date. Patients were classified according to CCI scores of 0, 1, 2, or ≥3.

Table 1.

Characteristics of patients with venous thromboembolism, Denmark, 1996-2022.

Variables	VTE n (%)
Total	138,049 (100)
Sex
Female	71,483 (51.8)
Male	66,566 (48.2)
Age (y), median (Q1;Q3)	66 (51;77)
Age group (y)
<50	33,175 (24.0)
50-69	48,876 (35.4)
≥70	55,998 (40.6)
Calendar year of VTE diagnosis
1996-2000	17,998 (13.0)
2001-2005	22,018 (15.9)
2006-2010	25,880 (18.7)
2011-2016	34,438 (24.9)
2017-2022	37,715 (27.3)
Years of follow-up, median (Q1;Q3)	5.5 (1.8;11.3)
Type of VTE
DVT	71,436 (51.7)
Femoral	3016 (4.2)
Distal	13,888 (19.4)
Unclassified	54,532 (76.3)
PE	54,312 (39.3)
SVT	12,301 (8.9)
Provoked VTE	60,496 (43.8)
Major	28,807 (20.9)
Trauma/fractures	12,589 (9.1)
Surgical procedure	21,474 (15.6)
Minor	31,689 (23.0)
Pregnancy	1395 (1.0)
Hormone replacement therapy	3313 (2.4)
Hormonal contraceptive	6916 (5.0)
Hospitalizationa	44,328 (32.1)
Preadmission comorbidities
Myocardial infarction	8133 (5.9)
Heart failure	8531 (6.2)
Peripheral vascular disease	9345 (6.8)
Cerebrovascular disease	14,488 (10.5)
Dementia	3043 (2.2)
Chronic pulmonary disease	17,792 (12.9)
Connective tissue disease	7517 (5.4)
Ulcer disease	7128 (5.2)
Mild liver disease	2747 (2.0)
Diabetes mellitus	8960 (6.5)
Hemiplegia	671 (0.5)
Moderate to severe renal disease	4927 (3.6)
Diabetes with end-organ damage	4663 (3.4)
Moderate to severe liver disease	766 (0.6)
AIDS	218 (0.2)
CCI score
0	81,644 (59.1)
1	29,924 (21.7)
2	13,506 (9.8)
≥3	12,975 (9.4)

CCI, Charlson Comorbidity Index; DVT, deep venous thrombosis; PE, pulmonary embolism; Q, quartile; SVT, superficial venous thrombosis; VTE, venous thromboembolism.

^aPatients hospitalized due to surgery or trauma/fracture were counted only in the “major provoking factor” group. The total number of patients categorized as having a minor provoking factor does not equal the sum of the counts of each minor provoking factor.

2.5. Statistical analysis

Patient characteristics were tabulated for patients with VTE. Categorical variables are presented as counts and percentages, and continuous variables as median, quartile (Q)1, and Q3.

Patients were followed from the VTE index date until a cancer diagnosis, death, emigration, or December 31, 2022, whichever came first.

To obtain cumulative incidence proportions (a measure of absolute risk) for cancer risk, we employed the Aalen–Johansen estimator, treating death as a competing event. We calculated cumulative incidence proportions after 1, 2, 5, 10, and 20 years of follow-up. Using the Kaplan–Meier estimator, we calculated the cumulative incidence proportion of all-cause mortality after 1, 2, 5, 10, and 20 years of follow-up.

We used standardized incidence ratios (SIRs) to assess the risk of cancer among patients with VTE relative to that of the general population [22]. We calculated the SIR as the ratio between observed cancer cases among patients with VTE and the number of cancer cases that would be expected if patients with VTE had the same cancer incidence rate as the general population. The expected number of cancer cases was calculated by multiplying the observed person-years at risk among patients with VTE by national cancer incidence rates from the Danish Cancer Registry, stratified by 5-year age group, sex, and calendar year. The 95% CIs were calculated assuming the observed number of cancer cases followed a Poisson distribution. Exact limits were used when the observed number of cases was fewer than 10; otherwise, Byar’s approximation was used [23]. To examine our first aim (ie, the long-term relative risk of cancer following VTE), we calculated the SIR for 7 follow-up periods: 0 to <1 year, 1 to <2 years, 2 to <5 years, 5 to <10 years, 10 to <15 years, 15 to <20 years, and ≥20 years. To examine our second aim (ie, whether cancer risk following VTE pertains particularly to cancers with a certain etiology), we calculated the SIR for each cancer group. The latter analysis was performed using 0 to <1 year and ≥1 year as categories of follow-up time.

