Abstract
Purpose:
Bariatric surgery is associated with a greater venous thromboembolism (VTE) risk in the weeks following surgery, but the long-term risk of VTE is incompletely characterized. We evaluated bariatric surgery in relation to long-term VTE risk.
Materials and Methods:
This population-based retrospective matched cohort study within three United States-based integrated health care systems included adults with body mass index (BMI) ≥35 kg/m2 who underwent bariatric surgery between January 2005 and September 2015 (n=30,171), matched to nonsurgical patients on site, age, sex, BMI, diabetes, insulin use, race/ethnicity, comorbidity score, and health care utilization (n=218,961). Follow-up for incident VTE ended September 2015 (median 9.3, max 10.7 years).
Results:
Our population included 30,171 bariatric surgery patients and 218,961 controls; we identified 4,068 VTE events. At 30 days post-index date, bariatric surgery was associated with a 5-fold greater VTE risk (HRadj=5.01; 95% CI: 4.14, 6.05) and a nearly 4-fold greater PE risk (HRadj=3.93; 95% CI: 2.87, 5.38) than no bariatric surgery. At 1-year post-index date, bariatric surgery was associated with a 48% lower VTE risk and a 70% lower PE risk (HRadj=0.52; 95% CI: 0.41, 0.66, and, HRadj=0.30; 95% CI: 0.21, 0.44, respectively). At 5-years post-index date, lower VTE risks persisted, with bariatric surgery associated with a 41% lower VTE risk and a 55% lower PE risk (HRadj=0.59; 95% CI: 0.48, 0.73, and HRadj=0.45; 95% CI: 0.32, 0.64, respectively).
Conclusion:
Although in the short-term, bariatric surgery is associated with a greater VTE risk, in the long-term it is associated with a substantially lower risk.
Keywords: Bariatric surgery, Venous thromboembolism, Obesity
Graphical Abstract
Introduction
Bariatric surgery is associated with a greater risk of venous thromboembolism (VTE, including pulmonary embolism [PE] and deep vein thrombosis [DVT]) in the short-term, with rates of VTE occurring in the first few weeks following bariatric surgery ranging from approximately 1 to 3%.(1-5) Broadly, surgery is a well-characterized risk factor for VTE(6), working through mechanisms including endothelial damage and venous stasis, and administration of anticoagulant prophylaxis is common in the early postsurgical phase by parenteral anticoagulation followed by longer-term vitamin K antagonist use in patients requiring anticoagulation.(7-9) However, whether the elevated risk of VTE following bariatric surgery is sustained in the longer-term post-surgery is not well-characterized. Given that higher body mass index (BMI) is positively associated with VTE risk, with obesity associated with an estimated 2 to 3-fold greater risk of incident VTE(10-13), the reduction in BMI following bariatric surgery has the potential to decrease VTE risk in the long-term. Bariatric surgery was associated with a 40% lower risk of VTE after a median follow-up of over 10 years in a retrospective matched cohort study using data from the United Kingdom’s Clinical Practice Research Datalink(14), but to our knowledge, no studies have evaluated the long-term risk of incident VTE after bariatric surgery in the United States.
Within a population-based cohort of patients with severe obesity, we conducted a retrospective matched cohort study to evaluate bariatric surgery in relation to long-term risk of incident VTE. We hypothesized that bariatric surgery would be associated with a lower long-term risk of VTE than no bariatric surgery. However, we also anticipated non-proportional hazards for this association over time, given that bariatric surgery was likely to be adversely associated with VTE risk in the first few weeks post-surgery. Our secondary aim was to separately evaluate the association of bariatric surgery with incident PE risk.
Materials and Methods
Design and Setting
We used data from a large, retrospective matched cohort study of adults with severe obesity enrolled in three regions of the United States-based Kaiser Permanente integrated healthcare delivery system in Northern California (KPNC), Southern California (KPSC), and Washington (KPWA).(15, 16) As an integrated healthcare system, Kaiser Permanente is a provider of both healthcare and insurance coverage, with rich electronic data about their members including electronic clinical, pharmacy, and administrative data. Approval was obtained from the Institutional Review Board at each site, allowing the research to be conducted without explicit participant consent.
Data Sources
At each participating site, electronic medical records, insurance claims, and other data systems were used to extract enrollment, insurance coverage, demographics, blood pressure, height, weight, laboratory values, medications dispensed, deaths, outpatient, inpatient, and emergency department use, and diagnosis and procedure codes of all surgical and non-surgical patients.
Study Population: Inclusion and Exclusion Criteria
Surgical Patients.
The population of bariatric surgery patients included adults (19-79 years) who had a primary (first observed, non-revisional) Roux-en-Y Gastric Bypass (RYGB) (n=18,080) or sleeve gastrectomy (SG) (n=14,590) operation between January 2005 and September 2015 (total n=32,670). Consistent with our previous work,(18-21) we identified bariatric operations using a combination of bariatric surgery registries for clinical and research use available at one site, chart review, and International Classification of Diseases (ICD-9) and Current Procedure Terminology (CPT) procedure codes. We excluded patients who had less than one full year of continuous enrollment, a history of cancer (except non-melanoma skin cancer), or missing pre-operative BMI data.
