Importance of Findings on the Initial Evaluation for Cancer in Patients with Symptomatic Idiopathic Deep Venous Thrombosis

Methods

Patient Sample

All adults who had venous ultrasonography at Brigham and Women's Hospital (Boston, Massachusetts) between January 1990 and May 1993 because deep venous thrombosis in a lower extremity was suspected were eligible. The study cohort was selected from the daily log of the vascular diagnostic laboratory. Patients were excluded if they presented with a risk factor for deep venous thrombosis [12, 13], such as preexisting cancer; had recently had surgery (within 3 months); had had prolonged immobilization (≥ 10 days); had recently had trauma to a lower limb (within 3 months); had a documented hypercoagulable state (either deficiency of antithrombin III, protein C, or protein S or the presence of a lupus anticoagulant); were pregnant; had recently given birth (within 2 weeks); or were using oral contraceptives. We also excluded patients who had had a procedure (such as computed tomography of the abdomen) to identify potential underlying cancer before venous ultrasonography, even though the diagnosis of cancer was not known at the time when venous ultrasonography was done. Patients who had had venous ultrasonography as part of the work-up for suspected pulmonary embolism were also excluded because our primary goal was to assess the work-up for cancer in patients with symptomatic idiopathic deep venous thrombosis (such as those with leg pain or swelling), not in patients with chest pain or shortness of breath.

If a patient had venous ultrasonography more than once between January 1990 and May 1993, only the first ultrasonogram was considered. The exclusion criteria were chosen before any data were collected. We did not exclude patients with a history of deep venous thrombosis because a recent study [6] showed an increased risk for cancer in patients with recurrent deep venous thrombosis. Of the 5797 venous ultrasonographic examinations, 3445 (59%) were excluded because they had been done in patients who had one or more risk factors (other than cancer) for deep venous thrombosis. Four hundred sixty (8%) of the patients in whom venous ultrasonography was done as part of the work-up for suspected pulmonary embolism were also excluded. Eight hundred thirty patients (14%) were excluded because they had ultrasonography as a follow-up procedure. Fourteen patients (0.2%) were excluded either because a work-up for cancer had been done before venous ultrasonography or because they refused to receive a mailed questionnaire about their medical situation. Finally, 62 patients (1%) were excluded because follow-up information was missing (Figure 1). None of these 62 patients received a diagnosis of deep venous thrombosis. The final cohort comprised 986 patients. The study was approved by the Human Research Committee of Brigham and Women's Hospital.

Venous Ultrasonography and Definition of Deep Venous Thrombosis

In all patients, duplex, color-assisted venous ultrasonography was done in each leg; 5.0 and 7.0 MHz transducers were used as described elsewhere [14]. Diagnosis of deep venous thrombosis was based on the lack of apposition on the venous walls during compression maneuvers. The presence or absence of venous flow augmentation was assessed by color-flow imaging and Doppler waveform analysis (done while the calf muscles were squeezed) to identify the deep venous thrombosis as occlusive or nonocclusive. Findings were confirmed by two observers. Deep venous thrombosis was classified as calf (tibial, peroneal, gastrocnemius, or soleal veins) or proximal (popliteal or femoral veins) venous thrombosis. Venous ultrasonography was used as the gold standard.

Data Collection

A standardized form was used to collect the following data from the charts of all study patients: date of birth, sex, place of residence, date of procedure, and current or previous use of tobacco. For all patients hospitalized with deep venous thrombosis, a second standardized form was used to collect clinical data from the charts about the initial clinical evaluation. This data included the results of medical history, physical examination, and laboratory tests. This 86-item form was based on the current recommendations for early cancer detection [15, 16] and included 1) findings in the medical history suggestive of underlying cancer, such as weight loss, fatigue, unusual pain, recent modification of bowel or bladder habits, unusual bleeding, or recent modification of cough or hoarseness; 2) findings on the physical examination suggestive of underlying cancer, such as abnormalities on abdominal or breast palpation (abnormal mass), oral cavity examination (ulcerative lesion), skin examination (atypical pigmentation), digital rectal examination (prostatic nodule or irregularities on the rectal walls), or lymphadenopathy or chest auscultation [rales or dullness]; 3) laboratory findings suggestive of underlying cancer, such as positive results on a test for fecal occult blood, abnormal values on the complete blood count and SMA 20, abnormal erythrocyte sedimentation rate, or abnormal results on urinalysis; 4) abnormalities on chest radiography suggestive of underlying cancer, such as lung nodules, pleural effusion, or mediastinal adenopathy; and 5) abnormal results of other procedures, such as computed tomography, ultrasonography, endoscopy, and biopsy. We also recorded the results of laboratory tests that were done after discharge but were related to the work-up done during hospitalization. Knowledge of previous laboratory test results can affect the observation or the recording of clinical findings [17], but we minimized this potential ascertainment bias by collecting data from the medical record only for the first clinical examination. This chart review form also recorded data on alcohol consumption and history of deep venous thrombosis. The person who reviewed the records was blinded to the occurrence of cancer during follow-up.

