Use of post-discharge heparin prophylaxis and the risk of venous thromboembolism and bleeding following bariatric surgery

Abstract

Introduction

Venous thromboembolism (VTE) is a significant cause of morbidity and mortality after bariatric surgery. Roughly 80% of VTEs occur post-discharge. The frequency of post-discharge heparin (PDH) prophylaxis use is unknown, and evidence about benefits and risks is limited. We aimed to determine the rate of use of PDH prophylaxis and evaluate its relationship with VTE and bleeding events.

Methods

Using the Truven Health MarketScan® database, we performed a retrospective cohort study (2007–2015) of adult patients who underwent sleeve gastrectomy or gastric bypass. We determined PDH prophylaxis from outpatient pharmacy claims, and post-discharge 90-day VTE and bleeding events from outpatient and inpatient claims. We used propensity score-adjusted regression models to mitigate confounding bias.

Results

Among 43,493 patients (median age 45 years; 78% women; 77% laparoscopic gastric bypass, 17% laparoscopic sleeve gastrectomy, 6% open gastric bypass), 6% received PDH prophylaxis. Overall, 224 patients (0.52%) experienced VTEs, and 806 patients (1.85%) experienced bleeding. The unadjusted VTE rate did not differ between patients who did and did not receive PDH prophylaxis (0.39% vs. 0.52%, respectively; p = 0.347). The unadjusted bleeding rate was higher for the PDH prophylaxis group (2.74% vs. 1.80%, p < 0.001). In our adjusted analysis, a 23% lower risk of VTE in the PDH prophylaxis group was not statistically significant (odds ratio [OR] 0.77, 95% confidence interval [CI] 0.41 to 1.46), whereas the 47% higher risk of bleeding was statistically significant (OR 1.47, 95% CI 1.14 to 1.88).

Conclusions

PDH prophylaxis after bariatric surgery is uncommon. In our analysis, use was not associated with a lower VTE risk but was associated with a higher bleeding risk.

Over one-third of adults in the United States (US) are obese [1]. Bariatric surgery is the most effective and durable treatment available for obesity and its related comorbidities [2, 3]. Although post-operative mortality for bariatric surgeries has decreased considerably over the past decade [4-6], venous thromboembolism (VTE) remains a primary contributor [7].

Reported incidence rates of VTEs vary from 0.4 to 6.4% [4, 7-11]. Although rare, these events are fatal in roughly one out of ten patients [7, 10-13]. Several studies demonstrate that approximately 80% of VTEs in bariatric surgery patients occur after discharge [8, 14-16], leading to the theory that post-discharge heparin prophylaxis (henceforth referred to as PDH prophylaxis) may reduce the risk of these events.

The American Society of Metabolic and Bariatric Surgery (ASMBS) advises “consideration” of PDH prophylaxis for patients deemed high risk for VTE, but they cite a lack of high-level evidence to support uniform recommendations [17]. Survey data from a small sample of bariatric surgeons indicate increasing support for the use of PDH prophylaxis [18]; however, little is known about nationwide patterns of use. Additionally, the risk of major bleeding in patients who receive PDH prophylaxis is poorly understood. Two recent cohort studies found no difference in the risk of bleeding with PDH prophylaxis, but these studies were limited by small sample sizes [19-21].

The primary objective of this study was to characterize the frequency of PDH prophylaxis use and its relationship, if any, with subsequent VTE and bleeding events. We hypothesized that a minority of bariatric surgery patients receive PDH prophylaxis and that it is associated with a lower risk of VTE but a higher risk of bleeding.

Materials and methods

This study adheres to reporting standards described by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement [22]. The University of Washington Institutional Review Board waived the need for consent and approved this investigation (IRB exemption number 00007898).

Study design, population, and data source

We performed a retrospective cohort study of patients who underwent bariatric surgery by using commercial insurance claims data in the Truven Health MarketScan® Commercial Claims and Encounters databases. These databases include inpatient, outpatient, and pharmacy claims. The initial cohort was defined as enrollees with a Current Procedural Terminology (CPT) code for an open or laparoscopic gastric bypass or a laparoscopic sleeve gastrectomy occurring between July 1, 2007, and June 30, 2015.