We further excluded 1,046 (3.2%) with a maximum preoperative BMI <35 kg/m2 (resulting in a population with obesity class II or higher [BMI ≥35 kg/m2]),(15, 17) 241 (0.74%) with pregnancy in the year prior to surgery, 25 (0.08%) who could not be matched to at least one non-surgical patient, and 1,187 (3.6%) with a history of VTE prior to surgery, resulting in 30,171 eligible surgical patients.
Matched Non-Surgical Patients.
For each patient who underwent bariatric surgery, we identified up to 10 matched non-surgical controls from our general medical population via a two-step process. First, among all patients with at least one BMI ≥35 kg/m2 who did not undergo bariatric surgery during the study period (N=1,635,897), we identified a pool of potential controls who were enrolled at the time of the surgery, satisfied the study inclusion/exclusion criteria, and matched the surgical patients to these controls on study site, sex, baseline age category (19-44, 45-64, 65-69 years), BMI category (35-39.9, 40-49.9, 50.0+ kg/m2), diabetes status (presence/absence based on laboratory, pharmacy, and diagnosis data), race and ethnicity, combined Charlson/Elixhauser comorbidity score(21), and insulin use. Second, for each control in the pool we calculated their Mahalanobis distance from the bariatric patient on the basis of age, BMI, Charlson/Elixhauser comorbidity score, and the number of days of health care utilization in the 7-12 months prior (a marker of comorbidity that is unaffected by utilization related to preparation for bariatric surgery).(22) Finally, up to 10 controls were selected based on the shortest Mahalanobis distance; each non-surgical patient was used as a control for only one surgical patient. These non-surgical patients all received usual medical care, which usually included no specific treatment for obesity.
Venous Thromboembolism
For our primary analysis, we identified the first occurrence of incident VTE using ICD-9 codes for PE and/or DVT (see Supplemental Digital Content Table 1). For sensitivity analyses, we required patients to have ICD-9 codes for PE and/or DVT and at least one anticoagulation prescription fill within 7 days after the ICD-9 code. As a secondary outcome of interest, we separately evaluated risk of incident PE with or without concurrent DVT.
Statistical Analyses
Participants contributed person-time from the index date (the date of bariatric surgery or, for non-surgical patients, the date of surgery for the patient to whom they had been matched) until death or a censoring event (disenrollment, end of the study period [September 30, 2015], incident cancer, pregnancy). We used Cox models to estimate multivariable-adjusted hazard ratios (HRs) for incident VTE associated with bariatric surgery vs. usual care (no surgery). All analyses were adjusted for factors used to match study cases with controls (site of surgery, age categories, most recent BMI pre-index date (continuous), sex, diabetes status, race and ethnicity, Charlson/Elixhauser comorbidity score, insulin use in the prior year, health care utilization in the 7 to 12 months prior to index date) and potential confounders identified a priori (dyslipidemia, hypertension diagnosis, current hormone therapy [HT] use as of the index date, current oral contraceptive [OC] use as of the index date, and smoking status [current/former vs. never]). Given that we hypothesized non-proportional hazards a priori, we considered several functional forms to permit non-proportional hazards in the effect of bariatric surgery, including interactions with log(time), and natural cubic splines [NCS] ranging from 1 to 6 degrees of freedom. Candidate models were compared via Bayesian information criterion [BIC]. In our primary analysis, an NCS with 5 degrees of freedom had the lowest BIC, while an NCS with 4 degrees of freedom provided the best BIC for the secondary analysis examining just PE. This modeling framework allows us to estimate multivariable-adjusted HR at any point over the course of follow-up, with 30 days, 1-year, and 5-years post-index date being of particular interest.
In secondary analyses, we separately evaluated the risk of incident PE. In sensitivity analyses, we employed our stricter definition of VTE that required participants to have ICD-9 codes for PE and/or DVT and at least one anticoagulation prescription fill within 7 days after the ICD-9 code.
In exploratory analyses, we evaluated time-varying interaction on the multiplicative scale of the association between bariatric surgery treatment and VTE incidence separately by BMI (≥50 vs. <50 kg/m2), age (<65 vs. ≥65 years), and sex (male vs. female), by including an interaction with the NCS for the effect of surgery over time along with main effects in a model and evaluated the presence of interaction using a likelihood ratio test p-value of p<0.05. A priori, we had also intended to evaluate potential interaction by racial and ethnic categories, but small samples sizes within racial and ethnic subgroups precluded us from this evaluation.
Finally, we conducted a sensitivity analysis to assess the influence of unmeasured confounding using the E-value methodology of VanderWeele and Ding.(23, 24) The E-value quantifies the minimum strength of association an unmeasured confounder must have with both bariatric surgery and VTE risk, while simultaneously considering the measured confounders, to negate the observed association between bariatric surgery and VTE risk in this study.(24) All analyses were performed in R version 4.1.3 on the on the FAS Secure Environment research computing cluster at Harvard University.