Follow-up and Definition of Cancer

The follow-up information for study patients in whom cancer was not diagnosed at the time of referral for ultrasonography was obtained through hospital chart review (38%), computerized record extraction (in patients who were members of a health maintenance organization in the Boston, Massachusetts, area) (54%), mailed questionnaires (6%), telephone interviews (1%), or the death registry of the Massachusetts Registry of Vital Records and Statistics (1%). Follow-up information included occurrence of cancer by site and date (including cancer found at autopsy), date of death or date on which the patient was last known to be alive, cause of death, and occurrence of deep venous thrombosis. Only the first diagnosis of cancer for each patient was considered. All cases of cancer were confirmed by histologic (95%) or cytologic (5%) examination, with the exception of one case of pancreatic cancer for which a clinical diagnosis was accepted. The patient who had this case had a pancreatic mass and liver metastases detected by computed tomography and died 3 weeks after discharge. Cases of nonmelanoma skin cancer were not included.

One hundred twenty-two of the 142 patients with deep venous thrombosis (the deep venous thrombosis group) and all of the 844 patients who did not have deep venous thrombosis (the comparison group) who had not received a diagnosis of cancer when deep venous thrombosis was suspected were followed for a median of 34 months (minimum, 3 months [unless the patient died within the first 3 months]; maximum, 60 months). This yielded 304 person-years of follow-up for the deep venous thrombosis group and 2548 person-years of follow-up for the comparison group (Figure 1). Ninety-five percent of the patients in the deep venous thrombosis group and 96% of those in the comparison group were followed for at least 1 year, until death, or until the occurrence of cancer; 2% of the deep venous thrombosis group and 4% of the comparison group were followed for 3 to less than 6 months; and 3% of the deep venous thrombosis group and 2% of the comparison group were followed for 6 to less than 12 months. No follow-up information was obtained for 4 of the 126 patients (3%) in the deep venous thrombosis group. Three of these 4 patients had no findings suggestive of cancer on the initial clinical evaluation when deep venous thrombosis was diagnosed, and the fourth (a 75-year-old man) had only mild anemia, moved to another state after receiving treatment for deep venous thrombosis, and could not be located. Follow-up information was obtained for the 6 nonhospitalized patients with deep venous thrombosis.

Statistical Analysis

The chi-square statistic, the Fisher exact test, the Student t-test, and the Wilcoxon rank-sum test were used as appropriate to compare patients who had deep venous thrombosis with patients who did not have deep venous thrombosis and to compare patients who had cancer with patients who did not have cancer. For the appropriate clinical evaluation for cancer, the analysis was restricted to data that were not missing; the exception was chest radiographs, which were available for 97% (132 of 136) of the patients. Data were considered to be missing if they were not recorded in the medical chart; otherwise, they were considered to be present. Patients with deep venous thrombosis were classified according to the number of components among the four components of the initial clinical evaluation in which findings were suggestive of cancer. The first component was a medical history, which was considered to be suggestive of cancer if it showed that the patient had had at least one of the following symptoms during the previous 6 months: weight loss (defined as an involuntary loss >5% of body weight), fatigue, any unusual pain, fever (defined as an elevation of body temperature to more than 99.5 °F [37.5 °C] with no evidence of relation to an infectious disease), recent change in bowel or bladder habits, unusual bleeding, or recent modification of cough or hoarseness. A symptom was considered to be present if it was noted in the chart; otherwise, it was considered to be absent. The second component was a physical examination, which was considered to be suggestive of cancer if it showed that the patient had had at least one of the following findings: abnormal abdominal palpation, cervical lymphadenopathy, abnormal results on skin examination, or abnormal results of chest auscultation. The third component was a complete blood count, which was considered to be suggestive of cancer if it showed anemia (hemoglobin concentration < 11.0 g/L for women or <13.0 g/L for men) or leukocytosis (leukocyte count > 12.0 × 10⁹/mL). The fourth component was a chest radiograph, which was considered to be suggestive of cancer if it showed the presence of a lung nodule, pleural effusion, or mediastinal adenopathy. We assigned equal weight to each of these four components. We could not assess the yield of three potential relevant abnormalities in the first clinical evaluation—digital rectal examination, breast examination, and urinalysis—because those examinations were too often either done after the first 24 hours of hospitalization or not recorded.

