Coffee consumption is associated with a reduced risk of venous thrombosis that is mediated through hemostatic factor levels

Authors

  • R. E. J. ROACH,

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    Search for more papers by this author
  • B. SIEGERINK,

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center
    Search for more papers by this author
  • S. le CESSIE,

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    2. Department of Medical Statistics and Bioinformatics, Leiden University Medical Center
    Search for more papers by this author
  • F. R. ROSENDAAL,

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center
    3. Thrombosis and Hemostasis Research Center, Leiden University Medical Center, Leiden, the Netherlands
    Search for more papers by this author
  • S. C. CANNEGIETER,

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center
    Search for more papers by this author
  • W. M. LIJFERING

    1. Department of Clinical Epidemiology, Leiden University Medical Center
    2. Einthoven Laboratory for Experimental Vascular Medicine, Leiden University Medical Center
    Search for more papers by this author

Frits R. Rosendaal, Department of Clinical Epidemiology, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, the Netherlands.
Tel.: +31 71 526 4037; fax: +31 71 526 6994.
E-mail: f.r.rosendaal@lumc.nl

Abstract

Summary.  Background: Coffee consumption is associated with a lower risk of venous thrombosis, but the role of confounding and the pathophysiology behind these findings are unclear. Objective: To assess the role of hemostatic factors in the relationship between coffee consumption and venous thrombosis. Methods: From a large case–control study, 1803 patients with a first venous thrombosis and 1803 partner controls were included. With conditional logistic regression, odds ratios (ORs) and 95% confidence intervals (CIs) for venous thrombosis were calculated for coffee consumption vs. no coffee consumption. In addition, mean differences in hemostatic factor levels between these groups were calculated in the controls. Results: Coffee consumption yielded a 30% lower risk of venous thrombosis than no coffee consumption (OR 0.7, 95% CI 0.5–0.9). Adjustment for several putative confounders (age, sex, body mass index, smoking, hormonal factors, statin, aspirin, alcohol, malignancy, and chronic disease) yielded an OR of 0.8 (95% CI 0.6–1.1). Results were similar for provoked and unprovoked events, and for deep vein thrombosis and pulmonary embolism. In controls, von Willebrand factor levels were 11 (3–19) IU dL−1 lower and factor (F) VIII levels were 11 (1–21) IU dL−1 lower in coffee consumers than in non-consumers. After adjustment of the risk estimates for these hemostatic factors, the inverse association between coffee consumption and venous thrombosis diminished (OR 1.0, 95% CI 0.7–1.4). There was no association between coffee consumption and anticoagulant proteins, fibrinogen levels, or fibrinolytic markers. Conclusions: Coffee consumption is associated with a lower risk of venous thrombosis, which seems to be mediated through von Willebrand factor and FVIII.

Introduction

Coffee is one of the most widely consumed beverages in the world [1]. It is known to have a stimulating effect on blood pressure and heart rate, owing to its caffeine content, and is often associated with an unhealthy lifestyle [1–3]. However, coffee also contains many active components that may be beneficial to human health, including caffeine, diterpenes, and polyphenols, the most notable of which are chlorogenic acids [3–7]. Previous research has found regular coffee intake to be inversely associated with insulin resistance, inflammatory markers, and endothelial dysfunction [3–5,7]. A weakly protective effect of moderate coffee consumption on cardiovascular disease has also been consistently shown [8,9].

The relationship between coffee consumption and venous thrombosis is less well studied. The Iowa Women’s Health Study found a 14% (95% confidence interval [CI] − 6 to 31) lower risk of venous thrombosis associated with coffee consumption in women over 55 years of age [10]. More recently, the Tromsø study confirmed these findings for participants drinking five or six cups of coffee per day (odds ratio [OR] 0.67, 95% CI 0.45–0.97) [11]. However, although the results of these two prospective studies were similar, the pathophysiology behind their findings was not elucidated. Furthermore, both of these were long-term cohort studies in which coffee consumption was measured at inclusion, whereas the venous event often occurred many years later. Therefore, temporal changes in lifestyle-related factors could not be taken into account.

The aims of our study were to investigate the effect and possible underlying causal mechanism of coffee consumption on the risk of venous thrombosis in a large case–control study in which all information was obtained shortly after the event occurred. As venous thrombosis is associated with abnormal levels of coagulant factors and fibrinolytic activity [12–15], we hypothesized that a decreased risk of venous thrombosis in coffee consumers could be explained by a more favorable coagulation profile.

Methods

Study design

This study was performed using data from the Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study, a large case–control study on risk factors for venous thrombosis. Details of this study have been described previously [16]. Briefly, between March 1999 and September 2004, 4956 consecutive patients aged 18–70 years with a first episode of deep vein thrombosis or pulmonary embolism were included from six anticoagulation clinics in the Netherlands (Amersfoort, Amsterdam, The Hague, Leiden, Rotterdam, and Utrecht). Detailed diagnostic information was obtained from hospital discharge reports and general practitioners. Deep vein thrombosis was objectively confirmed with Doppler ultrasonography. Pulmonary embolism was confirmed with a ventilation perfusion lung scan, spiral computed tomography, or angiogram. Partners of patients were invited to participate as controls if they were aged 18–70 years and had no history of venous thrombosis. In total, 3297 partners participated. Although, from January 2002 to September 2004, 3000 additional controls were recruited by random digit dialing, these were not included in the present analysis, as they were not queried on their coffee intake. For the same reason, patients and partners enrolled after 2002 could not be included. All participants gave written informed consent. The study was approved by the Medical Ethics Committee of the Leiden University Medical Center, Leiden, the Netherlands.

Data collection

Participants completed detailed questionnaires on demographic and lifestyle-related factors, as well as risk factors for venous thrombosis. Items covered in the questionnaires included surgery, pregnancy, plaster cast immobilization and hospitalization in the 3 months before the index date, oral contraceptive or postmenopausal hormone therapy use at the time of the index date, and malignancy in the 5 years before the index date. The index date was defined as the date of diagnosis of venous thrombosis for patients and their partners. Self-reported information was also obtained on weight, height, smoking status, and daily coffee consumption. Body mass index (BMI) was calculated by dividing weight (in kg) by height squared (m2). Smokers were divided into current, previous and never smokers. If the difference between the age at the index date and the age of smoking cessation was ≤ 1 year, the person was considered to be a current smoker. Participants were asked to classify their average coffee consumption in the year before the index date as: no coffee, 1–4 cups per day, 5–9 cups per day, or > 10 cups per day.

Blood collection

At least 3 months after anticoagulation therapy was discontinued, or during anticoagulation therapy in patients who were treated for > 1 year, patients and their partners visited the anticoagulation clinic for an interview and blood sampling. From the blood sample, levels of major natural anticoagulants (antithrombin, protein S and protein C), levels of procoagulant factors (fibrinogen, von Willebrand factor, and factor (F) VIII) and fibrinolytic potential (clot lysis time) were assessed. A detailed description of the assays can be found elsewhere [17–19]. All assays were performed in automated machines by laboratory technicians who were unaware of the case–control status of the samples. As blood was collected after the venous thrombotic event, one could argue that levels of coagulation factors (FVIII and von Willebrand factor) in our study may be affected by acute-phase reactions. However, blood was collected at least 3 months after the venous thrombotic event, by which time the effects of the acute-phase reaction would have worn off [20–23].

Inclusion criteria

Of the 4956 patients, 3202 were included before 2002, and were thus queried on their coffee consumption (Fig. 1). Of these 3202 eligible patients, 1870 had a partner who met the inclusion criteria and was willing to participate. After exclusion of couples for whom information on coffee consumption was missing for one or both individuals (n = 67), 1803 complete couples remained.

Figure 1.

 Selection of eligible couples in the MEGA study.

Statistical analysis

Conditional logistic regression was used to calculate ORs and 95% CIs as measures of relative risks for venous thrombosis in coffee consumers as compared with non-consumers. This method fully takes matching into account, with adjustment for all unmeasured factors for which couples tend to be similar (e.g. socioeconomic class) [24]. Analyses were additionally adjusted for age, sex, BMI, and smoking status. To avoid any apparent effect of coffee consumption on the risk of venous thrombosis being attributable to residual lifestyle-related confounders with which coffee consumption is associated [2,25], we also included statin use, aspirin use, hormonal factors (defined as oral contraception, postmenopausal hormone therapy, or pregnancy), alcohol consumption, malignancy and chronic disease (defined as liver disease, kidney disease, rheumatoid arthritis, multiple sclerosis, heart failure, angina pectoris, or intermittent claudication) as potential sources of confounding. Hormonal factors were added to the model as a dichotomous variable in which all men were classified as unexposed. This method has been explained previously [26,27]. For similar reasons, we analyzed levels of coffee consumption for individuals with and without malignancy and individuals with and without chronic diseases separately. In addition, as pregnant women are advised to reduce their coffee consumption, we did the same for pregnant and non-pregnant women. Patients with deep vein thrombosis and pulmonary embolism and patients with provoked and unprovoked venous thrombotic events were combined in most analyses, but also analyzed separately. Provoked venous thrombosis was defined as venous thrombosis preceded by surgery, plaster cast immobilization, bed rest, hospitalization or leg injury in the 3 months before the index date or long-distance travel in the 2 months before the index date.