Results
Our study included 30,171 patients who underwent a bariatric procedure (44.6% SG; 55.4% RYGB) (Table 1) and 218,961 matched controls who did not. Participants were followed for a median of 9.3 years and a maximum of 10.7 years. After 5 years, 71.8% of SG patients, 70.8% of RYGB patients, and 65.4% of nonsurgical patients remained in follow-up. In the bariatric surgery group, the average age was 45 years, 81.7% were female, and 31.8% of Hispanic ethnicity (Table 1). One third of both surgical and non-surgical patients had diabetes, with 10% of these patients using insulin.
Table 1.
Baseline characteristics of bariatric surgery patients and patients with severe obesity who did not undergo bariatric surgery, January 1, 2005 to September 30, 2015.
| Non-Surgical Controls |
Surgical Patients |
|
|---|---|---|
| (n=218,961) | (n=30,171) | |
| Sleeve Gastrectomy, n (%) | - | 13,468 (44.6%) |
| Roux-en-Y Gastric Bypass, n (%) | - | 16,703 (55.4%) |
| Year of Surgery (i.e. Index Date), n (%) | ||
| 2005-2007 | 26,672 (12.2%) | 3,861 (12.8%) |
| 2008-2010 | 53,672 (24.5%) | 8,005 (26.5%) |
| 2011-2013 | 87,678 (40.0%) | 12,254 (40.6%) |
| 2014-2015 | 50,939 (23.3%) | 6,051 (20.1%) |
| Health Care Site, n (%) | ||
| Kaiser Permanente Northern California (KPNC) | 87,358 (39.9%) | 9,554 (31.7%) |
| Kaiser Permanente Southern California (KPSC) | 122,257 (55.8%) | 19,364 (64.2%) |
| Kaiser Permanente Washington (KPWA) | 9,346 (4.3%) | 1,253 (4.2%) |
| Days Health Care Utilization in 7-12 Months Prior to Index Date, mean (SD) | 5.4 (4.9) | 7.9 (6.8) |
| Days Hospitalized in Year Prior to Index Date, mean (SD) | 0.73 (2.9) | 0.24 (1.3) |
| Age (years), mean (SD) | 45.6 (11.7) | 45.0 (11.0) |
| Age Categories, n (%) | ||
| <45 y | 103,859 (47.4%) | 14,705 (48.7%) |
| ≥45 y to <65 y | 106,196 (48.5%) | 14,436 (47.8%) |
| ≥65 y to <80 y | 8,906 (4.1%) | 1,030 (3.4%) |
| Female, n (%) | 174,651 (79.8%) | 24,652 (81.7%) |
| Race and Ethnicity, n (%) | ||
| Hispanic | 67,487 (30.8%) | 9,580 (31.8%) |
| Non-Hispanic Black | 33,555 (15.3%) | 5,041 (16.7%) |
| Non-Hispanic White | 106,446 (48.6%) | 14,175 (47.0%) |
| Other | 7,644 (3.5%) | 892 (3.0%) |
| Unknown/missing | 3,829 (1.7%) | 483 (1.6%) |
| BMI, kg/m2, mean (SD) | 42.6 (5.99) | 44.3 (6.84) |
| BMI Categories, kg/m2, n (%) | ||
| ≥35-<40 | 82,158 (37.5%) | 8,923 (29.6%) |
| ≥40-<50 | 110,889 (50.6%) | 15,896 (52.7%) |
| ≥50 | 25,914 (11.8%) | 5,352 (17.7%) |
| Charlson/Elixhauser Comorbidity Score, n (%) | ||
| −1 | 40,118 (18.3%) | 5,709 (18.9%) |
| 0 | 94,005 (42.9%) | 12,030 (39.9%) |
| 1 | 54,028 (24.7%) | 7,436 (24.6%) |
| 2+ | 30,810 (14.1%) | 4,996 (16.6%) |
| Diabetes Statusa, n (%) | 69,004 (31.5%) | 10,071 (33.4%) |
| Insulin Use, n (%) | 21,549 (9.8%) | 3,189 (10.6%) |
| Dyslipidemia Diagnosis, n (%) | 73,003 (33.3%) | 13,775 (45.7%) |
| Statin Use, n (%) | 49,852 (22.8%) | 7,205 (23.9%) |
| Hypertension Diagnosis, n (%) | 96,051 (43.9%) | 16,540 (54.8%) |
| History of Cardiac Event or Ischemic Heart Disease | 916 (0.4%) | 32 (0.1%) |
| History of Ischemic or Hemorrhagic Stroke | 384 (0.2%) | 4 (0.0%) |
| Current/Former Smoker, n (%) | 70,464 (32.2%) | 10,395 (34.5%) |
| Oral Contraceptive Use at Index Date, n (%) | 13,873 (6.3%) | 1,661 (5.5%) |
| Hormone Therapy Use at Index Date, n (%) | 3,773 (1.7%) | 562 (1.9%) |
| Anticoagulant Use at Index Date, n (%) | 2,157 (1.0%) | 238 (0.8%) |
| Anticoagulant Use in 30 Days Prior to Index Date, n (%) | 2,123 (1.0%) | 193 (0.6%) |
Diabetes status is the presence/absence of diabetes determined using laboratory, medication, and diagnosis data.
