Kaplan-Meier analysis of time to recurrence among patients with and without vena caval filters.
Elting LS, Escalante CP, Cooksley C, Avritscher EBC, Kurtin D, Hamblin L, Khosla SG, Rivera E. Outcomes and Cost of Deep Venous Thrombosis Among Patients With Cancer. Arch Intern Med. 2004;164(15):1653-1661. doi:10.1001/archinte.164.15.1653
Although deep venous thrombosis (DVT) often complicates the clinical course in patients with cancer, few studies of the outcomes of DVT in this population have been published. Furthermore, the cost of DVT is largely undescribed. We herein report the largest study of DVT in this population to date.
We reviewed the medical records of 529 consecutive cancer patients in whom DVT developed from January 1, 1994, through December 31, 1997, and followed up these patients through December 31, 2000, for outcomes. The cost of hospitalization was obtained from our hospital's cost-accounting system and inflated to 2002 US dollars using the Consumer Price Index for Medical Care. Logistic regression was used to identify factors that were associated with a high risk of poor outcomes.
The most common complication of DVT was bleeding, which occurred in 13% of patients. Pulmonary embolus occurred in 4%. Five patients (1%) died of complications of DVT and 5 (1%) of complications of anticoagulation. Recurrence of DVT was common (17% overall), particularly among those who had inferior vena cava filters (32%; P<.001) or a previous episode of DVT (P = .03). All but 4 patients were hospitalized for initial anticoagulation therapy, for a mean of 11 days. The mean cost of hospitalization was 2002 US $20 065.
Among patients with cancer, DVT frequently is associated with serious clinical outcomes. Its treatment is resource intensive and costly. More effective agents and less costly management strategies could have a significant impact on the outcomes and cost of DVT in this population.
Deep venous thrombosis (DVT) is a common complication of cancer and its therapy.1- 10 In the case of some advanced cancers, DVT is the presenting clinical sign. In most studies of DVT and its treatment, a diagnosis of cancer is a poor prognostic sign.1- 21 However, many studies do not report a separate analysis for cancer patients. Large epidemiological studies in this population are rare, and studies of the cost of DVT are very uncommon indeed. We herein report, to our knowledge, the largest study of the outcomes of DVT in cancer patients to date and describe the resources used in the management of DVT and their costs. We also examine risk factors for serious clinical outcomes such as DVT recurrence and pulmonary embolism (PE).
The first episode of DVT that developed in each of 529 consecutive patients from January 1, 1994, through December 31, 1997, was included in the study. Potential subjects were identified by searching an electronic claims database for all inpatient and outpatient encounters with an International Classification of Diseases, Ninth Revision diagnosis code of 451.xx (451.1-451.9). Those who also met the requirements for a diagnosis of DVT (as described in "Definitions") were included in the study. The DVT was managed at the discretion of the attending physician. No guidelines for DVT therapy were used during this time, and there were no clinical trials for prophylaxis or therapy for DVT.
The subjects' medical records were reviewed by physician and nurse abstractors for clinical outcomes information and resource utilization. A 5% sample of records was subjected to a second, masked review to estimate the accuracy of outcomes (100%) and other clinical data (96%). This study was approved by the Institutional Review Board at the University of Texas M. D. Anderson Cancer Center, Houston.
Patients with clinical evidence of DVT (eg, pain, erythema) and/or evidence of radiologic confirmation (eg, venogram, ultrasonogram, ventilation-perfusion scans) were considered to have DVT. Those with clinically suspected DVT but no definitive radiologic evidence of DVT were excluded.
Eligible patients were followed up through December 31, 2000, for the outcomes of interest, resolution of symptoms, recurrence of DVT, and complications. Resolution of symptoms was characterized as complete, partial, or unresolved. Partial resolution was defined as improvement but not complete resolution of symptoms, ie, pain or edema, and unresolved symptoms as no appreciable improvement. Recurrence was defined as the presence of new radiologic evidence of DVT. Complications were characterized as early or late and by their relationship to DVT or to anticoagulation therapy. Early complications were those that occurred during initial anticoagulation therapy. Late complications were those that occurred during maintenance anticoagulation therapy. Embolic events (ie, PE, superior vena cava syndrome, cerebral thrombosis, or disseminated intravascular coagulation) were considered to be due to the DVT. Thrombocytopenia and bleeding were considered to be related to anticoagulation therapy. Bleeding was characterized as minor (World Health Organization grades 1-2) or major (World Health Organization grades 3-4).
