Scientific Papers

Venous thromboembolic disease in admitted blunt trauma patients: what matters? | Thrombosis Journal


The aim of this study was to further define blunt trauma patients at risk for VTE. Our findings augment the trauma literature in noting that: (1) While VTE developed throughout the range of ISS, the lowest ISS score group (ISS 4–9) accounted for 27.7% of VTE patients; 2) Heparin and enoxaparin were associated with reduced VTE development; and 3) Approximately 7.3% of VTE occurred within 1 day of admission.

Increased ISS scores correlated with increased risk of VTE events in agreement with existing literature, however, approximately half of our VTE patients had an ISS < = 16, including 27.7% of VTE patients with an ISS < = 9. Of note, 81.6% of all patients in this study had ISS < = 16. Current literature posits that severe injury is a risk factor for VTE, with VTE incidence increasing with injury severity. A German trauma registry study reported a 1.2% incidence rate in patients with ISS < 25; 2.1% in ISS 25–34; 2.8% in ISS 35–49; and 4.1% in ISS 50–75 [15]. The study excluded patients with ISS < 9 [15]. Another study reported a VTE incidence of 1.0% in patients with ISS < 16, 4.5% in patients with ISS > = 16; and 7.7% in patients with ISS > = 25 [16]. A retrospective analysis of trauma patients with lower extremity fractures found that major trauma patients (ISS > 15, mean ISS = 26) had an approximately six-fold increased rate (6.8% vs. 1.17%) of VTE as compared with minor trauma patients (ISS < 15, mean ISS = 4) [17]. The most common serious injury (i.e. AIS > = 3) in VTE patients with ISS < = 9 in our study was in the extremities (63.3%). Research supports the importance of extremity injuries in VTE events. In a National Trauma Database databank study, extremity AIS > = 3 had an OR of 1.96 for VTE events in a multivariable model [18]. Another study found that any extremity fracture had 2.4 OR for VTE in multivariable analyses [19].

Thus, less severely injured patients, especially those with extremity injuries, are also at elevated risk for VTE events, and any surveillance and prophylactic strategies should include these patients or will risk omitting a large proportion of the target population. This finding underlines the importance of early mobility for this group, which is likely more able to comply with increased mobility expectations. Because these patients might traditionally be considered low-risk for VTE among the broader blunt trauma population, this finding also raises challenging questions about surveillance strategies and surveillance bias through increased venous duplex ultrasound use [7, 13].

This study was not designed as a non-inferiority study. Hence, we can only conclude that both heparin and enoxaparin reduced the adjusted odds ratio for VTE development. Previous research has generally indicated that heparin chemoprophylaxis is inferior to enoxaparin chemoprophylaxis in general trauma patients [20,21,22]. The orthopedic trauma literature, while acknowledging practice pattern variation, also indicated limited heparin efficacy in VTE prevention [23, 24]. In contrast, a recent randomized trial in trauma patients determined that thrice daily unfractionated heparin may be non-inferior to low molecular weight heparin in VTE prevention [25]. Anti-Xa level based enoxaparin dosing has been suggested to improve enoxaparin efficacy [2, 6, 26, 27]. In contrast, a randomized study of trauma patients found no benefit to anti-Xa level-based enoxaparin dosing [28]. During the study time period, our institution infrequently used anti-Xa level based enoxaparin dosing. Rather than chemoprophylaxis agent alone, a multi-center study indicated that it may be the sum of all prophylactic measures that determine VTE incidence, where only an expectation and/or culture of mobility within institutions was associated with reduction in VTE incidence [11]. To this end, physical and occupational therapy are routinely ordered for trauma patients at our institution. Finally, the baseline VTE rate may influence the efficacy of chemoprophylaxis, where VTE incidence is dependent on duplex screening strategies, type of injury, and institutional VTE prophylaxis protocols, among others [2, 4, 8, 11, 13, 29, 30]. Our overall 1.6% VTE incidence rate was on the lower end of the broad range described in trauma literature [2,3,4,5,6,7,8,9,10].

It should be noted that our institution’s chemoprophylaxis dosing regimen during the study period of 2012–2019 reflects practices prior to those recently published by WTA and AAST that generally recommend 40 mg enoxaparin every 12 h, with considerations for age, creatinine clearance, and BMI, amongst others [2, 6]. Our patients received lower doses of enoxaparin, with a caveat that 52.3% of our patients were age 65 years or older and patients receiving enoxaparin had a lower ISS than those receiving heparin.