In a mediation analysis, we explored whether the protective effect of coffee consumption on the risk of venous thrombosis could be explained by a more favorable coagulation profile. We calculated the mean differences in levels of major natural anticoagulant factors (protein S, protein C, and antithrombin) and procoagulant factors (fibrinogen, von Willebrand factor, and FVIII) and fibrinolysis (clot lysis time) with regard to overall coffee consumption, 1–4 cups of coffee per day and 5–9 cups of coffee per day vs. no coffee consumption, and adjusted for all of the above-mentioned confounders. Controls who had been using anticoagulation therapy at the time of blood drawing (n = 9) were excluded from the analysis of protein S, protein C, and clot lysis time, as these factors are vitamin K-dependent. Finally, we repeated the earlier conditional logistic regression analysis in patients and their partners, this time including hemostatic factors associated with coffee consumption in the regression models to assess whether these factors mediate the effect of coffee consumption on the risk of venous thrombosis. All statistical analyses were performed with spss for Windows, release 17.0 (SPSS, Chicago, IL, USA).

Results

A total of 3606 participants (1803 patients and their partners) were included in this study. Their characteristics are summarized in Table 1. The mean age at enrollment was 50 years (range, 18–70 years) in both patients and controls. The majority of participants, 1627 patients (90%) and 1675 controls (93%), were coffee consumers. Of these participants, 1162 patients and 1164 controls drank 1–4 cups of coffee per day, 401 patients and 433 partners drank 5–9 cups of coffee per day, and 64 patients and 78 partners drank ≥ 10 cups of coffee per day. As patients were matched to their partners, only exposure-discordant couples (i.e. couples in whom coffee consumption differs between patients and partners) were relevant to the univariate risk analyses [28]. In total, there were 98 couples in whom the patient drank coffee without the partner, and 146 couples in whom the partner drank coffee without the patient. In the category of 1–4 cups vs. no coffee consumption, there were 66 couples in whom the patient drank coffee without the partner, and 98 couples in whom the partner drank coffee without the patient. For 5–9 cups vs. no coffee, these numbers were 25 and 42, respectively. In the category of ≥ 10 cups of coffee per day, there were only seven and six discordant couples. These numbers were too small to analyze separately. Therefore, the latter two categories were pooled into a single category of ≥ 5 cups of coffee per day. Participants who did not drink coffee were younger and less likely to be overweight than the other participants. In coffee consumers, the prevalence of oral contraception use and pregnancy was lower than in non-consumers, whereas the rates of postmenopausal hormone therapy use, alcohol consumption and malignancy were higher than in non-consumers. There was no marked association between coffee consumption and other participant characteristics.

Table 1.   Clinical characteristics of the MEGA case–control study
 PatientsPartners
OverallNo coffee consumption1–4 cups per day≥ 5 cups per dayOverallNo coffee consumption1–4 cups per day≥ 5 cups per day
  1. HRT, hormone replacement therapy; OC, oral contraception; VT, venous thrombosis. Continous variables are presented as mean (range); categorical variables are presented as number (%). Data were missing for some participants in some subgroups. *Classical risk factors include surgery, malignancy, immobilization, trauma, plaster cast, and recent travel. †Defined as liver disease, kidney disease, rheumatoid arthritis, multiple sclerosis, heart failure, angina pectoris, or intermittent claudication.

