Accuracy of Ultrasound for the Diagnosis of Deep Venous Thrombosis ...

REVIEW

Accuracy of Ultrasound for the Diagnosis of Deep Venous Thrombosis in Asymptomatic Patients after Orthopedic Surgery A Meta-Analysis Philip S. Wells, MD; Anthonie W.A. Lensing, MD, PhD; Bruce L. Davidson, MD, MPH; Martin H. Prins, MD, PhD; and Jack Hirsh, MD

• Objective: To evaluate, by meta-analysis, the accuracy of ultrasound screening for deep venous thrombosis in patients after orthopedic surgery. • Data Sources: The MEDLINE database from January 1982 to October 1993. Bibliographies of retrieved articles and recent journal publications were searched independently and using Current Contents. • Study Selection: All articles evaluating the use of venous ultrasound imaging (B-mode, duplex, and color Doppler) compared with standard contrast venography for detecting deep venous thrombosis. We excluded abstracts, early reports of studies later reported in full, and studies in which venography was not done in all patients. Seventeen of 30 identified studies were eligible. • Data Extraction: Eligible articles were reviewed for the presence of three key criteria necessary for evaluating the accuracy of the diagnostic tests: 1) previously established objective criteria for venography and ultrasound, 2) independent blinded comparisons of venography and ultrasound, and 3) prospective evaluations of consecutive patients. Studies including all three key criteria were defined as level 1 (minimized bias) studies; otherwise, they were defined as level 2 studies. • Data Synthesis: In level 1 studies, ultrasonography had a sensitivity of 62% (95 of 153; 95% CI, 54% to 70%), a specificity of 97% (CI, 96% to 98%), and a positive predictive value of 66% (95 of 144; CI, 58% to 74%) for detecting proximal thrombi. For level 2 studies, the sensitivity was 95% (CI, 87% to 99%), the specificity was 100% (CI, 99% to 100%), and the positive predictive value was 100% (CI, 94% to 100%). Differences between level 1 and level 2 studies appeared to toe related to bias in study design. • Conclusions: Venous ultrasound imaging has only moderate sensitivity and a moderate positive predictive value when used to screen for deep venous thrombosis in patients after orthopedic surgery; thus, ultrasound imaging may have limitations as a screening test.

Ann Intern Med. 1995;122:47-53. From Ottawa Civic Hospital, Ottawa, Ontario, Canada; University of Amsterdam, Amsterdam, the Netherlands; Hahnemann University, Philadelphia, Pennsylvania; and Hamilton Civic Hospitals Research Centre, Hamilton, Ontario, Canada. For current author addresses, see end of text.

x atients who have major orthopedic operations have an increased risk for deep venous thrombosis; without prophylaxis, the incidence of deep venous thrombosis is about 50% after hip replacement and about 65% after major knee surgery (1). In both of these groups, the incidence of the more dangerous proximal venous thrombosis (thrombosis in the popliteal or more proximal veins) is approximately 20%. Most of these thrombi are asymptomatic. Nevertheless, the risk for pulmonary embolism from asymptomatic proximal venous thrombosis is substantial, about 25%, and fatal pulmonary embolism occurs in 1% to 2% of this group (2). Several effective prophylactic methods are available. However, even with the most effective methods, the incidence of postoperative thrombosis is 15% to 20% for elective hip surgery and the incidence of thrombosis is 20% to 30% for major knee surgery detected by routine venography done at the time of discharge from hospital (1). Because of this relatively high incidence of thrombosis despite primary prophylaxis, some authorities (3) advocate routine venography before hospital discharge in addition to primary prophylaxis to detect silent deep venous thrombosis in patients who have major orthopedic procedures. Thrombi that are detected are usually treated with anticoagulant agents. Venography is expensive, can be painful, and can produce other side effects (4); it is therefore not an ideal screening test. Radioactive fibrinogen leg scanning and impedance plethysmography have been used as screening tests but are much less sensitive than venography (5-7). More recently, venous ultrasound imaging has been evaluated as a screening test after hip surgery and has been recommended as a substitute for venography (8). The initial studies with venous ultrasound used realtime B-mode imaging and lack of venous compressibility with gentle probe pressure as the diagnostic criterion for venous thrombosis (9-11). Subsequently, a Doppler component (duplex) and then a color Doppler component were added as adjuncts to the original B-mode imaging. These modifications facilitate the identification of veins, but the definitive diagnostic criterion with both of these newer techniques is generally considered to be noncompressibility of the vein under gentle probe pressure. Studies (9-17) in symptomatic patients have consistently shown a high sensitivity and specificity (97% and 97%, respectively) for all three methods of ultrasound imaging. In contrast, studies evaluating the sensitivity of venous ultrasound imaging for detecting thrombi in asymptomatic © 1995 American College of Physicians

Downloaded From: http://annals.org/ by a McMaster University User on 05/17/2013

47

Table 1. Summary of Trials Included in the Meta-analysis* Criteria Satisfiedf

Mean Patient Age,y

Type of Surgery

Day Ultrasonography Done

DVT Prophylaxis

Prevalence of Proximal DVT, %

Level 1 studies Bonis et al. (36) Borris et al. (37) Agnelli et al. (38) Ginsberg et al. (7) Tremaine et al. (41) Froehlich et al. (42)