ACE = angiotensin converting enzyme; ARB = angiotensin receptor blocker; BMI = body mass index; SD = standard deviation; y = years.
In the leftmost plots within Figure 1, we present Kaplan-Meier cumulative incidence curves for time since index date to incident VTE event and, separately, PE. As shown in these curves, lines indicating the cumulative incidence associated with bariatric surgery vs. control appear to cross over time (p<0.0001 in VTE analyses; p=0.034 in PE analyses). In Figure 1‘s rightmost plots, we present the HR at any given point over time associated with bariatric surgery in relation to incident VTE risk, and separately, PE. As shown in these plots of HRs over time, although bariatric surgery is initially associated with HRs above 1 (indicating a greater VTE and PE risk associated with bariatric surgery), in the plots for both VTE and PE risk, the HR drops to <1 in the first 6 months following bariatric surgery.
Figure 1. Kaplan-Meier Cumulative Incidence Curves and Hazard Ratios over Time for VTE and PE identified by ICD-9 code.
Plots on the leftmost side of Figure 1 present Kaplan-Meier cumulative incidence curves for time to incident VTE (identified by ICD-9 code) (top left) and time to PE with or without DVT (identified by ICD-9 code) (bottom left). Plots on the rightmost side of Figure 1 present hazard ratios over time since index for the association between bariatric surgery and incident VTE (identified by ICD-9 code) (top right) and PE with or without DVT (identified by ICD-9 code) (bottom right).
DVT: deep vein thrombosis; PE: pulmonary embolism; w/: with; w/out: without.
As reported in Table 2, at 30 days post-index date, patients who had undergone bariatric surgery were at a 5-fold greater risk of any VTE (HRadj=5.01; 95% CI: 4.14, 6.05) and a nearly 4-fold greater risk of PE (HRadj=3.93; 95% CI: 2.87, 5.38) relative to patients who had not undergone bariatric surgery. However, at 1-year post-index date, bariatric surgery patients were at a 48% lower risk of any VTE and a 70% lower risk of PE (HRadj=0.52; 95% CI: 0.41, 0.66, and HRadj=0.30; 95% CI: 0.21, 0.44, respectively). At 5-years post-index date, bariatric surgery patients were at a 41% lower risk of any VTE and a 55% lower risk of PE (HRadj=0.59; 95% CI: 0.48, 0.73, and HRadj=0.45; 95% CI: 0.32, 0.64, respectively).
Table 2:
Long-term risk of incident VTE, defined using ICD-9 codes, among bariatric surgery patients in relation to patients with severe obesity who did not undergo bariatric surgery.
| Non-Surgical Controls | Surgical Patients | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Cumulative Person- Years |
Cumulative # of Events |
Rate/ 1000 Person- Years |
Cumulative Person-Years |
Cumulative # of Events |
Rate/ 1000 Person-Years |
Adjusted HR1 | 95% CI | p-value | |
| Any VTE | |||||||||
| 30 Days Post Index Date | 17,768 | 99 | 5.6 | 2,453 | 185 | 75.4 | 5.01 | (4.14, 6.05) | <0.001 |
| 1 Year Post Index Date | 190,558 | 953 | 5.0 | 27,148 | 328 | 12.1 | 0.52 | (0.41, 0.66) | <0.001 |
| 5 Years Post Index Date | 578,131 | 2,927 | 5.1 | 88,284 | 565 | 6.4 | 0.59 | (0.48, 0.73) | <0.001 |
| PE | |||||||||
| 30 Days Post Index Date | 17,770 | 45 | 2.5 | 2,459 | 56 | 22.8 | 3.93 | (2.87, 5.38) | <0.001 |
| 1 Year Post Index Date | 190,841 | 325 | 1.7 | 27,324 | 82 | 3.0 | 0.30 | (0.21, 0.44) | <0.001 |
| 5 Years Post Index Date | 581,408 | 1,056 | 1.8 | 89,317 | 142 | 1.6 | 0.45 | (0.32, 0.64) | <0.001 |
Adjusted for matching factors (site of surgery, age categories, most recent BMI pre-index, sex, diabetes status, race and ethnicity, Charlson/Elixhauser comorbidity score, insulin use in the prior year) and potential confounders identified (health care utilization in the 7 to 12 months prior to index date, dyslipidemia, hypertension diagnosis, current hormone therapy use as of index date, current oral contraceptive use as of the index date, smoking status).
CI: confidence interval; HR: hazard ratio; PE: pulmonary embolism; VTE: venous thromboembolism.
