Sex difference in risk of recurrent venous thrombosis and the risk profile for a second event

Authors

  • S. C. CHRISTIANSEN,

    1. Department of Clinical Epidemiology, Leiden University Medical Centre, the Netherlands
    2. Department of Internal Medicine, Ålesund Hospital, Ålesund, Norway
    Search for more papers by this author
  • W. M. LIJFERING,

    1. Department of Clinical Epidemiology, Leiden University Medical Centre, the Netherlands
    Search for more papers by this author
  • F. M. HELMERHORST,

    1. Department of Clinical Epidemiology, Leiden University Medical Centre, the Netherlands
    2. Department of Gynaecology and Reproductive Medicine, Leiden University Medical Centre
    Search for more papers by this author
  • F. R. ROSENDAAL,

    1. Department of Clinical Epidemiology, Leiden University Medical Centre, the Netherlands
    2. Department of Thrombosis and Haemostasis, Leiden University Medical Centre, the Netherlands
    Search for more papers by this author
  • S. C. CANNEGIETER

    1. Department of Clinical Epidemiology, Leiden University Medical Centre, the Netherlands
    Search for more papers by this author

Suzanne C. Cannegieter, Department of Clinical Epidemiology, C7-P Leiden University Medical Centre, P.O. Box 9600, 2300 RC Leiden, the Netherlands.
Tel.: +31 71 5261508; fax: +31 71 5266994.
E-mail: S.C.Cannegieter@lumc.nl

Abstract

Summary Background: The risk of recurrent venous thrombosis is higher in men than in women, and this is so far unexplained. We set out to determine the influence of age, time between first and second event, type of first event, oral contraception, pregnancy and surgery. Methods: We performed a prospective follow-up study of 474 patients with a first objective diagnosis of deep vein thrombosis, aged 18–70 years (Leiden Thrombophilia Study cohort). Results: During 3477 person-years of follow-up, 90 recurrences occurred. The overall incidence rates of recurrence (IRs) were 40.9 per 1000 person-years in men and 15.8 per 1000 person-years in women. Men with an unprovoked first event had the highest risk of recurrence, with almost one-third experiencing a second unprovoked event within 8 years (IR 41.2 per 1000 person-years). This risk was three-fold lower in women [IR 14.2 per 1000 person-years; hazard ratio 2.8 (95% confidence interval 1.4–5.7)]. Age at diagnosis had little effect on recurrence rate, and nor had time elapsed since the first event. In women, almost half of the recurrences were provoked and were mainly related to oral contraceptive use or pregnancy. Conclusions: The higher recurrence rate in men than in women is not the result of differences in the environmental or transient risk factors that we studied. The risk profile for a second thrombotic event is clearly different from that of a first.

Introduction

The occurrence of a first venous thrombosis depends on several factors, which can be categorized into clotting disorders (both inherited and acquired) and as environmental risk factors (e.g. surgery, pregnancy and oral contraceptive use) [1]. A combination of several of these risk factors will enhance a patient’s prothrombotic state, until a threshold is reached and a thrombus is formed [2].

After a first event, a patient has a 20–30% risk of a recurrence within the following decade [3–6]. Recent reviews have pointed out the importance of clinical risk factors, which probably contribute more to the risk of recurrence than do coagulation defects [2,7,8]. Several studies found a higher recurrence rate after a first idiopathic than after a first provoked event, with relative risks of 2–3 [5,9–11]. Also, three large cohort studies and a cross-sectional study showed a sex difference in the risk of recurrence, with men having a two-fold to three-fold increased risk of recurrent venous thrombosis relative to women [6,12–14]. It has been proposed that the difference in recurrence risk between the sexes can partly be explained by exposure to hormonal factors in women, which contributed to their first event [2,15]. When these women are included in the comparison of women with men for recurrence risk, and their exposure to oral contraceptives or pregnancy has been discontinued, the comparison may not be fair, as their risk will have been reduced. Age could also play a role, as these women were generally younger at the time of their first event than men [6,12,15]. Furthermore, the role of exposure to clinical risk factors (oral contraceptives, pregnancy and surgery) in the development of a recurrence is not well established [16–18]. Most reports advise withholding the use of oral contraceptives in women with a personal history of venous thrombosis, but until now prospective data supporting this advice have been lacking [19,20]. Some advise the use of progestin-only oral contraceptives, again without clear data supporting their safety.

We set out to analyze in detail the effect of sex, age and provoking factors on recurrence rate in a consecutive cohort of patients with a first objective deep vein thrombosis (DVT) [Leiden Thrombophilia Study (LETS)] who were followed for over a decade.

