Mitochondria-DNA copy-number and incident venous thromboembolism among middle-aged women: a population-based cohort study

The Women´s Health In the Lund Area study (WHILA) is a prospective cohort study that was conducted in southern Sweden. Sample characteristics, data collection and clinical definitions for WHILA have been described previously [29, 30]. Briefly, the study invited all women (n = 10,766) living in any of the five municipalities around the city of Lund by December 1995, and who were born between December 1935 and December 1945, to a screening procedure that took place from December 1995 until February 2000. A total of 6917 women consented to participate in the study, of which 6916 had complete datasets. Blood samples for DNA analyses were available at baseline only, and were collected midway through the study and therefore were available only for the last 3062 included women. Out of the 3062 blood sampled women, there were 541 samples with poor quality DNA as observed during droplet digital PCR analysis of reference gene. Thus, only 2521 women with sufficient quality of DNA were analyzed. These 541 women was excluded from analysis. Women (n = 13) treated with anticoagulants at baseline were also excluded (ATC code; B01AA03, B01AA01, and B01AB02). Women with prevalent VTE (defined below) or any of the following prevalent diagnoses at baseline were excluded: Stroke [ICD-7; 331, 332, 334.09. ICD-8; 430–434, 436. ICD-9; 430–434, 436. ICD-10; I60-I64 (not I63.6)]; Coronary heart disease (CHD) (ICD-7; 420. ICD-8; 410–414. ICD-9; 410–414. ICD-10; I20-I25.); and Cancer (ICD-7; 140–209). In the cohort, there were 22 women with prevalent stroke, 26 women with prevalent CHD, 232 women with prevalent cancer, and 120 women with VTE before baseline (self-reported and/or by register). Moreover, 14 women who were living in nursing homes were excluded. Some of the women had more than one reason for exclusion, which is why the numbers do not add up. After exclusions, a total of 2117 women remained of the 2521 women with sufficient quality DNA. After exclusions of prevalent stroke, prevalent CHD, prevalent cancer and prevalent VTE, 126 women remained with incident VTE. Of these, 39 women with VTE were censored in the cox regression due to incident CHD, stroke, or cancer before VTE during follow-up. Thus, after censoring for 39 incident cases of stroke, CHD, and cancer, we had 87 incident VTE cases in total.

The cohort was identified via the Swedish population registry that comprised all inhabitants in the Lund area. A health-screening program included laboratory examinations, blood samples and a basic baseline questionnaire that was mailed along with the invitation and collected in conjunction with the first examination. The baseline questionnaire included 104 questions concerning education, household, working status, perimenopausal status, medical history, drug treatment, personal and family history of diabetes or cardiovascular disease (myocardial infarction, stroke, deep venous thrombosis or hypertension in parents or siblings with an event before the age of 60 years). It also contained questions about habits like smoking, alcohol consumption, physical activity, general dietary habits, quality of life, as well as subjective physical and mental symptoms. This questionnaire was a composite of several validated questionnaires.

The mtDNA-CN was determined by droplet digital PCR (ddPCR) based method for quantification of absolute copy number of mtDNA in whole blood, as described in detail [31]. Compared to real-time PCR, ddPCR has greater precision and improved reproducibility and provides ultrasensitive and absolute nucleic acid quantification [32]. The used ddPCR method for mtDNA determination has been well-optimized with intra- and inter-assay coefficient variances as 3.1% and 4.2% respectively [31].