Clinical factors were examined for their association with serious outcomes. These included host-related, disease-related, and treatment-related factors. Performance status as defined by Zubrod et al,22 ability to ambulate, and time spent in bed were measured at presentation with DVT. The type of underlying malignancy and its status were also measured at presentation. The international normalized ratio was measured at baseline and the end of all therapy and recorded at its highest value.
The number and type of health care resources used for diagnosis and management of DVT and its complications were collected from the medical record. These included hospital days, clinic visits, blood products, diagnostic tests, and pharmaceuticals. Health care services that were used for diagnosis and management of the underlying malignancy were excluded. The cost (not the charge) of each hospitalization was obtained from our hospital's accounting system, which uses cost-charge ratios to compute cost of care. These costs are expressed in 1997 US dollars. To estimate the current cost of care, we inflated the 1997 estimates to 2002 US dollars using the Consumer Price Index for Medical Care.23
To determine statistical significance, we tested differences among categorical variables using the 2-tailed Pearson χ2 test,and differences among continuous variables using 2-tailed t tests. We developed multiple variable models of risk of serious outcomes using logistic regression of variables shown to be statistically significant in univariate analysis. Kaplan-Meier analysis was used to compare differences in time to recurrence. All statistical analyses were computed using BMDP Dynamic software (BMDP Statistical Software, Inc, Los Angeles, Calif).
Deep venous thrombosis developed in a total of 529 patients during the 4-year study. Two-hundred seventy-seven (52%) were women, and the mean age was 55 years (95% confidence interval [CI], 54-57 years). Although 406 patients (77%) presented as outpatients and 387 (73%) were able to walk without assistance, 325 (61%) had a poor performance status (Zubrod performance status, ≥2). Eighty patients (15%) had experienced episodes of DVT or PE in the past. Overall, DVT developed in 70 patients (13%) within 30 days of a surgical procedure. Among these, DVT developed in 22 patients within 1 week, in 23 patients at 1 to 2 weeks, and in 25 patients at 3 to 4 weeks postoperatively.
Lymphomas (15%) and genitourinary (15%), breast (13%), gynecologic (12%), lung (11%), and gastrointestinal tract (11%) malignancies predominated (data not shown). Approximately half of the patients (51%) had progressive disease despite appropriate therapy, 24% were responding to therapy, and 19% had new diagnoses. Ninety-nine patients (19%) had liver dysfunction (serum bilirubin levels at least 2 times normal values), primarily due to metastases (n = 84) or chemotherapy (n = 9). Thirty-three patients (6%) were bone marrow transplant recipients. One hundred six patients (20%) had inferior vena cava (IVC) filters. Among these, 89% were inserted during this DVT episode, and the remainder during a previous episode of DVT. Virtually all filters (91%) were bird's nest filters. All but 14 (13%) of the patients with filters received anticoagulation therapy.
Although 60% of outpatients presented to the emergency department an average of 4 days after the onset of symptoms (95% CI, 3-5 days), 82 (20%) had experienced symptoms for at least 2 weeks. In contrast, a diagnosis of DVT was established among most inpatients (71%) within 24 hours of the onset of signs and symptoms (P = .03). The most common signs and symptoms among outpatients were edema (62%), pain (23%), and/or heat (21%) at the site of DVT. These symptoms were somewhat less common among inpatients (51%, 18%, and 11%, respectively). Eight percent of patients were asymptomatic and received a diagnosis during the cancer staging workup. Overall, DVT developed in a lower extremity in 50% of patients and in a major central vein in 43%, with or without the involvement of an extremity, as shown in the following tabulation.
The distribution of sites was similar for inpatients and outpatients (data not shown). However, DVT involving a major central vein was significantly more common among the 229 patients (43%) who had central venous catheters at the onset of symptoms than among those without central venous catheters (51% vs 35%; P<.001) (Table 1).