We excluded patients with pharmacy claims for anticoagulation medication (defined as warfarin, unfractionated and low molecular weight heparin products, direct thrombin inhibitors, and direct Xa inhibitors) or a diagnosis of VTE in the 90 days prior to surgery, a diagnosis of VTE during index hospitalization, or an index stay greater than 30 days. We included patients between 18 to 64 years of age (we did not have access to Medicare claims), as well as those with 90 days of continuous enrollment prior to surgery and following index discharge (Fig. 1).

Fig. 1.

Flow diagram of study population

Variable definitions

We determined PDH prophylaxis use from outpatient pharmacy claims for heparin, including both low molecular weight and unfractionated products. We did not include antiplatelet or novel anticoagulant medications. Patients with more than a 35 day supply of outpatient heparin were excluded, as treatment beyond five weeks is not consistent with any known prophylaxis protocol [23, 24].

We defined post-discharge VTE as follows: any claim with an International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9) diagnostic code for VTE and/or both a CPT code for VTE diagnostic imaging and a subsequent outpatient pharmacy claim for anticoagulation occurring up to 90 days after discharge. For every patient with a VTE event, we reviewed claims manually to ensure that only claims for PDH that preceded the diagnosis of VTE were classified as prophylaxis. We defined post-discharge bleeding events as follows: either an ICD-9 diagnostic code for hemorrhage and/or a CPT code for transfusion or endoscopy with control of bleeding occurring up to 90 days after index discharge.

We selected the following a priori covariates to include in bivariate and multivariate analyses: age, sex, index admission length of stay, insurance plan type (high-deductible, low-deductible, or managed care/capitated), operation type (laparoscopic sleeve, laparoscopic or open gastric bypass), geographic region, and calendar year. Comorbidities were identified from ICD-9 diagnostic codes attached to outpatient and inpatient claims in the 90 days before surgery. The Quan modification [25] of the Charlson Comorbidity Index (CCI) was determined and included as a covariate. (See Online Appendix 1 for a complete list of ICD-9 and CPT codes used in this study).

Statistical analysis

Medians were used to summarize the non-normally distributed continuous variables. Frequencies were used to summarize categorical variables (sex, insurance plan type, operation type, region, CCI, year). Baseline data were stratified by receipt of PDH prophylaxis, and tests of differences between prophylaxis groups were performed. Bivariate tests of association between categorical variables and outcomes were compared using chi-squared tests, whereas continuous variables and outcomes were compared using the Mann–Whitney U test.

We performed an unadjusted comparison of the risk for VTE or bleeding events by PDH prophylaxis status using Chi-squared tests. We then performed an adjusted logistic regression analysis incorporating a propensity score to mitigate selection bias [26-28]. The propensity score was calculated using age, sex, index admission length of stay, insurance plan type, operation type, region, year, and CCI. We chose to use a propensity score-adjusted logistic regression over direct covariate adjustment using logistic regression due to the large number of degrees of freedom in our model relative to the number of rare events.

We conducted a case complete analysis. Of the 133,615 patients in the cohort prior to the application of exclusion criteria, 2494 (1.9%) were missing covariate data (specifically, insurance plan type and/or geographic region). No patients were missing exposure or outcome data. The rate of both outcomes did not vary across individuals with and without missing covariate data (0.44% vs. 0.52% with VTE events, respectively, p = 0.714; 2.20% vs. 1.85% with bleeding events, p = 0.128); however, the rate of PDH prophylaxis was lower among individuals with missing covariate data compared to those with complete data (4.9% vs. 5.9%, p = 0.013).

We defined statistical significance as a p-value below 0.05. All analyses were performed using STATA statistical software, version 14.2 IC (Stata Corp LP, College Station, TX).

Results

Our cohort consisted of 43,493 patients with a median age of 45 years, 22% of whom were men. Most patients underwent a gastric bypass (77% laparoscopic and 6% open), and 17% underwent a laparoscopic sleeve gastrectomy (Table 1).

