Risk Factors for Venous Thromboembolism in Hospitalized Patients With Acute Medical Illness: Analysis of the MEDENOX Study

Background  There is limited information about risk factors for venous thromboembolism (VTE) in acutely ill hospitalized general medical patients.

Methods  An international, randomized, double-masked, placebo-controlled trial (MEDENOX) has previously been conducted in 1102 acutely ill, immobilized general medical patients and has shown the efficacy of using a low-molecular-weight heparin, enoxaparin sodium, in preventing thrombosis. We performed logistic regression analysis to evaluate the independent nature of different types of acute medical illness (heart failure, respiratory failure, infection, rheumatic disorder, and inflammatory bowel disease) and predefined factors (chronic heart and respiratory failure, age, previous VTE, and cancer) as risk factors for VTE.

Results  The primary univariate analysis showed that the presence of an acute infectious disease, age older than 75 years, cancer, and a history of VTE were statistically significantly associated with an increased VTE risk. Multiple logistic regression analysis indicated that these factors were independently associated with VTE.

Conclusions  Several independent risk factors for VTE were identified. These findings allow recognition of individuals at increased risk of VTE and will contribute to the formulation of an evidence-based risk assessment model for thromboprophylaxis in hospitalized general medical patients.

Venous thromboembolism (VTE) remains an important cause of morbidity and mortality in a wide range of patients and conditions.¹ The rationale for providing thromboprophylaxis is that prevention is clinically and financially beneficial compared with treatment of a thromboembolic event once it has occurred. Extensive data have been collected to support this position for surgical patients. Until recently, there was a paucity of data on the benefits of thromboprophylaxis in general medical patients,² despite evidence showing that VTE is a substantial problem in medical populations.³ Although thromboprophylaxis studies had been performed, they tended to focus on the use of thromboprophylaxis in high-risk patients, such as those with myocardial infarction or stroke.¹ In contrast, the limited data in general medical patients were based on small, open-label studies, some of which used diagnostic methods now considered obsolete to detect VTE.⁴

Recent randomized trials^5,6 have helped define the level of thromboembolic risk in well-characterized acutely ill medical patients, particularly those with cardiorespiratory disease. The MEDENOX (prophylaxis in MEDical patients with ENOXaparin) trial⁵ was a randomized, double-masked, placebo-controlled study designed to clarify the actual thromboembolic risk in clearly defined populations of medical patients and to evaluate the benefit-risk ratio of 2-dose regimens of a low-molecular-weight heparin. The study showed that these patients were at moderate risk of VTE according to accepted criteria (a 15% incidence in the placebo arm), and these findings are now reflected in the American College of Chest Physicians and International Consensus Statements that recommend thromboprophylaxis.^1,7 However, there remains limited data on the independent nature of either disease- or patient-related VTE risk factors in medical patients.⁸ This uncertainty, we suggest, has resulted in certain empiricism in the selection of patients who require thromboprophylaxis. We aim to identify independent risk factors for VTE using logistic regression analyses in the well-defined, hospitalized, general medical patients in the MEDENOX study.

The MEDENOX study was a multicenter, randomized, double-masked trial that compared low-molecular-weight heparin (20 or 40 mg of enoxaparin sodium subcutaneously once daily) with placebo, as previously described elsewhere.⁵ The primary outcome event was the occurrence of documented VTE between day 1 and day 14. Venous thromboembolism was defined as deep vein thrombosis identified by either mandatory lower limb contrast venography or (1) compression ultrasound (mean ± SD day 10 ± 4), (2) clinical suspicion of deep vein thrombosis, (3) pulmonary embolism confirmed by pulmonary angiography or high-probability lung scan, or (4) fatal pulmonary embolism.

Patients were older than 40 years with a planned hospital stay of at least 6 days and had been recently immobilized for 3 days or less. Patients were hospitalized for acute heart failure, New York Heart Association class III or IV, or acute respiratory failure that did not require immediate ventilatory support. Other patients included had 1 of 3 medical conditions (acute infectious disease without septic shock, an acute rheumatic disorder, or an active episode of inflammatory bowel disease) and at least 1 predefined VTE risk factor (age >75 years, cancer, previous history of VTE, obesity, varicose veins, hormone therapy, and chronic heart or respiratory failure). Chronic respiratory disease and chronic heart failure were not assessed as risk factors if the patient was hospitalized with the corresponding acute medical illness.

Aventis Pharmaceuticals (Bridgewater, NJ) made the complete MEDENOX database available. A total of 1102 patients were enrolled in 60 medical centers in 9 countries. All analyses were performed on this database using the intention-to-treat population, defined as all patients with an evaluable primary end point assessment (n = 866). Evaluable patients were divided equally among the 3 treatment groups: placebo (n = 288), 20 mg of enoxaparin (n = 287), and 40 mg of enoxaparin (n = 291).

In the initial step of building a regression model, 8 explanatory variables (age >75 years, cancer, history of VTE, obesity, varicose veins, hormone therapy, chronic heart failure, and chronic respiratory failure) were given binary values (ie, present or not present), and their effect on the response variable (the presence or not of VTE) was considered individually (univariate analysis). In addition, the disease groupings defined in the MEDENOX trial were examined in an identical manner.