Methods

For the original LETS, a case–control study on the etiology of venous thrombosis, 474 consecutive patients with a first episode of symptomatic proximal DVT in the leg (= 453) or arm (= 21) were enrolled between 1988 and 1992. Patients older than 70 years or with a recent malignancy were excluded. Ninety per cent of the invited patients agreed to participate. The study was approved by the medical ethics committee of the Leiden University Medical Centre [21]. At least 3 months after cessation of the initial anticoagulation treatment, all patients were seen in the Leiden outpatient clinic for an extensive interview covering circumstances in the time window preceding the index event. At the same time, a blood-draw was performed, with a median time between this venepuncture and the index event of 19 months (range 6–68 months).

All 474 patients gave informed consent for a follow-up study to establish the risk of a recurrent event. Follow-up was performed by sending questionnaires (four times during follow-up) to the patients. These enquired after the occurrence of recurrent events, anticoagulation, surgery, immobilization, trauma, pregnancy and the use of oral contraception or hormonal replacement therapy. Patients were further interviewed by telephone if they responded positively on any item from the questionnaire or if they did not respond to a questionnaire. Subsequently, confirmation of all relevant clinical information pertaining to recurrent thrombotic events or risk situations was obtained from the treating physicians. All patients gave informed consent for discharge letters to be obtained from all relevant inpatient and outpatient visits to confirm initial and subsequent events. The time span of follow-up was from the end of the initial anticoagulation period (90 days) until 1 January 2000. One hundred and ninety-five individuals stopped using anticoagulant treatment after the initial period of 90 days. The others either continued for a prolonged period or restarted later during follow-up. For details on anticoagulant use during the follow-up, we refer to our original publication [6]. Major outcomes of the follow-up part of LETS were published in 2005 [6]; the present study forms a subanalysis, with a particular focus on risk factors for recurrence and the unexplained sex difference.

Initial venous thrombosis had to be confirmed by ultrasound, venography or impedance plethysmography. Recurrent thrombotic events were confirmed and reports on diagnostic methods were obtained by collecting the discharge letters from the treating hospitals. Recurrent thrombotic events were adjudicated when they were objectively confirmed with compression ultrasound, venography, or impedance plethysmography. Distal DVT was not classified as an initial or recurrent venous thrombotic event. When the recurrence was an episode of pulmonary embolism, this required positive findings from perfusion lung scanning (≥ 1 segmental perfusion defect), ventilation–perfusion lung scanning (≥ intermediate probability) or computerized tomography to be adjudicated as such. Recurrent thromboses in two patients were rejected as recurrent events because they occurred within the initial anticoagulation treatment for the first event.

A provoked event, both for first and recurrent events, was defined as occurring during pregnancy or puerperium (defined as a period of 6 weeks after delivery), during (or within 30 days after the cessation of) oral contraceptive use, during immobilization (i.e. a period of ≤ 3 months of immobilization because of hospitalization or because of being bedridden at home) or within a period of 30 days after major trauma or surgery. All other events were classified as idiopathic.

Major surgery was defined as a procedure that required general or spinal anesthesia. Brands of oral contraceptive were carefully recorded and subsequently arranged in groups according to their ability to elevate sex hormone binding globulin (SHBG) levels [22] and increase activated protein C (APC) resistance [23]. Preparations containing triphasic levonorgestrel (LNG) were analyzed as a separate group. Preparations containing monophasic third-generation progestogens [gestodene (GSD) or desogestrel (DSG)] were combined with preparations containing cyproterone acetate (CPA) into one group, known to have the highest ability to increase SHBG levels and APC resistance. Of all women who discontinued the use of oral contraception after their first event, three started using hormonal replacement therapy during the follow-up. One woman used a compound during follow-up containing estradiol and norgestrel, which are mostly prescribed for patients with dysmenorrhea. These four women were excluded from the analysis of the effect of oral contraceptives. Furthermore, we only included women between 16 and 48 years of age for this analysis.

Statistical analysis

Observation time was calculated as the time from start of follow-up (set at 90 days after the first event, even if anticoagulation lasted longer) until the end of follow-up, that is, date of the first recurrence, death, emigration or end of the study, whichever occurred first.

During the follow-up, time periods of increased risk were defined as the sum of exposure to (the different classes of) oral contraceptive, pregnancy (plus 6 weeks of the postpartum period), and postoperative periods (30 days after the date of surgery). Total follow-up time during which patients were not exposed to any clinical risk factor was calculated as the total follow-up time minus the sum of the periods of increased risk.