Table 1.

Characteristics of study population, overall and by post-discharge prophylaxis status

	Overall (n = 43,493)	Received post-hospital discharge prophylaxis		p-value
		No (n = 40,906)	Yes (n = 2587)
Female, n (%)	33,788 (77.7%)	31,885 (78.0%)	1903 (73.6%)	< 0.001
Median age, years (IQR)	45 (38–53)	45 (38–53)	46 (37–54)	0.105
Insurance plan type, n (%)				< 0.001
Managed care	7646 (17.6%)	7359 (18.0%)	287 (11.1%)
Standard/lower-deductible	32,225 (74.1%)	30,121 (73.6%)	2,104 (81.3%)
High-deductible	3622 (8.3%)	3426 (8.4%)	196 (7.6%)
Operation type, n (%)				< 0.001
Laparoscopic gastric bypass	33,597 (77.3%)	31,511 (77.0%)	2086 (80.6%)
Laparoscopic sleeve	7381 (17.0%)	7025 (17.2%)	356 (13.8%)
Open gastric bypass	2515 (5.8%)	2370 (5.8%)	145 (5.6%)
Median index length of stay, days (IQR)	2 (1–2)	2 (1–2)	2 (1–2)	0.001
Region (%)				< 0.001
Northeast	7245 (16.7%)	6905 (16.9%)	340 (13.1%)
North Central	9,652 (22.2%)	8,697 (21.3%)	955 (36.9%)
South	17,303 (39.8%)	16,185 (39.6%)	1118 (43.2%)
West	9293 (21.4%)	9119 (22.3%)	174 (6.7%)
Comorbidity index, n (%)				0.016
0	27,158 (62.4%)	25,578 (62.5%)	1580 (61.1%)
1	12,954 (29.8%)	12,189 (29.8%)	765 (29.6%)
2	2,442 (5.6%)	2272 (5.6%)	170 (6.6%)
≥ 3	939 (2.2%)	867 (2.1%)	72 (2.8%)
Year, n (%)				< 0.001
2007* to 2009	7.793 (17.9%)	7,368 (18.0%)	425 (16.4%)
2010 to 2012	23,986 (55.2%)	22,490 (55.0%)	1496 (57.8%)
2012 to 2015*	11,714 (26.9%)	11,048 (27.0%)	666 (25.7%)

IQR inter-quartile range

^*Only 6 months of data included from each of these years

Overall, 6% of patients received PDH prophylaxis. All covariates, except age, were significantly different between patients who did and did not receive PDH prophylaxis. The largest differences in the rates of PDH prophylaxis use were between the West versus North Central geographic regions (1.9% versus 9.9%, respectively), and between patients with managed care versus low-deductible insurance plans (3.7% versus 6.5%, respectively). Absolute differences in the rate of PDH prophylaxis were less than 2% across sex, operation type, Charlson score, and length of stay.

A total of 224 (0.52%) patients experienced a post-discharge VTE event within 90 days of index discharge. The VTE rate was lower in patients who received PDH prophylaxis compared with those who did not, but this difference was not statistically significant (0.39% vs. 0.52%, p = 0.347; see Table 2). Of note, no patient experienced both a bleeding event and a VTE event.

Table 2.

Unadjusted association of PDH prophylaxis to 90-day VTE events, with breakdown by defining criteria

	Overall (N = 43,493)	Received post-hospital discharge prophylaxis		p-value
		No (N = 40,906)	Yes (N = 2587)
Post-discharge VTE events, n (%)	224 (0.52%)	214 (0.52%)	10 (0.39%)	0.347
VTE event ascertained based on:
A. LE duplex and/or CTA, followed by AC, n	112 (0.26%)	108 (0.26%)	4. (0.15%)
B. Diagnosis code for DVT and/or PE, n	103 (0.24%)	99 (0.24%)	4 (0.15%)
C. Both A. and B., n	9 (0.02%)	7 (0.02%)	2 (0.08%)

LE lower extremity, CTA computed tomography angiography, AC anticoagulation, DVT deep venous thrombosis, PE pulmonary embolism

A total of 806 (1.85%) patients experienced a post-discharge bleeding event within 90 days. The unadjusted bleeding event rate was higher among patients who received PDH prophylaxis compared with those who did not (2.74% vs. 1.80%, p < 0.001; see Table 3). Of note, all but one patient who received a transfusion also underwent an endoscopy with control of bleeding.

