Venous thrombosis is associated with hyperglycemia at diagnosis: a case–control study

Authors


Danny M. Cohn, Academic Medical Centre, Department of Internal Medicine, F4-139, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands.
Tel.: +31 20 566 7516; fax: +31 20 566 9343.
E-mail: d.m.cohn@amc.uva.nl

Summary.

Background: Patients with (undiagnosed) diabetes mellitus, impaired glucose tolerance or stress-induced hyperglycemia may be at greater risk for venous thrombosis and present with relative hyperglycemia during the thrombotic event. Objectives: To assess whether venous thrombosis is associated with hyperglycemia at diagnosis. Patients/methods: We performed a case–control study, derived from a cohort of consecutive patients referred for suspected deep vein thrombosis. Cases were patients with confirmed symptomatic venous thrombosis of the lower extremity. Controls were randomly selected in a 1 : 2 ratio from individuals in whom this diagnosis was excluded. We measured plasma glucose levels upon presentation to the hospital. Results: In total, 188 patients with thrombosis and 370 controls were studied. The glucose cut-off levels for the first to fourth quartiles were as follows: first quartile, < 5.3 mmol L−1; second quartile, 5.3–5.7 mmol L−1; third quartile, 5.7–6.6 mmol L−1; and fourth quartile, ≥ 6.6 mmol L−1. When adjusted for body mass index, a known history of diabetes mellitus, age, sex, ethnicity and whether known risk factors for deep vein thrombosis were present, the odds ratios for deep vein thrombosis in the second, third and fourth quartiles of glucose levels as compared with the first quartile were 1.59 [95% confidence interval (CI) 0.89–2.85], 2.04 (95% CI  1.15–3.62) and 2.21 (95% CI  1.20–4.05), respectively; P for trend = 0.001. Conclusions: Increased glucose levels measured at presentation were associated with venous thrombosis. Experimental evidence supports a potential causal role for hyperglycemia in this process. As this is the first report on the association between (stress) hyperglycemia and venous thrombosis, confirmation in other studies is required.

Introduction

Venous thromboembolism (VTE) is a common disease with an annual incidence of 2–3 per 1000 inhabitants [1]. Both acquired and inherited risk factors are known to play a role in the development of thrombosis [2]. Nevertheless, in approximately 25% of patients with VTE, neither an acquired nor an inherited risk factor can be demonstrated [3,4]. Evidence is growing that classic risk factors for arterial disease are also involved in the development of VTE [5,6]. Hyperglycemia is associated with arterial thrombosis [7–9] and, indeed, patients with diabetes mellitus or the metabolic syndrome [10] also have an increased risk of VTE [11,12]. This increased risk of VTE can, in part, be explained by the platelet activation and hypercoagulability present in diabetes mellitus [13]. Activation of the coagulation system has also been observed in acute experimentally induced hyperglycemia in healthy male volunteers [14]. During acute illness such as myocardial infarction, a phenomenon called stress hyperglycemia may occur independently of the presence of known diabetes.

As hyperglycemia stimulates coagulation, we hypothesized that higher glucose levels – independently of known diabetes mellitus – would be more common in patients presenting with acute VTE.

Materials and methods

A case–control study was performed. Cases and controls were selected from the Amsterdam Case–Control Study on Thrombosis, which was initiated in 1999 to identify new risk factors for deep vein thrombosis (DVT). All consecutive outpatients older than 18 years referred to the Academic Medical Centre in Amsterdam between September 1999 and August 2006 with clinically suspected DVT of the lower extremity were eligible for this study. The study protocol was approved by the Medical Ethics Review Committee, and all participants provided written informed consent. At presentation, the patient’s medical history was obtained through a standardized questionnaire including specific questions about symptom duration, presence of known risk factors [concomitant malignancy, pregnancy, use of hormone replacement therapy, oral contraceptives or selective estrogen receptor modulators, recent trauma (within the last 60 days), being bedridden for > 3 days, uncommon travel (> 6 h) within the last 3 months, paralysis of the symptomatic leg, or surgery within the last 4 weeks], concomitant diseases, and medication use. In addition, body mass index (BMI) was calculated. Cases were patients with thrombosis of the lower extremity confirmed by compression ultrasonography, including proximal DVT (i.e. proximal thrombosis of the iliac or superficial femoral vein, or calf vein thrombosis, involving at least the upper third part of the deep calf veins), symptomatic calf vein thrombosis, and superficial thrombophlebitis. The diagnosis was confirmed following a diagnostic management strategy, based on the Wells criteria [15] and a Tinaquant D-dimer assay (Roche Diagnostics, Basel, Switzerland), followed by compression ultrasound if indicated as validated and described previously [16]. Controls were selected in a 1 : 2 ratio from those individuals in whom thrombosis was ruled out using the above-mentioned strategy. Selection was performed randomly, only taking into account the male/female ratio of the cases.