1,2,3 1,2,3 1,2,3 1,2,3 1,2,3 1,2,3

67 67 73 NR 66 82

TH TH TH TH,TK TH HF

NR NR SH or LMWH W or SH or LMWH NR GCS/IPC or W

40 18 11 10 3 13

Elliott et al. (43) Barnes et al. (44)

1,2,3 1,2,3

68 NR

TH, TK TH

8 to 10 8 to 10 10 10 to 14 5 to 8 Every third day and day 11 8 NR

4 6

Woolson et al. (45)

1,2,3

70

TK

8

Davidson et al. (49) Mattos et al. (50) Level 2 studies Woolson et al. (40)

1,2,3 1,2,3

NR 68

TH, TK TH, TK

10 3 to 9

W or LMWH GCS plus dextran or LDH-DHE GCS plus IPC with or without ASA W or LMWH NR

1

63

TH

8

Dorfman et al. (39)

1,3

81

HF

Comerota et al. (46) White et al. (47) Cronan et al. (48)

3 1,2 1

NR NR 81.6

TH, TK LEF HF

Every third day and day 10 6 or 7 NR Every third day and day 8

Study (Reference)

17 7 5

GCS plus IPC with or without ASA GCS/IPC or W

13

NR NR GCS, IPC, W

19 38 16

16

* ASA = aspirin; DVT = deep venous thrombosis; GCS = graduated compression stockings; HF = hip fracture; IPC = intermittent pneumatic compression; LDH-DHE = dihydroergotamine-heparin; LEF = pelvic and lower extremity fractures; LMWH = low molecular weight heparin; NR = not reported; SH = standard heparin; TH = total hip arthroplasty; TK = total knee arthroplasty; US = ultrasound; and W = warfarin. t Criteria: 1 = previously established objective criteria for venography and ultrasound; 2 = independent, blinded comparison of venography and ultrasound; and 3 = prospective evaluation of consecutive patients.

patients after surgery have produced inconsistent results, with reported sensitivities ranging from 38% to 100%. The reason for the marked differences in the sensitivity among studies evaluating venous ultrasound imaging for asymptomatic proximal venous thrombosis is uncertain. Possible explanations for the observed differences in sensitivity among studies include 1) falsely high estimates of sensitivity because of bias resulting from shortcomings in the study design, 2) falsely low estimates of accuracy because of background noise caused by inadequate technique, 3) dependence of accuracy on the type of ultrasound method used, and 4) chance. In a previous study (5) addressing the variation in the sensitivity of radioactive fibrinogen leg scanning as a screening test for postoperative deep venous thrombosis, we provided evidence that wide differences were probably caused by bias in study design. Bias can result either from inappropriate patient selection or from diagnostic suspicion bias. To avoid a biased selection of patients, a study should include consecutive patients. To avoid diagnostic suspicion bias, the diagnostic tests should be done and their results interpreted by blinded observers using validated and explicit diagnostic criteria to ensure that the results can be reproduced by other investigators (18). To critically evaluate the accuracy of ultrasound screening for deep venous thrombosis, we did a systematic overview of the literature. Studies in which the potential for bias was minimized were evaluated separately from those in which bias was not minimized. We determined the accuracy for detecting asymptomatic proximal venous thrombosis of each of the three ultrasound imaging methods (real-time B-mode, duplex, 48

and color Doppler ultrasonography) by doing a metaanalysis (combining the results of studies regardless of the modality used and then analyzing the results separately for each of the three modalities). We also examined the accuracy of ultrasound as a screening test for isolated calf venous thrombosis.

Methods The review was initiated by a computer search of the Englishlanguage medical literature using the MEDLINE database from January 1982 to October 1993. We used combinations of the medical subject headings "ultrasound," "orthopedics," "postoperative period," and "thrombophlebitis" to identify all articles that evaluated screening with venous ultrasound imaging. Bibliographies of retrieved articles were checked for any additional studies. Recent journals were searched independently and using Current Contents to find new reports that were not identified in the computer search. Early reports of data that were later published in full were excluded from the analysis. Abstracts were also excluded because it is usually not possible to completely evaluate the methods and data. Remaining articles were then critically reviewed for the presence of three key methodologic criteria for the evaluation of the accuracy of diagnostic tests. Two of the authors independently checked the articles for the following methodologic standards: 1) previous establishment of objective criteria for normal and abnormal venographic and ultrasonographic results, 2) an independent comparison of the ultrasound result with contrast venography (the reference standard for diagnosis of venous thrombosis) by investigators blinded to the other test result, and 3) the prospective evaluation of consecutive eligible patients. A study was considered to have included consecutive patients if this was explicitly mentioned in the article or if the article stated that patients were excluded only if they refused consent or were allergic to contrast medium. Reports satisfying all three methodologic standards were classi-