Eighty-seven percent of patients experienced complete or partial resolution of the signs and symptoms of DVT (Table 2). Most patients (88%) were treated initially with unfractionated (UF) heparin sodium, whereas 4% received warfarin sodium, 1% received low-molecular-weight (LMW) heparin, and the remainder (7%) did not receive anticoagulants. Bleeding developed in 66 patients (12%) during anticoagulation therapy (Table 2). Forty-three bleeding episodes were considered major, including 6 episodes of central nervous system hemorrhage. Recurrence of DVT developed in 91 patients (17%) a median of 5 weeks after the initial episode (mean, 18 weeks; 95% CI, 10-25 weeks), most involving the lower extremities. Three of the 55 recurrences in patients without IVC filters and none of the 36 recurrences among patients with filters involved the IVC. Among the 529 patients, 55 died during DVT, although most of the deaths (45/55) were related to the underlying malignancy rather than to DVT. Five patients died as a result of DVT, whereas 5 died as a result of complications of anticoagulation.
In some respects, patients with postoperative DVT fared better than their counterparts who had not undergone surgery. Sixty-seven (96%) of 70 patients with postoperative DVT, but only 394 (86%) of the other 459 patients, had complete resolution of symptoms (P = .04). No deaths occurred among patients with postoperative DVT. However, the incidence of complications among patients in whom DVT developed within 1 week of surgery was double that observed among all other patients (7/22 [32%] vs 85/507 [17%]; P = .08). Half of the complications among postoperative patients involved bleeding during initial anticoagulation therapy.
Six patients (1%) experienced thrombocytopenia during DVT, although these episodes were probably unrelated to heparin administration (Table 3). Three episodes occurred in patients who did not receive heparin. The remaining 3 patients had preexisting chemotherapy-induced thrombocytopenia that worsened during UF heparin therapy. All episodes of thrombocytopenia resolved.
Pulmonary embolism developed in 23 patients during DVT a median of 11 days after initiation of anticoagulation therapy (Table 3). Four (17%) of these patients died. The risk of PE did not vary significantly by site of DVT, occurring in 3% of episodes involving the upper extremities and 5% of those involving the lower extremities, head and neck, or central veins. Ten episodes occurred during and 13 occurred after initial anticoagulation therapy. In 3 patients who experienced PE during initial anticoagulation therapy, recurrent PE developed later. Cerebrovascular accident occurred in 3 patients and superior vena cava syndrome developed in 3 patients. Two patients experienced septic thromboses and 5 patients experienced disseminated intravascular coagulation during therapy for DVT.
Five patients died as a result of complications of DVT and 5 as a result of complications of therapy (Table 3). One patient died as a result of PE during initial anticoagulation therapy. Four patients died after initial anticoagulation therapy, 3 as a result of PE during warfarin therapy, and 1 as a result of septic thrombosis and overwhelming infection. Five additional patients died as a result of anticoagulant-induced hemorrhage, 3 during heparin therapy, and 2 during warfarin therapy.
In univariate analysis, unresolved symptoms were most common among patients who had limited mobility (P<.001) and poor performance status (P = .07) and among those whose cancers were progressing on therapy (P<.001) (Table 4). Symptoms also were less likely to resolve among patients with gastrointestinal tract malignancies (28%) than among patients with other malignancies (10%; P<.001). Among those treated with UF or LMW heparin, symptoms resolved in 88% and 100%, respectively. However, among those who received warfarin and those who received no anticoagulation therapy, symptoms resolved in only 68% and 79%, respectively. Only 23 patients received thrombolytic agents. Among these, symptoms resolved in 21 patients (91%) compared with 440 (87%) of the 506 patients not treated with thrombolytic agents (P = .26). Unresolved symptoms were unrelated to international normalized ratio (14% for ≤2.0 compared with 11% for >2.0; P = .49). Multivariate analysis revealed that those with limited mobility (odds ratio [OR], 3.75; P<.001), those with progressive malignancy (OR, 2.92; P<.001), and those with gastrointestinal tract malignancies (OR, 2.23; P = .03) were at high risk of unresolved symptoms.
