Deep vein thrombosis and major adverse limb events in diabetic patients: a population-based cohort study

Po-Chang Wang Chang Gung Memorial Hospital Chiayi Branch: Chiayi Chang Gung Memorial Hospital Tien-Hsing Chen Chang Gung Memorial Hospital Chang-Min Chung Chang Gung Memorial Hospital Chiayi Branch: Chiayi Chang Gung Memorial Hospital Mei-Yen Chen Chang Gung Institute of Technology: Chang Gung University of Science and Technology Jung-Jung Chang Chang Gung Memorial Hospital Chiayi Branch: Chiayi Chang Gung Memorial Hospital Yu-Sheng Lin Chang Gung Memorial Hospital Chiayi Branch: Chiayi Chang Gung Memorial Hospital Pao-Hsien Chu Chang Gung Memorial Hospital Linkou Main Branch: Chang Gung Memorial Hospital Yun-Shing Peng Chang Gung Memorial Hospital Chiayi Branch: Chiayi Chang Gung Memorial Hospital Ming-shyan Lin (  mingshyan@gmail.com ) Chiayi Chang Gung Memorial Hospital https://orcid.org/0000-0001-9348-8153


Introduction
Diabetes mellitus affects 463 million people worldwide with a global prevalence estimated to be 9.3% in 2019 (1). Diabetes has been associated with a 2 ~ 4-fold increased risk of critical limb ischemia (2), and diabetic patients with critical limb ischemia have been reported to have a 7 ~ 15-fold higher risk of undergoing major limb amputation (3,4) and consequently a high mortality rate (5). The recent COMPASS (6) and VOYAGER studies (7) showed that adding antithrombotic therapy could reduce the incidence of major adverse limb events (MALEs) including amputation. Major amputation due to vascular causes is common in patients with a history of peripheral artery disease (PAD), revascularization and recurrent atrial thrombus (8,9). Even though arterial and venous thrombosis are generally considered to be separate entities due to differences in vasculature and clot formation. Moreover, body of evidence have shown associations between venous thromboembolism (VTE), including pulmonary embolism (PE) and deep vein thrombosis (DVT), share clusters of CV risks [10], subsequent arterial cardiovascular events [11][12][13][14] and mortality [15]. Chang et al. were the rst to report a signi cant association between hemostatic varicose and peripheral vascular disease, including DVT and PAD (16); however, the exact mechanisms have yet to be identi ed.
Deep vein thrombosis is prevalent in patients with diabete s (17), and it is regarded to be a prothrombotic state with complications including pulmonary embolism and post-thrombotic syndrome (PTS). Prandoni et al. reported a signi cantly higher rate of subsequent arterial events in patients with DVT and residual thrombus (18), and that this could lead to PTS complicated with edema, venous stasis, ulceration, and thrombophlebitis with focal in ammation. Most previous studies have investigated associations between VTE and atherosclerotic cardiovascular disease in the general population, however few studies have focused on potentially life-threatening limb arterial complications in patients with diabetes in this context.
We hypothesized that DVT may be associated with an increased risk of mortality due to a higher rate of subsequent MALEs and major adverse cardiovascular events (MACEs). Therefore, the aim of this study was to assess the in uence of DVT on MALEs and MACEs in patients with a new diagnosis of type 2 diabetes.

Data source
This study was a retrospective cohort study using data from the National Health Insurance Research Database (NHIRD) of Taiwan (19). The NHIRD is derived from the claims data of the Taiwan National Health Insurance (NHI) program, a mandatory and single-payer system that covers more than 99% of the 23 million residents in Taiwan. The NHIRD prospectively records all submitted standardized data of healthcare services, including details of demographic characteristics, out-patient medical record, procedures, hospitalizations, drug prescriptions, medical diagnoses of diseases, and vital status.

Diseases are registered using International Classi cation of Diseases, Ninth Revision, Clinical
Modi cation (ICD-9-CM) codes. Validation of the NHIRD data is routinely conducted by the NHI Bureau, and all information in the NHIRD is de-identi ed to protect privacy. The Research Ethics Committee of Chang-Gung Memorial Hospital approved the study protocol (201900915B1).

Patient identi cation
Patients with a diagnosis of type 2 diabetes in ≥ 5 outpatient visits de ned as ICD-9-CM code 250.xx combined with the use of any oral hypoglycemic agents between 2001 and 2013 were identi ed. The accuracy of a diagnosis of diabetes in the NHIRD has previously been validated (20,21). The date of the rst diagnosis of diabetes was de ned as the index date. The patients were divided into two groups: those with coexisting deep vein thrombosis (DVT group), and those without DVT (non-DVT group). A history of DVT was identi ed using the discharge diagnosis (ICD-9-CM code 453) combined with the use of anticoagulant therapy during an admission before the index date. The accuracy of a diagnosis of DVT has been validated in previous NHIRD studies (22). The exclusion criteria were: 1) missing demographic data; 2) age < 20 years; 3) type 1 diabetes; 4) history of malignancy, autoimmune diseases, and systemic thromboembolism; 5) history of foot ulcers and gangrene; and 6) medical records of percutaneous transluminal angioplasty (PTA) and amputation. In order to focus on the effect of a history of DVT, patients who had received oral anticoagulant (OAC) treatment 3 months before the index date were also excluded ( Fig. 1).

