Venous thromboembolism (VTE), including deep vein thrombosis and pulmonary embolism (PE), is often accompanied by acute hyperglycemia or ‘stress-hyperglycemia’ . Conversely, growing evidence suggests that both chronic and acute hyperglycemia contribute to coagulation activation and hypofibrinolysis, resulting in a prothrombotic state [2,3].
The clinical consequences of the presence of stress-hyperglycemia during VTE have been assessed in different patient settings. In an outpatient population, hyperglycemia at presentation was shown to be associated with VTE, with a clear linear relation between glucose levels and risk of VTE . Furthermore, in orthopedic patients undergoing total hip arthroplasty there was an association between stress-hyperglycemia and VTE . In addition, a recently published large retrospective cohort study in >13 000 patients with acute PE has shown that elevated admission glucose levels were present in the majority of patients and were independently associated with 30-day mortality .
What could explain the presence of stress-hyperglycemia during VTE? First, elevated glucose levels during a VTE can result from the physical stress response (inflammatory and counter-regulatory hormone action) induced by the VTE event itself. Second, undiagnosed impaired glucose tolerance may be present in a proportion of patients before the VTE itself and may therefore have contributed to the development of thrombosis. Both may be operative in different subjects, but may also coincide in the same subject. As stress-hyperglycemia may be considered as a manifestation of impaired glucose tolerance, which in itself frequently evolves into diabetes mellitus (DM), one would expect an increase in incidence rate of DM in patients after a diagnosis of VTE. In this study we tested the hypothesis that the risk of DM in subjects with PE is increased compared with subjects without PE.
Data were derived from the PHARMO Record Linkage System (RLS), which consists of multiple observational databases linked on a patient level, covering over three million individuals in defined areas of the Netherlands. For the purpose of this study, data on drug prescribing from the community pharmacy database and on hospitalization from the Dutch National Medical register were used. Drugs are coded according to the Anatomical Therapeutic Chemical (ATC) classification. The hospital admission and discharge codes are coded according to the International Classification of Diseases, 9th Revision (ICD-9-CM).
All subjects with a first hospitalization for PE (ICD 4151) between 1998 and 2009 were identified (PE-cohort). Exclusion criteria were: known DM defined as prescription of antidiabetic medication (ATC A10A-A10B-A10X); known malignancy defined as hospitalization (ICD 1400-2400) within 5 years prior to the diagnosis of PE; and systemic use of glucocorticosteroids (ATC H02AB-QH02AB) within 6 months before the diagnosis of PE. Subjects without a diagnosis of PE (the non-PE cohort) were derived from the same source population from which the PE patients were identified. Selection was performed randomly, taking into account the male/female ratio, date of birth (±1 year) and geographical region of the PE subjects. Subsequently, the same exclusion criteria were applied in the non-PE cohort as were used in the PE-cohort. The date of hospital discharge for PE was considered to represent the start of follow-up (i.e. index date). Non-PE subjects were assigned the same index date as their matched PE subject. Each subject was followed for 5 years from their index date to the occurrence of the study outcome or censoring (last available prescription or admission in PHARMO RLS or in the Dutch registry for mortality), whichever came first.
The main outcome of the study was the onset of DM within 5 years after the index date. DM was defined as the prescription of glucose-lowering therapy, either orally or subcutaneously (ATC A10A-A10B-A10X) as registered in the PHARMO RLS.
The association between PE and study outcome was explored by means of the Kaplan–Meier method and formally tested using the log rank test. Subsequently, a Cox proportional-hazards regression model was used to adjust for age. Data management and statistical analyses were performed with SAS software version 6.12 (SAS Institute Inc., Cary, NC, USA) and SPSS software version 18.0 (SPSS Inc., Chicago, IL, USA).
The PE-cohort consisted of 5045 subjects, whereas the non-PE cohort contained 6785 subjects. The mean age ± SD of the studied cohorts at baseline was 58 ± 18 years in the PE cohort and 56 ± 18 years in the control cohort. In both cohorts, sex was distributed similarly (44% male subjects). During 5 years of follow-up, DM occurred in 168 (3.3%) subjects with PE and in 234 (3.4%) subjects without PE (P = 0.717). After 5 years of follow-up, the Kaplan–Meier estimate of DM-free survival (standard error) was similar in both groups; 0.952 (0.004) in the PE cohort and 0.952 (0.003) in the non PE-cohort (P = 0.543), as can be seen in Fig. 1. Using a Cox proportional-hazards regression model adjusting for age, PE was not associated with an increased risk of developing DM (HR 1.01, 95% CI 0.8–1.2; P = 0.93).
In this large population-based registry study, we could not confirm the hypothesis that a diagnosis of PE was associated with an increased risk of DM. We found a similar 5-year incidence of DM of 3% in the studied cohorts.
To our knowledge, this is the first study examining the association of a diagnosis of PE and the risk of developing DM. The incidence of DM we found in the non-PE cohort (i.e. control population), 7 per 1000 persons per year, was similar to the estimated mean incidence of DM in the Netherlands based on five general practice records described by Baan et al. .
Our findings contribute to ongoing discussions about whether stress hyperglycemia solely results from the physical stress induced by the venous thrombo-embolic event itself or whether it may reflect a pre-existent disturbed glucose homeostasis [4,8]. Although disturbed glucose homeostasis will be present in a proportion of patients with PE, the present findings suggest that overall, PE patients do not carry an elevated risk of developing DM. Screening of this population for DM, as is advocated after clinical presentations of atherosclerotic diseases such as myocardial infarction and stroke, does not seem warranted [9,10].
The main strength of this study was the large cohort of patients with PE, who were followed for a relatively long period. The study design has some obvious limitations, which are inherent to all population-based registry studies. The diagnosis of pulmonary embolism was derived from ICD-codes, which could raise concern about accuracy and may contribute to selection bias. However, Casez et al.  recently showed that ICD discharge diagnosis codes yield sufficient sensitivity for identifying objectively confirmed PE. Furthermore, subjects diagnosed with DM and treated with diet only are not identifiable as DM subjects in our prescription drug-based database. This may have contributed to an underestimation of the 5-year incidence of DM in both cohorts. Also, information on the pathogenesis of PE is lacking. It might be that in a subgroup of patients with a specific etiology of PE (i.e. surgery), hyperglycemia is predominantly explained by physical stress, whereas in patients with unprovoked events a pre-existent disturbed glucose homeostasis may play a more prominent role. Unfortunately, glucose levels and other parameters were not available for this investigation. Finally, because patients were retrieved through hospital admissions, we might have missed patients that may have been treated as outpatients. In the Netherlands, however, guidelines clearly recommend in-hospital treatment of PE .
To conclude, the results of the present study show a similar 5-year incidence of DM for subjects with and without PE. These findings suggest that PE patients in general are not at increased risk of developing DM, although further investigation in subgroups with a specific pathogenesis of PE would be welcome.