In sensitivity analyses that required an anticoagulation prescription fill within 7 days of VTE diagnosis to consider a VTE to be valid (Supplemental Digital Content Table 2), the number of eligible events that had occurred at 5 years post-index date decreased (2,927 to 1,232 VTE events; 1,056 to 656 PE events). Hazard ratio estimates shifted further from the null at 30-days (HRadj for any VTE=11.93; 95% CI: 8.38, 16.98; HRadj for PE=7.45; 95% CI: 4.64, 11.97) and at 5-years post-index date (HRadj for any VTE=0.64; 95% CI: 0.48, 0.85; HRadj for PE=0.56; 95% CI: 0.37, 0.85) but closer to the null at 1-year post-index date (HRadj for any VTE=0.50; 95% CI: 0.37, 0.66; HRadj for PE=0.29; 95% CI: 0.18, 0.46), with no meaningful difference in interpretation of results. In Supplemental Digital Content Figure 1 we present Kaplan-Meier cumulative incidence curves and HRs over time for time since index date (in years) to incident VTE and PE, with this event requirement of an anticoagulation prescription fill.
In exploratory analyses, we found no evidence that the association between bariatric surgery and risk of any VTE or PE differed by age, BMI categories, or sex (all likelihood ratio p-values >0.05).
In sensitivity analyses that assessed the potential influence of unmeasured confounding using the E-value methodology, we found that the observed 5-year HR of 0.59 for incident VTE could only be explained by an unmeasured confounder that was associated with both receipt of bariatric surgery and risk of VTE by a risk ratio of more than 2.76 above and beyond that of the confounders that were measured in this study (upper confidence bound, 2.08). (Supplemental Digital Content Table 3)
Discussion
In this large, population-based multisite cohort study, we found that although bariatric surgery is associated with a greater risk of VTE in the short-term, this increased risk does not persist long-term, with bariatric surgery being associated with a 48% and 41% lower risk of any VTE in the 1-year and 5-years post-index date, respectively.
To our knowledge, only one prior study has evaluated the long-term risk of incident VTE after bariatric surgery, and ours is the first US-based study.(14) Our findings are consistent with this prior retrospective matched cohort study which used data from the United Kingdom’s Clinical Practice Research Datalink, in which bariatric surgery was associated with a 40% lower risk of VTE (HR=0.60; 95% CI: 0.43, 0.84) after a median follow-up of 10.7 years.(14) In this UK-based study, results differed by VTE outcome, with bariatric surgery associated with a lower risk of DVT but not PE long-term (HR=0.52; 95% CI: 0.35, 0.78, and, HR=0.88; 95% CI: 0.51, 1.52, respectively). This difference in association by VTE outcome differs from our study, in which bariatric surgery was also significantly associated with a lower risk of PE long-term (1-year post-index date HR=0.30; 95% CI: 0.21, 0.44, and, 5-year post-index date HR=0.45; 95% CI: 0.32, 0.64). However, our study was much larger in size than the UK-based study(14) (n=4,073 bariatric patients and n=4,073 controls in the UK-based study vs. n=30,171 bariatric patients and n=218,961 controls in our study), making it more likely that we would have sufficient sample size in our study to detect associations specific to PE.
Our findings highlight that while in the short-term, bariatric surgery is associated with a greater risk of VTE, this balance of risks and benefits shifts strongly in the direction of a lower VTE risk in the long-term. This result has important implications to the discussion of risks and benefits related to bariatric surgery. It supports consideration of not only the greater risk of VTE following bariatric surgery in the short term(1-4), but also of potential long-term reductions in VTE risk that may accompany reductions in BMI. Obesity is associated with a 2 to 3-fold greater risk of VTE(10-13) with estimated population attributable risks suggesting that, in the United States, obesity may explain 30% of all VTEs.(25) There are multiple proposed mechanisms linking obesity with VTE risk, including obesity’s association with venous stasis, changes in levels of hemostatic and inflammatory markers that are adversely associated with VTE risk, and downstream reductions in other risks that are strongly associated with VTE risk, such as cancer.(25, 26) Therefore, this long-term lower risk of VTE associated with bariatric surgery seems meaningful and biologically plausible. It is also important to note that surgery is an established VTE risk factor(11) that is non-specific to bariatric surgery, with magnitudes of this association varying by surgery type.(11) Furthermore, surgery among patients with higher BMI has been associated with VTE risks of a greater magnitude than surgeries among patients with lower BMI.(27, 28)
Our sensitivity analyses that required both an ICD-9 code for VTE and an anticoagulation prescription fill within 7 days after the event markedly reduced the number of eligible events, from 2,927 VTE and 1,056 PE events at 5 years post-index date to 1,232 VTEs (42% of all VTEs) and 656 PEs (62% of all PEs). Also, our VTE incidence was notably higher than the general population's VTE incidence rate (1 to 2 events per 1,000 person-years(11)) when we used the ICD-9 code-based definition of VTE (5.1 and 6.4 VTE events/1000 person-years among non-surgical controls and cases, respectively). However, it was only slightly higher when using both ICD-9 codes and anticoagulation prescription fills to define VTE events (2.1 and 2.5 VTE events/1000 person-years among non-surgical controls and cases, respectively). The decision not to mandate an anticoagulation fill within 7 days post-event in primary analyses was based on the anticipation of prophylactic use of anticoagulants in the adult population with severe obesity undergoing the study. This approach might lead to misclassification of events and include ICD-9 codes not related to actual VTE events. However, our sensitivity analysis, though ruling out certain true VTEs among adults receiving prophylactic anticoagulation, indicates that the 'true' estimates might lie between the primary analyses (Table 2) and sensitivity analyses (Supplemental Digital Content Table 2). However, the interpretations of results stemming from either our primary or sensitivity analyses are consistent that surgery is strongly associated with a lower VTE risk long-term.