The probability of cancer-free survival in patients with and without deep venous thrombosis was calculated by using the Kaplan-Meier method, and significance was assessed by using the log-rank test. In addition, the Cox proportional-hazards model was used to compare incidence rates for cancer between patients with deep venous thrombosis and patients without deep venous thrombosis while adjusting for differences in age, sex, and smoking status (current or former smoker; never smoker). Information on smoking status was available for 97% of the 142 patients with deep venous thrombosis and 72% of the 844 patients without deep venous thrombosis.

We also compared the number of cases of cancer seen during follow-up in the comparison group with the expected number of cases of cancer. We determined the expected number from the age- and sex-specific incidence rates for total cancer in the general population (from the Surveillance, Epidemiology, and End Results [SEER] Program, based in the surveillance program at the National Cancer Institute from 1987 to 1991 [[18]). Confidence intervals were based on the binomial distribution (exact CIs) or its normal approximation, as appropriate. The SAS statistical package (SAS Institute, Cary, North Carolina) was used for the analyses.

To measure interobserver agreement between two independent reviewers of the clinical evaluation, the κ statistic was calculated from a random sample of 21 patients with deep venous thrombosis (15%). The result of this test was 0.84 (95% CI, 0.75 to 0.92), which suggests excellent interobserver agreement [19].

Results

Study Patients

From January 1990 to May 1993, 986 consecutive patients who were suspected of having idiopathic deep venous thrombosis and who had venous ultrasonography were included in the study. The mean age (±SD) was 53 ± 17 years (range, 16 to 92 years); 637 patients (65%) were female, 904 (92%) were outpatients, and 903 (92%) had no history of deep venous thrombosis. Deep venous thrombosis was diagnosed in 142 patients (14%) (Figure 1), 119 of whom (84%) had proximal deep venous thrombosis (deep venous thrombosis was not diagnosed in the remaining 844 patients). Compared with the patients who did not have deep venous thrombosis, patients with deep venous thrombosis were older (56 compared with 52 years), less likely to be female (45% [64 of 142] compared with 68% [574 of 844] were female), and more likely to be current or former smokers (51% [70 of 138] compared with 41% [246 of 601]). Of the patients without deep venous thrombosis, 9 (1%) received a diagnosis of superficial phlebitis and 54 (6%) were hospitalized for a leg sign or symptom.

Clinical Evaluation for Cancer in Patients with Deep Venous Thrombosis

Of the 142 patients with deep venous thrombosis, 136 (96%) were hospitalized (Figure 1). Of the 6 who were not hospitalized when venous thrombosis was diagnosed, 5 had an isolated calf deep venous thrombosis and 1 refused to be hospitalized. Sixteen of the 136 hospitalized patients (12% [CI, 6% to 17%]) received a diagnosis of cancer during the hospitalization (Table 1). All 16 patients had had one or more abnormalities on the history, physical examination, basic laboratory tests, or chest radiography. Compared with the patients in whom cancer was not diagnosed during hospitalization (Table 2), the patients in whom cancer was newly diagnosed during hospitalization for deep venous thrombosis were older (P = 0.05) and more likely to be current or former smokers (P = 0.09). Patients with cancer were less likely to have previously had deep venous thrombosis (6% compared with 33%; P = 0.04). Their history, as taken during the initial clinical evaluation, was suggestive of underlying cancer more often (88% compared with 13%), and they were more likely to have a physical finding suggesting cancer (56% compared with 7%), an abnormal complete blood count (88% compared with 37%), or a suspicious chest radiograph (38% compared with 2%). The probability of detecting cancer at the time of the deep venous thrombosis increased as the number of findings suggestive of cancer on the four components of the clinical evaluation increased (Table 3). Of the patients who had no findings suggestive of cancer on the medical history, physical examination, laboratory tests, or chest radiography (n = 56), none received a diagnosis of cancer at the time of deep venous thrombosis. Patients with cancer were more likely to have low albumin concentrations (81% compared with 24%; P < 0.001) and elevated alkaline phosphatase levels (58% compared with 25%; P = 0.02). The laboratory tests for alkaline phosphatase levels and albumin concentrations, which were done in 93% of the patients, did not add information to that provided by the four components of the initial evaluation. The results of urinalysis, which was done in 84% of the patients, were abnormal in 4 of 14 patients with cancer (2 of the 4 had hematuria; 1 of these 2 had bladder cancer, and the other had prostate cancer) and 1 of 99 patients without cancer (P < 0.001). However, the 2 patients who had cancer and hematuria also had other abnormalities on the other components of the clinical evaluation.