Incidence rates of recurrent thrombosis were calculated as the number of events over the accumulated person-time. We estimated cumulative incidences with Kaplan–Meier survival analysis. To obtain hazard ratios (HRs) of recurrence, we used a Cox proportional hazards model to correct for age differences. For the analysis of the effect of oral contraceptives and pregnancy, we used a Poisson regression model to adjust for age. For this analysis, person-time was split for several women, as during follow-up they could be exposed to (several brands of) oral contraceptives, be pregnant, or not be exposed to hormonal risk factors. Thus, some women could contribute to several exposure groups.

The effect of anticoagulation on recurrence risk was published in our first article on the LETS cohort [6]. For the current article, we were not primarily interested in exact recurrence rates but in the effect of several clinical risk factors on recurrence with respect to the sex difference. As the use of anticoagulation did not differ for the sexes [there were 121 (60%) men and 158 (58%) women who used anticoagulation for more than 3 months; median duration of initial anticoagulant treatment time in both men and women was 4 months], and as the HR for recurrence in men as compared with women remained the same after observation was restricted to the times when patients did not use anticoagulants {2.6 [95% confidence interval (CI) 1.7–3.9] and 2.7 (95% CI 1.8–4.1) respectively}, we decided not to include oral anticoagulant use in the subgroup analyses.

We used the stata statistical software package (Stata Corp., College Station, TX, USA) and spss version 14.0 (SPSS, Chicago, IL, USA).

Results

Clinical characteristics of the study population are provided in Table 1. Four hundred and seventy-four patients were followed for a total follow-up of 3477 person-years (mean, 7.3 years; range, 7 days to 11.7 years). Follow-up was complete for 94% of the patients; 27 patients were lost to follow-up and therefore included until their last observation [6].

Table 1.   Characteristcs of men and women with a first episode of venous thrombosis
 Women, n = 272Men, n = 202All, n = 474
  1. Continous variables are given as median (range), and categorical variables as number (%).

First venous thrombosis
 Age at onset (years)43 (15–69)48 (16–69)45 (16–69)
 10–3059 (22)22 (11)81 (17)
 30–4052 (19)26 (13)78 (16)
 40–5077 (28)61 (30)138 (29)
 50–6041 (15)45 (22)86 (18)
 60–7043 (16)48 (24)91 (19)
Risk factors for first venous thrombosis
 Idiopathic96 (35)163 (81)259 (55)
 Provoked176 (65)39 (19)215 (45)
 Oral contraceptives128 (47)128 (47)
 Pregnancy/puerperium21 (8)21 (8)
 Surgery, trauma,   immobilization27 (10)39 (19)66 (14)
Duration of anticoagulation (months)
 Total4 (3–116)4 (3–93)4 (3–116)

There were slightly more women (n = 272) than men (n = 202). Men were a little older at the start of follow-up, with a mean age of 48.5 years; the mean age of women was 43.0 years.

When the follow-up ended at 1 January 2000, 90 recurrences had occurred. Of these patients, 57 were men and 33 were women. Seventy-three patients had a DVT in the leg (49 men and 24 women), 12 patients had a pulmonary embolism (seven men and five women), four had a thrombosis in the arm (one man and three women), and one woman had Budd–Chiari syndrome with an extension into the vena cava inferior.

The overall rate of recurrence was 25.9 per 1000 person-years (95% CI 20.8–31.8 per 1000 person-years). In men, the overall rate of recurrence was 40.9 per 1000 person-years (95% CI 31.0–53.0 per 1000 person-years), as compared with a rate in women of 15.8 per 1000 person-years (95% CI 10.9–22.1 per 1000 person-years) [6].

In women, most first events were provoked (176/272, 65%), mainly by oral contraceptive use (128 of these 176 women), whereas only 19% (39/202) of men had a provoked first event. The recurrence rate in women was similar following provoked (15.4 per 1000 person-years) and idiopathic (16.7 per 1000 person-years) first events, and in both cases it was clearly lower than in men (29.2 per 1000 person-years in men with a first provoked event and 43.8 per 1000 person-years in men with a first idiopathic event). These results were published in our earlier article [6], but are repeated here as a basis for the following analyses on the effect of several clinical and transient risk factors on the recurrence rates.