Strengths and Limitations
Strengths of our study include its population-based setting, large sample size, and long-term follow-up for VTE. An additional strength is our study’s high proportion and large number of SG operations, representing current bariatric operative trends for which survival/mortality data have been lacking. In addition, we used rigorous statistical methods to carefully match bariatric patients to their non-surgical controls and control for factors that might account for the differences in mortality between these groups.
This study also has several limitations including loss to follow-up of 29%-35% at 5-years. Also a limitation, we did not have access to data regarding indication for anticoagulation use at or before index. Although we excluded adults with history of VTE prior to index, making it unlikely that an anticoagulant prescription fill at or before index was for VTE treatment, it is possible that it was for prophylactic use or treatment of another condition such as atrial fibrillation. However, the association between bariatric surgery and short-term VTE risk is well-characterized(1-4), and the primary aim of our study was to evaluate long-term VTE risk. We do not believe that prophylactic anticoagulation around the time of surgery is likely to bias the long-term association between bariatric surgery and VTE risk. We also acknowledge that there are many interesting questions that one could address in future studies (such as bariatric surgery vs. surgeries of other types in relation to short-term VTE risk, or bariatric surgery in relation to long-term VTE risk above and beyond post-index BMI). By focusing our study’s scope to bariatric surgery vs. no surgery in relation to long-term VTE risk, we aimed to provide evidence of clinical relevance to patients with obesity considering bariatric surgery vs. no surgery who may be weighing the known short-term VTE risk associated with bariatric surgery with the potentially long-term decreases in BMI and the downstream health benefits. In addition, although bariatric surgery is increasingly shifting from solely inpatient to commonly outpatient, our study is likely to be most reflective of risks associated with inpatient bariatric surgery, given practices during the 2005 to 2015 time period. We believe that this shift from inpatient-only to a mix of inpatient and outpatient procedures is likely to reduce the short-term risk of VTE associated with bariatric surgery but is unlikely to change long-term VTE risk. The observational design of the study also precludes causal inference, and unmeasured confounding may have persisted despite model adjustment for many VTE risk factors. However, our sensitivity analyses using the E-value methodology indicate that each follow-up time point had an E-value greater than 2.0 for the upper bound of the CI; this suggests that these results could only be explained by an unmeasured confounder that was associated with both bariatric surgery and VTE risk by a risk ratio of 2.0. Currently there are no long-term randomized studies investigating the impact of bariatric operations on risk of VTE.
Conclusion
In summary, although bariatric surgery is associated with a greater VTE risk in the months immediately following surgery, it is strongly associated with a lower risk of VTE long-term, in the years following surgery. As patients and their clinicians weigh the risks and benefits of bariatric surgery, our finding provides additional evidence to consider not only the procedure’s risks in the short-term, but to balance these with potentially strong long-term benefits.(29)
Supplementary Material
Key Points.
Long-term VTE risk post-bariatric surgery is incompletely characterized.
Our large population-based matched cohort included adults with severe obesity.
Bariatric surgery was associated with a significantly lower risk of VTE long-term.
This supports balanced consideration of short-term risks and long-term benefits.
Funding/Support and Role of Funder/Sponsor:
The study was funded by NIH/NIDDK R01DK105960-01. Dr. Harrington was funded by a grant from the NHLBI (K01HL139997). Sponsors did not play a role in the collection, management, analysis or interpretation of data, the preparation, review, or approval of the manuscript, or the decision to submit the manuscript for publication.
Footnotes
Conflict of Interest Statement: Anita P. Courcoulas had a research grant from Allurion Inc. David E. Arterburn has grants from NIH and PCORI; a contract from Sharecare, Inc.; and received reimbursements for travel expenses from the American Society of Metabolic and Bariatric Surgery. Karen J. Coleman has funding for research from NIDDK, NHLBI, NIMH, and FDA and is paid a stipend for reviewing grants for NIH (outside of the submitted work). No authors have spouses, partners, or children that have financial relationships that may be relevant to the submitted work. Laura B. Harrington, Luke Benz, Sebastien Haneuse, Eric Johnson, Robert A. Li, Mary Kay Theis, Julie Cooper, Philip L. Chin, Gary G. Grinberg, Christopher R. Daigle, Julietta H. Chang, Scott S. Um, Panduranga R. Yenumula, and Jorge Zelada Getty have nothing to declare.
Ethical Approval Statement: For this type of study formal consent is not required.
Informed Consent Statement: Informed Consent does not apply.
Data Access, Responsibility, and Analysis:
Luke Benz, Sebastien Haneuse, and David Arterburn had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Data Sharing:
Study participants did not give written consent for their data to be shared publicly, and the data contain protected health information governed by the US Health Insurance Portability and Accountability Act (HIPAA). Because of this, we do not plan to publicly share data; however, authors will respond to reasonable requests, with permission of all health systems involved and a fully executed data use agreement. In the interest of reproducibility, code used for all analyses is available on GitHub at https://github.com/lbenz730/arterburn_vte/.