Sixteen percent of patients treated initially with UF heparin experienced recurrence of DVT compared with 32% of those treated with warfarin, 29% of those treated with LMW heparin, and 23% of those who received no anticoagulation therapy (Table 4). The number of patients treated with agents other than UF heparin was too small to support precise estimates of the recurrence rate. Furthermore, patients who received no anticoagulation therapy typically had more advanced malignancies and were, therefore, at risk for recurrence a shorter time owing to their shorter survival. A recurrent episode of DVT was more common among patients who had experienced DVT in the past (P = .005). However, unlike unresolved symptoms, recurrence was more common among those who were ambulatory (P = .02) and those with good performance status (P = .01), most likely because they were at risk longer owing to their superior survival. Similarly, recurrence was more common among patients younger than 50 years (20%) than those older than 70 years (12%), also likely owing to shorter time at risk. However, after accounting for varying time at risk using Kaplan-Meier analysis, the median time to recurrence was significantly shorter among patients with IVC filters than among those without (P<.001) (Figure 1). The median time to recurrence was greater than 60 months for patients without IVC filters, irrespective of whether they had a history of DVT. The median time to recurrence was 43 months for patients with a filter and no history of DVT and 36 months for those with a filter and a history of DVT. Multivariate analysis showed that patients with IVC filters (P<.001), those who had a history of DVT (P = .03), and those with gynecologic tumors (P = .002) were at highest risk for recurrence.
Because of the small number of deaths due to DVT (n = 10), no clear pattern of risk factors emerged (Table 4). However, DVT-related death was significantly more common among patients who had a history of DVT compared with those in whom this was the first episode (6% vs 1%; P = .02). Six of the 10 DVT-related deaths occurred among patients whose cancers were progressing despite appropriate therapy.
Management of DVT consumed considerable health care resources. All but 4 patients were hospitalized for initial anticoagulation therapy. Their mean length of DVT-attributable hospitalization was 11 days (95% CI, 11-13 days). Early and late complications increased the mean length of hospitalization by 7 and 11 days, respectively, among those in whom complications developed. Patients in whom DVT developed during their hospitalization had significantly longer DVT-attributable days of hospitalization (21 days) compared with patients who presented as outpatients (8 days; P<.001). A mean of 8 coagulation studies were performed during the initial DVT episode, reflecting the frequent use of UF heparin and warfarin. Early and late complications led to an average of 3 red blood cell transfusions per episode with complications.
The average cost of hospitalization for the index DVT episode was $16 482 ($1465 per hospital day) in 1997 US dollars. The estimated mean cost in 2002 US dollars of these episodes is $20 065 per episode ($1784 per hospital day) (Table 5). Patients in whom PE or bleeding developed during the initial hospitalization had longer hospital stays and higher costs. However, the cost per day of hospitalization for DVT episodes with bleeding was 38% higher than the cost per day for DVT episodes without bleeding episodes.
Our study demonstrates the clinical and financial importance of DVT among patients with cancer. The study's principal strength is its large sample of patients with cancer (N = 529), which is to our knowledge the largest series of cancer patients published to date. Studies of DVT are rarely limited to cancer patients and typically do not present a separate analysis for cancer patients. For this reason, precise estimation of rates of serious clinical outcomes is difficult for this population. Our large sample of consecutive cases of DVT supports such estimates. Furthermore, our study focuses on the outcomes and cost of DVT management in general oncology practice, not health care administered as part of a clinical trial. This design provides realistic estimates of the resources used and the outcomes that can be expected in everyday management of DVT. However, the retrospective design introduces 2 important limitations. First, some information may not have been recorded in the medical record. Because this occurs most often in the case of minor problems (ie, minor bleeding), the rate of minor complications may be underestimated in this study. Second, although our focus on standard practice provides realistic estimates of outcomes and cost, it also includes a wide range of management strategies, not all of which would be considered standard. Clinical trials with controlled treatment protocols are common in the literature. Thus, our report of everyday management makes a unique contribution, despite the diversity of management strategies.