Self-rated health (SRH) was assessed only at baseline from a single question, in which the participants were asked to rate their perceived health in a 7-graded Likert scale from “Very poor” to “Excellent, could not be better” (1 = very poor, 7 = Excellent). In this study, alternative 1–4 was classified as poor SRH and 5–7 as good SRH. Weight and height were rounded off to the nearest 0.1 kg and 0.5 cm. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared (kg/m²). Varicose veins were defined by ICD-7; 460, ICD-8; 454, ICD-9; 454, and ICD-10; I83. Women were categorized as current smokers (i.e., those who smoked regularly or occasionally), non-smokers and former smokers. Non-alcohol drinkers were defined as not drinking any alcohol (0 g/week), normal consumption more than 0–12 g/week, and high consumption as over 12 g/week. Low physical activity, i.e., less than 30 min vigorous activity 5 days a week [33] was defined by the three lowest answering alternatives (i.e., very low, low and light activity) (Table 1). Overall diet was defined by questions about intake of fat, sugar, fruit, or dietary fiber. If high intake of sugar or fat or low intake of dietary fiber or fruit were reported, the diet was classified as less healthy [29]. If low intake of sugar and fat and high intake of fruit and dietary fiber was reported, the diet was classified as healthy. Education was categorized into three groups: comprehensive school (9 years), upper secondary school (12 years), and university degree.

Information about diseases and medication was obtained from the Swedish Hospital Discharge Register, the Hospital Outpatient Register and the Swedish Cancer Register in addition to self-reported data from WHILA study baseline questionnaires and measurements.

The definition of VTE includes deep venous thrombosis (DVT), pulmonary embolism (PE), and other VTE types by the following ICD-codes from the Swedish Hospital Discharge Register and the Hospital Outpatient Register:

DVT (deep venous thromboembolism): ICD-7 463, 682. ICD-8 451, 671 (not 671.00). ICD-9 451 (not 451A) 671D 671E. ICD-10 I80 (not I80.0) O22.3, 0871. PE (pulmonary embolism): ICD-7 465, 684. ICD-8 451, 671 (not 671.00). ICD-9 415B, 416 W, 639G, 673C. ICD-10 I26, O88.2, O082. Other VTE types: ICD-7 464, 466, 583.00, 334.40, 334.50. ICD-8 321, 452,453. ICD-9 437G, 451A, 452,453,671C, 671F,6171X. ICD-10 I63.6, I676, I800, I81, I82, O222, O225, O229, O870, O873, O879, O087.

Levels of mtDNA-CN were used as a dichotomized variable according to median, though mtDNA-CN was also used as a continuous variable in the analysis. In the multivariate models, waist circumference was used instead of BMI; this was due to previous studies showing waist circumference as a valid predictor for VTE [34]. A power analysis revealed that a risk ratio of 1.8 could be detected at a significance level of 5% (two-sided) with a power of 82% in the studied population when using mtDNA-CN as a dichotomous predictor variable (https://​www.​openepi.​com/​Power/​PowerCohort.​htm). P-values were calculated with two-sided Student’s t test for continuous variables, with χ²-test for dichotomized and normal distributed categorical variables, and Wilcoxon rank sum test for not normally distributed categorical variables. Cox proportional hazards regression was used to analyze the relationship between different variables and time to VTE. Hazard ratios (HR) with 95% confidence intervals (CI) were calculated. Women who were affected by stroke, CHD, or cancer diagnosis before the VTE during the follow-up time were censored. Well-known risk factors for increased VTE-risk were used as confounding variables in the multivariate analysis. At first, in the multivariable model, there was an adjustment for mtDNA-CN and age. In the second model, smoking, waist circumference, physical activity, and varicose veins were added. Among the categorical variables, the same category was used as reference through all analysis. Regarding smoking, “non-smoker” was used as reference, in physical activity the reference was “low physical activity”, in varicose veins it was “no varicose veins” reference, and in mtDNA-CN median, under the median (< 117) was reference. The mtDNA-CN median and mtDNA-CN as a continuous variable were analyzed separately in the different models by Cox regression using the same adjustments. In the main analysis, we excluded all the patients with prevalent VTE, CHD, stroke, cancer, and women treated with anticoagulants at baseline or living in nursing homes. During follow-up, those with incident CHD, stroke, or cancer before VTE were censored. At the next step, a sensitivity analysis was conducted; the only exclusions in the sensitivity analysis were those with poor quality DNA. All other excluded and censored individuals were included in the sensitivity analysis. P-values < 0.05 were considered significant. Analyses were performed in STATA 16.1.

During the follow-up time of 17.9 years, 87 women were diagnosed with VTE. The sum of follow-up time was 31,073.123 years corresponding to a VTE incident rate at 2.8 (95% CI 2.3–3.4) per 1000 person-years. Unadjusted Cox-regression analysis was conducted (Table 2) to determine any associations between possible risk factors and incident VTE. The univariate Cox-regression showed no significant association between increased mtDNA-CN and lower risk for VTE (HR = 0.99, 95% CI 0.98–1.00, p = 0.13). There was a non-significant decreased risk of VTE with mtDNA-CN over the median (HR = 0.76, 95% CI 0.50–1.17, p = 0.21) compared with mtDNA-CN under median. Among potential VTE risk factors, statistically significant associations were observed with weight (HR = 1.02, 95% CI 1.00–1.04 p = 0.01), waist circumference (HR = 1.02, 95% CI 1.00–1.04, p = 0.01), hip circumference (HR = 1.03 95% CI 1.00–1.06 p = 0.02), and varicose veins (HR = 2.89, 95% CI 1.17–7.13, p = 0.02). A significant association with sugar intake was observed. No significant association was shown with incident VTE and current smokers (HR = 1.30, 95% CI 0.78–2.18, p = 0.32) or former smokers (HR = 0.74, 95% CI 0.41–1.33 p = 0.31). Regarding physical activity (HR = 0.87, 95% CI 0.56–1.34 p = 0.52), there were no significant associations with incident VTE. Likewise, there was not shown any significant associations between incident VTE and self-rated health (HR = 1.74, 95% CI 0.24–12.49 p = 0.58) (Table 2).

In the second step, a multivariable Cox regression analysis was performed, at first when adjusted for age (Table 3) dichotomized mtDNA-CN over median showed a non-significant decreased risk for VTE (HR = 0.76, 95% CI 0.50–1.77 p = 0.22) compared to mtDNA-CN under the median (Table 3). With age adjusted Cox regression and mtDNA-CN as a continuous variable, a non-significant decreased risk was shown with increasing mtDNA-CN (HR = 0.99, 95% CI 0.98–1.00 p = 0.14). The last model 3, adjusted for age, smoking, waist circumference, physical activity, and varicose veins showed mtDNA-CN over the median a non-significant decreased risk of incident VTE (HR = 0.89, 95% CI 0.57–1.38 p = 0.60) using mtDNA-CN under the median as reference. Using the same adjustments but with mtDNA-CN as a continuous variable showed no effect of neither decreased nor increased risk for incident VTE (HR = 1.00, 95% CI 0.99–1.00 p = 0.34) (Table 3).

The only exclusions done in the sensitivity analysis were those with poor quality DNA (541 women). In the sensitivity analyses, a total of 2521 women were remaining and out of these, 142 women were affected with incident VTE during follow-up. Moreover, in the sensitivity analyses, no censoring was performed for women with incident stroke, CHD, or cancer. The sum of follow-up time in the sensitivity analysis was 38,805.318 years corresponding to a VTE incident rate at 3.7 (95% CI 3.1–4.3) per 1000 person-years total analysis time at risk under observation. The result of the univariate sensitivity analysis (Supplementary Table 1) is approximately the same as the univariate main analysis regarding hazard ratios and confidence intervals. The same variables showing significant association in main analysis were significant in the sensitivity analysis, except for knowledge of family history of VTE, which showed an association with incident VTE in the sensitivity analysis (HR = 1.67, 95% CI 1.05–2.68, p = 0.03) (Supplementary Table 1). Regarding varicose veins, it was the opposite with no significant associations in the sensitivity analysis (Supplementary Table 1). The multivariable sensitive analysis (Table 4) showed the same pattern as the main multivariable analysis. There were minor differences in hazard ratios and level of significance between the main analysis and the sensitivity analysis.

This is the first study aiming to investigate the association between mtDNA-CN and incident VTE, which is a strength of this study. Performing the study in a small cohort containing only middle-aged women with a narrow age span is a limitation.