Covariates
The covariates in this study were age, sex, comorbidities (hypertension, dyslipidemia, ischemic heart disease, heart failure, atrial brillation, PAD, chronic obstructive pulmonary disease, gouty arthritis, chronic kidney disease (CKD) or dialysis, liver cirrhosis), Charlson's Comorbidity Index score, history of events (prior ischemic stroke and previous myocardial infarction), hyperglycemia lowering drugs (metformin, sulfonylurea, thiazolidinedione, dipeptidyl peptidase-4 inhibitors [DPP4is] and insulin) and other medications (antiplatelet, anticoagulation, angiotensin converting enzyme inhibitors / angiotensin receptor blockers [ACEIs/ARBs], ß-blockers, dihydropyridine calcium channel blockers, diuretics, statins and cilostazol). Comorbidities were identi ed according to at least two clinic visits or any inpatient record in the previous year before the index date and were validated after review by one of the authors [Hsieh, Cheng-Yang]. Historical events were identi ed using any inpatient diagnosis before the index date up to 1997. The use of medications was extracted within 6 months before the index date. All information on medications was extracted from the claims data of outpatient visits or re lls for chronic illnesses at the pharmacy using Anatomical Therapeutic Chemical codes or Taiwan NHI reimbursement codes.

Outcomes
The primary outcome was MALE, which was de ned as a composite of foot ulcer, gangrene, PTA, and amputation. Individual MALE components were also investigated. The occurrence of ulcer and gangrene was based on ICD-9-CM diagnostic codes in at least two outpatient visits or any inpatient diagnosis (Supplemental Table S1). The occurrence of PTA and amputation was identi ed using Taiwan NHI reimbursement codes in inpatient claims data with validation according to a prior study (23).
The secondary outcomes were all-cause mortality, cardiovascular death, ischemic stroke, acute myocardial infarction, heart failure hospitalization and systemic thromboembolism. All-cause mortality was de ned as withdrawal from the NHI program (24

Statistical analysis
For comparisons of the risk of outcomes between the DVT and non-DVT groups, propensity score matching was performed to reduce possible confounding. The propensity score was the predicted probability to be in one group (i.e., DVT) given the values of covariates using multivariable logistic regression without considering interaction effects. The variables selected to calculate propensity score are listed in Table 1, where the follow-up duration was replaced with the index date. Each subject in the DVT group was matched with ve counterparts in the non-DVT group to increase the statistical power.
The matching was processed using a greedy nearest neighbor algorithm with a caliper of 0.2 times the standard deviation of the logit of the propensity score, with random matching order and without replacement. The balance after propensity score matching was checked using the absolute value of standardized difference (STD) between the groups, where a value < 0.1 was considered to be a negligible difference. The risks of fatal outcomes (all-cause mortality and cardiovascular death) between the groups were compared using a Cox proportional hazard model. The incidence rates of non-fatal outcomes (MALEs or acute myocardial infarction) between groups were compared using a Fine and Gray subdistribution hazard model which considered all-cause mortality as a competing risk. The study group (DVT vs. non-DVT) was the only explanatory variable in the survival analyses. Finally, post-hoc subgroup analysis of the primary outcomes with statistical signi cance (including MALEs, amputation, and systemic thromboembolism) were conducted to assess whether the effect of DVT on outcomes was consistent across different levels of subgroup variables. Selected subgroup variables were age (dichotomized by 75 years), sex, hypertension, dyslipidemia, ischemic heart disease, heart failure, PAD, CKD, prior stroke, and Charlson's Comorbidity Index score (dichotomized by 2 points).
A two-sided P value < 0.05 was considered to be statistically signi cant, and no adjustment of multiple testing (multiplicity) was made in this study. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC), including the 'psmatch' procedure for propensity score matching, 'phreg' for survival analysis, and a macro of '%cif' to generate the cumulative incidence function under the Fine and Gray subdistribution hazard method.