2.6. Stratified analyses

We performed a number of stratified analyses as follows: by age (<50 years, 50-69 years, and ≥70 years) and sex (female or male); by type of VTE (SVT, DVT, or PE); by type of DVT (femoral, distal, or unclassified); by type of VTE diagnosis (main or secondary); by CCI score (0, 1, 2, and ≥3); and whether the VTE was provoked or unprovoked. We defined provoked VTE based on major and minor provoking factors, following the guidance document published by the International Society on Thrombosis and Haemostasis [24] and previous literature [25,26]. Major provoking factors included surgery or fracture/trauma within the past 3 months. Minor provoking factors included pregnancy, the use of postmenopausal hormonal replacement therapy, the use of hormonal contraceptives, or hospitalization within the past 3 months. For hormonal replacement therapy and hormonal contraceptives, we included only agents associated with increased VTE risk (ie, excluding transdermal and vaginal administration of drugs) [27,28]. In a supplemental analysis, we stratified by site-specific cancers and calculated the SIR for each cancer group for 0 to <1 year and ≥1 year of follow-up.

Data management and statistical analyses were performed using SAS version 9.4 (SAS Institute). Data visualization was performed in R version 4.3.2 (R Foundation for Statistical Computing).

2.7. Ethics, approval, and registration

According to Danish law, ethical approval and informed consent were not required for this registry-based study. The study was reported to the Danish Data Protection Agency by Aarhus University (number 2016-051-000001/812).

3. Results

3.1. Patient characteristics and mortality

We identified 138,049 patients with VTE. Of these, 51.7% had DVT, 39.3% had PE, and 8.9% had SVT. Patients with provoked VTE accounted for 43.8% of the study population, of whom 47.6% had a major provoking factor. The median age was 66 years, and 51.8% of patients were female. CCI scores were 0 in 59.1% of patients, 1 in 21.7%, 2 in 9.8%, and ≥3 in 9.4%. The all-cause mortality risk following VTE was 11.1% after 1 year of follow-up, 14.9% after 2 years, 23.8% after 5 years, 35.6% after 10 years, and 53.2% after 20 years (Supplementary Table S3).

3.2. Overall cancer risk

During a median follow-up of 5.5 years (Q1, 1.8; Q3, 11.3), 19,260 cases of incident cancer were identified, of which 5504 occurred within the first year. The overall absolute cancer risk following VTE was 4.0% after 1 year of follow-up, 5.3% after 2 years, 8.6% after 5 years, 13.4% after 10 years, and 20.4% after 20 years (Table 2). Compared with the general population, the SIR was 3.45 (95% CI, 3.36-3.55) during the first year of follow-up and between 1.09 (95% CI, 1.06-1.13) and 1.16 (95% CI, 1.10-1.22) during subsequent years, up to 20 years of follow-up. Thereafter, the SIR was 1.00 (95% CI, 0.90-1.11; Figure 1).

Table 2.

Absolute cancer risk following a first-time venous thromboembolism according to follow-up, Denmark, 1996-2022.

	1 y	2 y	5 y	10 y	20 y
Cancer	% Risk (95% CI)	% Risk (95% CI)	% Risk (95% CI)	% Risk (95% CI)	% Risk (95% CI)
All cancers	4.0 (3.9-4.1)	5.3 (5.2-5.4)	8.6 (8.4-8.7)	13.4 (13.2-13.6)	20.4 (20.1-20.7)
Smoking-related	1.6 (1.5-1.6)	2.0 (1.9-2.0)	3.0 (2.9-3.1)	4.5 (4.4-4.6)	6.7 (6.5-6.9)
Obesity-related	1.5 (1.4-1.5)	1.9 (1.9-2.0)	3.2 (3.1-3.3)	5.1 (5.0-5.2)	7.9 (7.7-8.1)
Alcohol-related	0.2 (0.1-0.2)	0.2 (0.2-0.3)	0.4 (0.4-0.4)	0.7 (0.6-0.7)	1.1 (1.0-1.2)
Hormone-related	0.7 (0.6-0.7)	1.0 (1.0-1.1)	1.8 (1.8-1.9)	3.2 (3.1-3.3)	5.0 (4.9-5.2)
Infection-related	0.2 (0.2-0.3)	0.3 (0.3-0.3)	0.5 (0.5-0.6)	0.8 (0.7-0.8)	1.2 (1.2-1.3)
All other cancers	0.9 (0.8-0.9)	1.2 (1.1-1.2)	1.9 (1.8-1.9)	2.9 (2.8-3.0)	4.3 (4.2-4.5)

Figure 1.