Table 3.

Unadjusted association of PDH prophylaxis to 90-day bleeding events, with breakdown by defining criteria

	Overall (N = 43,493)	Received post-hospital discharge prophylaxis		p-value
		No (N = 40,906)	Yes (N = 2587)
Post-discharge bleeding events, n (% total cohort)	806 (1.85%)	735 (1.80%)	71 (2.74%)	0.001
Bleeding event ascertained based on:
A. Procedure code for transfusion and/or endoscopy with control of bleeding, n	465 (0.07%)	429 (1.05%)	36 (1.39%)
B. Diagnosis code for hemorrhage, n	111 (0.26%)	93 (0.23%)
C. Both A. and B., n	230 (0.53%)	213 (0.52%)	17 (0.66%)

In our propensity score-adjusted regression model, PDH prophylaxis was associated with a 23% lower adjusted odds of a VTE, but this finding was not statistically significant (odds ratio [OR] 0.77, 95% confidence interval [CI] 0.41 to 1.46). There was a statistically significant 47% higher adjusted odds of bleeding for patients who received PDH prophylaxis (OR 1.47, 95% CI 1.14 to 1.88; see Table 4).

Table 4.

Propensity score-adjusted regression comparing patients who received post-discharge heparin prophylaxis with those who did not Propensity score model included age, sex, index admission length of stay, insurance plan type, operation type, geographic region, calendar year, and Charlson Comorbidity Index

	Odds ratio [95% confidence interval]
VTE risk (PDH versus no PDH)	0.77 [0.41–1.46]
Bleeding risk (PDH versus no PDH)	1.47 [1.14–1.88]

Discussion

Little is known about patterns of use and associated risks and benefits of PDH prophylaxis in the U.S. Using a large sample of adults and their dependents with commercial insurance, we found that PDH prophylaxis was infrequently utilized after bariatric surgery. Furthermore, PDH prophylaxis was not associated with a statistically significant decrease in risk of VTE, but it was associated with a statistically significant increase in bleeding risk.

Although we observed our hypothesis that only a minority of patients would receive PDH prophylaxis, we did not expect the use of PDH prophylaxis to be so rare. A small sample of U.S. bariatric surgeons self-report using PDH prophylaxis ten times more often than we observed [18], suggesting a discrepancy in reported versus actual behaviors and/or a non-representative sample of surgeons. Alternatively, patients may choose to not fill prescriptions, resulting in an apparent underuse of PDH prophylaxis when using claims to measure utilization. However, this explanation is unlikely to account for a ten-fold discrepancy in surgeon and patient behavior. Another explanation for low PDH prophylaxis utilization in the US may reflect the ASMBS position that providers should “consider” its use, rather than making an explicit recommendation. In contrast, in France and the United Kingdom, professional society guidelines recommend a minimum of one week of heparin prophylaxis following bariatric surgery [29-31]. A recent French study using national insurance claims data [14] reported that 75% of bariatric surgery patients received PDH prophylaxis. The lack of a strong guideline recommendation in the US may explain the rarity of PDH prophylaxis use after bariatric surgery.