1 January 1995 • Annals of Internal Medicine • Volume 122 • Number 1


fied as level 1 studies; otherwise, reports were classified as level 2 studies. Sensitivity, specificity, and positive predictive values for proximal and isolated calf venous thrombosis were calculated for the studies individually and were then calculated for the results of the pooled level 1 and pooled level 2 studies. Separate analyses were done for each of the three different ultrasound modalities (real-time B-mode, duplex, or color Doppler ultrasonography). A statistical test of homogeneity was calculated for the sensitivity and specificity of the three modalities in the pooled level 1 analysis. The 95% CIs for sensitivity and specificity were calculated according to the binomial distribution, adjusting for heterogeneity among studies using a random-effects model. The calculation of the 95% CIs for positive predictive value took into account sampling error in the sensitivities and specificities (19). Likelihood ratios, and their 95% CIs, for abnormal ultrasound results were calculated for each study using the modification of adding 0.5 to each cell when zero entries occurred in the 2 X 2 table (20, 21). Accuracy data were compared between level 1 and level 2 studies by using the normal approximation to the binomial distribution and by adjusting for between-study heterogeneity with a random-effects model (22). Two-tailed P values are reported, and values of less than 0.05 were considered to be statistically significant.

color Doppler ultrasonography was evaluated in 2 reports (49, 50). The last eligible report evaluated color Doppler ultrasonography only for the detection of isolated calf venous thrombosis, and, consequently, this report is included only in that analysis (51). When level 1 and level 2 studies were combined, 2001 patients were studied. Two hundred seventeen proximal deep venous thrombi were detected by venography for an overall prevalence of no more than 10.8%. (Some studies reported deep venous thromboses by limbs only.) Table 1 shows the characteristics of the 16 analyzed studies, including documentation of the presence or absence of the three criteria necessary to minimize bias when evaluating the accuracy of diagnostic tests. In general, the timing of the ultrasound assessment, mean patient age, and prevalence of proximal venous thrombosis were similar. Studies with the highest rates of deep venous thrombosis did not provide information about whether or not prophylaxis was used or about the method of prophylaxis.

Results

Accuracy of Ultrasonography for Proximal Venous Thrombosis

We identified 30 studies, all in patients who had had orthopedic surgery. Thirteen of these studies were excluded from analysis: Two studies were excluded because venography was not done or was done only in patients with abnormal ultrasound results (23, 24), 4 studies were excluded because it was impossible to distinguish the data on asymptomatic patients from those on symptomatic patients (25-28), and 7 studies were excluded because they were abstracts or early reports of studies later reported in full (29-35). Sixteen of the remaining 17 reports evaluated proximal venous thrombi; real-time B-mode ultrasonography was evaluated in 7 reports (7, 36-41), duplex ultrasonography was evaluated in 7 reports (42-48), and

In the level 1 studies (Table 2), ultrasonography detected 95 of 153 proximal thrombi, for a sensitivity of 62% (95% CI, 54% to 70%). A falsely abnormal ultrasonographic result was found for 49 of 1463 venograms, for a specificity of 97% (CI, 96% to 98%). Accordingly, the positive predictive value was 66% (95 of 144; CI, 58% to 74%). The accuracy values, likelihood ratios, and 95% CIs for level 1 studies were also calculated separately for each of the three ultrasound modalities (Table 2). The receiver operator characteristic curve for level 1 studies is almost a vertical line (Figure 1). Thus, in contrast to the usual situation, the receiver operator characteristic curve

Table 2. Accuracy of Ultrasonography for Detecting Proximal Deep Venous Thrombosis in Asymptomatic Patients from Studies Minimizing the Potential for Bias (Level 1) Study (Reference)

Sensitivity

Specificity Of„{r,lw,\

Prevalence

Likelihood Ratios (95% CI)

%

63 (15/24) 73(8/11) 57 (12/21) 52 (11/21) 100 (2/2) 61 (49 to 72)

94 96 99 99 95

(34/36) (48/50) (165/166) (184/186) (55/58) 98 (96 to 99)

88 (15/17) 80 (8/10) 92 (12/13) 85 (11/13) 40(2/5) 83 (72 to 90)

40 18 11 10 3 13.7

11.2 (4.2 to 29.8) 18.2 (4.5 to 74.0) 94.9 (13.0 to 693.0) 48.7 (11.6 to 205.3) 19.3 (5.0 to 74.4) 30.1 (15.9 to 57.1)

100 (5/5) 100 (4/4) 79 (15/19) 67 (10/15) 79 (64 to 90)

97 92 98 99

(34/35) (85/92) (283/289) (72/73) 97 (95 to 98)

83 (5/6) 36(4/11) 71 (15/21) 91 (10/11) 69 (57 to 79)

13 4 6 17 8.1

35.0 (5.1 to 241.1) 13.1 (6.4 to 26.7) 38.0 (16.7 to 86.7) 48.7 (6.7 to 352.4) 25.7 (15.2 to 43.3)

92 (275/298) 99 (179/180) 95 (93 to 97) 97 (1414/1463)

26 (8/31) 83 (5/6) 35 (23 to 49) 66 (95/144)

7 5 6.1

4.9 (2.5 to 9.6) 89.9 (11.6 to 698.0) 8.4 (4.7 to 14.7)

9.5

18.5 (13.7 to 25.7)