Standardized incidence ratio of cancer comparing patients with venous thromboembolism with the general population, by length of follow-up.

3.3. Cancer-specific risk

When we examined cancer risk by specific cancer group, we found that the absolute cancer risk was highest for obesity- and smoking-related cancers for all follow-up intervals (Table 2). In the first year of follow-up, the SIR increased across all cancer groups. After 1 year of follow-up, the SIR decreased in all cancer groups but remained elevated (Figure 2).

Figure 2.

Standardized incidence ratios of cancers following venous thromboembolism compared with the general population. Cancers are stratified by etiology. The upper panel shows results within 1 year of follow-up; the lower panel shows results for the subsequent years.

3.4. Stratified analyses

When stratified by age, sex, VTE type (SVT, DVT, and PE), DVT type (femoral, distal, and unclassified), diagnosis type (primary or secondary), CCI score, and provoked vs unprovoked VTE, the pattern observed in the main analysis persisted: the highest relative cancer risk was observed within 1 year of follow-up, followed by a decline to an excess relative risk of approximately 9% to 16%, up until 20 years of follow-up (Figures 3 and 4).

Figure 3.

Standardized incidence ratios of cancer following venous thromboembolism compared with the general population, by length of follow-up, and stratified by age group, sex, and the Charlson Comorbidity Index (CCI) score.

Figure 4.

Standardized incidence ratios of cancer following venous thromboembolism (VTE) compared with the general population, by length of follow-up, and stratified by type of VTE, type of deep venous thrombosis (DVT), VTE as primary or secondary diagnosis, and presence of provoking factors. DVT, deep venous thrombosis; PE, pulmonary embolism; SVT, superficial venous thrombosis; VTE, venous thromboembolism.

Regarding etiology-specific cancers, the long-term increased relative cancer risk in the stratified analyses pertained largely to the same cancer types as those reported in the main analysis (Supplementary Tables S4–S10).

Cancer-specific SIRs are provided in Supplementary Table S11. The site-specific cancers driving the majority of the increased cancer risk in the long term were particularly associated with smoking, alcohol, and obesity (lung, pharynx, larynx, esophagus, stomach, liver, pancreas, and colon), as well as hematological cancers.

4. Discussion

In this analysis, adjusted for age, sex, and calendar year, we found that patients with VTE had an elevated relative risk of a first-time cancer diagnosis compared with the general population. Our findings revealed a clear dichotomous risk pattern during follow-up; the risk was particularly increased during the first year following VTE, followed by a marked decline in excess cancer risk. However, the cancer risk remained 9% to 16% higher than expected, up to 20 years of follow-up. This pattern was consistent across various cancer subgroups. The long-term increased cancer risk was elevated across all cancer etiology groups, but driven by site-specific cancers often related to obesity, alcohol, and smoking, as well as hematological cancers.

Studies on the long-term risk of cancer after VTE have contradictory conclusions, despite findings largely comparable to those of the present study. Several reported an increased relative risk of cancer, ranging from 1.1 to 1.3, after at least 1 year of follow-up until a maximum of 24 years of follow-up [[6], [7], [8], [9], [10],26,29]. Another study concluded that patients with VTE had the same long-term cancer risk as the general population, with a hazard ratio of 1.09 (95% CI, 0.59-1.34) [11]. In the latter study, 1495 patients with VTE who were cancer-free after 6 months were matched by age and sex to a comparison cohort drawn from the general population. After 30 months of follow-up, 48 patients in the VTE cohort were diagnosed with cancer compared with 43 in the general population cohort. The low number of cancer events resulted in a wide CI. Notably, the point estimate of 1.09 in that study [11] aligns with previous studies [6,7,10,29] and with our findings from the long-term follow-up in the present study. Overall, our results corroborate those previously published, reinforcing that patients with VTE have an increased cancer risk beyond the first year of follow-up.

Our study demonstrated both an increased short- and long-term relative risk of cancer among patients with VTE. Several factors likely contribute to the elevated short-term cancer risk, with converted cancers being highly likely due to several mechanisms. First, cancer-associated hypercoagulability can result in VTE as the first sign of cancer [30]. Second, solid tumors may compress the vasculature, resulting in ensuing venous obstruction and thrombus formation (eg, pelvic tumors) [31]. Third, clinical awareness that VTE may be a marker of undetected cancer can lead to enhanced diagnostic surveillance.

