High levels of glucose at time of diagnosing venous thrombosis: a case-control study


Vladimir Tichelaar, Division of Hemostasis and Thrombosis, Department of Hematology, University Medical Center Groningen, Hanzeplein 1, 9713 GZ, Groningen, the Netherlands.
Tel.: + 31 50 3610225; fax: + 31 50 3611790.
E-mail: y.tichelaar@onco.umcg.nl

Venous thrombosis (VT) is a multicausal disease [1]. In about 50% of cases no trigger can be found [2]. Recent studies have focused on arterial cardiovascular risk factors as potential risk factors for VT, including diabetes mellitus [3]. High glucose levels have been associated with activation of the coagulation system, in healthy men [4] as well as in patients with diabetes mellitus [5].

Recently, Hermanides et al. [6] demonstrated that patients with deep vein thrombosis (n = 188) had higher glucose levels when presenting at the emergency department, compared with controls without deep vein thrombosis (n = 370). They concluded that the association between deep vein thrombosis and higher levels of glucose at time of diagnosis was not influenced by pre-existing diabetes mellitus. A possible drawback of that study was that the authors could not adjust for acute phase reactions with appropriate statistical methods (e.g. regression or stratification methods), which might have been due to small numbers.

The goal of the present study was to replicate the findings of Hermanides et al. [6]. Therefore, we performed a case–control study and studied glucose levels in patients with acute VT and compared them with controls in whom the disease was ruled out.

Consecutive patients, suspected of deep vein thrombosis or pulmonary embolism, visiting our emergency department from April 2008 till January 2010, were included after informed consent was obtained. Clinical data were collected before objective diagnosis was made, using a standardized questionnaire [7]. All patients subsequently underwent vena puncture and were examined by compression ultrasound (CUS) or spiral computerized tomography scan. Cases were subjects with objectively confirmed VT (deep vein thrombosis, pulmonary embolism or both), and controls were subjects in whom this disease was ruled out. Calf vein thrombosis and thrombophlebitis were not considered VT. In this way, 397 patients were eligible for this study. Of these, 46 were excluded because blood sampling was not performed before diagnosis was made. To avoid mixing of patients with first and recurrent VT we also excluded cases with recurrent VT (n = 49) and controls with previous VT (n = 58) for this analysis. Single non-fasting glucose and CRP levels were measured at time of enrollment, which means for cases that these levels were measured shortly after VT occurred. All patients were categorized by quartiles of glucose levels of the controls. As estimates of relative risks, odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated. With a multiple logistic regression model we adjusted for age (continuous variable), sex and CRP, which was divided into five categories (zero till 5 mg L−1, and quartiles of the controls) to correct for the skewedness of the distribution. Subgroup analyses were performed for unprovoked, provoked and non-diabetic patients separately. Statistical analyses were performed in spss version 16.0 (SPSS Inc., Chicago, IL, USA).

We included 244 patients in this analysis. Cases numbered 163 and controls 81 (Table 1). There were slightly more women in the control group (53% vs. 44%). About 56% of cases had known risk factors for VT, compared with 46% of the controls. Almost all patients were Caucasian (> 99%). Median levels of non-fasting glucose were 6.2 vs. 5.8 mmol L−1 and median CRP levels were 29 vs. 7 mg L−1, in cases and controls. Data on glucose were missing in 11 cases and two controls. Ten per cent of cases had diabetes mellitus, as did 11% of controls. Crude odds ratios (ORs) for VT compared with the lowest quartile of glucose were 1.7 (95% CI, 0.7–4.0) for the second quartile, 2.2 (95% CI, 0.9–5.0) for the third quartile and 2.3 (95% CI, 1.0–5.2) for the highest quartile (Table 2). Adjustment for age and sex did not materially affect these risk estimates (data not shown). After adjustment for age, sex and CRP, estimates declined. Compared with the lowest quartile of glucose, the ORs for VT were 0.7 (95% CI, 0.2–2.3) for the second quartile, 0.6 (95% CI, 0.2–2.2) for the third quartile and 1.1 (95% CI, 0.4–3.2) for the highest quartile. It should be remembered here that both the assessment of hyperglycemia and CRP relate to post-VT glucose and CRP levels. Restricting analysis to patients without diabetes mellitus did not materially affect risk estimates (data not shown). In patients with unprovoked VT, relative risks of VT increased for increasing quartiles of glucose (crude OR 5.2 [95% CI, 1.2–23.4] for the highest quartile compared with the lowest quartile). Adjustment for age, sex and CRP lowered risk estimates (adjusted OR 3.0 [95% CI, 0.4–22.9] for the fourth quartile compared with the lowest quartile of glucose). In patients with provoked VT, relative risk for the highest quartile of glucose compared with the lowest quartile was 1.8 (95% CI, 0.6–5.5). Adjustment for age, sex and CRP clearly lowered relative risks of VT (adjusted OR 0.8 [95% CI, 0.2–3.4] for the fourth quartile of glucose compared with the lowest quartile).