Sample storage and laboratory analysis

On admission and prior to diagnostic testing, blood samples were drawn and collected in tubes containing 0.109 mol L−1 trisodium citrate. Within 1 h after collection, platelet-poor plasma was obtained by centrifugation twice for 20 min at 1600 × g and 4 °C. The plasma was stored in 2-mL cryovials containing 0.5 mL of plasma at − 80 °C.

Glucose was measured using the HK/G-6PD method (Roche/Hitachi, Basel, Switzerland) and corrected for the 10% dilution with sodium citrate. To assess whether elevated glucose levels were related to an acute-phase response induced by the thrombotic event itself, we analyzed C-reactive protein (CRP) levels from 20 cases and 20 controls, randomly selected from each quartile, giving a total of 160 patients.

Statistical analysis

Results are presented as mean ± standard deviation or median with interquartile range (IQR), depending on the observed distribution. The primary objective of this study was to assess the relationship between glucose levels at presentation and VTE, which was expressed as odds ratios (ORs) with 95% confidence intervals. Glucose levels of the controls were divided into quartiles, as glucose levels are typically non-normally distributed. Subsequently, the cases were assigned to these quartiles according to their admission glucose values. Binary logistic regression was used. The regression model was created on the basis of clinically relevant potential confounders (BMI, concomitant known diabetes mellitus, sex, ethnicity, age at diagnosis, and whether known risk factors for VTE were present). Furthermore, ORs were calculated for cases in which the diagnosis of thrombosis was restricted to DVT only, excluding calf vein thrombosis and superficial thrombophlebitis. To assess the correlation between glucose levels and CRP levels, a scatter plot was performed and the correlation coefficient (expressed as r) was calculated. All statistical analyses were performed in spss version 15.0 (SPSS Inc., Chicago, IL, USA).

Results

In total, 188 patients with confirmed thrombosis and 370 controls were included in this study, as blood samples were not available for two cases and 10 controls. The baseline characteristics of the two study groups are shown in Table 1. Mean age and gender distribution were comparable. The control group consisted of a smaller proportion of white patients and a greater proportion of black patients as compared with the cases. The mean BMI was higher in the control group, as tended to be the number of patients with known diabetes mellitus. Median glucose levels were 5.9 mmol L−1 (IQR  5.3–6.6) in patients with thrombosis, and 5.6 mmol L−1 (IQR  5.2–6.6) in the controls.

Table 1.   Baseline characteristics of the two study groups
 Cases (n = 188)Controls (n = 370)
  1. BMI, body mass index; IQR, interquartile range; SD, standard deviation. Known risk factors: concomitant malignancy, pregnancy, use of hormone replacement therapy, oral contraceptives or selective estrogen receptor modulators, recent trauma (within last 60 days), being bedridden for > 3 days, uncommon travel (> 6 h) within the last 3 months, paralysis of the symptomatic leg, or surgery within the last 4 weeks.