38 (8/21) 50 (5/10) 42 (25 to 61) 62 (95/153)

* Only percentages are provided for the totals along with their 95% CIs. 1 January 1995 • Annals of Internal Medicine • Volume 122 • Number 1 Downloaded From: http://annals.org/ by a McMaster University User on 05/17/2013

49

thrombosis (41, 51). These studies were both level 1; real-time B-mode ultrasonography was used in one study, and color Doppler ultrasonography was used in the other. Overall, 14 of 29 isolated calf venous thrombi were detected, for a sensitivity of 48% (CI, 29% to 67%). One false-positive result was reported, yielding a positive predictive value of 93% (CI, 68% to 100%) when ultrasonography was used to detect isolated calf venous thrombosis. Discussion

Figure 1. The receiver operator characteristic curve for level 1 studies. did not suggest a tradeoff between sensitivity and specificity. The test for homogeneity among the three different ultrasound modalities in the level 1 studies was significant for both sensitivity (P = 0.008) and specificity (P = 0.003). Therefore, comparison of the sensitivity and specificity between level 1 and level 2 studies was done after adjustment for heterogeneity using a random-effects model. A significant difference was noted between level 1 and level 2 results (P < 0.001 for sensitivity, specificity, and positive predictive value). Accuracy values and likelihood ratios for level 2 studies were all considerably higher (Table 3); the sensitivity was 95% (CI, 87% to 99%), the specificity was 100% (CI, 99% to 100%), and the positive predictive value was 100% (CI, 94% to 100%). Accuracy of Ultrasonography for Isolated Calf Venous Thrombosis Most of the studies did not evaluate the deep veins of the calf. Of four studies (7, 41, 49, 51) that reported results in calf veins, only two studies addressed the accuracy of ultrasonography for detecting isolated calf venous

Even with the most effective methods of prophylaxis available, the rate of asymptomatic deep venous thrombosis after major orthopedic surgery is substantial, with proximal deep venous thrombosis occurring in 5% to 12% of patients (52). Although most of these proximal venous thrombi are asymptomatic, they are considered to be the source of most clinically important pulmonary emboli (1), and if these thrombi are detected at routine venography, they are usually treated with anticoagulant agents. On the basis of the results of our pooled analysis of level 1 studies, ultrasonography is only moderately sensitive for detecting proximal deep venous thrombosis, and ultrasonography would fail to detect proximal venous thrombi in 4 of every 10 patients. In addition, even though specificity is high, many patients who receive prophylaxis against thrombi formation after orthopedic surgery and who have an abnormal ultrasound will not have deep venous thrombosis because with effective primary prophylaxis, the prevalence of thrombosis is relatively low. Thus, on the basis of the specificity reported in the level 1 studies and of the observed summary likelihood ratio (18.5), the probability of proximal deep venous thrombosis occurring with an abnormal ultrasound would be only 50% if the prevalence of thrombosis was 5%. The probability of proximal deep venous thrombosis with an abnormal ultrasound would be 80% if the prevalence of thrombosis was 20%. This finding suggests that when ultrasonography is used to screen patients after major orthopedic surgery, it would be prudent to confirm an abnormal result with venography, particularly if effective prophylaxis is used. Further, our analysis indicates that insufficient data are available to determine whether ultrasonography is a reliable screening method for detecting isolated calf venous thrombi. It is unclear whether our conclusions apply to other patients who have different types of surgery. We found only one study (53) done in asymptomatic patients after

Table 3. Accuracy of Ultrasonography for Detecting Proximal Deep Venous Thrombosis in Asymptomatic Patients from Studies not Minimizing the Potential for Bias (Level 2) Study (Reference)

Sensitivity

Specificity

Positive Predictive Value

Prevalence

/&./„/„ \

%

/u{n/nj

Real-time B-mode Woolson et al. (40) Dorfman et al. (39) Duplex ultrasound Comerota et al. (46) White et al. (47) Cronan et al. (48) Total (all studies)

50

Likelihood Ratios (95% CI)

89 (17/19) 100 (14/14)

100 (133/133) 100 (75/75)

100 (17/17) 100 (14/14)

12.5 15.7

234.5 (201 to 273.5) 146.9 (20.7 to 1042.9)

100 (7/7) 92(11/12) 100 (12/12) 95 (61/64)

100 (29/29) 100 (20/20) 100 (64/64) 100 (321/321)

100 (7/7) 100(11/11) 100 (12/12) 100 (61/61)

19.4 37.5 15.8 16.6

56.2 (7.9 to 399.0) 37.0 (31.2 to 43.9) 125.0 (17.6 to 887.4) 609.3 (577.3 to 643.1)