General characteristics
 Total1803 (100)176 (10)1162 (64)465 (26)1803 (100)128 (7)1164 (65)511 (28)
 Men902 (50)37 (21)545 (47)320 (69)902 (50)62 (48)489 (42)351 (69)
 Age at enrollment (years)50 (18–70)37 (19–69)52 (18–70)48 (23–70)50 (18–70)39 (20–70)51 (18–70)49 (20–70)
 Body mass index (kg m2)26 (16–57)25 (17–42)26 (17–57)27 (16–52)26 (17–48)25 (17–43)26 (17–48)26 (17–42)
Classical VT risk factors
 Present*1030 (57)102 (58)680 (59)248 (53)364 (20)27 (21)216 (19)121 (24)
 OC use (% in women)487 (54)89 (65)327 (53)71 (49)157 (18)20 (30)114 (17)23 (14)
 Pregnancy (% in women)80 (9)28 (20)45 (7)7 (5)16 (2)2 (3)14 (2)0 (0)
 HRT use (% in women)46 (5)4 (3)38 (6)4 (3)62 (7)1 (2)50 (7)11 (7)
Arterial cardiovascular risk factors
 Overweight778 (44)57 (34)502 (44)219 (48)715 (41)35 (29)476 (42)204 (41)
 Obesity333 (19)34 (21)203 (18)96 (21)258 (15)20 (17)165 (15)73 (15)
 Previous smoking607 (34)42 (24)430 (37)135 (30)527 (29)34 (27)359 (31)134 (27)
 Current smoking611 (34)37 (21)340 (29)234 (51)599 (34)31 (25)304 (26)264 (52)
 Alcohol consumption1442 (80)98 (56)951 (82)393 (85)1515 (84)93 (73)976 (84)446 (88)
 Alcohol abstinence357 (20)77 (44)209 (18)71 (15)284 (16)34 (27)186 (16)64 (13)
Comorbidity
 Malignancy162 (9)11 (6)124 (11)27 (6)24 (1)2 (2)16 (1)6 (1)
 Other chronic diseases†141 (8)13 (7)103 (9)25 (5)102 (6)7 (6)71 (6)24 (5)
 Statin use61 (3)5 (3)40 (3)16 (3)94 (5)6 (5)65 (6)23 (5)
 Aspirin use54 (3)4 (2)37 (3)13 (3)58 (3)1 (1)42 (4)15 (3)

Overall, daily coffee consumption was associated with a lower risk of venous thrombosis than no coffee consumption (OR 0.66, 95% CI 0.50–0.86) (Table 2). Adjustment for age, sex, BMI, and smoking, as well as hormonal factors, statin use, aspirin use, alcohol intake, malignancy, and chronic disease, yielded an OR of 0.75 (95% CI 0.55–1.04). The fully adjusted ORs for 1–4 cups of coffee and ≥ 5 cups of coffee per day were 0.75 (95% CI 0.54–1.03) and 0.77 (95% CI 0.54–1.10), respectively. Restricting the overall analysis to deep vein thrombosis or pulmonary embolism and analyzing patients with and without classical provoking risk factors for venous thrombosis separately did not change the results (Table 3).

Table 2.   Risk of venous thrombosis by categories of coffee consumption, adjusted for lifestyle-related factors
  Patients, n (%)P+/C−Controls, n (%)P−/C+Odds ratio (95% CI)*Odds ratio (95% CI)†Odds ratio (95% CI)‡Odds ratio (95% CI)§Odds ratio (95% CI)¶
  1. CI, confidence interval. P+/C−: the number of patients (P) who drank coffee without the partner control (C). P−/C+: the number of controls who drank coffee without the patient. Odds ratios were calculated as the ratios of discordant pairs. *Adjusted for partnership. †Adjusted for partnership, age, and sex. ‡Adjusted for partnership, age, sex, body mass index (BMI), and smoking. §Adjusted for partnership, sex, age, BMI, smoking, hormonal factors (oral contraception, postmenopausal hormone therapy, and pregnancy), statin, and aspirin. ¶Adjusted for partnership, age, sex, BMI, smoking, hormonal factors (oral contraception, postmenopausal hormone therapy, and pregnancy), statin, aspirin, alcohol consumption, malignancy, and other chronic diseases.

