The relationship between lifestyle factors and venous thromboembolism among women: a report from the MISS study

Authors


Pelle G. Lindqvist, Department of Obstetrics and Gynaecology, Karolinska University Hospital, Huddinge, SE-14186 Stockholm, Sweden. E-mail: pelle.lindqvist@ki.se

Summary

There has been a great advance in our knowledge of the role that thrombophilic factors play in the risk of venous thromboembolic events (VTE). However, the effect of lifestyle factors on VTE has been inadequately explored in large scale studies of women. This cohort study comprised one thousand native Swedish women for each age year between 25 and 64 inclusive (total = 40 000) drawn from the South Swedish population registry for 1990 (= 40 000), who were followed for a mean of eleven years. Seventy-four percent completed a questionnaire at the inception of the study (= 29 518) and 24 098 women responded to a follow-up inquiry between the years 2000–2002. The main outcome was the relationship between VTE and physical exercise, smoking habits, and alcohol consumption. Moderate drinkers of alcohol (10–15 g/d) and women engaged in strenuous exercise were at half the risk of VTE compared to those who consumed little or no alcohol or lived a sedentary life. Heavy smoking was associated with a 30% increased risk of VTE. Lifestyle factors have a major impact on the risk of VTE. Women non-smokers who were physically active and who consumed alcohol in moderation were at a lower risk of VTE.

Venous thromboembolic events (VTEs) are a major cause of female morbidity and mortality. Life expectancy after a VTE is considerable shortened (Heit et al, 1999). The yearly risk of VTE is reported to be around 1/1000, and is lower for young women and higher for old (Rosendaal, 1999). Risk factors for VTE include advancing age, the presence of inherited or acquired thrombophilias, hypofibrinolysis, surgery, hormonal use (combined contraceptives or hormone replacement therapy), immobilization, overweight, pregnancy, and malignancy (Dahlbäck, 2000; Lisman et al, 2005).

Lifestyle patterns may affect the risk of VTE. Several large cohort studies have shown smoking to be related to an increased risk of VTE (Goldhaber et al, 1997; Hansson et al, 1999; Juul et al, 2004). Smoking has also been associated with an increased risk of VTE in gravidae (Lindqvist et al, 1999a; Larsen et al, 2007), which was found to be dose-dependent (Lindqvist et al, 1999a). Increased body mass index (BMI) was also found to be a risk factor for VTE among women (Goldhaber et al, 1983, 1997; Juul et al, 2004; Canonico et al, 2006). Light to moderate alcohol consumption (<30 g/d) and exercise have been related to a lower risk of cardiovascular events, but little is known of the risk of thromboembolism on the basis of large cohort studies.

In the Melanoma Inquiry of Southern Sweden (MISS) study we followed 40 000 women prospectively for a mean period of eleven years and obtained information on their physical activity, smoking habits, alcohol consumption, as well as such established risk factors for VTE as age, BMI, malignancy, parity, and hormonal treatment.

This longitudinal cohort study was carried out in order to assess how lifestyle factors influence the risk of VTE among women.

Material and methods

The study was approved by the Ethics Committee of Lund University (LU 632–03). The MISS study was initiated in 1990. One thousand native Swedish women for each age year between 25 and 64 inclusive (total = 40 000) and no history of malignancy were chosen from the general population registry of the South Swedish Health Care Region by computerized random selection. Twenty-seven women could not be contacted, leaving 39 973 as the study cohort. This represents 20% of the South Swedish female population in the selected age groups. The women were invited to take part in a standardized written questionnaire concerning risk factors for malignant melanoma. The initial inquiry was made between 1990 and 1992 with a written follow-up conducted between 2000 and 2002. The questionnaire inquired into several items of potential interest for thrombosis risk, such as parity, participation in sports, smoking habits, alcohol consumption, use of combined oral contraceptive (COC), age at menopause, marital status, and educational level. The follow-up questionnaire posed similar questions, also including physical activity, weight, height, and use of medication for more than one month.

The unique personal identification number assigned to each Swedish resident enabled us to determine all deaths and causes of death in the National Cause of Death (NCD) register. The incidence of VTE was determined both through responses to the follow-up questions regarding disease and long-term medication, and by means of entries in the National Patient Registry (NPR), which records all women who have been hospitalized.

Those women who were invited to participate in the MISS study (n = 39 974) were sought by International Classification of Diseases, Ninth Revision (ICD 9) code numbers for VTE; 634G or H, 635G or H, 636G or H, 637G or H, 638G or H, 639G or H, 671D, 671E, 671F, 673C, 451B, 452, 325, 437G, 572B, 453 (C, D, W, or X) or 415B or the corresponding ICD 10 codes O082, O087, O223, O871, O873, O879, O225, O229, O882, I802, I803, I81*, I82*, I636, I676, K550, I26*. The unique personal identification number and the above-mentioned diagnosis codes were cross-matched in the NCD and NPR. Thus, the registered number of VTEs was established for both those women included and those not included in the study. The register was used to identify all VTEs that occurred up to December 31, 2002. Malignancy was defined as a diagnosis of any type of malignancy prior to that date. Information was gathered from both Regional and National Cancer Registries. Vital statistics were determined up to December 31, 2002.

Weight and height were recorded at the second interview, and BMI was calculated as kg/m2. BMI was classified into three groups: <25 (reference), 25 to <30 (overweight), and 30+ (obese).

Smoking habits were recorded at the inception of the study. Smoking habits were grouped both as never- or ever-smoked, and categorized into the following subgroups: non-smokers (reference); those who had smoked less than 100 000 cigarettes in their lifetime; and those who had smoked at least 100 000 cigarettes on the basis of how participants had categorized their cigarette smoking in mean consumption at five-year intervals.

Drinking habits were collected at the initial questionnaire in terms of beer, wine, and liquor consumed per month. Alcohol consumption was categorized into five subgroups depending on equivalent amounts of alcohol consumed, no consumption (reference); <5 g/d; 5 to <10 g/d; 10 to 15 g/d (moderate consumption); and >15 g/d. Alcohol consumption was also dichotomized into those who characterized themselves as total abstainers and those who did not (reference). Parity was initially categorized as nulliparous, 1-para, 2-para, or at least 3-para. Since there was no difference in risk between 1- and 2-para, the categorization was changed to nulliparous, 1- or 2-para (reference), or at least 3-para.

Combined oral contraceptive use was assessed using data from both inception and follow-up interview. At inception they were asked; ‘Have you or are you using COC?’ If the answer was yes, they were asked to specify when they commenced COC use and when they stopped COC. The follow-up interview participants were also asked to specify the years of COC use. Women negating or not answering any of these questions were classified as never-users, those that indicated use only prior to the study period were classified as prior users, and those indicating use during the study period by any question were classified as current users.

The level of regular exercise was obtained through the question –“In addition to your usual work, do you regularly exercise?” No; Do you go for a walk once a day; Do you go for a walk several times a week; Do you bicycle, swim, Participate in gymnastics, dancing, or similar activities one or more times a week (i.e., strenuous exercise). Physical exercise was than divided into three categories: none, go for a walk one or more times a week, or strenuous exercise.

Statistics

The characteristics of the women in this study are presented as mean and standard deviation (SD) or by number and percentage. Analysis of selected variables was performed with Cox regression analysis using 95% confidence intervals (CI). The presence of VTE was used as a dependant variable and “time-at-risk” as a time variable. Time-at-risk was defined as time from initial participation in the study to VTE, death, or December 31, 2002, whichever came first. As increasing age is a strong risk factor for VTE, it was included for adjustment in all risk estimates. Age was introduced as a categorized variable. All calculations were performed using Statistical Package for the Social Sciences (spss) software (spss Inc., Chicago IL, USA) and P values <0·05 were considered statistically significant.

Results

Table I presents the demographic characteristics of the women in the study cohort and indicates the increased incidence of VTE among women with less than nine years of schooling, those who were unmarried or widows, and both nulli- and multiparous (≥3) individuals.

Table I.   Demographic characteristics of women with and without VTE from inquiry at inclusion.
 Women with VTEWomen without VTEAge-adjusted
(n = 312)(n = 29 206)RR95% CI
  1. All information, except cancer (follow-up data) was gathered at inclusion of study and analysed with Cox regression analysis and presented with age-adjustment.

  2. RR, relative risk.

Education (years)
 <99554881·61·2–2·2
 93927341·30·9–2·0
 10–126575941·30·9–1·8
 ≥127496231·0reference
 Other3937671·10·7–1·6
Marital status
 Unmarried3025331·61·1–2·4
 Married22622 7841·0reference
 Divorced2226450·80·5–1·2
 Widowed3211211·71·2–2·6
 Missing information21231·50·4–5·9
Parity
 05849141·61·2–2·2
 1–214416 3131·0reference
 ≥311079791·41·1–1·8
Menarche (years)    
 ≤1064711·70·7–3·8
 11–1314115 7931·0reference
 ≥1414912 4261·10·9–1·4
 Missing information165162·91·8–4·9
Cancer during study period
 Yes7016044·23·2–5·5
 No24227 6021·0reference

Women who were diagnosed with cancer during the study period were at a fourfold increased risk of VTE. Data was gathered from 29 518 women (74%) (317 290 women-years) at the first interview and from 24 098 women at the second interview (262 429 women-years).