Incidence of Cancer

The probability of a diagnosis of cancer during a median follow-up period of 34 months was similar in patients with (2.5% [3 of 22]) and patients without (2.7% [23 of 844]) deep venous thrombosis (P > 0.2; log-rank test) (Table 4). After adjustment for age, sex, and smoking status in a Cox proportional-hazards model, the incidence rate of cancer was similar in the two groups (incidence rate ratio, 1.0 [CI, 0.4 to 2.2]). Nineteen patients died during the follow-up period: 9 in the deep venous thrombosis group (7%) and 10 in the comparison group (1%). Even if the patient with deep venous thrombosis who had anemia and was missing follow-up information when deep venous thrombosis occurred is assumed to have developed cancer during follow-up, the percentage of cancer would still be similar in the deep venous thrombosis group and the comparison group (3.2% and 2.7%, respectively).

The 3 patients in the deep venous thrombosis group who received a diagnosis of cancer during follow-up included a 46-year-old man in whom a spinal cord tumor (myxopapillary ependymoma of the filum terminale, with no metastases) was diagnosed 6 months after deep venous thrombosis, a 37-year-old man in whom superficial melanoma was diagnosed 22 months after deep venous thrombosis, and a 56-year-old women in whom breast cancer (lymph nodes positive) was diagnosed 24 months after deep venous thrombosis (Table 4). In these 3 patients, cancer was diagnosed primarily because low back pain persisted for 3 months, a primary care physician suspected melanoma during a health maintenance visit, and a lump suggestive of cancer was found during self-examination of the breast, respectively. At the time of venous thrombosis, the patient with the spinal cord tumor had anemia that was not evaluated further. The other 2 patients had no findings on the medical history, physical examination, or laboratory tests that were suggestive of cancer at the time of the deep venous thrombosis, and none of the 3 patients had recurrent deep venous thrombosis after the index hospitalization. Using follow-up information on cancer diagnoses as the gold standard, the sensitivity of the routine initial clinical evaluation (done at the time of hospitalization for venous thrombosis) was still 100% (CI, 82% to 100%) (19 of 19 patients) 6 months after presentation and was 89% (CI, 67% to 99%) (17 of 19 patients) 24 months after presentation. In the unlikely event that all 4 of the patients who were lost during follow-up developed cancer, the sensitivity of the initial clinical findings would still be 74% (CI, 52% to 89%) (17 of 23 patients).

In the group without deep venous thrombosis, 6 cases of lung cancer, 4 cases of breast cancer, 2 cases of prostate cancer, 2 cases of non-Hodgkin lymphoma, and various other types of cancer were diagnosed (Table 4). The mean period (±SD) between the referral visit and the diagnosis of cancer was 19 ± 13 months. The mean age of these patients with cancer was 64 ± 15 years, and 15 of the patients (65%) were female.

Because the patients in the comparison group did not have the same clinical evaluation as the patients in the deep venous thrombosis group, some of the cases of cancer detected during the follow-up period might have been detected if the patients had been hospitalized. To ensure equal diagnostic ability in these two groups, all seven patients without deep venous thrombosis in whom cancer was diagnosed within the 12 months after the venous examination were excluded. The probability of cancer-free survival (P > 0.2; log-rank test) and the incidence rate ratio (0.9 [CI, 0.2 to 4.4]) were not substantially modified by such exclusion. Finally, 21.1 cases of cancer would have been expected to occur in the comparison group during the follow-up period compared with the 23 cases of cancer seen, yielding a relative risk for cancer in the comparison group of 1.09 (CI, 0.6 to 1.8) compared with the age- and sex-adjusted U.S. population.