There are several possible explanations for why we did not find an association between PDH prophylaxis and a lower risk of VTE. One explanation is that there is no true effect of PDH prophylaxis on VTE risk. Alternatively, the association may be small, but we failed to observe it because of limited statistical power. Statistical power is not only a function of the overall population size but also the event rate and frequency of exposure groups (e.g., prophylaxis versus none). In our case, we had an unexpectedly low rate of VTE prophylaxis and a rare event, which may have contributed to insufficient statistical power despite a large sample size. A French study with a similar data source and design as ours showed a similar magnitude of reduction in the odds of VTE (approximately 20%) with PDH prophylaxis use and a statistically significant result; however, their study likely had more power to detect this association given that VTE prophylaxis was more common [14]. A third explanation is that we underestimated the magnitude of the relationship between PDH prophylaxis and VTEs because of unmeasured confounding. For instance, if patients at a higher risk of VTE were more likely to receive PDH prophylaxis, then our inability to adjust for baseline VTE risk using claims data might bias our observations toward the null or even a paradoxically higher rate of VTE in the PDH prophylaxis group. Examples of unmeasured variables that may have biased our results toward a non-significant association between PDH prophylaxis and VTE risk include higher body mass index, inherited or acquired syndromes of hypercoagulability, lower functional status and mobility, and more severe comorbid conditions. Lastly, patients may ill the prescription of a self-administered heparin injection, but not take it. Bias resulting from unmeasured non-compliance with medication could have biased our findings toward the null. Even if we assume a true benefit from PDH prophylaxis, its potential benefit needs to be considered along with its risks.

Unlike prior studies, we found a statistically significant increase in bleeding risk associated with PDH prophylaxis. Two small studies concluded that PDH prophylaxis was safe, based on the absence of an association with bleeding [19, 32]. In the previously mentioned French study, Thereaux and colleagues reported a lower rate of re-hospitalization for bleeding among the group of patients who received PDH prophylaxis, a finding that is inconsistent with the pharmacologic mechanism of heparin products and lacks an alternative biologic explanation. Importantly, our findings were again likely biased by unmeasured confounding, but in a manner that may have resulted in underestimation of the risk of bleeding in the PDH group. Specifically, patients at a higher risk of bleeding may be less likely to receive PDH prophylaxis, biasing our results because patients in the non-prophylaxis group have an unmeasured higher risk of bleeding (a conservative bias). Also, non-compliance with prescribed medication, to the extent it occurs, would lead to an underestimation of the bleeding risk due to misclassification of individuals as experiencing PDH prophylaxis exposure, when in fact they did not take the prescribed medication (again a conservative bias). In light of this new information, guidelines should consider the potential benefits of PDH prophylaxis along with the risk of bleeding.

If PDH prophylaxis after bariatric surgery truly reduces the risk of VTE, it is also important to consider this benefit in absolute terms. If, as we found in our study, there is a roughly 20–25% reduction in the odds of VTE with PDH prophylaxis, and the baseline rate of VTEs is assumed to be 1 in 200 bariatric surgery patients, we would need to treat 800–1000 patients with PDH prophylaxis to prevent one VTE. Assuming a 10% case-fatality rate among individuals with VTE, the number needed to treat to avoid one death is between 8000 and 10,000. With respect to bleeding risk, however, we found a roughly 50% increased odds with the use of PDH prophylaxis, such that approximately 1 in 100 patients could be harmed by its use. Although the clinical severity of the bleeding risk cannot be adequately described using claims data, the finding of its existence is important to weigh against the potential benefit of PDH prophylaxis use. An increased risk of bleeding associated with PDH prophylaxis is a novel finding and likely the most important observation from this study.

In addition to the limitations discussed previously, our findings may only be generalizable to privately-insured adult workers and their dependents. There is no reason to believe, however, that the relationship between PDH prophylaxis and VTE or bleeding would be different for publicly-insured bariatric surgery patients. Lastly, we did not measure dose or duration of heparin after discharge, and these could be important factors affecting the rate of VTE and bleeding.

In light of the uncertain benefit and new evidence of increased bleeding risks, we cautiously suggest that professional societies recommend against routine use of heparin prophylaxis after hospital discharge until a higher level of evidence is available to support the conclusion that the benefits outweigh risks. It is possible that the possible benefit of PDH prophylaxis may be greater in subgroups of bariatric surgery patients at higher risk for VTEs [16, 33]. Future research efforts should examine the outcomes of PDH prophylaxis for these higher-risk patients, defined using validated individualized risk prediction models, such as the Caprini Score [34, 35]. The benefits of post-discharge heparin prophylaxis for bariatric surgery patients should be weighed cautiously against its bleeding risks.