Age, years (mean ± SD)57 ± 1756 ± 16
Female (%)57.458.1
Ethnicity (%)
 White78.769.2
 Black10.118.1
 Asian/Pacific islander2.74.2
 Other8.58.5
BMI, kg m−2 [median (IQR)]26.6 (23.9–29.1)27.2 (24.2–31.3)
Diabetes type 1 or 2 (%)3.710.0
Glucose, mmol L−1 [median (IQR)]5.6 (5.2–6.6)5.9 (5.3–6.6)

In thrombosis patients, 38% had one or more acquired risk factors, and the distribution of thrombosis was as follows: 82% had DVT, 6% had calf vein thrombosis, and the remaining 12% had superficial thrombophlebitis. The cut-off glucose levels were as follows: first quartile, < 5.3 mmol L−1 (n = 141); second quartile, 5.3–5.7 mmol L−1 (n = 134); thirrd quartile, 5.7–6.6 mmol L−1 (n = 139); and fourth quartile, ≥ 6.6 mmol L−1 (n = 144) (Table 2).

Table 2.   Odds ratios (ORs) for venous thrombosis
Quartile (glucose, mmol L−1)CasesControls Crude OR (95% CI) Adjusted OR* (95% CI) Cases Controls Crude OR (95% CI) Adjusted OR (95% CI)
  1. CI, confidence interval. *Adjusted for body mass index (BMI), known diabetes mellitus, sex, age, ethnicity, and known risk factors. Analyses for deep vein thrombosis only. Analyses for deep vein thrombosis only, adjusted for BMI, diabetes mellitus, sex, age, ethnicity, and known risk factors.

First: < 5.339102ReferenceReference28102ReferenceReference
Second: 5.3–5.746881.37 (0.82–2.28)1.40 (0.82–2.38)37881.53 (0.87–2.70)1.59 (0.89–2.85)
Third: 5.7–6.652871.56 (0.94–2.59)1.69 (1.00–2.87)46871.93 (1.11–3.34)2.04 (1.15–3.62)
Fourth: ≥ 6.651931.43 (0.87–2.37)1.94 (1.12–3.39)43931.68 (0.97–2.93)2.21 (1.20–4.05)
   P for trend = 0.14P for trend = 0.01  P for trend = 0.05P for trend = 0.001

After adjustment for BMI, concomitant known diabetes mellitus, sex, ethnicity, age at diagnosis, and whether known risk factors for VTE were present, the ORs for thrombosis in the second, third and fourth quartiles of glucose levels as compared with the first quartile were 1.40 (95% CI  0.82–2.38), 1.69 (95% CI  1.00–2.87) and 1.94 (95% CI  1.12–3.39), respectively; P for trend = 0.01. The same trend of an increasing OR could be observed when adjusting only for BMI, age, sex, and known diabetes mellitus. As most risk factors predominantly relate to DVT, a separate analysis was planned for DVT only, thus excluding patients with superficial thrombophlebitis or calf vein thrombosis, leaving 154 cases and 370 controls. Here also, the OR increased with glucose levels: 1.59 (95% CI  0.89–2.85), 2.04 (95% CI  1.15–3.62) and 2.21 (95% CI  1.20–4.05) for the second, third and fourth quartiles of glucose level, respectively (see Table 2; P for trend = 0.001). Finally, the Spearman rank correlation coefficient for CRP and plasma glucose was 0.09, with a P-value of 0.27.

Discussion

In this case–control study, increased glucose levels measured at the time of presentation were associated with venous thrombosis. This could be a relevant clinical concept, as the general population is becoming increasingly glucose intolerant. The relationship between glucose and DVT was not readily explained by an acute-phase reaction due to the thrombotic event itself.

Whereas our results indicate that increased glucose levels and VTE coincide, it is impossible with the current design to demonstrate a causal relationship. However, a causal relationship seems plausible. To assess this, one can apply the diagnostic criteria for causation [17].

A causal relationship is supported by the available biological evidence from experiments in humans. Stegenga et al. [14] showed that experimentally induced acute hyperglycemia activates the coagulation system in healthy volunteers. From a pathophysiologic point of view, hyperglycemia is known to induce coagulation activation through glycocalyx damage [18], upregulation of tissue factor [19,20], non-enzymatic glycation, and the development of increased oxidative stress [21]. Long-term exposure to hyperglycemia, such as occurs in diabetes mellitus, is a known risk factor for VTE [22]. In addition, the effect of hyperglycemia on coagulation seems to be modifiable in diabetes patients, as treating hyperglycemia among these patients led to downregulation of coagulation activation in several randomized controlled trials [23,24]. Furthermore, our results are in line with the findings of Mraovic et al. [25], who demonstrated that hyperglycemia increases the risk of pulmonary embolism after major orthopedic surgery. Thus, direct and indirect evidence supports a possible association for acute and chronic hyperglycemia in the development of VTE. Furthermore, the association is consistent from this study to other studies, which is in line with the criterion for repetitive demonstration of causality.