Baseline characteristics
A total of 1,628,675 patients with type 2 diabetes were included, of whom 2,020 (0.12%) had a history of DVT hospitalizations before propensity score matching (PSM). The patients in the DVT group were older (66.3 vs. 58.3 years), were more predominantly female (54.3% vs. 47.5%), had higher prevalence rates of several comorbidities (including hypertension, ischemic heart disease, heart failure, atrial brillation, PAD, chronic obstructive pulmonary disease, gout, CKD and liver cirrhosis), and a higher Charlson's Comorbidity Index score than the patients in the non-DVT group. In addition, more were likely to have had a previous stroke and myocardial infarction, and to have more prescriptions for anti-platelet agents, ACEIs/ARBs, beta-blockers, dihydropyridine calcium channel blockers, diuretics, cilostazol and insulin compared to the patients in the non-DVT group. However, the DVT group was less frequently prescribed with metformin or sulfonylurea. After PSM, the distribution of baseline characteristics between groups was well balanced, with all standardized difference absolute values < 0.  (Fig. 3B). The risks of cardiovascular death, acute myocardial infarction, ischemic stroke, and heart failure hospitalization were not signi cantly different between the two groups (supplemental Table 2).

Subgroup analysis
We further conducted subgroup analysis of the primary outcomes with statistical signi cance (including MALEs, amputation, and systemic thromboembolism). The association between a history of DVT and an increased risk of MALEs was more obvious in the patients without CKD (P for interaction < 0.05; supplemental Figure A). In contrast, the association between a history of DVT and risks of amputation and systemic thromboembolism were consistent across different subgroups (supplemental Figure B and C).

Discussion
This longitudinal cohort study identi ed signi cant associations between DVT and MALEs including major amputation and severe wound conditions (ulcer and gangrene) in patients with type 2 diabetes. To the best of our knowledge, this is the rst study to show that a history of prior DVT was independently associated with a 1.6-fold higher risk of MALEs and 1.4-fold higher risk of major amputation. These critical events may contribute to functional disability, future mortality, and insurance costs for long-term diabetic care.
In this study, more than 21% of the enrolled diabetic patients with DVT were very elderly (age > 75 years) and had a history of PAD and various comorbidities, both of which tend to be risk factors for MALEs (25).
Female patients have also previously been reported to have worse outcomes after revascularization, however male sex was shown to independently increase the risk of PTS ulcers in the RIETE registry (26). During long-term follow-up (5.2 years), the average 5-year mortality rate in the current study was 38.1%, which is higher than that in the study of Chang et al. (17.97% at 5 years) (16). The proportion of patients with diabetes in Chang's study was 18.43% lower than in our study; however, they found that male patients with diabetes and unprovoked VTE had a higher risk of mortality. Our results suggest that DVT increased the incidence of lower limb complications, arterial thrombosis, and amputation, thereby contributing to poor survival in our patients with type 2 diabetes.
Factors reported to contribute to major limb amputation include wound infection, microvascular dysfunction (3), and recurrent arterial thrombosis with chronic luminal occlusion (8,9). Our ndings indicate that DVT also has a signi cant in uence on foot ulcer and gangrene in patients with diabetes. The EUCLID trial showed reported an overall rate of major amputation of 8.4% in patients with critical limb ischemia (4), while the rate of amputation in the DVT group was 2.5% in our study. The rate of revascularization for PAD is low in patients with diabetes and DVT because advanced endovascular interventions are increasingly being used in Taiwan (27). The COMPASS and VOYAGER studies (6,7) concluded that thrombosis should not be ignored in PAD, because the addition of anticoagulants can successfully reduce MALEs and mortality. Therefore, systemic thromboembolism may be a major risk factor for MALEs as supported by our results. Even though there were more below the knee amputations than above the knee amputations (38 vs. 17 in the DVT group), the risk of incidental above the knee amputations was still signi cantly higher (1.97-fold) in the DVT group. Mao et al. reported a high incidence of above the knee amputations in patients with atrial brillation receiving PTA, and they concluded that intra-cardiac or up-stream intra-arterial thrombus was the major source (28). In addition, the patients receiving above the knee amputations had a poor survival rate because of poor limb salvage, functional disability, and ambulance, which is compatible with our analysis. Therefore, DVT may increase the risk of incidental above the knee amputations and mortality in patients with diabetes.
Several studies have reported associations between VTE and cardiovascular disease with cardiovascular risk burden (10) and arterial thrombosis (11). However, the in uence of DVT on ischemic stroke and myocardial infarction may be inconsistent due to previous studies focusing on unprovoked VTE or pulmonary embolism (16), in contrast to our study population (diabetic patients with DVT). A 20-year cohort study with a large number of patients reported that DVT and pulmonary embolism were risk factors for acute arterial cardiovascular events; however, the relative risk decreased during 1 ~ 5 years (11). Consuelo et al. further reported that VTE was not associated with subsequent acute myocardial infarction but death (29). Furthermore, the severity of diabetes, proteinuria, and other CVD-associated risk factors (dyslipidemia and hypertension) have been reported to be major risk factors for future MACEs and cardiac dysfunction. Heart failure is also a known risk factor for DVT (32), increasing atrial brillation burden (31), physical inactivity and impaired tissue perfusion due to lower cardiac output. Nevertheless, incidental heart failure hospitalizations and cardiovascular events were not prevalent in our DVT group. Taken together, we suggest that the association between DVT and all-cause mortality is closed related to MALEs and systemic thromboembolism rather than traditional cardiovascular risk factors and MACEs.
Moreover, our ndings suggest that MALEs are mainly due to DVT-associated pathomechanisms, a prothrombotic state and systemic thromboembolism.
Venous thrombosis can induce recurrent thromboembolism, peripheral edema, and in ammation as well as accelerated diabetic atherosclerosis. Those patients with DVT had 2.7 ~ 7.1% PTS for 1 ~ 3 years observation, and diabetes is also a risk factor of PTS (2.3-fold) (26). Residual thrombus venous hypertension has been shown to play a central pathogenetic role in PTS, and in turn to cause dilatation of the capillaries and increased endothelial permeability to plasma, proteins, and erythrocytes (32). The consequences of PTS can result in valve insu ciency, chronic edema, in ammation, hyperpigmentation of the skin, and stasis dermatitis, or even the development of a venous ulcer. Persistent edema has been shown to affect wound healing (33) and induce an in ammatory cascade as evidenced by increasing levels of CRP (C-Reactive Protein), platelet aggregation (34), endothelial dysfunction (35), coagulation disturbance and cytokine activation.