Discussion

Sixteen of 136 patients with idiopathic deep venous thrombosis (12% [CI, 6% to 17%]) received a new diagnosis of cancer when they were hospitalized for their thrombotic episode. All patients with cancer had at least one finding suggestive of cancer on a comprehensive medical history, physical examination, complete blood count, or chest radiography. After those patients were excluded, the likelihood that cancer would subsequently be diagnosed during a 34-month median follow-up period did not differ in the group of patients with deep venous thrombosis and in the comparison group (incidence rate ratio, 1.0). Furthermore, the incidence of cancer in the comparison group was similar to that in the general population. Given these findings, a substantially more extensive cancer screening done when deep venous thrombosis is diagnosed does not appear to be warranted. We believe that decisions about further testing—invasive procedures, such as bronchoscopy, or expensive tests, such as computed tomography—should be guided by the abnormalities detected by this clinical, radiologic, and laboratory evaluation. This evaluation should probably include the examination of stool specimens for occult blood [20], but our own data do not show this. Screening mammography should be done in women in the appropriate age range who have not had mammography in the past year. The addition of screening blood chemistry tests and urinalysis is supported by general clinical practice and is our preference, but data for or against this recommendation are lacking. In our small cohort, urinalysis and screening blood chemistry tests, such as those for alkaline phosphatase levels and albumin concentrations, did not add information to that provided by the four components; they might, however, be helpful as screening tests in other patients and are certainly recommended as a way to search for specific types of cancer in patients who had other findings suggestive of cancer.

The potential benefit of the early detection of cancer in patients with deep venous thrombosis is difficult to assess [21, 22], but early detection is probably worthwhile. In our cohort, as many as 4 of the 16 patients in whom cancer was newly diagnosed when deep venous thrombosis was diagnosed might be labeled as having had early-stage cancer: a 74-year-old man with transitional-cell carcinoma of the bladder (stage I), a 48-year-old woman with sarcoma of the uterine stroma (stage I), a 74-year-old man with prostate cancer (Gleason 3), and possibly a 70-year-old man with colon cancer (Dukes stage B2).

Our results extend the findings of two studies that partially addressed our two objectives. One was a recent Scandinavian study [7], which showed that extensive screening for cancer might be not required in patients with deep venous thrombosis; the other was a cohort study [10], which showed that risk for subsequent cancer is not greatly increased in patients with deep venous thrombosis who are apparently free of cancer when deep venous thrombosis is diagnosed. Our data support the current pragmatic recommendations [2, 19] to look for cancer only in patients who have signs or symptoms suggestive of cancer. The proportion of both prevalent and incident cancers in our group of patients with deep venous thrombosis (13%) was within the range reported in the literature for patients with venous disorder [2-10, 21]. In patients with deep venous thrombosis, we also confirmed 1) the association of underlying cancer with age [3, 6] and decreased hemoglobin concentration [3] and 2) the importance of doing chest radiography [5].

Our results partly contradict those of some previous studies, in which 1) routine clinical examination was unable to detect underlying cancer in patients with deep venous thrombosis or 2) the incidence of cancer during follow-up was higher in the group with deep venous thrombosis than in the comparison group [5-7]. Differences in study samples, secular trends, and statistical methods might explain this discrepancy. First, the comparison groups in those studies comprised either patients with secondary deep venous thrombosis or patients in whom idiopathic or secondary deep venous thrombosis was suspected but not confirmed. Our comparison group was limited to patients in whom idiopathic deep venous thrombosis had been clinically suspected but was ruled out. Such a comparison group seems appropriate because neither leg pain nor a swollen leg, the most frequent indication for venous examination, could be viewed as a sign or symptom of underlying cancer. Second, our study cohort was assembled between 1990 and 1993, after well-designed studies reported the association between some occult cancers and deep venous thrombosis [3, 4] and during the period when subsequent reports confirmed such an association [5-7]. Those reports and the recommendations in the medical literature stating that physicians should search for underlying cancer in patients with idiopathic deep venous thrombosis might have led the physicians at our teaching hospital to be particularly concerned about the possibility of cancer in patients with deep venous thrombosis and to report more findings suggestive of cancer. This high index of suspicion for cancer might explain the discrepancy between our results and previous findings [6]. Finally, previous studies did not adjust for tobacco use (a major risk factor for cancer and a potential confounder) in comparisons of cancer incidence in patients with deep venous thrombosis with that in the comparison groups.

The high sensitivity of clinical findings for underlying cancer (89% to 100%) was also associated with low specificity (for example, only 63% of the patients with deep venous thrombosis and no cancer had normal hemoglobin values), resulting in poor test efficiency and many unnecessary work-ups. However, high sensitivity is essential for an effective “screening test” (which identifies patients who deserve further testing); high specificity, which avoids the misclassification of patients without disease, is less necessary.