Table II indicates the age-adjusted analysis of selected lifestyle variables and the VTE risk. Those who completely abstained from alcohol and women who had smoked more than 100 000 cigarettes over the course of their lifetime were at a 30% increased risk of VTE. Women living a sedentary life were at twice the risk of those who were active. As compared to women who drank 10–15 g alcohol a day (the equivalent of between a half to one glass of wine per day), women drinking <5 g/d had a more than doubled risk. There was a statistically significant lower risk among those who previously used COC (RR 0·7, 95% CI 0·5–0·9).

Table II.   Lifestyle factors and risk of VTE, bivariate age-adjusted analysis.
 Women with VTEWomen without VTEAge-adjusted
RR95% CI
  1. COX regression analysis with age-adjustment.

  2. RR, relative risk.

Data from inclusion 1990–1992
 Smoker
  Ever smoked17717 1681·10·9–1·4
  Never smoked13511 8311·0reference
 Total number of cigarettes
  Never smoked13712 1801·0reference
  <100 0008395701·00·7–1·3
  ≥100 0008667381·31·0–1·7
  Missing information67180·80·3–1·7
 Competing sport (present or prior)
  No23719 2101·0reference
  Yes3325961·20·8–1·7
  Missing information4274001·20·6–2·5
 Abstain from alcohol
  No25225 7451·0reference
  Yes5932831·41·0–1·8
  Missing information11780·50·06–3·2
 Alcohol consumption (g/d)
  No6135541·0reference
  <516214 6740·80·6–1·1
  5 to <103748580·60·4–1·0
  10 to <15818120·40·2–0·8
  ≥15912790·60·3–1·1
  Missing information3530290·80·5–1·2
 Combined oral contraceptives
  Never used17010 2221·0reference
  Use during study-period3437411·30·8–2·1
  Prior use10815 2430·70·5–0·9
Data from follow-up interview 2000–2002
 Present physical activity
  No2516311·0reference
  Once in a while4960150·70·4–1·1
  More than 1 h/week7611 4460·50·3–0·8
  Missing information6947870·70·4–1·1
 Present regular exercise
  No2620331·0reference
  A walk one to two times a week10110 7050·70·5–1·1
  Strenuous exercise4276580·40·3–0·7
  Missing information5034831·10·7–1·8
 Body mass index (BMI)
  <258213 0011·0reference
  25 to <306967111·41·0–2·0
  >305024292·82·0–4·0
  Missing information1817381·20·7–2·0

Table III shows the results of multivariate analysis of the risk of VTE. Since information on BMI and physical exercise were gathered at the second interview, we present two different models. The multivariate analysis changed the estimates only marginally compared to only age-adjustment alone. Sedentary women were at twice as much risk as those who were physically active. Heavy smokers (i.e., >100 000 cigarettes smoked) were at a 30% significantly increased risk in comparison to non-smokers. The age-related risk of VTE increased by about 6% (95% CI 4–7%) for each additional year of age. Cancer during the study period was an independent risk factor with RR between 3·3 and 4·3 in all calculations. Nulliparity was significantly related to VTE in all models with a RR between 1·4 and 1·7.

Table III.   Multivariate analysis of selected lifestyle factors and the risk of venous thromboembolism (VTE).
 Model 1Model 2
RR95%CIRR95%CI
  1. Model 1: Each variable adjusted for diagnosis of cancer during the study period, age, and parity.