Univariate analyses were performed on 2 populations: (1) the complete MEDENOX patient population for whom primary outcome data were available (n = 866) and (2) patients who had not received active or effective antithrombotic treatment (patients who received placebo or 20 mg of enoxaparin) (n = 575). Relative risks were calculated as the risk of events in patients with the characteristic divided by the risk of events in patients without the characteristic. Evidence of association was calculated using the Fisher exact test. Characteristics identified by the univariate analyses as potential risk factors were considered for inclusion in a multivariate logistic regression analysis. To ensure selection of the best combination of explanatory variables, only those with a P<.20 were included in the model, and they remained in the model only if they were significantly related to the response variable (P<.10).

A multiple logistic regression analysis was conducted to determine the independent nature of the risk factors while adjusting for other characteristics. A stepwise selection model was used to look for other characteristics, and a stepwise selection procedure was used that started with all covariates in the model.

The software package used for the analyses was SAS version 6.12 (SAS Institute Inc, Cary, NC).

The incidence of VTE by day 14 (the primary outcome event) was significantly lower in the group assigned to receive 40 mg of enoxaparin (16 [5.5%] of 291 patients) than in the placebo group (43 [14.9%] of 288 patients), a risk reduction of 63% (relative risk, 0.37; 97.6% confidence interval [CI], 0.22-0.63; P<.001). There was no significant difference in the incidence of VTE between the group that received 20 mg of enoxaparin (43 [15.0%] of 287 patients) and the placebo group (P = .90).⁵ Considering the disease groups separately, the highest incidence of VTE occurred in patients with New York Heart Association class IV heart failure (Table 1). Patients who had experienced a previous thromboembolic episode or cancer where found to have experienced the highest incidence of VTE (Table 2).

Among the predefined risk factors, age older than 75 years, cancer, and previous VTE were significantly associated with an increased risk of VTE (Table 3). The only acute medical illness significantly associated with an increased risk of VTE was an acute infectious disease (Table 4). Acute respiratory failure and chronic respiratory disease had the lowest risk of VTE (9.2% and 9.0%, respectively) compared with other acute medical illnesses and risk factors (Table 3 and Table 4).

Comparing the complete patient population with the patients who did not receive effective antithrombotic prophylaxis, there was no difference in the risk factors (Table 5) or type of acute medical illness (Table 6) associated with an increased risk of VTE (age >75 years, cancer, previous VTE, and the presence of an acute infectious disease).

The multiple regression model showed that age older than 75 years, cancer, previous VTE, acute infectious disease, and chronic respiratory disease were all independently related to risk of VTE (Table 7). Previous VTE had the highest odds ratio (OR, 2.06; 95% CI, 1.10-3.69).

The number of VTE events was categorized by the number of risk factors present in each patient (Table 8). The proportion of patients experiencing a thromboembolic event increased with the number of risk factors. The increase in the number of VTE events as the number of risk factors increased from zero to more than 2 was not statistically significant (χ² test for linear trend: OR, 1.47; P = .47).

The MEDENOX study offered a unique opportunity to examine the independent nature of specific patient- and disease-related risk factors in acutely ill, hospitalized medical patients. The MEDENOX study comprised 3 treatment populations—those who received effective antithrombotic therapy (40 mg of enoxaparin once daily), those who received ineffective therapy (20 mg of enoxaparin once daily), and those who received placebo.⁵ We considered that effective antithrombotic therapy may exert a confounding effect on the risk findings and therefore performed univariate analyses on 2 populations: the complete MEDENOX population and the subset of patients treated with placebo or ineffective thromboprophylaxis. Our findings were similar in each group and in the choice of factors to enter into the multivariate logistic regression analyses made from the complete MEDENOX population.

Our analysis shows that the presence of an acute infectious disease or cancer and a previous history of VTE or age older than 75 years are independent risk factors for objectively confirmed VTE. The MEDENOX study, in keeping with most thromboprophylaxis trials, used accurate objective testing by ascending contrast venography to detect deep vein thrombosis as the main efficacy finding. The thrombi detected were largely asymptomatic, an important point to consider when reviewing the independent risk factors found in this study as, in risk studies previously described,⁹ clinically apparent VTE frequently was the variable analyzed. A relationship between venographically detected deep vein thrombosis and symptomatic VTE has been shown in surgical patients,^10,11 but there is a paucity of data in medical patients.

The risk of VTE was highest among patients who had previously experienced a thromboembolic event and was independently associated with confirmed VTE (OR, 2.06; 95% CI, 1.10-3.69). In a French study¹² of hospital inpatients admitted to a medical ward for deep vein thrombosis, multivariate analysis showed that a personal history of VTE was an independent risk factor (OR, 4.7; 95% CI, 2.4-8.9). In addition, the study also showed that a family history of VTE was independently related (OR, 3.3). In an analysis of medical outpatients, the Sirius study⁸ showed a very strong association with previous VTE, showing that it was independently associated with symptomatic VTE (OR, 15.6; 95% CI, 6.77-35.89).