No coffee consumption128 (7)146176 (10)98ReferenceReferenceReferenceReferenceReference
Coffee consumption1675 (93)981627 (90)1460.66 (0.50–0.86)0.65 (0.50–0.85)0.62 (0.47–0.82)0.70 (0.51–0.94)0.75 (0.55–1.04)
 1–4 cups per day1164 (65)661162 (64)980.68 (0.52–0.89)0.67 (0.51–0.88)0.65 (0.49–0.86)0.70 (0.52–0.95)0.75 (0.54–1.03)
 ≥ 5 cups per day511 (28)32465 (26)480.62 (0.47–0.83)0.59 (0.44–0.80)0.55 (0.40–0.75)0.67 (0.48–0.95)0.77 (0.54–1.10)
Table 3.   Risk of venous thrombosis according to coffee consumption; subgroup analyses
  Patients, n (%)P+/C−Controls, n (%) P−/C+Odds ratio (95% CI)*Odds ratio (95% CI)†Odds ratio (95% CI)‡Odds ratio (95% CI)§Odds ratio (95% CI)¶
  1. CI, confidence interval; DVT, deep vein thrombosis; PE, pulmonary embolism; VT, venous thrombosis. P+/C−: the number of patients (P) who drank coffee without the partner control (C). P−/C+: the number of partner controls who drank coffee without the patient. Odds ratios were calculated as the ratios of discordant pairs. *Adjusted for partnership. †Adjusted for partnership, age, and sex. ‡Adjusted for partnership, age, sex, body mass index (BMI), and smoking. §Adjusted for partnership, sex, age, BMI, smoking, hormonal factors (oral contraception, postmenopausal hormone therapy, and pregnancy), statin, and aspirin. ¶Adjusted for partnership, age, sex, BMI, smoking, hormonal factors, statin, aspirin, alcohol intake, malignancy, and other chronic diseases.

Provoked VT
 No coffee consumption102 (10)48128 (7)891 (Reference)1 (Reference)1 (Reference)1 (Reference)1 (Reference)
 Coffee consumption928 (90)1675 (93)0.54 (0.38–0.77)0.54 (0.38–0.77)0.54 (0.37–0.77)0.63 (0.42–0.92)0.61 (0.40–0.93)
Unprovoked VT
 No coffee consumption74 (10)50128 (7)571 Reference1 (Reference)1 (Reference)1 (Reference)1 (Reference)
 Coffee consumption697 (90)1675 (93)0.88 (0.60–1.28)0.86 (0.58–1.27)0.70 (0.46–1.07)0.70 (0.46–1.08)0.75 (0.49–1.14)
DVT only
 No coffee consumption99 (10)54128 (7)851 Reference1 (Reference)1 (Reference)1 (Reference)1 (Reference)
 Coffee consumption936 (90)1675 (93)0.64 (0.45–0.89)0.63 (0.44–0.88)0.58 (0.40–0.83)0.59 (0.41–0.85)0.60 (0.41–0.87)
PE ± DVT
 No coffee consumption77 (10)44128 (7)611 (Reference)1 (Reference)1 (Reference)1 (Reference)1 (Reference)
 Coffee consumption691 (90)1675 (93)0.72 (0.49–1.06)0.72 (0.48–1.06)0.64 (0.42–0.98)0.64 (0.42–0.98)0.70 (0.45–1.09)

To further examine whether our results could be explained by residual confounding factors, we compared levels of coffee consumption between individuals with and without malignancy, individuals with and without chronic diseases, and pregnant and non-pregnant women. Of the individuals with malignancy, 93% consumed coffee, whereas 91% of individuals without malignancy were coffee consumers. For chronic diseases, these percentages were 92% and 91%. However, as expected, the percentage of coffee consumers was higher among non-pregnant women (90%) than among non-pregnant women (69%). To test whether this difference in coffee consumption could explain our findings, we repeated our main analysis after excluding pregnant women. This analysis gave results similar to those of our overall analysis: a 25% lower risk of venous thrombosis (OR 0.75, 95% CI 0.54–1.04), which did not change after adjustment for all of the above-mentioned potential confounders (OR 0.69, 95% CI 0.46–1.04).

Next, we calculated the mean differences in levels of major natural anticoagulant factors (protein C, protein S, and antithrombin) and procoagulant factors (fibrinogen, von Willebrand factor, and FVIII) and fibrinolysis (clot lysis time) in control subjects for different categories of coffee consumption (Table 4). After adjustment for age, sex, BMI, smoking, hormonal factors, aspirin use, statin use, alcohol intake, malignancy, and chronic disease, no association was found between coffee consumption and anticoagulant factor levels, fibrinogen levels, or fibrinolysis. However, von Willebrand factor and FVIII levels remained lower in coffee consumers than in non-consumers.