  2. Model 2: Adjusted for age, cancer, parity and all above variables, analysed in population that answered interview 2.

  3. Cox-regression analysis.

  4. RR, relative risk.

Cigarettes
 Never smoked1·0reference1·0reference
 <100 0001·00·7–1·31·10·8–1·5
 ≥100 0001·31·0–1·71·31·0–1·9
 Missing information0·70·3–1·60·80·3–2·5
Alcohol consumption (g/d)
 No alcohol1·0reference1·0reference
 <50·90·6–1·20·90·6–1·3
 5 to <100·60·4–1·00·60·4–1·1
 10 to <150·40·2–0·80·40·2–0·96
 ≥150·50·3–1·10·70·3–1·5
 Missing information0·80·5–1·20·80·5–1·3
Combined oral contraceptives
 Never used1·0reference1·0reference
 Current during study-period1·30·8–2·11·40·8–2·3
 Prior use0·70·5–0·90·70·5–1·0
Present regular exercise
 No1·0reference1·0reference
 Go for a walk a couple of times   a week0·70·4–1·10·80·5–1·2
 Strenuous exercise every week0·40·3–0·70·50·3–0·9
 Missing information0·70·4–1·10·70·4–1·2
Body mass index (BMI)
 <251·0reference1·0reference
 25–301·41·0–2·01·31·0–1·9
 >302·92·0–4·12·51·8–3·6
 Missing information1·20·7–2·01·10·6–1·8

Women who completely abstained from alcohol were older than those who drank (52 years vs. 44 years, < 0·001), had a higher BMI (26·1 vs. 24·8, < 0·001), a sedentary lifestyle (17·5% vs. 7·8%, < 0·001), and had smoked fewer cigarettes (< 0·001).

Figure 1 presents the incidence of VTE by age groups for those women included in the study (participants, n = 29 518) and for those who were not (non-participants, n = 10 455). There were 312 VTE cases for 317 290 women-years among the study participants and 165 cases for 132 161 women-years in the non-participants. Thus, the non-participants were at a 27% increased risk of VTE (OR 1·27 95% CI 1·05–1·5). The difference was mainly attributed to age-groups above 60.

Figure 1.

 Incidence of venous thromboembolism in study participants and non-participants.

Discussion

The present study found that several lifestyle factors were related to the risk of VTE. A sedentary life was linked to a doubled risk of VTE as compared to those who took walks more than twice a week. Heavy smoking and being overweight were associated with a 30% increased risk of VTE, while the risk for obese women increased by threefold. Moderate alcohol drinkers were at half the risk as those with little or no alcohol consumption. Women who had previously used COCs were at a 30% lower risk than those who had never done so.

The twofold lower risk among active women (moderate exercisers) might be due to an increase in endothelial nitric oxide synthase, which has a beneficial effect on vessel walls (Indolfi et al, 2002). Plasma viscosity is determined largely by fibrinogen, and a negative relationship between fibrinogen and physical activity has been established (Folsom et al, 1991; Stratton et al, 1991). Cross-sectional studies have shown an “antithrombotic” effect of long term regular exercise, i.e., lower plasma levels of fibrinogen and increased fibrinolytic capacity (increased tissue plasminogen activator (tPA) activity, decreased plasminogen activator inhibitor type-I, and tPA mass concentration) (Eliasson et al, 1996; el-Sayed, 1996).

The increased risk of VTE among heavy smokers is similar to that reported for gravida (Lindqvist et al, 1999a), but smaller than reported among middle-aged Swedish men (Hansson et al, 1999) and in the Nurses Health Study (Goldhaber et al, 1997). In the Nurses Health Study, the focus was on pulmonary embolism and the classification scheme was different (Goldhaber et al, 1997). The increased risk of VTE among smokers may be due to an increased level of homocysteine (Nygård et al, 1995). Another mechanism that has been suggested is inhibited or defective fibrinolysis among smokers (Mercelina-Roumans et al, 1996). Smoking has been shown to be related to higher fibrinogen levels (Eliasson et al, 1995; Yarnell et al, 2000).

The increased risk of VTE among those who abstain from alcohol might be due to a protective effect caused by alcohol or by selection bias in the study group. Since the data regarding people who abstain is prospective, there is no recall bias. However, those women who drink no alcohol whatsoever comprise a heterogeneous group: some do so because of religious beliefs, some for other medical conditions, some for personal reasons, and others due to prior alcohol or drug abuse, i.e., one runs the risk of wrong classification. We could not differentiate between these groups in our study. Since we have adjusted for age in all the analyses, we do not have any bias due to lower consumption with advancing age. In addition, women who abstained from alcohol had more risk due to their BMI and sedentary lifestyle, although they smoked less. This profile is in agreement with Naimi et al (2005). Our findings are in agreement with cross-sectional studies that found lower levels of fibrinogen, plasma viscosity, white blood count, von Willebrand factor, and factor VII as well as lower platelet count and activity among moderate drinkers (Yarnell et al, 2000; Mukamal et al, 2001). Based on these possible relationships, it seems plausible that an inverse relationship exists between alcohol consumption and VTE.