Patients with cancer are at increased risk of VTE, which is related to factors associated with the cancer, the interventions used (such as central venous catheters, chemotherapy, and radiotherapy), and surgery.¹³ Our analysis shows that in general medical patients, a history of or the presence of an active cancer is statistically significantly related to high risk of VTE (OR, 1.62; 95% CI, 0.93-2.75). These findings closely mirror those of other studies. Cancer has been shown to be an independent risk factor in a population case-control series of outpatients with symptomatic VTE,⁹ and in the HERS (Heart and Estrogen/progestin Replacement Study),¹⁴ cancer was independently associated with first-episode VTE.

This analysis shows that acute infection was an independent risk factor in the generation of venous thrombosis. An early study of VTE in medical patients revealed a benefit of heparin use in immobilized patients with heart failure or chest infection. Low-dose unfractionated heparin therapy statistically significantly reduced deep vein thrombosis from 26% to 4%.¹⁵ This study and the MEDENOX study confirm that elderly immobilized medical patients frequently have multiple comorbidities, including infection. In the MEDENOX study, a substantial proportion of patients had infection concomitant with cardiorespiratory disease. This is the first time that infection has been reported as an independent risk factor in hospitalized medical patients. The SIRIUS study⁸ found infectious disease to be a risk factor for VTE in medical outpatients (OR, 1.95; 95% CI, 1.31-2.92). The importance of identifying infection as an independent risk factor for VTE in medical patients has also been demonstrated by an association recently described between fatal pulmonary embolism and infection (R. Alikhan, MBBS, F. Peters, MBBS, R. Wilmott, A.T. Cohen, MD, unpublished data, 2002).

The risk of thrombosis increases sharply with age, from approximately 1 in 10 000 people per year for those younger than 40 years to 1 in 100 people per year for those 75 years and older,¹⁶ and is an independent risk factor for VTE.¹ Although it is not certain why risk is dependent on age, it has been suggested that it is related to a combination of decreased mobility and muscle tone with increased morbidity and degenerative vascular changes.¹⁷ In a population-based cohort study¹⁸ that examined independent predictors of death within 7 days after a VTE and during a 25-year period in 2218 patients, the OR of death was 1.3 times higher in the 75-year-old patients compared with the 60-year-olds. A further study¹⁹ showed that the incidence of VTE is markedly raised with age, with patients older than 75 years experiencing an annual incidence of 1% and VTE rates twice those for 60- to 74-year-olds. A previous analysis²⁰ of the MEDENOX study identified individuals older than 75 years as gaining marked benefit from taking 40 mg of enoxaparin in reducing the incidence of VTE (78% relative risk reduction).

Obesity was not found to be a risk factor for VTE in the present study, in keeping with the findings from HERS, which showed that, in 2763 postmenopausal women with no previous history of VTE, there was no association between obesity and symptomatic VTE (relative risk, 1.0).¹⁴ However, the results of 3 other studies—the SIRIUS study,⁸ a prospective nurses health study,²¹ and a prospective Swedish cohort study²²—suggested that obesity was linked to VTE.

The link between varicose veins and VTE is a contentious one with somewhat conflicting findings. In our analysis, varicose veins were not an independent risk factor for VTE. This is in keeping with findings from the Framingham Study,²³ which did not identify varicose veins as an independent predictor of pulmonary embolism.

In a group of medical outpatients, but not patients with chronic respiratory disease, the SIRIUS study⁸ showed that chronic heart failure was independently associated with VTE (OR, 2.93; 95% CI, 1.55-5.56). In our analysis of acutely ill immobilized medical patients, chronic heart failure was not independently related to VTE, and patients with chronic respiratory disease had statistically significantly less VTE relative to the other higher-risk disease groups and risk factors in the model.

In 1995, the American College of Chest Physicians added a "grade A" recommendation for "general medical patients with clinical risk factors for venous thromboembolism, particularly those with congestive heart failure and/or chest infections" and recommended low-dose unfractionated heparin or low-molecular-weight heparin therapy.²⁴ In 2001, the American College of Chest Physicians extended the recommendations to include medical patients with the following risk factors: severe lung disease, cancer, and bed rest.¹ Our analysis extends the findings of the MEDENOX study, revealing that certain disease and patient factors are independently related to the genesis of VTE in acutely ill medical patients. This analysis shows that a previous history of VTE in medical patients seems to be the most important predictor of future events. Currently, there is no precise patient risk profile for medical illness that would precipitate a mandatory recommendation for thromboprophylaxis, and we believe that these findings will facilitate the formulation of a risk assessment model based on clinical evidence.

Corresponding author and reprints: Alexander T. Cohen, MBBS, MSc, MD, FRACP, Academic Department of Surgery, Guy's, King's, and St Thomas' School of Medicine, Bessemer Road, London SE5 9PJ, England (e-mail: alexander.cohen@kcl.ac.uk).

Accepted for publication June 24, 2003.

This analysis was supported by a grant from Aventis Pharmaceuticals.