We have shown that DVT often results in serious clinical outcomes. Although death due to DVT was relatively uncommon (2%), complications and recurrences were quite common (18% and 17%, respectively). Our findings are at the highest extremes of rates observed in previous studies24- 42 (Table 6). In these studies, the frequency of bleeding varied from less than 1% to 17%, depending on whether UF heparin or LMW heparin was used. The frequency of recurrence varied from 0% to 21% when heparin therapy was used and from 5% to 37% when IVC filters were used. The rates of bleeding and recurrence varied considerably across these studies, but the finding of bleeding and recurrence as being most common among patients with cancer was reported consistently. Our observations parallel these studies and a recently published study of patients with and without cancer that compared the risks of bleeding (12.4% vs 4.9%, respectively) and recurrence (20.7% vs 6.8%, respectively).41 Our estimates of the incidence of bleeding (12%) and recurrence (17%) among cancer patients are very similar. Therefore, although our bleeding and recurrence rates were quite high when compared with previously published rates from the general medical population, they are consistent with previous reports of high rates among cancer patients.
Our data show that the most significant predictor of recurrent thromboembolism is the presence of an IVC filter. This association has been reported previously, although data to the contrary have also been published34,39,40 (Table 6). The excess thromboembolic events may be due to migration of very small clots through the filter or, alternatively, formation of clots around the filter itself due to reduced venous flow (although the lack of recurrences in the IVC suggests that this was not a major mechanism in our patients). Irrespective of the pathogenesis, this observation illustrates the trade-off between the early benefits of prevention of PE (which occurs in 1%-10% of patients without filters but may be fatal in 10%-25% of cases)43,44 and the long-term risks of subsequent thromboembolic events (which occur in 5%-37% of patients with filters). These risks and benefits must be carefully weighed when decisions are made for individual patients. Some authors have reported success with retrievable IVC filters.45 This new technology may hold promise for providing early benefits of PE prevention without the added risk of late thromboembolic events.
The Agency for Healthcare Research and Quality reported that the national bill for thromboembolic events in the United States was $1 861 013 599 in 2000, the most recent year for which such data are available.46 These estimates are based on claims for all discharges for thromboembolic events in the general medical population, including but not limited to patients with cancer. In that data set, PE discharges cost an average of $15 412 for 6.3 days of hospitalization, whereas DVT discharges cost an average of $8937 for 5.2 days of hospitalization. Overall, the mean cost of a thromboembolic episode in 2000 was $11 141. Even after inflating this cost to 2002 US dollars, our estimates of cost exceed the national average charge by 64%. Why do these estimates differ so significantly?
The cost of DVT is driven by 2 major factors: development of complications (ie, bleeding) and length of hospitalization. Since complications are more common among cancer patients, the average cost per day would be expected to be higher. Similarly, the average length of stay for cancer patients exceeds that for patients with most other medical conditions. The overall result of these factors is more hospital days at a higher cost per day. However, the high cost of DVT in cancer patients is not inevitable.
Although it may not be possible to reduce the incidence of bleeding among cancer patients who have multiple risk factors for bleeding, the duration of hospitalization among cancer patients without complications probably can be reduced by using LMW heparin in the outpatient setting. Several studies in which approximately 20% of subjects had cancer have reported successful outpatient therapy of DVT with LMW heparin.31,32,47,48 In these studies, mean hospital stays for therapy for DVT were reduced by more than 50% among patients who received LMW heparin in the outpatient setting. Our cases were identified between 1994 and 1997, with follow-up through 2000. During that time, no studies had documented the safety of outpatient therapy in cancer patients, and studies in the general medical population were just being published. However, practice has now begun to change. Currently, more than 40% of patients with lower extremity DVT are treated as outpatients in our institution. This change in practice will likely result in a lower mean cost of an episode of DVT overall, but an increase in the cost per episode with hospitalization, because hospitalization will be limited to patients who are very ill or in whom complications develop. We anticipate publication of studies of outpatient therapy with large samples of cancer patients in the near future and look forward to the important contributions that these studies may make to a reduction in the national bill for thromboembolic disease among cancer patients.
Correspondence: Linda S. Elting, DrPH, Department of Biostatistics, The University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd, Box 196, Houston, TX 77030-4009 (email@example.com).
Accepted for publication September 4, 2003.
This study was supported in part by a grant from Pharmacia, Inc, Peapack, NJ (Drs Elting and Escalante).
This study was presented in part at the annual meeting of the American Society of Clinical Oncology; May 17, 1999; Atlanta, Ga.