1 January 1995 • Annals of Internal Medicine • Volume 122 • Number 1


nonorthopedic surgery. This study was done in patients who had had craniotomy, and the findings were reported after we did our meta-analysis; therefore, the study was not included in our analysis. However, this article (53), which is a level 1 study using our criteria, had similar results when compared with our pooled level 1 data; the sensitivity was 38%, and the positive predictive value was 56%. Our analysis suggests that differences in the reported accuracy of ultrasonography are largely due to differences in study methods. The sensitivity, specificity, and positive predictive value of ultrasound for proximal deep vein thrombosis are clinically and statistically different for studies that minimized the potential for bias (level 1 studies) when compared with studies that did not minimize bias (level 2 studies). These differences are also seen in summary likelihood ratios that were markedly discrepant between level 1 and level 2 studies. Although we did not do a blinded evaluation of the studies and bias might therefore exist in our interpretation of the data, the potential for bias is at least partially overcome because we did two independent evaluations of all studies. A similar difference in sensitivity between level 1 and level 2 studies was observed in a pooled analysis of studies comparing 125I-fibrinogen leg scanning and venography for detecting asymptomatic deep venous thrombosis, thereby supporting our conclusion that bias in the study design accounts for observed differences (5). Our decision to divide studies into level 1 and level 2 was based on well-established methodologic criteria for the evaluation of a new diagnostic test (18, 54). These criteria are selected to minimize the possibility of bias in the evaluation of the new test. Diagnostic suspicion bias is avoided by ensuring that the experimental test and the reference standard are evaluated independently by observers who have no knowledge of the other test result. Explicit objective criteria should be previously established to optimize the accuracy and reproducibility of the results. Selection bias is minimized by ensuring that the evaluation is prospective and involves consecutive patients; if enrollment of patients is not consecutive, the reported accuracy may apply to a patient population that is not representative of those usually assessed postoperatively (18). Duplex and color Doppler imaging may have theoretical advantages when compared with B-mode imaging because the Doppler component facilitates localization of vessels. However, results from pooled level 1 studies were compared with results from pooled level 2 studies because only a few studies included each modality. Means of sensitivities, specificities, and positive predictive values for each of the three ultrasound modalities were similar for the level 1 studies, but means and confidence intervals for these accuracy indices differed among individual studies; as a result, the test for homogeneity was statistically significant. Because of this heterogeneity, we made an adjustment, using a random-effects model, in our comparison of level 1 with level 2 studies, and the difference in sensitivity and specificity was still statistically significant. Thus, the heterogeneity of level 1 studies does not account for their lower sensitivity and specificity. Although the interpretation of ultrasonography may be operator dependent and it

is possible that the heterogeneity may reflect this, it is not likely that level 1 studies would have been done by less experienced technicians than those doing level 2 studies. In addition, the only study (11) assessing the reproducibility of diagnostic ultrasound showed 100% agreement. Thus, operator dependence is unlikely to be the cause of the differences between level 1 and level 2 studies. In addition, results of our analysis are unlikely to be related to patient characteristics because the age of the patient, the type of orthopedic surgery, and the day ultrasound was done were similar in level 1 and level 2 studies (Table 1). The variable prevalence of deep venous thrombosis was largely due to differences among the studies in the methods of prophylaxis used. In general, if the types of surgical procedure and methods of prophylaxis are considered, rates of deep venous thrombosis are similar in level 1 and level 2 studies. The practical implications of our study are uncertain. It is clear that venous ultrasonography is less sensitive when used as a screening test in asymptomatic patients after orthopedic surgery than has been reported (55) in symptomatic patients with suspected venous thrombosis. Similar differences between sensitivities of impedance plethysmography in symptomatic and asymptomatic patients were observed for the diagnosis of proximal venous thrombosis (6, 7, 56-58). It is likely that the lower sensitivity of both of these noninvasive tests for detecting proximal venous thrombosis in patients after orthopedic surgery occurs because many thrombi are smaller and more often nonocclusive in asymptomatic patients than are thrombi in symptomatic patients (59, 60). It is possible that some of the small thrombi that are not detected by venous ultrasonography done at the time of hospital discharge will grow and produce clinically important complications. However, it is more likely that most of the small thrombi missed by venous ultrasonography will not become symptomatic. Thus, the decreased sensitivity of ultrasonography in asymptomatic patients may not be an important limitation, and screening high-risk patients with ultrasonography would be both clinically effective and cost-effective. Resolution of this issue related to the course of asymptomatic thrombi that escape detection by ultrasonography requires long-term follow-up studies of patients who have received effective primary prophylaxis after hip replacement and who have a negative venous ultrasound test at discharge. Such studies are currently being done. In the meantime, physicians using venous ultrasound imaging to screen patients after orthopedic surgery should be aware of the potential limitations of this modality. Acknowledgments: The authors thank Janette Hatten and Diane Cullinane for expert technical assistance. Grant Support: Dr. Wells was a recipient of a McLaughlin Scholarship from the University of Ottawa at the time of this study. Dr. Hirsh is a Distinguished Professor of the Heart and Stroke Foundation of Canada, and he is a Trillium Award recipient from the Ministry of Health. Requests for Reprints: Philip Wells, MD, Civic Parkdale Clinic, Room 455 Fourth Floor, 737 Parkdale Avenue, Ottawa, Ontario K1Y 1J8, Canada. Current Author Addresses: Dr. Wells: Ottawa Civic Hospital, Fourth Floor, Civic Parkdale Clinic, Ottawa, Ontario K1Y 1J8, Canada. Dr. Lensing: Centre for Haemostasis, Thrombosis, Atherosclerosis, and Inflammation Research, University of Amsterdam, Academic Medical Centre, Meibergdreefg 1105AZ, Amsterdam, the Netherlands.