References
- 1.Finks JF, English WJ, Carlin AM, Krause KR, Share DA, Banerjee M, et al. Predicting risk for venous thromboembolism with bariatric surgery: results from the Michigan Bariatric Surgery Collaborative. Annals of surgery. 2012;255(6):1100–4. Epub 2012/05/09. doi: 10.1097/SLA.0b013e31825659d4. [DOI] [PubMed] [Google Scholar]
- 2.Froehling DA, Daniels PR, Mauck KF, Collazo-Clavell ML, Ashrani AA, Sarr MG, et al. Incidence of venous thromboembolism after bariatric surgery: a population-based cohort study. Obes Surg. 2013;23(11):1874–9. Epub 2013/09/12. doi: 10.1007/s11695-013-1073-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Steele KE, Schweitzer MA, Prokopowicz G, Shore AD, Eaton LC, Lidor AO, et al. The long-term risk of venous thromboembolism following bariatric surgery. Obes Surg. 2011;21(9):1371–6. Epub 2011/06/01. doi: 10.1007/s11695-011-0445-7. [DOI] [PubMed] [Google Scholar]
- 4.Stein PD, Matta F. Pulmonary embolism and deep venous thrombosis following bariatric surgery. Obes Surg. 2013;23(5):663–8. doi: 10.1007/s11695-012-0854-2. [DOI] [PubMed] [Google Scholar]
- 5.Winegar DA, Sherif B, Pate V, DeMaria EJ. Venous thromboembolism after bariatric surgery performed by Bariatric Surgery Center of Excellence Participants: analysis of the Bariatric Outcomes Longitudinal Database. Surg Obes Relat Dis. 2011;7(2):181–8. Epub 20101229. doi: 10.1016/j.soard.2010.12.008. [DOI] [PubMed] [Google Scholar]
- 6.Sweetland S, Green J, Liu B, Berrington de Gonzalez A, Canonico M, Reeves G, et al. Duration and magnitude of the postoperative risk of venous thromboembolism in middle aged women: prospective cohort study. Bmj. 2009;339:b4583. Epub 2009/12/05. doi: 10.1136/bmj.b4583. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Mechanick JI, Apovian C, Brethauer S, Timothy Garvey W, Joffe AM, Kim J, et al. Clinical Practice Guidelines for the Perioperative Nutrition, Metabolic, and Nonsurgical Support of Patients Undergoing Bariatric Procedures - 2019 Update: Cosponsored by American Association of Clinical Endocrinologists/American College of Endocrinology, The Obesity Society, American Society for Metabolic and Bariatric Surgery, Obesity Medicine Association, and American Society of Anesthesiologists. Obesity (Silver Spring). 2020;28(4):O1–o58. doi: 10.1002/oby.22719. [DOI] [PubMed] [Google Scholar]
- 8.Martin KA, Lee CR, Farrell TM, Moll S. Oral Anticoagulant Use After Bariatric Surgery: A Literature Review and Clinical Guidance. Am J Med. 2017;130(5):517–24. Epub 20170201. doi: 10.1016/j.amjmed.2016.12.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Martin KA, Beyer-Westendorf J, Davidson BL, Huisman MV, Sandset PM, Moll S. Use of direct oral anticoagulants in patients with obesity for treatment and prevention of venous thromboembolism: Updated communication from the ISTH SSC Subcommittee on Control of Anticoagulation. J Thromb Haemost. 2021;19(8):1874–82. Epub 20210714. doi: 10.1111/jth.15358. [DOI] [PubMed] [Google Scholar]
- 10.White RH. The epidemiology of venous thromboembolism. Circulation. 2003;107(23 Suppl 1):I4–8. Epub 2003/06/20. doi: 10.1161/01.cir.0000078468.11849.66. [DOI] [PubMed] [Google Scholar]
- 11.Lutsey PL, Zakai NA. Epidemiology and prevention of venous thromboembolism. Nature reviews Cardiology. 2023;20(4):248–62. Epub 20221018. doi: 10.1038/s41569-022-00787-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Stein PD, Beemath A, Olson RE. Obesity as a risk factor in venous thromboembolism. Am J Med. 2005;118(9):978–80. Epub 2005/09/17. doi: 10.1016/j.amjmed.2005.03.012. [DOI] [PubMed] [Google Scholar]
- 13.Mahmoodi BK, Cushman M, Anne Naess I, Allison MA, Jan Bos W, Braekkan SK, et al. Association of Traditional Cardiovascular Risk Factors With Venous Thromboembolism: An Individual Participant Data Meta-Analysis of Prospective Studies. Circulation. 2017;135(1):7–16. Epub 2016/11/11. doi: 10.1161/circulationaha.116.024507. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Moussa O, Ardissino M, Tang A, Lazzari L, Millar O, Ziprin P, et al. Long-term Impact of Bariatric Surgery on Venous Thromboembolic Risk: A Matched Cohort Study. Annals of surgery. 2019. Epub 2019/12/20. doi: 10.1097/sla.0000000000003750. [DOI] [PubMed] [Google Scholar]
- 15.Arterburn DE, Johnson E, Coleman KJ, Herrinton LJ, Courcoulas AP, Fisher D, et al. Weight Outcomes of Sleeve Gastrectomy and Gastric Bypass Compared to Nonsurgical Treatment. Annals of surgery. 2021;274(6):e1269–e76. doi: 10.1097/sla.0000000000003826. [DOI] [PubMed] [Google Scholar]
- 16.Courcoulas AP, Johnson E, Arterburn DE, Haneuse S, Herrinton LJ, Fisher DP, et al. Reduction in Long-term Mortality After Sleeve Gastrectomy and Gastric Bypass Compared to Nonsurgical Patients With Severe Obesity. Annals of surgery. 2023;277(3):442–8. Epub 20210813. doi: 10.1097/sla.0000000000005155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Center for Disease Control and Prevention. Defining Adult Overweight & Obesity 2023. [cited 2023 October 29]. Available from: https://www.cdc.gov/obesity/basics/adult-defining.html#:~:text=Obesity%20is%20frequently%20subdivided%20into,BMI%20of%2040%20or%20higher.