Strengths and Limitations
DVT is less prevalent in Asia than in European countries, and the NHIRD is a good resource for such investigations. However, retrospective surveys have inherent limitations including a lack of data on DVT/PTS severity, echo reports, laboratory ndings, body weight, lifestyle habits (such as smoking), infected wound condition and physical activity. The RIETE registry observed that the characteristics of obesity and DVT treatment (duration or drugs) failed to predict the risk PTS after acute DVT (26). Hyperglycemia affects diabetic vascular complications, however, A1c level is not a predictor associated with amputation (5) or recurrent VTE (36). Inadequate self-care, living alone without family support, lower social-economic status, and infection with a high white blood cell count have also been shown to be independent risk factors for adverse outcomes in long-term wound care. This may also suggest that we underestimated ulcer/gangrene and severity because less severe foot cases may have been treated at clinics. However, in-hospital records of ulcer and gangrene may be considered to be critical limb events.
Thrombophilia and hyperhomocysteinemia have been reported to contribute to both venous and arterial thrombosis (37). Many factors involving hypercoagulation status are likely to be lost in general practice. We excluded patients with malignancy and autoimmune diseases associated with unprovoked DVT to decrease the effects of hypercoagulation status and recurrent DVT under inadequate therapy.
In conclusion, we found independent associations between DVT and MALE outcomes including major limb amputation, systemic thromboembolism, and mortality in patients with type 2 diabetes. The impact of a history of DVT highlights the importance of thrombotic prevention in diabetic foot care. We needed a further large-scale randomized control trial to prove the relationship.

Declarations
Acknowledgments: The authors thank Mr. Alfred Hsing-Fen Lin, MS and Mrs. Bing-Yu Chen PhD, for their assistance of the statistics analysis.
Author contributions: All authors contributed to statistical analysis and writing of the study. PCW, THC and MSL and participated in study design, acquisition of the data, critical review and wrote the manuscript; MYC, CMC, JJC participated in acquisition of data and designed the research; YSL, PHC, YSP participated in the analysis and interpretation of data; MSL collected data and contributed to study direction.
Funding and Duality of Interest: The authors declared they do not have anything to disclose regarding conflicts of interest with respect to this manuscript. This work was also supported by grants from the Chang Gung Memorial Hospital, Taiwan (CLRPG2G0082).
Availability of data and material: The data used to support the ndings of this study are available from the corresponding author upon request.

Ethics approval and consent to participate
The hospital identi cation number of each subjects was encrypted and de-identi ed to protect their privacy because of retrospective nature of the database study. Therefore, informed consent could be waived for this research. The institutional review board of Chang Gung Memorial Hospital had approved the study protocol.
Availability of data and materials All data will be available upon reasonable request.

Consent for publication
All authors have read the manuscript and consent for publication.   Cumulative event rate of all-cause mortality (A) and cumulative incidence function of systemic thromboembolism (B) of the patients with type 2 diabetes strati ed by a history of deep vein thrombosis hospitalization during follow-up in the propensity score-matched cohort.

Figure 3
Cumulative event rate of all-cause mortality (A) and cumulative incidence function of systemic thromboembolism (B) of the patients with type 2 diabetes strati ed by a history of deep vein thrombosis hospitalization during follow-up in the propensity score-matched cohort.