1 January 1995 • Annals of Internal Medicine • Volume 122 • Number 1 Downloaded From: http://annals.org/ by a McMaster University User on 05/17/2013

51

Dr. Davidson: Division of Allergy, Pulmonary-Critical Care Medicine Division, MS 107, Hahnemann University, Broad and Vine Streets, Philadelphia, PA 19102. Dr. Prins: Department of Clinical Epidemiology and Biostatistics, University of Amsterdam, Academic Medical Centre, Meibergdreefg 1105AZ, Amsterdam, the Netherlands. Dr. Hirsh: Hamilton Civic Hospitals Research Centre, Henderson General Division, 711 Concession Street, Hamilton, Ontario L8V 1C3, Canada.

References 1. Claggett GP, Anderson FA, Levine MN, Salzman EW, Wheeler HB. Prevention of venous thromboembolism. In: Dalen JE, Hirsh J, eds. Third ACCP consensus conference on antithrombotic therapy. Chest. 1992;102:391S-407S. 2. Nurmohamed MT, Rosendaal FR, Duller HR, Dekker E, Hommes DW, Vandenbroucke JP, et al. Low-molecular-weight heparin versus standard heparin in general and orthopedic surgery: a meta-analysis. Lancet. 1992;340:152-6. 3. Hull R, Raskob G, Pineo G, Rosenbloom D, Evans W, Mallory T, et al. A comparison of subcutaneous low-molecular-weight heparin with warfarin sodium for prophylaxis against deep-vein thrombosis after hip or knee implantation. N Engl J Med. 1993;329; 1370-6. 4. Bettmann MA, Robbins A, Braun SD, Wetzner S, Dunnick NR, Finkelstein J. Contrast venography of the leg: diagnostic efficacy, tolerance, and complication rates with ionic and non-ionic contrast media. Radiology. 1987;165:113-6. 5. Lensing AW, Hirsh J. 125I-fibrinogen leg scanning: reassessment of its role for the diagnosis of venous thrombosis in post-operative patients. Thromb Haemost. 1993;69:2-7. 6. Cruickshank MK, Levine MN, Hirsh J, Turpie AG, Powers P, Jay R, et al. An evaluation of impedance plethysmography and 125I-fibrinogen leg scanning in patients following hip surgery. Thromb Haemost. 1989; 62:830-4. 7. Ginsberg JS, Caco CC, Brill-Edwards PA, Panju AA, Bona R, Demers CM, et al. Venous thrombosis in patients who have undergone major hip or knee surgery: detection with compression US and impedance plethysmography. Radiology. 1991;181:651-4. 8. Nicolaides A, Kalodiki E. Duplex scanning in post-operative patients. Haemostasis. 1993;23(Suppl l):72-9. 9. Baxter GM, McKechnie S, Duffy P. Colour Doppler ultrasound in deep venous thrombosis: a comparison with venography. Clin Radiol. 1990;42:32-6. 10. Dauzat MM, Laroche JP, Charms C, Blin B, Domingo-Faye MM, Sainte-Luce P, et al. Real-time B-mode ultrasonography for better specificity in the noninvasive diagnosis of deep venous thrombosis. J Ultrasound Med. 1986;5:625-31. 11. Lensing AW, Prandoni P, Brandjes D, Huisman PM, Vigo M, Tomasella G, et al. Detection of deep-vein thrombosis by real-time B-mode ultrasonography. N Engl J Med. 1989;320:342-5. 12. Cronan JJ, Dorfman GS, Scola FH, Schepps B, Alexander J. Deep venous thrombosis: US assessment using vein compressibility. Radiology. 1987;162:191-4. 13. Habscheid W, Hohmann M, Wilhelm T, Epping J. Real-time ultrasound in the diagnosis of acute deep venous thrombosis of the lower extremity. Angiology. 1990;599-608. 14. Vogel P, Laing FC, Jeffrey RB Jr, Wing VW. Deep venous thrombosis of the lower extremity: US evaluation. Radiology. 1987;163:747-51. 15. Elias A, Le Corff G, Bouvier JL, Benichou M, Serradimigni A. Value of real-time B-mode ultrasound imaging in the diagnosis of deep vein thrombosis of the lower limbs. Int Angiol. 1987;6:175-82. 16. Rose SC, Zwiebel WJ, Nelson BD, Priest DL, Knighton RA, Brown JW, et al. Symptomatic lower extremity deep venous thrombosis: accuracy, limitations, and role of color duplex flow imaging in diagnosis. Radiology. 1990;175:639-44. 17. Mitchell DC, Grasty MS, Stebbings WS, Nockler IB, Lewars MD, Levison RA, et al. Comparison of duplex ultrasonography and venography in the diagnosis of deep venous thrombosis. Br J Surg. 1991; 78:611-3. 18. Ransohoff DF, Feinstein AR. Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. N Engl J Med. 1978;299:926-30. 19. Heckerling PS. Confidence in diagnostic testing. J Gen Intern Med. 1988;3:604-6. 20. Anscombe FJ. On estimating binomial response relations. Biometrika. 1956;43:461-4. 21. Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763-70. 22. Donner A, Donald A. The statistical analysis of repeated binary measurements. J Clin Epidemiol. 1988;41:899-905. 23. Flinn WR, Sandager GP, Cerullo LJ, Havey RJ, Yao JS. Duplex venous scanning for the prospective surveillance of perioperative venous thrombosis. Arch Surg. 1989;124:901-5. 24. Kraay MJ, Goldberg VM, Herbener TE. Vascular ultrasonography for deep venous thrombosis after total knee arthroplasty. Clin Orthop. 1993;Jan(286):18-26. 52