In this study, an adjusted OR of 2.21 (95% CI  1.20–4.05) for DVT was observed in the highest quartile, which suggests a strong relationship. In comparison, the OR for the well-established risk factor for VTE, the prothrombin 20210A mutation, is 2.8 (95% CI  1.4–5.6) [26]. The OR for venous thrombosis increases with increasing glucose levels, from 1.40 (95% CI 0.82–2.38) in the second quartile to 1.69 (95% CI  1.00–2.87) in the third quartile and 1.94 (95% CI  1.12–3.39) in the fourth quartile. We tested for differences in ORs among the quartiles of glucose levels, and found a significant linear trend (= 0.01). This is in concordance with a dose–response gradient, another criterion for causality.

The question arises of whether elevated glucose levels during a VTE result from the inflammatory and counter-regulatory hormone action initiated by the VTE event itself, or whether hyperglycemia preceded the VTE event. Although a significant proportion of the patients with hyperglycemia during an episode of VTE will have an undisturbed glucose tolerance at follow-up [27], undiagnosed impaired glucose tolerance is likely to have been present in a proportion of patients before the VTE event itself, and may therefore have contributed to the development of thrombosis. We therefore suspect a temporal relationship. In addition, no correlation was found between the acute-phase reaction, measured by CRP, and glucose levels. The presence of stress hyperglycemia during a thrombotic event, independent of its cause, could be relevant: it has been shown to have evident clinical consequences in patients with myocardial infarction and patients admitted to the intensive care unit (especially without known diabetes mellitus) [28,29], although results from recent intervention trials have been disappointing [30,31].

Interestingly, we found a greater proportion of patients with diabetes in the control group than in the case group. In fact, this higher rate can be caused by referral bias. Patients with diabetes are usually under chronic medical care and are prone to leg and foot problems that can resemble DVT, such as erysipelas. Thus, they are more easily referred for suspicion of DVT. We had no information on the use of antidiabetic drugs. Differences in the distribution of diabetes treatment in both cases and controls could be a source of bias. However, we believe this effect to be limited. The model was adjusted for diabetes mellitus as a potential confounder.

Our study has several limitations. First, the control group consisted of patients with complaints in their legs instead of healthy controls. Consequently, it may be possible that glucose levels were increased in the controls because of an underlying disease such as infection. However, this would have led to an underestimation of the association between glucose and DVT. Second, the glucose levels were measured on admission, and it was unknown whether these were fasting or non-fasting samples. However, owing to the study design, there were no differences between cases and controls with respect to the time of presentation, and larger dispersion of glucose levels would therefore have affected cases as much as controls.

Third, citrated plasma is not the plasma of choice for determining glucose and CRP levels, because of the dilution with sodium citrate. Also, sodium citrate does not inhibit ex vivo glycolysis. However, we corrected glucose levels for the dilution, and the obtained blood samples were centrifuged and stored within 1 h, thereby minimizing glycolysis. Again, as blood samples of both cases and controls were obtained and processed in the same manner, the results for glucose levels were affected equally. As this is the first report on the association between stress and hyperglycemia and VTE, confirmation in other studies is required.

In conclusion, our findings suggest that higher glucose levels are a risk factor for the development of venous thrombosis. It will therefore be of importance to analyse whether disturbed glucose homeostasis persists after the acute phase of venous thrombosis.

Acknowledgements

We would like to acknowledge M. M. Levi for assessing the manuscript and for his significant intellectual contribution. In addition, we would like to acknowledge the ‘vasculists’, a group of medical students responsible for execution of this study, ranging from design of the study to recruitment and data collection.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary