Clinical Factors Associated With Catheter-Related VTE in Patients Undergoing Hematopoietic Cell Transplantation: A Multi-Center Study

Abstract

Venous thromboembolism (VTE) is a clinically significant complication that occurs in patients undergoing allogeneic hematopoietic cell transplantation (HCT). Due to the prolonged need for indwelling central venous catheters (CVCs), the incidence of catheter-related deep venous thrombosis (CR-DVT) appears higher at 3%–4% than that of pulmonary embolism (PE) or lower extremity deep vein thrombosis (LE-DVT) at 1%–4% in the first year post-transplant [1-3]. Previous studies have limited power to assess risk factors associated with CR-DVT due to single-center design with uniform institutional practice. In the present study, we analyzed the incidence and risk factors associated with isolated CR-DVT from two large allogenic HCT centers.

We performed a retrospective cohort study for patients undergoing first allogeneic HCT at MD Anderson Cancer Center (MDACC) 2016–2020 and Fred Hutchinson Cancer Center (FHCC) 2006–2019. Baseline patient characteristics included demographics, body mass index (BMI), pre-transplant disease, donor match, conditioning regimen, Karnofsky Performance Status (KPS), prior autologous HCT, common laboratory values, and CVC type. We defined CVC as a catheter that extended into the superior vena cava, with further sub-classification into peripherally inserted central catheter (PICC), non-tunneled CVC (placed by a vascular access team), and tunneled CVC (placed by Interventional Radiology). Time-varying variables included the status and duration of post-transplant hospitalizations and development of acute graft-versus-host disease (GVHD).

The primary outcome of isolated CR-DVT was defined as isolated, symptomatic or incidentally found, acute upper extremity DVT associated with an ipsilateral CVC that was documented by either venogram, contrasted CT scan, or compression ultrasound. CR-DVT events concurrent with PE or LE-DVT were classified as the latter. Diagnostic imaging was performed based on clinical symptoms. There was no formal thrombosis risk stratification, thromboprophylaxis, surveillance, or screening program. The electronic medical records (EMR) of eligible patients were examined using ICD 9 or 10 codes to identify possible VTE events, while radiology reports were probed with a natural language processing (NLP) algorithm for the same purpose. All patients with possible new VTE events, including those with recurrent events, were individually confirmed on chart review.

All patients were assessed from the time of transplant cell infusion until first thrombosis, death, loss to follow-up, or 366 days post-transplant. Isolated CR-DVT incidence was assessed by a cumulative incidence competing risk model, with death as a competing cause. Unadjusted and multivariable Cox proportional hazards models were used to measure the association between patient- and transplant-specific factors and time to CRT using a shared frailty model to account for clustering of patients from each site. Acute GVHD and inpatient hospitalization status were treated as time-varying covariates. To differentiate VTE diagnosed prior to admission versus hospital-acquired, we characterized the time-varying exposure as inpatient after 48 h of admission.

A total of 4250 patients (2879 FHCC and 1371 MDACC) were included in the analysis. Baseline characteristics are shown in Table S1, and site-specific characteristics are shown in Table S2. The median age of the overall cohort was 54.8 years, 41.8% were female, and 80.7% were White. The indication for HCT was 66.7% myeloid leukemia, 16.0% lymphoid leukemia, 10.2% lymphoma, 4.1% myeloma, and 3.1% other. In the combined cohort, 73.6% had matched donors, 64.1% received myeloablative conditioning, and 11.4% had prior autologous HCT. Most patients (78%) had KPS ≥ 80. The median pre-conditioning white blood cell count (WBC), hemoglobin, and platelet count of the overall cohort were 3.5 × 10⁹/L, 10.6 g/dL, and 114 × 10⁹/L, respectively, while the median pre-conditioning creatinine, total bilirubin, and LDH were 0.9 mg/dL, 0.5 mg/dL, and 191 U/L, respectively. The CVC type for venous access included 68.5% tunneled CVC, 26.6% non-tunneled CVC, and 4.9% PICC. Notably, all patients from FHCC had tunneled CVC per institutional policy. Of the 12.3% of patients with prior VTE before HCT, 51.3% had a history of CR-DVT, and 48.7% had a history of PE or LE-DVT. There were variable anticoagulation prophylactic and management strategies peri-transplant at both sites.

With a median follow-up of 366 days, 274 developed isolated CR-DVT and 191 patients developed PE or LE-DVT (including five patients with concurrent CR-DVT and PE). The 100-day and 1-year cumulative incidence of isolated CR-DVT was 4.3% (n = 185) and 6.6% (n = 274), respectively (Figure S1). The corresponding incidence for PE/LE-DVT was 1.8% (n = 76) and 4.7% (n = 191), respectively.

The strongest risk factor associated with isolated CR-DVT was the type of venous access (Tables 1 and S3). The use of PICC and non-tunneled CVC were strongly associated with CR-DVT when compared with tunneled CVC (HR 3.36 [95% CI 2.18–5.18] and 2.82 [95% CI 2.17–3.67], respectively). History of CR-DVT and PE/LE-DVT were both associated with future CR-DVT events (HR 1.70 [95% CI 1.13–2.57] and 2.14 [95% CI 1.43–3.20], respectively). Additionally, hospitalization status and acute GVHD status (grade 3–4) were both time-varying risk factors for CR-DVT (HR 3.06 [95% CI 2.31–4.06] and 1.93 [95% CI 1.35–2.77], respectively). Pre-conditioning thrombocytosis (defined as platelet count > 350 × 10⁹/L) was a risk factor for CR-DVT (HR 2.78 [95% CI 1.64–4.71]), while leukocytosis (defined as > 11 × 10⁹/L) showed an inverse association with CR-DVT (HR 0.39 [95% CI 0.16–0.95]).

The strongest risk factor associated with PE/LE-DVT was a history of prior PE/LE-DVT (HR 2.47 [95% CI 1.62–3.74]). Among the time-varying risk factors, hospitalization status and acute GVHD (grade 3–4) were again found to be significant (HR 1.91 [95% CI 1.28–2.83] and HR 1.80 [95% CI 1.20–2.70], respectively). Unique risk factors for PE/LE-DVT included lower KPS (80 vs. 100) (HR 2.01 [95% CI 1.27–3.19]) and obesity (BMI ≥ 35 kg/m²) (HR 1.70 [95% CI 1.17–2.46]). Venous access line type and history of CR-DVT were not associated with incident PE/LE-DVT.

In summary, we identified 465 total incident VTE within the first-year post-transplant in this multicenter study of 4250 patients. The cumulative incidence was clinically significant at 6.6% for isolated CR-DVT and 4.7% for PE/LE-DVT at 1 year. Our study provides new insights into potentially modifiable CVC-related risk factors for the prevention of CR-DVT. Furthermore, we have identified other patient- and HCT-related factors that uniquely contribute to the risks of CR-DVT and PE/LE-DVT.

Currently, there is not a unified standard of care for CVC type post-transplant, and clinical practices of the use of CVC vary significantly across different institutions. Pharmacologic thromboprophylaxis after allogeneic HCT remains controversial and is not routinely implemented, especially before platelet engraftment [4, 5]. Our two-center study design provides the opportunity to compare different practices for CVC use after HCT and their impact on CR-DVT risk. We found that the use of indwelling non-tunneled CVC or PICC had a significantly higher risk of CR-DVT when compared with tunneled CVC (HR 2.82 and 3.36, respectively). This is consistent with findings from a prior meta-analysis where PICC was associated with a higher risk of DVT (OR 2.55) in critically ill patients or those with cancer [6]. Our findings suggest that the selection of CVC type could be a strategy to reduce the risk of CR-DVT.

We have further identified overlapping risk factors for CR-DVT and PE/LE-DVT in HCT, namely acute GVHD, prolonged hospitalization, and history of VTE. GVHD and hospitalization are well-known time-varying risk factors for overall VTE development in both HCT and non-HCT settings due to their association with acute inflammatory, infectious, and immobilization states. Interestingly, a prior history of CR-DVT contributed to the risk of new CR-DVT but not to the risk of PE/LE-DVT. Conversely, a prior history of PE/LE-DVT also contributed to the risk of new CR-DVT. This suggests that CR-DVT is seldom a source of PE and that underlying thrombophilia can also contribute to the risk of CR-DVT. Furthermore, risk factors unique to CR-DVT and PE/LE-DVT suggest that not all VTE developments are mechanistically similar. For example, factors unique to CR-DVT included high platelet and low WBC count. In contrast, obesity (> 35 kg/m²) and poor KPS were uniquely associated with PE/LE-DVT.

Strengths of our study include the evaluation of two large cohorts, which allows a broader generalization when compared to previous studies. We assessed the impact of CVC-related practice discrepancies, unveiling potential practice change to mitigate the risk of CR-DVT. Limitations to our study include those inherent to a retrospective analysis. We did not assess the impact of other baseline risk factors, including family history, underlying thrombophilic states, and site of catheter insertion, or other time-varying risk factors, including acute infection and chronic GVHD for CR-DVT. There were also significantly fewer patients with PICC compared to the other CVC types, as well as differing VTE prophylaxis/monitoring strategies between the two sites.

In conclusion, our study provides the largest to date investigation of strategies for a multimodal approach for CR-DVT prevention. There are overlapping and unique risk factors for isolated CR-DVT and PE/LE-DVT. CVC-type selection appears to be an attainable intervention to implement, while the selection of populations with high risk of VTE post-transplant (acute GVHD and prolonged hospitalization) may be of interest for future studies to test the clinical net benefit of pharmacological thromboprophylaxis.

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board (IRB) at both MDACC and FHCC.

The authors declare no conflicts of interest.