Table 1.   Clinical characteristics of 244 subjects
 Cases (163)Controls (81)
Age, years, median (range)53 (18–92)54 (14–93)
Female, n (%)71 (44)43 (53)
Deep vein thrombosis, n (%)84 (52) 
 With pulmonary embolism, n (%)4 (2) 
Pulmonary embolism, n (%)75 (46) 
Unprovoked venous thrombosis, n (%)71 (44) 
Known risk factors, n (%)92 (56)37 (46)
 Surgery, trauma, immobility n (%)41 (25)18 (22)
 Oral contraceptives, n (%)30 (43)2 (5)
 Pregnancy or puerperium, n (%)6 (4)2 (3)
 Malignancy, n (%)22 (14)8 (10)
Diabetes mellitus, n (%)16 (10)9 (11)
Glucose, mmol L−1, median (range)6.2 (4.1–29.4)5.8 (4.1–22.3)
C-reactive protein, mg L−1, median (range)29 (0–268)7 (0–160)
Table 2.   Relative risks of venous thrombosis for quartiles of glucose for incident cases
Glucose quartiles (mmol L−1)Cases
n = 152
n = 79
Crude odds ratio (95%CI)Adjusted odds ratio* (95%CI)
  1. *Adjusted for age, sex and C-reactive protein.

Total population
 First quartile (< 5.1)1817ReferenceReference
 Second quartile (5.1–5.8)36201.7 (0.7–4.0)0.7 (0.2–2.3)
 Third quartile (5.8–6.8)48212.2 (0.9–5.0)0.6 (0.2–2.2)
 Fourth quartile (≥ 6.8)50212.3 (1.0–5.2)1.1 (0.4–3.2)
Patients with unprovoked venous thrombosis
 First quartile (< 5.1)37ReferenceReference
 Second quartile (5.1–5.8)16103.7 (0.8–17.9)1.6 (0.1–32.5)
 Third quartile (5.8–6.8)20133.6 (0.8–16.4)1.6 (0.2–13.3)
 Fourth quartile (≥ 6.8)29135.2 (1.2–23.4)3.0 (0.4–22.9)
Patients with provoked venous thrombosis
 First quartile (< 5.1)1510ReferenceReference
 Second quartile (5.1–5.8)20101.3 (0.4–4.0)0.6 (0.1–2.8)
 Third quartile (5.8–6.8)2882.3 (0.8–7.2)0.4 (0.1–2.1)
 Fourth quartile (≥ 6.8)2181.8 (0.6–5.5)0.8 (0.2–3.4)

In this case–control study we found that elevated post-VT glucose levels might be associated with an increased risk of VT in the acute setting. However, this association is at least partially a result of the acute phase reaction, as risk estimates declined considerably after adjustment for CRP. In patients with unprovoked VT the association was more robust (ORs declined from 5.2 to 3.0) than in patients with provoked VT (ORs declined from 1.8 to 0.8, comparing the highest quartile with the lowest quartile). That we did not demonstrate this association in patients with provoked VT could be explained by the fact that cases with known risk factors for VT may be over-similar compared with controls [8]. It is plausible that higher levels of glucose and CRP are also related to other diseases that cause the same complaints in the legs or lungs in controls (e.g. erysipelas or lymph-edema) as does VT in cases. This may lead to an increased incidence of the risk factor of interest in the control group, and thus our estimation of the ‘true’ risk of VT may be biased towards the null; that is, we might have under-estimated the effect. In this study, diabetes mellitus did not play a role in the association between higher levels of glucose and acute VT, which is in accordance with other studies [3]. In addition, metformin use was observed in 3% of cases and in 4% of the controls. Given the equal distribution of and small numbers for metformin use in cases and controls, it is unlikely that the potential (beneficial) influence that metformin has on coagulation factors [9] disrupted our association of interest.

The objective of this study was to replicate the findings from Hermanides et al. [6] They used a similar etiologic study design to ours and found similar results. Nevertheless, as glucose levels were measured shortly after the event, this impedes any statement about the causal meaning of glucose levels for VT risk, which is a major limitation of our study. To address this question sufficiently, one should perform a follow-up study, preferably by (non-human) experimentation. In addition, our data could be seen as results from a diagnostic study, identifying clinical characteristics associated with a true VT in those presenting in the emergency room with VT symptoms. This may be more convenient as its interpretation does not concern the problems of temporality (of measuring glucose levels, shortly after the event occurred) that our study and the study of Hermanides et al. had to deal with. However, it is probably too early to make conclusions on the clinical implications of our observations. For this we need confirmation by other studies and more data on the underlying mechanism.

Our study has a few other limitations. First, this study was performed in a university hospital setting, using referred patients without VT as controls. Consequently, our results cannot directly be translated to a general population, while the use of referred patients without VT as a control group contains the danger that it biased our outcomes towards the null [8]. Second, glucose levels were only measured once at presentation at the emergency department. The day to day variance of glucose levels is known to be about 10–15%, and could therefore have resulted in small physiological fluctuations among participants. However, due to the design of our study, this would have affected both cases and controls in a similar manner. Such random misclassification could at most have led to an under-estimation of the association we found. Third, we could not correct for body mass index, which is a known risk factor for VT [3]. However, after correction for body mass index, Hermanides et al. [6] found higher ORs of glucose quartiles for VT, so our risk estimates might be an under-estimation. A final limitation of our study is its small size. As a consequence, most risk estimates did not reach conventional levels of statistical significance, and our results should be handled with caution.

In conclusion, in this study glucose levels were higher in patients who were just diagnosed with VT compared with patients in whom VT was ruled out. However, this appears to be at least partially the result of an acute phase reaction. Follow-up studies, preferably by (non-human) experimentation, are needed to determine whether this relationship is causal or a post-hoc phenomenon.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.