The long-term increased risk of VTE is unlikely to be explained by an undetected cancer at the time of VTE diagnosis, given that the cancer became clinically overt only several years later. Other factors, therefore, must contribute to the long-term association. Detection bias may partly account for this: patients with VTE may have had more or longer outpatient clinic visits even after the initial treatment period. Additionally, routine laboratory testing or diagnostic procedures undertaken in case of bleeding events in anticoagulated patients could have led to an early cancer diagnosis [32]. However, if such lead-time bias were indeed the main driver, a compensatory deficit in later cancer risk would be expected [6]. Yet, this was not observed in our study. At the same time, the hypothesis of shared risk factors linking VTE to subsequent cancer is supported by our results, suggesting that lifestyle factors such as obesity, alcohol, and smoking could, at least in part, drive the association. Obesity is associated with low-grade systemic inflammation, increased thrombin formation, decreased fibrinolysis, and immobility – each of which may lead to venous thrombosis [14,33]. In addition to the cancers in the obesity-related cancer group, several hematological cancers are also associated with obesity [34]. Smoking may lead to VTE by sustaining a procoagulant state, reducing fibrinolysis, and increasing inflammation and blood viscosity [35,36]. However, conflicting results have been reported on the association between smoking and VTE. Some studies report a null association [14], while others report a positive dose-response relation [37,38] or an association only with provoked VTE [39,40]. A systematic review and meta-analysis showed a modest association between smoking and VTE, with a positive dose-response relation [41]. Moreover, smoking is a well-established strong risk factor for several cancers [42]. While the association between alcohol and cancer is firmly established [43], the association between alcohol and VTE has shown conflicting results, ranging from meta-analyses reporting a neutral or slightly protective effect [44,45] to an increased risk in patients with alcohol use disorders [46] and in those with high alcohol intake [47,48]. Importantly, it must be stressed that other pathways linking VTE to subsequent cancer exist. For instance, the increased risk of hematological cancers could be explained by genetics, such as a JAK2 mutation or other mutations associated with clonal hematopoiesis of indeterminate potential [49]. These mutations are associated with an increased risk of VTE and hematological cancers [49].

The findings of this study should be interpreted as a renewed impetus to explore the hypothesis that shared risk factors (eg, smoking and obesity) may partly account for the observed association between VTE and subsequent long-term cancer risk. Further research on the biological and behavioral pathways linking VTE to cancer, as well as the specific risk factors involved, is crucial to deepen our understanding of this relationship. However, our findings support the importance of identifying treatable risk factors and using the opportunity to target them with intervention at the time of VTE diagnosis.

4.1. Limitations

Several limitations should be considered when interpreting our results. The population-based design and virtually complete follow-up of all individuals minimized the risk of selection bias. However, some misclassification of a VTE diagnosis was likely. The validity of a first-time VTE diagnosis in the DNPR has been assessed against medical records, with positive predictive values of 72% [50] (1994-2006) and 88% [51] (2010-2012). Still, any exposure misclassification should be nondifferential, resulting in bias toward a null association. Misclassification of the outcome should be minimal, as data in the Danish Cancer Registry are 95% to 98% valid and complete [19,52]. Another concern is that adjudicating VTE events as provoked or unprovoked is challenging due to a lack of consensus on a comprehensive list of provoking factors. The etiology-based categorization of cancers aimed to explore the role of shared risk factors in the association between VTE and long-term cancer risk. However, categorizing cancers into overall groups based on likely etiology is complex, as cancer etiology is multifactorial, and several cancers would have fit into more than 1 group. Hence, factors beyond those suggested by a group’s label could have contributed to the increased cancer risk. Another shortcoming is the lack of exact information on potential lifestyle factors that may play a role in linking VTE to cancer (eg, smoking status and body mass index), thereby limiting our ability to examine the impact of these risk factors directly. The lack of data on ethnicity, and hence on potential ethnic differences in the association between VTE and cancer, warrants caution when applying these findings to populations with an ethnic composition different from that of Denmark. Finally, comparisons of cancer risk across subgroups based solely on the magnitude of the SIRs should be interpreted with caution, as differences in the underlying age, sex, and calendar year distributions may influence the estimates.

5. Conclusion

This study clearly shows that patients with VTE had an increased relative risk of cancer over 20 years of follow-up. Over the long term, the elevated cancer risk was increased across all cancer etiology groups, largely driven by site-specific cancers related to obesity, alcohol, and smoking, as well as hematological cancers. Shared risk factors, such as lifestyle-related factors, may partly explain the association between VTE and long-term cancer risk.