1 January 1995 • Annals of Internal Medicine

25. Mussurakis S, Papaioannou S, Voros D, Vrakatselis T. Compression ultrasonography as a reliable imaging monitor in deep venous thrombosis. Surg Gynecol Obstet. 1990;171:233-9. 26. Nix ML, Nelson CL, Harmon BH, Ferris EF, Barnes RW. Duplex venous scanning: image vs Doppler accuracy. J Vase Tech. 1989;8: 121-4. 27. Monreal M, Montserrat E, Salvador R, Bechini J, Donoso L, MaCallejas J, et al. Real-time ultrasound for diagnosis of symptomatic venous thrombosis and for screening of patients at risk: correlation with ascending conventional venography. Angiology. 1989;40:527-33. 28. Kerr TM, Cranley JJ, Johnson JR, Lutter KS, Riechmann GC, Cranley RD, et al. Analysis of 1084 consecutive lower extremities involved with acute venous thrombosis diagnosed by duplex scanning. Surgery. 1990;108:520-7. 29. Barnes RW, Nix ML, Barnes CL, Lavender RC, Golden WE, Harmon BH, et al. Perioperative asymptomatic venous thrombosis: rote of duplex scanning versus venography. J Vase Surg. 1989;9:251-60. 30. Agnelli G, Volpato R, Radicchia S, Rinonapoli E, Lupattelli, Nenci GG. Accuracy of real time B-mode ultrasonography in the diagnosis of asymptomatic deep vein thrombosis in hip surgery patients. Thromb Haemost. 1991;65(s):1175. 31. White RH, Daschbach MM, Bray TJ, Goulet JA, Hartling RP. Deep vein thrombosis following fracture of the pelvis [Abstract]. Clin Research. 1988;36:751A. 32. Piovella F, Carbone S, Siragusa S, Vicentini L, Ghidini M, Passera R, et al. B-mode serial compression ultrasonography in the early diagnosis of symptomatic proximal deep vein thrombosis after elective hip surgery. Thromb Haemost. 1993;69(s);1829. 33. Comerota AJ, Katz ML, Kerr R. The diagnosis of symptomatic and asymptomatic deep vein thrombosis with venous duplex imaging. Thromb Haemost. 1993;69(s):1813. 34. Wille-Jorgensen P, Lausen I,.Jorgensen LN, Jensen R, Rasmussen MS, Raaschou HO. Color doppler flow imaging as screening method for deep vein thrombosis following major general surgery. TMromb Haemost. 1993;69(s):279. 35. Lensing AW, Doris I, McGrath F, Prandoni P, Ginsberg J, Turpie AGG, et al. Color doppler versus real-time ultrasonography in the diagnosis of postoperative asymptomatic deep vein thrombosis. Thromb Haemost. 1993;69(s):2941. 36. Bonis LC, Christiansen HM, Lassen MR, Olsen AD, Schott P. Comparison of real-time B-mode ultrasonography and bilateral ascending phlebography for detection of postoperative deep vein thrombosis following elective hip surgery. The Venous Thrombosis Group. Thromb Haemost. 1989;61:363-5. 37. Bonis LC, Christiansen HM, Lassen MR, Olsen AD, Schott P. Realtime B-mode ultrasonography in the diagnosis of postoperative deep vein thrombosis in non-symptomatic high-risk patients. Eur J Vase Surg. 1990;4:473-5. 38. Agnelli G, Volpato R, Radicchia S, Veschi F, Di Filippo P, Lupattelli L, et al. Detection of asymptomatic deep vein thrombosis by real-time B-mode compression ultrasound in hip surgery patients. Thromb Haemost. 1992;68:257-60. 39. Dorfman GS, Froehlich JA, Cronan JJ, Urbanek PJ, Herndon JH. Lower-extremity venous thrombosis in patients with acute hip fractures: determination of anatomic location and time of onset with compression sonography. AJR Am J Roentgenol. 1990;154:851-5. 40. Woolson ST, McCrory DW, Walter JF, Mahoney WJ, Watt MJ, Cahill P. B-mode ultrasound scanning in the detection of proximal venous thrombosis after total hip replacement. J Bone Joint Surg. 1990*,72: 983-7. 41. Tremaine MD, Choroszy CJ, Gordon GH, Menking SA. Diagnosis of deep venous thrombosis by compression ultrasound in knee arthroplasty patients. J Arthroplasty. 1992;7:187-92. 42. Froehlich JA, Dorfman GS, Cronan JJ, Urbanek PJ, Herndon JH, Aaron RK. Compression ultrasonography for the detection of deep venous thrombosis in patients who have a fracture of the hip. A prospective study. J Bone Joint Surg. 1989;71:249-56. 43. Elliott CG, Suchyta M, Rose SC, Talbot S, Ford C, Raskob G, et al. Duplex ultrasonography for the detection of deep vein thrombi after total hip or knee arthroplasty. Angiology. 1993;44:26-33. 44. Barnes CL, Nelson CL, Nix ML, McCowan TC, Lavender RC, Barnes RW. Duplex scanning versus venography as a screening examination in total hip arthroplasty patients. Clin Orthop. 1990;271:180-9. 45. Woolson ST, Pottorff G. Venous ultrasonography in the detection of proximal vein thrombosis after total knee arthroplasty. Clin Orthop. 1991;273:131-5. 46. Comerota AJ, Katz ML, Greenwald LL, Leefmans E, Czeredarczuk M, White JV. Venous duplex imaging: should it replace hemodynamic tests for deep venous thrombosis? J Vase Surg. 1990;11:53-61. 47. White RH, Goulet JA, Bray TJ, Daschbach MM, McGahan JP, Hartling RP. Deep-vein thrombosis after fracture of the pelvis: assessment with serial duplex-ultrasound screening. J Bone Joint Surg. 1990;72: 495-500. 48. Cronan JJ, Froehlich JA, Dorfman GS. Image-directed Doppler ultrasound: a screening technique for patients at high risk to develop deep vein thrombosis. J Clin Ultrasound. 1991;19:133-8.

• Volume 122 • Number 1


49. Davidson BL, Elliott GC, Lensing AW. Low accuracy of color Doppler ultrasound to detect proximal leg vein thrombosis during screening of asymptomatic high-risk patients. Ann Intern Med. 1992;117:735-8. 50. Mattos MA, Londrey GL, Leutz DW, Hodgson KJ, Ramsey DE, Barkmeier LD, et al. Color-flow duplex scanning for the surveillance and diagnosis of acute deep venous thrombosis. J Vase Surg. 1992;15:36676. 51. Rose SC, Zwiebel WJ, Murdock LE, Hofmann AA, Priest DL, Knighton RA, et al. Insensitivity of color Doppler flow imaging for detection of acute calf deep venous thrombosis in asymptomatic postoperative patients. J Vase Interv Radiol. 1993;4:111-7. 52. Anderson DR, O'Brien BJ, Levine MN, Roberts R, Wells PS, Hirsh J. Efficacy and cost of low-molecular-weight heparin compared with standard heparin for the prevention of deep vein thrombosis after total hip arthroplasty. Ann Intern Med. 1993;119;1105-12. 53. Jongbloets LM, Lensing AW, Koopman MM, Buller HR, ten Cate JW. Limitations of compression ultrasound for the detection of symptomless postoperative deep vein thrombosis. Lancet. 1994;343:1142-4. 54. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. The interpretation of diagnostic data. In: Clinical Epidemiology: A Basic Science for Clinical Medicine. Boston/Toronto: Little, Brown and Company; 1991.

55. White RH, McGahan JP, Daschbach MM, Hartling RP. Diagnosis of deep-vein thrombosis using duplex ultrasound. Ann Intern Med. 1989; 111:297-304. 56. Hull R, van Aken WG, Hirsh J, Gallus AS, Hoicka G, Turpie AG, et al. Impedance plethysmography using the occlusive cuff technique in the diagnosis of venous thrombosis. Circulation. 1976;53:696-700. 57. Peters SH, Jonker JJ, de Boer AC, den Ottolander GJ. Home-diagnosis of deep venous thrombosis with impedance plethysmography. Thromb Haemost. 1982;48:297-300. 58. Agnelli G, Cosmi B, Ranucci V, Renga C, Mosca S, Lupattelli L, et al. Impedance plethysmography in the diagnosis of asymptomatic deep vein thrombosis in hip surgery. A venography-controlled study. Arch Intern Med. 1991;151:2167-71. 59. Cogo A, Lensing AW, Prandoni P, Hirsh J. Distribution of thrombosis in patients with symptomatic deep vein thrombosis. Implications for simplifying the diagnostic process with compression ultrasound. Arch Intern Med. 1993;153:2777-80. 60. Wells PS, Lensing AW, Anderson DR, Hirsh J. The distribution of deep vein thrombosis in asymptomatic post-operative patients and the implications for screening. Blood. 1992;80:(s):316.

The hurried scribbles in our laboratory notebooks are intelligible only to ourselves, and the lecture and the seminar have only a temporary and narrow influence. The pubic record, however, is permanent; it is there for all time for a source of pride or shame as the case may be. Hubert Bradford Vickery Journal of Biological Chemistry 1958;233:1249-1250 Submitted by: Saul Rosen, MD National Institutes of Health Bethesda, MD 20892

Submissions from readers are welcomed. If the quotation is published, the sender's name will be acknowledged. Please include a complete citation, as done for any reference.—The Editors

1 January 1995 • Annals of Internal Medicine • Volume 122 • Number 1 Downloaded From: http://annals.org/ by a McMaster University User on 05/17/2013

53