Effect of age on recurrence risk

Table 2 shows the overall recurrence rate stratified by age and sex. The incidence rates were higher for men than for women in all age categories, with the exception of the patients who were younger than 30 years at the time of their first event, where rates were about equal in both sexes. The recurrence rates in men varied somewhat over the age categories, and did not clearly increase or decrease with age. In women, the recurrence rate went down with age, from an incidence rate of 27.3 per 1000 person-years for women aged under 30 years at the time of the first event to 9.7 per 1000 person-years for women between 60 and 70 years of age. This pattern was the same when only the subgroup of subjects with a first idiopathic event was considered, that is, no effect of age in men and a decreasing rate with age in women. When we further limited the analysis to the group with a first and a second idiopathic event, no effect of age was found in women.

Table 2.   The absolute risk of recurrence according to sex and age, and further stratified according to first and second idiopathic events
Age (years)Men, allMen, first idiopathic VTWomen, allWomen, first idiopathic VT
n¹n²/FUy¹IR
95% CI
All recurrencesSecond idiopathicn¹n²/FUy¹IR
95% CI
All recurrencesSecond idiopathic
n¹n²/FUy¹IR
95% CI
n³/FUy²IR
95% CI
n¹n²/FUy¹IR
95% CI
n³/FUy²IR
95% CI
  1. n¹, number of patients; n², number of recurrences; CI, confidence interval; FUy¹, years of total follow-up; IR, incidence rate per 1000 person-years; FUy², years of idiopathic follow-up, i.e. total follow-up after exclusion of periods of surgery, pregnancy or oral contraception; n³, number of idiopathic recurrences; VT, venous thrombosis.

10–30224/17323.1
6.3–59.1
193/15319.6
4.0–57.2
3/15319.6
4.0–57.3
5912/44027.3
14.1–47.7
20/180
0–207.2
0/170
0–212.7
30–402611/16865.6
32.7–117.4
188/11371.0
30.6–139.8
8/11371.0
30.6–139.9
527/41316.9
6.8–34.9
102/8025.0
3.0–90.1
1/8012.5
0.3–69.8
40–506120/42547.1
28.8–72.5
4819/32159.2
35.7–92.4
17/32053.2
31.0–85.0
777/60311.6
4.7–23.9
244/18421.8
5.9–55.7
3/17517.1
3.5–50.1
50–60459/30329.8
13.6–56.5
367/24828.2
11.3–58.2
7/24728.3
11.4–58.3
414/31712.6
3.4–32.3
233/17117.6
3.6–51.3
3/17017.6
3.6–51.5
60–704813/32639.9
21.2–68.3
4212/28542.0
21.7–73.6
11/28438.7
19.3–69.2
433/3109.7
2.0–28.3
373/26511.3
2.3–33.1
3/26411.4
2.3–33.2
Total20257/139440.9
31.0–53.0
16349/112043.8
32.4–57.8
46/111741.2
30.2–54.9
27233/208315.8
10.9–22.2
9612/71816.7
8.6–29.3
10/70614.2
6.8–26.0

Risk of recurrence through time

Table 3 shows recurrence rates in men as compared with those in women over time. The absolute recurrence rates in men and women were highest within the first 2 years of follow-up, and slightly decreased thereafter. The relative risk was invariably higher in men than in women within all subgroups.

Table 3.   Recurrence rates of venous thrombosis in men and women according to time after first event
Time after first event (years)*n¹n²/FUyIR95% CIHR95% CIHR†95% CI
  1. n¹, number of patients; n², number of recurrences; FUy, years of total follow-up; IR, incidence rate per 1000 person-years; CI, confidence interval; HR, crude Hazard Ratio.

  2. *Follow-up until recurrent event, death, emigration or end of study, whichever occurred first.

  3. †Age-adjusted hazard ratio.

0–2
 Men20217/38144.626.0–71.42.00.9–4.12.11.0–4.4
 Women27212/52722.811.8–39.8Ref.Ref.
2–5
 Men18022/49744.327.7–67.04.11.8–9.14.41.9–10.0
 Women2558/73910.84.6–21.3Ref.Ref.
5–8
 Men15614/40134.919.1–58.62.21.0–5.02.51.1–5.7
 Women24010/63515.77.6–29.0Ref.Ref.
> 8
 Men874/11534.89.5–89.02.10.5–9.42.20.5–10.1
 Women1463/18216.53.4–48.2Ref.Ref.

Type of recurrence in relation to sex and type of first event

Table 4 shows the rates of idiopathic recurrences by sex and by type of first event. Fifty-three men suffered from an idiopathic recurrence over a follow-up time of 1390 person-years (which excludes all person-time during which a subject was exposed to risk-enhancing situations) [incidence rate of recurrence (IR) 38.1 per 1000 person-years, 95% CI 28.6–49.9), as compared with 17 women with an idiopathic recurrence over a follow-up period of 1832 person-years (IR 9.3 per 1000 person-years, 95% CI 5.4–14.9). This higher risk in men for an idiopathic recurrence was present both in those cases in whom the first event was provoked and in those in whom the first event was idiopathic (Table 3). The highest idiopathic recurrence rate was found in men with an idiopathic first event (IR 41.2 per 1000 person-years; 95% CI 30.2–54.9) and the lowest in women with a first provoked event (IR 6.2 per 1000 person-years; 95% CI 2.5–12.8).