Table 4.   Association between different amounts of coffee and hemostatic factors in partner controls
Hemostatic factorMeanOverall coffee consumption1–4 cups per day≥ 5 cups per day
No coffeeMean difference*Mean difference†Mean difference†Mean difference†
  1. CLT, clot lysis time. *Adjusted for age and sex. †Adjusted for age, sex, body mass index, smoking, hormonal factors (oral contraception, postmenopausal hormone therapy, and pregnancy), statin use, aspirin use, alcohol consumption, malignancy, and chronic disease. ‡Vitamin K antagonist use excluded.

Anticoagulant
 Protein S (IU dL−1)‡101−1 (−5 to 4)−1 (−5 to 3)−1 (−5 to 3)−1 (−6 to 3)
 Protein C (IU dL−1)‡1180 (−4 to 5)1 (−3 to 6)2 (−3 to 7)1 (−4 to 6)
 Antithrombin (IU dL−1)1061 (−1 to 3)1 (−2 to 3)1 (−2 to 3)1 (−1 to 4)
Procoagulant
 Fibrinogen (g L−1)3.30.0 (−0.2 to 0.1)0.0 (−0.2 to 0.1)0.0 (−0.2 to 0.1)0.0 (−0.2 to 0.1)
 Factor VIII (IU dL−1)113−11 (−19 to −3)−5 (−13 to 3)−4 (−12 to 4)−9 (−17 to 0)
 von Willebrand factor114−11 (−21 to −1)−7 (−17 to 4)−6 (−16 to 5)−9 (−21 to 2)
Fibrinolytic
 CLT (min)‡800.8 (−7.5 to 9.2)2.3 (−6.3 to 11.0)−1.1 (−2.3 to 0.0)2.3 (−2.3 to 9.7)

In a final analysis, we repeated our previous analysis, with calculation of the risk of venous thrombosis in coffee consumers as compared with non-consumers, this time including procoagulant factors as covariates (Table 5). The inverse association between venous thrombosis and coffee consumption adjusted for partnership alone (OR 0.66, 95% CI 0.50–0.86) was attenuated after adjustment for FVIII (OR 0.99, 95% CI 0.69–1.43). Additional adjustment for von Willebrand factor did not change these results (OR 0.95, 95% CI 0.65–1.35). Results were similar for 1–4 cups of coffee per day and ≥ 5 cups of coffee per day. Reversing the order in which the covariates were added to the model (i.e. starting with von Willebrand factor alone and then adding FVIII) did not change the results. Adjusting the risk estimates for FVIII, von Willebrand factor and all of the above-mentioned lifestyle-related factors also did not change the results (OR 0.95, 95% CI 0.61–1.48).

Table 5.   Risk of venous thrombosis by categories of coffee consumption, adjusted for hemostatic factors
  Patients, n (%)P+/C−Controls, n (%)P−/C+Odds ratio (95% CI)*Odds ratio (95% CI)†Odds ratio (95% CI)‡
  1. P+/C−: the number of patients (P) who drank coffee without the partner control (C). P−/C+: the number of controls who drank coffee without the patient. * Adjusted for partnership. †Adjusted for partnership and factor VIII ‡Adjusted for partnership factor VIII and von Willebrand factor.

No coffee consumption128 (7)146176 (10)98ReferenceReferenceReference
Coffee consumption1675 (93)981627 (90)1460.66 (0.50–0.86)0.99 (0.69–1.43)0.95 (0.65–1.35)
 1–4 cups per day1164 (65)661162 (64)980.68 (0.52–0.89)0.95 (0.66–1.38)0.92 (0.63–1.33)
 ≥ 5 cups per day511 (28)32465 (26)480.62 (0.47–0.83)1.07 (0.72–1.59)0.98 (0.66–1.46)

Discussion

In our study, we found that the risk of venous thrombosis was lower in coffee-consuming participants than in participants with no coffee consumption, even after adjustment for many lifestyle-related factors. This inverse association with coffee consumption seemed to be mediated through levels of von Willebrand factor and FVIII, which were lower in coffee consumers than in non-consumers in the control group. After adjustment of the risk estimates for these procoagulant factors, the inverse association between coffee consumption and venous thrombosis diminished.