Venous ultrasonography accurately diagnoses symptomatic proximal deep venous thrombosis, but its reliability for the diagnosis of distal deep venous thrombosis is uncertain [23]. Few studies [24, 25] have evaluated the accuracy of infrapopliteal venous ultrasonography exclusively in symptomatic patients. Our group recently reported that compression ultrasonography might be reliable for the evaluation of patients in whom symptomatic distal deep venous thrombosis is suspected [26]. However, we cannot exclude the possibility that some of the 844 patients who were classified as not having deep venous thrombosis actually had venous thrombosis, particularly those who were subsequently found to have cancer.

We planned an 18-month follow-up period for most study participants (>90% of the cohort was followed for 18 months) because previous cohort studies [5-7] have shown that most incident cancers are diagnosed within the first year after the diagnosis of a first episode of deep venous thrombosis. Because the patients in our comparison group were less likely to be hospitalized and did not have the same clinical evaluation as the patients in our deep venous thrombosis group, some of their subsequent cases of cancer may have been diagnosable at the original presentation (this would be an example of work-up bias), especially if the cancer was subsequently diagnosed within several months. However, even after we excluded the patients without deep venous thrombosis in whom cancer was diagnosed within 12 months after the venous examination, no statistically significant difference in cancer incidence was seen between the two groups: Three of 122 patients in the deep venous thrombosis group (2.5%) and 16 of 844 patients in the comparison group (1.9%) had cancer. Finally, by assessing the incidence of cancer after excluding patients whose cancer was newly diagnosed when deep venous thrombosis occurred, we avoided a falsely high estimate of cancer incidence during the follow-up period [21]. To eliminate selection bias (the selection of a group of patients who have an increased or decreased risk for cancer because of a factor other than thrombosis), we selected the comparison group as we selected the deep venous thrombosis group—because of clinical suspicion of idiopathic deep venous thrombosis. In clinical practice, the search for underlying cancer in patients with risk factors (other than cancer) for venous disorders is a different clinical issue because the presence of such a factor such as prolonged immobilization or recent surgery reduces the likelihood that a seriously unsuspected tumor is the cause of deep venous thrombosis.

Our study was done at one hospital, which limits the generalizability of our results. The retrospective collection of data from the clinical evaluation reflected routine care. Although the use of routine care data might be considered a strength, it was also a weakness because of missing data. We could not reliably assess the yield of some potential tools for early detection of cancer, such as erythrocyte sedimentation rate, urinalysis, breast examination, or digital rectal examination, because these tests were 1) done after the first clinical evaluation or 2) too often not recorded. Knowledge of laboratory results can sometimes affect the observation or recording of clinical findings. We minimized this potential ascertainment bias in two ways: by collecting data from the medical record only for the first clinical examination and then by dividing the clinical evaluation into four components of equal weight. The retrospective design of our study also means that the intensity and scrutiny with which the patients were investigated at the time of admission might have varied; our results should therefore be confirmed through prospective studies. Finally, although our study was relatively large, we encountered a limited number of cases of cancer; thus, we cannot ensure that every case of cancer will be detected by the screening methods we propose.

In summary, our study suggests that clinical evaluation, including a comprehensive medical history, physical examination, routine laboratory tests, and chest radiography, is appropriate in screening for cancer in these patients. Independent of our data, other data and consensus guidelines suggest that testing of stool specimens for occult blood, mammography, urinalysis, and probably routine blood chemistry tests should also be part of this initial screening evaluation. Additional testing should focus on abnormalities that are suggested by this initial clinical evaluation. Although patients with idiopathic deep venous thrombosis should be carefully followed after their initial evaluation, our data suggest that the incidence of cancer in these patients during the first several years of follow-up after the acute thrombotic episode is no different from that of the rest of the population. Follow-up in these patients should probably also focus on clinical signs and symptoms and routine screening for cancer rather than on the repetitive use of other tests.

Presented at the joint session SGIM/AFCR at the Annual Meeting of the Society of General Internal Medicine, San Diego, California, May 1995.

Dr. Pearson: Department of Ambulatory Care and Prevention, Harvard Community Health Plan, 126 Brookline Avenue, Boston, MA 02215.

Dr. Cook: Section for Clinical Epidemiology, Division of General Medicine, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Dr. Creager: Vascular Diagnostic Laboratory, Cardiovascular Division, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Dr. Goldman: University of California, San Francisco, Department of Medicine, 505 Parnassus Avenue, San Francisco, CA 94143-0120.