A longitudinal cohort study is not the most appropriate design for studying the relationship between COC use and the acute precipitation of a disease like venous thromboembolism. A nested case-control design would have been more appropriate. Thus, the data on current use of COC should, at best, be regarded as a rough estimate. However, we believe that the decreased risk among prior COC users is noteworthy. Having used COCs and not sustained a VTE, these women have shown themselves to be at low risk. The few that develop a VTE can take preventive measures in the future. A similar discussion might explain the lower risk among those with one or two prior deliveries, as compared to nulliparous women. Our finding of increased risk among nullieparous women is in agreement with studies previously conducted in Sweden (Lindqvist et al, 1999a), but it is partly in opposition to the generally accepted “fact” that multiparity is a risk factor for VTE.

The age-related increase in risk is small below 40 years of age (Fig 1). The basal risk was supposedly even lower if one adjusts for pregnancy and hormonal use. The risk increased somewhat more rapidly after age 40 years, and showed a sharp increase after 59 (Fig 1). We were not surprised to find that women who agreed to be included in the MISS study were at lower risk for VTE as compared to those who did not participate. The reason for this is probably that women in good health are more prone to participate, i.e., bias due to self-selection. Non-participants were at more than 25% increased risk for VTE. In the south of Sweden, some 10–15% of the population are carriers of Factor V (FV) Leiden (F5 R506Q) (Zöller et al, 1996; Lindqvist et al, 1999b), which is a higher percentage than Southern Europe (3%)(Rodeghiero & Tosetto, 1999) and Central Europe (7%) together (Ehrenforth et al, 1999), but in the same range as Denmark (Juul et al, 2004). Thus, one reason for the comparatively high incidence of VTE in our study cohort is the high prevalence of F5 R506Q. Thus, due to differences in F5 R506Q, we can expect the incidence of VTE to be some 30% higher in Sweden than in South Europe.

One of the reasons we undertook this study was that, we had asked women in our prior study of FV Leiden and pregnancy (Lindqvist et al, 1999b) how the awareness that they were carriers of F5 R506Q had changed their lives (Lindqvist & Dahlbäck, 2003). They had been informed that carriership of F5 R506Q was related to an increased risk of venous thromboembolism, but not to other complications of pregnancy. We recommended women to preferably use other alternative in place of COCs. In the event of surgery, immobilization, pregnancy, or other high-risk situations, they should inform their physician of their F5 R506Q status. We did not give any lifestyle recommendations. However, many women spontaneously informed us that after becoming aware that they were carriers of F5 R506Q, they started to eat healthier, exercised more, and lived a healthier life (Lindqvist & Dahlbäck, 2003).

Strength and weakness

One strength of this study is its use of a unselected cohort drawn from the population registry and the fact that most information was obtained at the inception of the study. Another strength is the availability of reliable information regarding the outcome of both participants and non-participants, enabling us to visualize the size of the self-selection bias, i.e., the women entering the study are not the same as those who did respond to the inquiry. The combination of administrative data and the questions put to women at follow-up about diseases and long-term medication is a strength and we believe most outpatient VTEs have been recorded. However, we have no data on the incidence of misclassification among the diagnosis numbers. Furthermore, we have no data on prior thromboembolic events. The assumption was made that behavior did not change over time and, consequently, information from one assessment only was used in the models. A shortcoming was that data on BMI and regular exercise was not prospectively gathered and was incomplete. Another potential shortcoming is that the lack of data regarding the method used for diagnosing the VTE. However, the use of objective methods is widely accepted in Sweden, since the cost of objectively verifying VTE is borne by the Swedish social security system. Thus, we do not believe there is a substantial “over diagnosis”. Clinically unrecognized VTE might have been missed (e.g., a sudden death due to pulmonary embolus might be missed without an autopsy). However, in order to minimize this shortcoming the study population was cross-matched against the cause of death registry.

We believe that our study offers supplementary epidemiological information on how lifestyle factors influence the risk of VTE. We conclude that women who declared themselves non-smoking, physically active, at their normal weight, and moderate in their alcohol consumption were at the lowest risk of VTE.

Acknowledgements

This study was supported by the ALF (Faculty of Medicine, Lund University, Region Skåne), the Swedish Cancer Society, the Swedish Medical Research Council, funds provided by Lund University Hospital and Region Skåne, the Gustav V Jubilee Fund, and the Kamprad Foundation.

Ancillary