Not surprisingly, a substantial number of patients in our study who developed VTE did not receive chemoprophylaxis prior to the index event (n = 74, 41.8%). As expected, these 74 patients were severely injured (median ISS 18.5), with the head/neck being the most frequent severely injured region (n = 48) and abdomen being the least frequent severely injured region (n = 4). In contrast, in the 2,657 patients without VTE who did not receive chemoprophylaxis, the median ISS was 9 [IQR 5–16)], p < 0.0001.

Because of the rarity of the VTE events (< 2% of the admitted blunt trauma patients), we used the Youden index to increase sensitivity and balanced accuracy for VTE classification in logistic regression and GBM models. We also created a fully balanced synthetic dataset to address the class imbalance. The synthetic dataset had limited effects on the performance of logistic regression models where classification thresholds were based on Youden index. The synthetic dataset had substantial benefits for random forest classification as measured by the F1 score. Albeit balanced accuracy could be improved by the methods used to address the imbalanced dataset, the relatively low F1 scores of all models indicate difficulty in increasing both sensitivity and positive predictive value. In terms of ease of interpretation, model simplicity and general performance, the multivariable logistic regression analysis of risk factors optimized by Youden index threshold for classification is favored.

In examining additional risk factors for VTE development in logistic regression we found older age; male sex; extremity AIS; mvc/mcc mechanism of injury; surgery; number of complications; and ICU stay to be important. The gender findings contrast to a study that demonstrated no difference in VTE rates in trauma patients grouped by sex [31]. Possible reasons for these discrepant findings include differing ages and injury severities between studies. However, our results aligned with another study that noted male sex as a VTE risk factor [32]. The CLOTT study identified age, major head injury, pelvic fractures, femoral vein lines, and major venous injury as DVT risk factors [5]. The failure of logistic regression analyses to identify head/neck injury as a major risk factor and its comparatively lesser importance in other models may in part be because of some collinearity between severe head/neck injury and lack of chemoprophylaxis. Central lines were not significant predictors in logistic regression but did have high influence in random forest and GBM models with synthetic data.

Several of the above as well as other risk factors have been bundled into risk assessment scoring systems for VTE for the trauma population, such as the Greenfield Risk Assessment Profile and the Trauma Embolic Scoring System (TESS) [1, 2, 12, 33]. TESS includes 5 risk factors (obesity, ventilator duration > 3 days, lower-extremity trauma, age, and ISS); our data affirms the importance of several of these findings (age, extremity AIS, mechanical ventilation, among others) [1, 2]. AAST guidelines, however, concluded that none of these scoring systems are necessary as most injured patients requiring hospitalization for over 24 h are at increased VTE risk and therefore chemical prophylaxis should be initiated promptly [2]. The high adjusted odds ratio for surgical patients may be partly explained as an injury requiring surgery and subsequent, especially multiple, surgeries requiring anesthesia could pose a double hit on VTE pathogenesis [15]. The elevated odds ratio for ICU stay is likely related to more severe injuries, need for mechanical ventilation, age, among others. Both of these factors had high importance levels in multiple models. The association of VTE with number of non-VTE complications deserves further evaluation. Albeit rather infrequent, significantly more VTE patients had a history of disseminated cancer and/or current receipt of chemotherapy in this study.

Finally, 7.3% (n = 13) of index VTE events were present within 1 day of admission, which has important implications in terms of quality metrics, as these would likely not be preventable, especially if present on day 0. This concern was raised in the CLOTT study, which demonstrated that 1/4 of pulmonary thromboses were identified on index CT scan [5]. An older study also found that 4/63 PE events occurred within 1 day of injury [34]. Further, a single-center retrospective study noted that of 142 trauma patients with DVT, 55 were noted on duplex scans performed within 48 h of admission [35]. Hence, a corollary question is the duration of chemoprophylaxis necessary to achieve a steady state concentration to not just optimize anti-Xa levels, but also to effect VTE event reduction, as therapeutic anti-Xa levels may not necessarily reduce VTE events [28]?

Limitations

This study has several important limitations. It is a single suburban trauma center study with a limited number of VTE patients, which may limit its generalizability to other institutions/situations. Classification of chemoprophylaxis as present among patients who received at least one dose on the day prior to VTE diagnosis may oversimplify variability and obscure delays in prophylaxis dosing or missed doses. This may affect comparisons between unfractionated heparin and enoxaparin groups. This study was not powered to detect differences between the groups.



Source link