Table 4.   Hazard ratios of the risk of idiopathic recurrence in men as compared with women, according to whether the first event was idiopathic or provoked
 First eventn¹Second eventn²FUyIR95% CIHR95% CIHR*95% CI
  1. n¹, number of patients; n², number of idiopathic recurrences; FUy, years of idiopathic follow-up, i.e. total follow-up after exclusion of periods of surgery, pregnancy or oral contraception; IR, incidence rate per 1000 person-years; CI, confidence interval; HR, crude hazard ratio.

  2. *Age-adjusted hazard ratio.

MenIdiopathic163Idiopathic46111741.230.2–54.92.91.5–5.72.81.4–5.7
WomenIdiopathic96Idiopathic1070614.26.8–26.1Ref.Ref.
MenProvoked39Idiopathic727325.610.3–52.83.91.4–11.04.01.4–11.8
WomenProvoked176Idiopathic711036.32.6–13.1Ref.Ref.

Figure 1 summarizes these differences as cumulative incidences of recurrence over time, stratified by sex and type of first event. Figure 1A shows all recurrent events, and is therefore equivalent to the incidence rates that we published earlier [6]. Figure 1B shows only the idiopathic recurrences, that is, those that could not have been prevented other than by prolonged anticoagulation therapy.

Figure 1.

 (A) Overall recurrence rate according to whether the first event was idiopathic or provoked. (B) Idiopathic recurrence rate according to whether the first event was idopathic or provoked.

Overall, 20 provoked recurrent events happened, 16 of which were in women. Of these, 11 occurred during or shortly after oral contraceptive use, three during or shortly after pregnancy, one after surgery, and one following a trauma. The total cumulative time for which women were exposed to risk-enhancing situations (surgery, pregnancy and oral contraceptive use) was 243.3 person-years, yielding an incidence rate of 65.8 per 1000 person-years (95% CI 37.6–106.5 per 1000 person-years) during those time windows. Of the four provoked events in men, two were related to surgery and two to trauma. Men were exposed to risk-enhancing situations for only 3.6 person-years in total, so this led to an incidence rate of 1123.6 per 1000 person-years (95% CI 305.6–2876.4 per 1000 person-years).

Of the 16 provoked events in women, 14 occurred in women whose first event had been provoked as well. Of these 14, 10 happened in women who were under 30 years of age, three in women who were between 30 and 40 years of age and one in a woman who was between 40 and 50 years of age at the time of the first event. Hence, no provoked second events happened in women who were over 50 years of age at the time of the first event.

Of the four provoked recurrences in men, one happened in a man whose first event had been provoked as well. No relationship with age could be found here.

Oral contraceptive use and pregnancy

A substantial number of women (n = 128, 47%) had used oral contraception at the time of the initial thrombosis [6], 112 of whom were between 16 and 48 years of age. In addition, the first thrombotic event was related to pregnancy in 21 women (eight antepartum and 13 postpartum). In these 133 women with a first event related to hormonal factors, we assessed the recurrence risk associated with oral contraceptive use and pregnancy during follow-up. Most of these women discontinued the use of oral contraception after the first venous thrombosis. However, 53 women continued, started or restarted oral contraceptive use during the follow-up. These women used five different types of oral contraceptive, and 11 of them (20.8%) suffered a recurrent event. With a total follow-up of 226 person-years of exposure to oral contraceptives, this led to a recurrence rate of 48.8 per 1000 person-years (95% CI 24.3–87.2) for oral contraceptive use. This was a 4.6-fold (95% CI 1.9–11.5) higher recurrence rate than in women in the same age group (16–48 years) who stopped using oral contraceptives after their first event, and who were not pregnant during follow-up (adjusted for age) (Table 5).