- 18.Fisher DP, Johnson E, Haneuse S, Arterburn D, Coleman KJ, O'Connor PJ, et al. Association Between Bariatric Surgery and Macrovascular Disease Outcomes in Patients With Type 2 Diabetes and Severe Obesity. Jama. 2018;320(15):1570–82. doi: 10.1001/jama.2018.14619. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.O'Brien R, Johnson E, Haneuse S, Coleman KJ, O'Connor PJ, Fisher DP, et al. Microvascular Outcomes in Patients With Diabetes After Bariatric Surgery Versus Usual Care: A Matched Cohort Study. Ann Intern Med. 2018;169(5):300–10. Epub 20180807. doi: 10.7326/m17-2383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Li RA, Liu L, Arterburn D, Coleman KJ, Courcoulas AP, Fisher D, et al. Five-year Longitudinal Cohort Study of Reinterventions After Sleeve Gastrectomy and Roux-en-Y Gastric Bypass. Annals of surgery. 2021;273(4):758–65. doi: 10.1097/sla.0000000000003401. [DOI] [PubMed] [Google Scholar]
- 21.Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011;64(7):749–59. Epub 20110105. doi: 10.1016/j.jclinepi.2010.10.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Mahalanobis PC. On the generalized distance in statistics. Proc Natl Acad Sci India 1936;2:49–55. [Google Scholar]
- 23.VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value. Ann Intern Med. 2017;167(4):268–74. Epub 20170711. doi: 10.7326/m16-2607. [DOI] [PubMed] [Google Scholar]
- 24.Haneuse S, VanderWeele TJ, Arterburn D. Using the E-Value to Assess the Potential Effect of Unmeasured Confounding in Observational Studies. Jama. 2019;321(6):602–3. doi: 10.1001/jama.2018.21554. [DOI] [PubMed] [Google Scholar]
- 25.Folsom AR, Cushman M. Exploring Opportunities for Primary Prevention of Unprovoked Venous Thromboembolism: Ready for Prime Time? J Am Heart Assoc. 2020;9(23):e019395. Epub 20201116. doi: 10.1161/jaha.120.019395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Schauer DP, Feigelson HS, Koebnick C, Caan B, Weinmann S, Leonard AC, et al. Bariatric Surgery and the Risk of Cancer in a Large Multisite Cohort. Annals of surgery. 2019;269(1):95–101. doi: 10.1097/sla.0000000000002525. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Pahlkotter MK, Mohidul S, Moen MR, Digney BW, Holmes S, Muertos K, et al. BMI and VTE Risk in Emergency General Surgery, Does Size Matter? : An ACS-NSQIP Database Analysis. Am Surg. 2020;86(12):1660–5. Epub 20200805. doi: 10.1177/0003134820940272. [DOI] [PubMed] [Google Scholar]
- 28.Zhang ZH, Shen B, Yang J, Zhou ZK, Kang PD, Pei FX. Risk factors for venous thromboembolism of total hip arthroplasty and total knee arthroplasty: a systematic review of evidences in ten years. BMC Musculoskelet Disord. 2015;16:24. Epub 20150210. doi: 10.1186/s12891-015-0470-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Courcoulas A, Coley RY, Clark JM, McBride CL, Cirelli E, McTigue K, et al. Interventions and Operations 5 Years After Bariatric Surgery in a Cohort From the US National Patient-Centered Clinical Research Network Bariatric Study. JAMA Surg. 2020;155(3):194–204. doi: 10.1001/jamasurg.2019.5470. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Luke Benz, Sebastien Haneuse, and David Arterburn had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study participants did not give written consent for their data to be shared publicly, and the data contain protected health information governed by the US Health Insurance Portability and Accountability Act (HIPAA). Because of this, we do not plan to publicly share data; however, authors will respond to reasonable requests, with permission of all health systems involved and a fully executed data use agreement. In the interest of reproducibility, code used for all analyses is available on GitHub at https://github.com/lbenz730/arterburn_vte/.