Our results are in line with previous studies that have shown that coffee consumption is associated with a more favorable cardiovascular profile [4,5,7]. However, to our knowledge, this is the first study to investigate the effect of coffee consumption on the risk of venous thrombosis while taking hemostatic factors into account. We found a lower risk of venous thrombosis in coffee consumers than in non-consumers, which seemed to be mediated through levels of von Willebrand factor and FVIII. These findings are plausible, as both epidemiologic and murine studies have shown that increased levels of these factors increase the risk of venous thrombosis [14,15]. It is not possible to determine the precise biological mechanisms underlying this observation from epidemiologic data. However, we can speculate on a possible explanation for our findings. Coffee contains active compounds, among which are polyphenols, which have anti-inflammatory properties and improve endothelial health [4,29,30]. Furthermore, polyphenols have been found to inhibit platelet activation [29,31]. As both endothelium and platelets are sources of von Willebrand factor, these polyphenols might constitute part of the underlying mechanisms explaining our results [21,32].

Our results differ from the findings of two previous studies on coffee and hemostatic factors [33,34]. A small, randomized controlled trial performed in the Netherlands in the 1990s showed no effect of 9 weeks of coffee consumption on the levels of FVIII [34]. In addition, an older experiment found an increase in fibrinolysis immediately after coffee consumption [33]. However, owing to differences in study design (e.g. measurements of FVIII antigen vs. FVIII activity), and the fact that coffee consumption has been shown to have an immediate negative effect and a beneficial long-term effect on endothelial function [7,35], these findings do not necessarily contradict our results. Nevertheless, as coffee is the leading worldwide beverage after water [1], it is important to ascertain the effect of coffee consumption before definite claims of a protective nature are made. Therefore, a randomized controlled trial would be helpful to confirm a long-term lowering effect of coffee consumption on procoagulant factors.

Owing to the observational nature of our study, we cannot rule out the possibility that our results may be partially explained by residual confounding. Indeed, it is possible that the participants in our study who did not drink coffee had poorer health than the coffee-consuming participants. If this poor health is associated with increases in von Willebrand factor and FVIII levels and an increased risk of venous thrombosis, the protective effect of coffee will be overestimated. However, from the clinical characteristics (Table 1), the non-consumers in our study appeared to be healthier (younger, and less likely to be overweight or to smoke) than the coffee-consuming participants. Furthermore, the inverse association between coffee consumption and the risk of venous thrombosis remained after adjustment for many lifestyle-related factors and in individuals with and without chronic diseases, making residual confounding an unlikely explanation for our results.

A possible source of bias in case–control studies is that information on coffee consumption is obtained after venous thrombosis has occurred. It is therefore possible that the reported coffee intake reflects current consumption, rather than coffee consumption before the index date. We reduced this problem by minimizing the time between the event and the questionnaire. In addition, as patients and their partners received, and are likely to have filled out, the questionnaire at the same time, we lessened recall bias.

A strength of our research is the large study size. Data were collected in the same manner for all participants, and all venous thrombotic events were objectively diagnosed. A limitation of our study was that it was not possible to study participants consuming ≥ 10 cups of coffee per day separately, owing to the small numbers of participants in this category. Another limitation was that adjustment for many potential confounders yielded imprecisely estimated results with wide CIs. However, all risk estimates point in the same direction, and our results are fully in line with the findings of the previous two prospective studies on this topic, reducing the likelihood that our results can be explained by chance. Levels of coagulation factors in patients were measured after the venous thrombotic event occurred. However, as the effect of coffee consumption on the levels of hemostatic factors was only studied in controls, this cannot have affected the lower levels found in coffee consumers. Finally, one of the apparent problems of partner controls is that they are (mostly) of the opposite sex. However, we analyzed our data with conditional logistic regression methods in which adjustment for sex is possible and was performed [26,27]. In addition, we controlled for the female-specific factors oral contraception, postmenopausal hormone therapy, and pregnancy. Therefore, we consider it unlikely that the results can be explained by choosing partner controls.

In conclusion, our findings show that daily coffee consumption is associated with a lower risk of venous thrombosis than no coffee consumption. This protective effect seems, at least partially, to be mediated through lower von Willebrand factor and FVIII levels.

Disclosure of conflict of interests

This research was supported by the Netherlands Heart Foundation (NHS 98.113), the Dutch Cancer Foundation (RUL 99/1992), and the Netherlands Organization for Scientific Research (912-03-033|2003). W. M. Lijfering is a Postdoctoral Researcher of the Netherlands Heart Foundation (2011 T 12). The funding organizations did not play a role in the design and conduct of this study, the collection, management, analysis and interpretation of the data, or the preparation, review or approval of the manuscript.

Ancillary