Table 5.   Risk of recurrence among 133 women 16–48 years of age with hormonal risk factors at their first event (oral contraceptive use or pregnancy), according to exposure to hormonal risk factors during follow-up
Exposure during follow-up (n1)n²FUyIR95% CIIRR¹95% CIIRR²95% CI
  1. LNG, levonorgestrel; GSD, gestodene; DSG, desogestrel; CPA, cyproterone acetate; LYN, lynestrenol; NET, norethindrone; n¹, number of patients [the numbers do not add up, as some women were exposed to several categories (e.g. both pregnancy and no hormonal exposure) during follow-up]; n², number of recurrences; FUy, years of total follow-up; IR, incidence rate per 1000 person-years; CI, confidence interval; IRR¹, crude incidence rate ratio; IRR², age-adjusted incidence rate ratio; NP, not possible to calculate, as almost all pregnancies were in the same age category.

  2. *Recurrence without oral contraceptive use or pregnancy within 1 month before event.

  3. †Recurrence with oral contraceptive use within 1 month before event.

No hormonal contraception or pregnancy (80)8*760.310.54.5–20.7Ref.Ref.Ref.Ref.
Hormonal contraception (53)11†225.648.824.3–87.24.61.9–11.54.31.7–11.1
Monophasic LNG (13)2†45.044.45.4–160.54.20.9–19.93.50.7–16.9
Monophasic GSD, DSG, CPA (26)5†99.750.216.3–117.14.81.6–14.64.41.4–14.0
Triphasic LNG (6)2†14.4138.916.8–501.713.12.8–61.613.02.8–61.5
Progestin-only (12)2†52.138.44.7–138.83.70.8–17.23.60.7–17.3
Monophasic LYN, NET (7)014.400–256.4
Pregnancy (31)334.387.618.0–255.98.32.2–31.4NP 

The recurrence rate in users of monophasic oral contraceptives containing LNG was similar to the rate found in users of monophasic contraceptives containing GSD, DSG or CPA [44.4 per 1000 person-years (95% CI 5.4–160.5 per 1000 person-years) and 50.2 per 1000 person-years (95% CI 16.3–117.1 per 1000 person-years)] (Table 5). Users of triphasic oral contraception with LNG had a higher risk, with two events occurring in six women, leading to an incidence rate of 138.9 per 1000 person-years (95% CI 16.8–501.7 per 1000 person-years). Twelve women used oral or injectable progestin-only preparations, and two suffered recurrences (both of whom used injectable medroxyprogesterone 150 mg mL–1), giving a rate of 38.4 per 1000 person-years (95% CI 4.7–138.8 per 1000 person-years), which was only slightly lower than that of the second/third-generation group. The age-adjusted HRs for all subtypes were similar, and showed an approximately four-fold increase relative to non-use in all women aged between 16 and 48 years (Table 5).

Of the 133 women who had been exposed to hormonal factors at the time of their first event, 31 had one or more pregnancies during follow-up (47 pregnancies all together), with a total exposure time of 34.3 person-years (including a 6-week postpartum period). During this period, three pregnancy-related recurrences occurred. One of the three recurrences occurred in the 31st week of pregnancy in a woman who had used oral contraception when she had her first venous thrombosis. The second patient had her first event a few weeks postpartum and a recurrence at day 3 postpartum of a later pregnancy. The third patient had both events during pregnancies. Overall, the recurrence rate during pregnancy/puerperium was 87.6/1000 person-years (95% CI 18.0–255.9), which corresponds to an 8.3-fold increased risk (95% CI 2.2–31.4) as compared with women with a first thrombotic event related to hormonal risk factors who were not exposed during follow-up.

Surgery

One hundred and four patients (32 men and 72 women) underwent 139 major surgical procedures during follow-up. Their mean age (45.7 years; range, 20.0–68.8 years) was comparable to that of the overall cohort. Three recurrences related to surgery occurred in two men (age: 44.9–65.3 years) and one woman (age: 34.2 years), all of whom had had an unprovoked first event. The three recurrences occurred 5 days after a laparotomy, 6 days after a knee arthroscopy, and 3 weeks after a diagnostic laparoscopy. We had only limited information regarding anticoagulant prophylaxis during these interventions.

The total exposure time for surgery added up to 11.2 person-years, which led to a recurrence rate of 267.1 per 1000 person-years (95% CI 55.0–780.1). Men’s risk of surgery-related recurrence was 561.8 per 1000 person-years (95% CI 68.0–2028.1), and that of women was 130.4 per 1000 person-years (95% CI 3.3–726.2).

Discussion

In our first article on the LETS cohort we reported, among several general aspects of recurrence risk for venous thrombosis, a higher recurrence rate for men than for women [6]. Some of the numbers and risks have been repeated in this second article, as a necessary basis for further analyses into the effect of several clinical and transient risk factors, which could possibly explain the sex difference in recurrence risk. In this study, we found that the clearly higher recurrence rate for men persisted through time. This higher risk could not be explained by age or by a sex difference in transient risk factors after the first event (such as oral contraceptives). Furthermore, the risk depended strongly on whether the first event was idiopathic or not, but men with a provoked first event still had a substantially higher overall recurrence rate than women. Of men with an idiopathic first event, almost one-third were likely to suffer a second idiopathic event within the next 8 years. This is therefore a group that is likely to benefit from long-term anticoagulant treatment. In women, the majority of the recurrences (48%) were provoked, and occurred mainly during hormonal contraceptive use or pregnancy or postpartum.

In 2004, Kyrle et al. [12] first reported a difference in recurrence risk between men and women (HR 3.6). In retrospect, several previous articles contained information on this issue, some of which had found no or only a small difference [24–27]. However, most subsequent studies confirmed an increased recurrence risk for men, which was also the conclusion of a meta-analysis [28]. This meta-analysis included 11 cohort studies [10–13,24,26,29–33] with clearly defined inception cohorts that achieved a > 90% complete follow-up. Five of these studies found an increased risk of recurrence in men (relative risk range: 1.7–6.3) [12,13,29,31,33], five studies found a weak or no effect (relative risk range: 1.2–1.5) [10,11,26,30,32], and one study found a decreased risk in men (relative risk: 0.5) [24]. Of two more recent studies, one did not find a difference with respect to sex [34], whereas the other clearly did [35]. None of these studies could provide an explanation for the sex difference.

Surprisingly, even though the strongest risk factor for a first event is age [2,36], age appears to have no effect on the risk of a second event. None of the studies that looked into the effect of age on recurrence found a strong effect. Some described a slightly increased risk (HR 1.36 per decade increase in age [37], and others found no relationship [12,34] or even a reduced risk with increasing age [38]. We found no effect of age in men, and a decreasing risk with age in women; this observation can be attributed to oral contraceptive use and pregnancy being a risk factor mainly in young women, whose risk reduces after removal of the risk factor. When we considered only idiopathic second events, no effect of age was seen in women either. It is of note that these subgroups contained small numbers of patients, so the exact values for the rates should be interpreted with caution. They serve predominantly to demonstrate the generally absent effect of age. Nevertheless, they are in line with recent reported findings from another prospective cohort study on recurrent venous thrombosis risk [39].

We reported earlier, as did others [5,34], that the recurrence risk is highest when the first event was unprovoked. This difference was mainly observed in men, whereas men with a provoked first event still had a higher recurrence risk than women with an unprovoked first event. The recurrence risk in all women was low, and half of the recurrences in women were related to well-known risk factors for thrombosis such as oral contraceptive use and pregnancy. Not many studies have been published on the effect of these factors on a second thrombosis, as patients are usually discouraged from using hormonal contraception after a first event (with the exception of progestin-only contraceptives), and because pregnancy is a relatively rare event in women with a history of thrombosis. We found a clear overall four-fold increased risk for oral contraceptive use, without an indication of a difference per type of contraception. This means that women who stop using hormonal contraception after a first event can reduce their recurrence risk from 5% to 1% per year, which seems worthwhile, although the risk of pregnancy will have to be considered as well. We could not demonstrate a difference between the risk of second-generation and third-generation pills, which may be related to the small number of subjects. However, it could also be that no such difference exists in these patients, whose risk profile is different from that of a first event in many respects. Progestin-only compounds did not appear to be safer than the other types in our study. In 1998, the World Health Organization (WHO) found a weakly increased risk of a first venous thrombosis in users of progestin-only contraceptives [40], and 1 year later another study initiated by the WHO found users of progestin-only preparations to have a two-fold increased risk of a first venous thrombosis [41]. Studies into the effect of progestin-only pills in women with a history of thrombosis are rare. A French series described two groups of 71 women with a previous venous thrombosis, one group using chlormadinone acetate (CMA) and the other without hormonal contraception. The recurrence risk was low in the CMA group, in which two women developed a recurrence as compared with six in the non-treated group [42]. Our results are based on only a few users of progestin-only compounds, and the two recurrences that occurred in this group were both related to a specific type, that is, injectable medroxyprogesterone. Therefore, our results suggest that the use of this preparation may have to be discouraged, although we cannot draw any conclusions about the safety of other types of progestin-only contraceptive. Interestingly, a recent study suggested that progesterone-only oral contraceptives do not increase the risk of first venous thrombosis, [43], whereas two other studies, presented as an abstract, showed that injectable progesterone agents increased the risk of first venous thrombosis two-fold to three-fold [44,45], which is in accordance with our results on recurrent risk.

It has been suggested that the discrepancy in recurrence risk between the sexes could be explained by an age difference at the time of the first event, with women generally experiencing a first event at younger ages, because of oral contraceptive use or pregnancy. In this study, we observed a somewhat higher relative risk of recurrence in men as compared with women in the first 5 years after the initial event than when follow-up time was restricted to more than 5 years. A previous study obtained similar results, and explained the sex difference by this lead time bias phenomenon [15]. However, our study showed that men were still at higher risk of recurrence than women when follow-up time was restricted to more than 5 years. Moreover, men still had higher recurrence rates than women when we restricted the analyses to idiopathic first events. Therefore, lead time bias may partially explain the sex difference, but not totally.

This study has several limitations in addition to the small numbers in some subgroups. First, we did not have detailed information on the anticoagulant use around each surgical procedure or during or after pregnancy. We calculated the overall risk of exposure to pregnancy and surgery, but it may well be that sufficient anticoagulant prophylaxis was lacking in some cases, and that the events occurred in this group. Second, we started the follow-up of the study somewhat arbitrarily at 90 days after the first event, but the duration of anticoagulant treatment after the first event was often longer than 3 months. It may have been that these subjects were considered to be at high risk for a recurrent event by their physician. However, this cannot explain the strong sex differences, as sex as a risk factor for recurrence was not expected at the time. Third, our study results may not be generalizable to subjects who were 70 years or older during their first event, because such patients were not included in the original LETS cohort. However, a large prospective study from Austria (AUREC) that included patients up to the age of ≥ 90 years found similar results for age and sex with regard to recurrent venous thrombosis risk [12,39]. Furthermore, for the results on contraceptives and pregnancy, this exclusion criterion is not relevant. Finally, an ipsilateral recurrent thrombotic event can be difficult to diagnose. Similarly, residual changes that were detected in some patients with recurrence may not have been detected initially (owing to impedance plethysmography diagnosis) but may have been present at the end of the period of initial anticoagulation therapy. Hence, some misclassification may have occurred. However, one should expect similar misclassification in men and women; that is, if misclassification occurred, it is expected to be non-differential. This type of misclassification would have led to an underestimation of the true relative risk estimates for men as compared with women [46].

It is very remarkable that risk factors for a first thrombosis are apparently different from those for a second thrombosis. Age, the strongest risk factor for a first event, does not appear to play a role in the development of a recurrence. Male sex, which is not a risk factor for a first event, is a strong risk factor for recurrence. The known thrombophilic defects, which clearly affect the incidence of a first event, have little or no effect on the risk of a second event [2,5,6]. In contrast, clinical risk factors that increase the risk of a first event appear to have a similar, if not even more pronounced, effect in causing a second event: oral contraceptive use, pregnancy and surgery. Explanations for these differences are hard to provide, and should be the focus of further studies.

This study may have implications for patient care. First, women should be encouraged to avoid the use of sex steroids after a first venous thrombosis. Progestin-only preparations may still be an exception here, although the two recurrences in users of medroxyprogesterone are not encouraging. Second, patients with a history of venous thrombosis need to receive strict anticoagulant prophylaxis around surgery or other procedures, and possibly pregnancies. It may be helpful to educate patients and have them carry an emergency medical information card. Finally, the risk of recurrence in men, especially in those whose first event occurred without any precipitating factors, is sufficiently high to consider prolonged duration of anticoagulation in this group. Guidelines on the duration of anticoagulation should provide different policies for men and women [35,47,48].

In summary, this study determined the influence of sex, age and provoking factors on the risk of recurrent venous thrombosis. Neither age, nor time between first and second event nor transient risk factors (oral contraceptives, pregnancy and surgery) could explain the different recurrent risks for the sexes. The risk profile for a second event is clearly different from that for a first, which should be kept in mind for clinical care and further research.

Addendum

S. C. Christiansen: collected data, analyzed and interpreted data, and wrote the article. W. M. Lijfering: analyzed and interpreted data. F. R. Rosendaal: designed the study, obtained funding, wrote the article and supervised the study. F. M. Helmerhorst: critically revised the article and supervised the study. S. C. Cannegieter: analyzed and interpreted data, wrote the article and supervised the study.

Acknowledgements

We are grateful to the personnel of the Anticoagulation Clinics of Leiden, Rotterdam and Amsterdam, who facilitated the inclusion of the patients. We thank A. Schreijer, I. de Jonge and I. Noordermeer for data management, and all patients for their collaboration. The LETS study was funded by the Netherlands Heart Foundation (NHS 89.063) and the follow-up by the Prevention Fund/ZonMW (2827170).

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary