Age-adjusted versus clinical probability-adjusted D-dimer to exclude pulmonary embolism

Accepted Manuscript Age-adjusted versus clinical probability-adjusted D-dimer to exclude pulmonary embolism

Sarah Takach Lapner, Scott M. Stevens, Scott C. Woller, Gregory Snow, Clive Kearon PII: DOI: Reference:

S0049-3848(18)30334-7 doi:10.1016/j.thromres.2018.05.003 TR 7024

To appear in:

Thrombosis Research

Received date: Revised date: Accepted date:

27 March 2018 28 April 2018 4 May 2018

Please cite this article as: Sarah Takach Lapner, Scott M. Stevens, Scott C. Woller, Gregory Snow, Clive Kearon , Age-adjusted versus clinical probability-adjusted D-dimer to exclude pulmonary embolism. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Tr(2017), doi:10.1016/ j.thromres.2018.05.003

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Age-adjusted versus Clinical Probability-adjusted D-dimer to Exclude Pulmonary Embolism

Sarah Takach Lapnera, MD, Scott M. Stevens b, MD, Scott C. Woller b, MD, Gregory Snow, PhD c, Clive Kearon, MD, PhD d,e

IP

T

Abstract Word Count: 249

CR

Body Word Count: 2,239 Tables: 4

US

References: 21

PT

ED

M

AN

Author Affiliations: a) Department of Medicine, University of Alberta, 4-112 Clinical Sciences Building, 11350 83 Avenue, Edmonton, AB, Canada T6G 2R3 b) Department of Medicine, Intermountain Medical Center, Murray, UT and Department of Internal Medicine, University of Utah, Salt Lake City, UT, USA c) Statistical Data Center, Office of Research, Intermountain Healthcare, Murray, UT, USA‡ d) Department of Medicine, McMaster University, 1280 Main Street West, Hamilton, ON, Canada, L8V 4K1 e) Thrombosis and Atherosclerosis Research Institute, McMaster University, 237 Barton Street East, Hamilton, Ontario, Canada L8L 2X2

AC

CE

Correspondence: Scott M. Stevens, MD Department of Medicine, Intermountain Medical Center 5169 South Cottonwood Street, Suite 303 Murray, UT 84107, USA [email protected]

1

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE ABSTRACT Introduction: A low D-dimer can exclude suspected pulmonary embolism (PE) in cases with low or intermediate clinical probability of disease. Yet D-dimer is nonspecific, so many cases without PE require imaging. D-dimer’s specificity is improved by increasing the threshold for a positive test with age (age x 10ng/mL; age-adjusted D-dimer; AADD)

IP

T

or clinical probability of PE (1000ng/mL if low and 500ng/mL if intermediate clinical

CR

probability; clinical probability-adjusted D-dimer; CPADD). It is unclear which approach is preferable.

AADD compared to CPADD in suspected PE.

US

Objectives: We report the sensitivity, specificity and negative predictive value (NPV) of

AN

Materials and Methods: A retrospective cohort of 3500 consecutive cases imaged for

M

suspected PE at two U.S. emergency departments was assembled. We analyzed

ED

cases with low or intermediate clinical probability of PE (Revised Geneva Score) who had a D-dimer. The outcome was acute PE on imaging at presentation.

PT

Results: Of the 3500 cases, 1745 were eligible. 37% were low, and 63% were intermediate clinical probability of PE. PE was present in 145 (8.3%) cases. Sensitivity

CE

of CPADD was 87.5% vs. 96.6% for AADD (difference 9.1%; 95% CI 4.3% to 14.0%).

AC

NPV of CPADD was 97.1% vs. 99.0% for AADD (difference 1.9%; 95% CI, 0.7% to 3.1%). Specificity of CPADD was 37.5% vs. 30.2% for AADD (difference –7.3%; 95% CI –9.4% to –5.1%). D-dimer was negative in 35.4% of cases using CPADD vs. 28.0% using AADD. Conclusions: CPADD modestly improved the specificity of D-dimer, but had a lower NPV than AADD. AADD appears preferable in this analysis. KEYWORDS: Pulmonary embolism; D-dimer; sensitivity; specificity; diagnosis 2

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Introduction D-dimer is a sensitive test for PE and a negative result can be used to safely exclude PE in patients with low or intermediate clinical probability, thereby avoiding the need for imaging.[1] However, D-dimer is a non-specific test that is often positive in the absence

T

of PE.[2, 3] Traditionally, a low D-dimer threshold has been used to categorize results

IP

as positive or negative to maximize sensitivity and negative predictive value for

CR

thrombosis. However, use of a low D-dimer threshold is associated with low specificity which limits the clinical usefulness of D-dimer testing as few patients will have a

US

negative test even in the absence of PE. This results in many patients without PE being

AN

exposed to the risks and expense of imaging to refute pulmonary embolism.

M

Two methods have been proposed to increase the D-dimer threshold that is used to

ED

exclude venous thromboembolism (VTE), and thereby increase specificity, without substantially reducing the test's negative predictive value for thrombosis: i)

PT

progressively increasing the D-dimer threshold with increasing age over 50 years (age-

CE

adjusted D-dimer interpretation; AADD); and ii) using a D-dimer threshold in patients with low clinical probability that is twice as high as the threshold used in patients with

AC

intermediate clinical probability (clinical probability-adjusted D-dimer interpretation; CPADD). Both strategies have been shown to increase specificity and the proportion of patients who can have VTE excluded by D-dimer testing while maintaining high negative predictive value.[4-6] Whether one of these two D-dimer interpretation strategies is better than the other is uncertain. Superiority would be established by showing that, compared to the other, one of the strategies: i) excluded VTE in a higher

3

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE proportion of patients while having the same negative predictive value; or ii) had a higher negative predictive value while excluding VTE in the same proportion of patients.

We previously showed that, compared to using the conventional threshold of 500 ng/ml,

T

use of an AADD threshold (age x 10 ng/ml) increased by 2.6-fold the number of patients

IP

over 50 years who could have PE excluded by the combination of a negative D-dimer

CR

test and a non-high clinical probability of PE.[7] The negative predictive value, however, may have been lower with the AADD threshold (98.5% versus 100%). The data we

US

collected for those analyses also provides us with an opportunity to determine the

AN

proportion of patients who would have PE excluded using CPADD interpretation, and the negative predictive value associated with that approach. In this paper, we compare

M

the proportion of patients with negative D-dimer results and the associated negative

Methods

CE

PT

patients with suspected PE.

ED

predictive values of the AADD and CPADD approaches to D-dimer interpretation in

AC

Patient Population

Data and D-dimer results were obtained from a retrospective review of patient records, as has been previously reported.[7, 8] However, whereas our original analysis assessing AADD interpretation was confined to patients older than 50 years with suspected PE, the current analysis includes adult patients of all ages. Electronic medical records at the Intermountain Medical Center (Murray, Utah) and the LDS

4

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Hospital (Salt Lake City, Utah) were used to identify consecutive patients undergoing computed tomographic pulmonary angiogram (CTPA) for suspected PE in the emergency department. Manual review of the emergency physician reports was performed to confirm that each CTPA was ordered on the suspicion of PE. Based on

T

retrospective review of medical records, all patients had clinical probability of PE

IP

estimated using the Revised Geneva Score (RGS; Table 1).[9] All patients with low

CR

(RGS <4) or intermediate (RGS 4-10) probability of PE who had a CTPA and D-dimer testing were included in the current analysis. Patients were classified as PE positive or

US

PE negative according to whether or not PE was identified on the CTPA that was done

AN

at initial presentation to the emergency department.

The protocol was approved by the institutional review boards/ethics review committees

ED

M

of the respective institutions that participated in the study.

Calculation of Revised Geneva Score

PT

We have previously described the methods used to gather the relevant information to

CE

calculate the RGS.[7, 8] In brief, the elements of the RGS score were collected via a combination of: i) a validated computer program designed to identify previous episodes

AC

of VTE; ii) extraction of surgical data from the Operating Room Management Information System; iii) extraction of fracture and cancer diagnoses from the International Classification of Diseases, Ninth Division codes; and iv) extraction of historical features (presence of unilateral leg pain, hemoptysis) and physical findings (pain on palpation of the deep veins, unilateral edema and maximal heart rate) from manual review of emergency department physician reports. All data gathered from

5

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE electronic sources was verified during manual review of the emergency physician’s report. Extraction of clinical data was performed by individuals who were unaware that a comparison of different D-dimer interpretation strategies would take place.

T

D-dimer measurement

IP

Whether patients had a D-dimer test was subject to physician discretion. D-dimer

CR

measurement was performed using the STA-Liatest D-dimer assay (Diagnostica Stago,

AN

D-dimer interpretation strategy definitions

US

Parsippany, NJ).

D-dimer levels were categorized as positive or negative for the two interpretation

M

strategies as follows:

ED

AADD Strategy: D-dimer results were categorized as negative if the D-dimer level was less than 500 ng/mL in patients 50 years or younger or, in patients older than 50 years,

PT

if it was less than the patient’s age multiplied by 10.

CE

CPADD Strategy: D-dimer results were categorized as negative if the D-dimer level was less than 1000 ng/mL in patients with low clinical probability (RGS 0 to 3) and less than

AC

500 ng/mL in patients with intermediate clinical probability (RGS 4 to 10).

Statistical Analysis A 2x2 table was constructed for both the AADD and CPADD strategies according to whether the D-dimer result in individual patients was categorized as positive or negative, and whether the patient was categorized as PE-positive or PE-negative based

6

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE on the CTPA result. Estimates of sensitivity, specificity, negative predictive value and the proportion of all patients who had a negative D-dimer test were calculated for each strategy. The Wilson score method was used to calculate 95% confidence intervals (CI) for each diagnostic accuracy parameter. We used exact binomial testing to perform

T

pairwise comparisons of sensitivity, specificity and the proportion of all patients with

IP

negative D-dimer results between the two strategies, and we used the Agresti and Min

CR

approach to calculate the 95% CIs for the absolute differences for each comparison.[10] For the negative predictive value, the generalized score statistic method for paired data

US

proposed by Leisenring, Alonzo and Pepe was used.[11] Comparisons were considered

AN

significant if the two-sided p-values were less than 0.05. Analyses were performed

ED

M

using SPSS version 24 (IBM Armonk, New York) and R version 4.2 (Vienna, Austria).

Study population

PT

Results

CE

A total of 3500 patients had a CTPA for suspected PE, of these 3333 had either a low or an intermediate clinical probability of PE, and of these, 1745 had D-dimer measured

AC

(646 with low clinical probability, 37%; 1099 with intermediate clinical probability, 63%) and are included in this analysis. Mean age was 52, and 52.9% were over 50 years of age (Table 2). PE was present in 8.3% overall, 6.2% of patients with a low clinical probability, 9.6% of patients with an intermediate clinical probability, 9.1% of patients 51 years or older and 7.4% of patients younger than 50 years (Table 2).

7

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Comparison of the age-adjusted strategy and clinical probability-adjusted strategy Sensitivity of the CPADD strategy was 87.6% and of the AADD strategy was 96.6% (difference 8.9%; 95% CI 4.3% to 14.0%, Table 3). Negative predictive value of the

T

CPADD strategy was 97.1% and of the AADD strategy was 99.0% (difference 1.9%;

IP

95% CI, 0.7% to 3.1%). Specificity of the CPADD strategy was 37.5% and of the AADD

CR

strategy was 30.2% (difference –7.3%; 95% CI –9.4% to –5.1%). The proportion of patients with a negative D-dimer result was 35.4% with the CPADD strategy and 28.0%

AN

US

with the AADD strategy (difference –7.4%; 95% CI –9.4% to –5.3%).

The prevalence of PE when the two interpretation strategies agreed that D-dimer results

M

were positive and negative, and when the two interpretation strategies disagreed,

ED

overall and according to age and clinical probability is shown in Table 4. D-dimer was positive by the CPADD strategy and negative by the AADD strategy in 101 (5.8% of

PT

total) patients, of which 0 (0%) had PE. D-dimer was positive by the AADD strategy and

CE

negative by the CPADD strategy in 230 (13.2% of total) patients, of which 13 (5.7%) had PE. Compared to using the conventional D-dimer threshold of 500 ng/mL in all

AC

patients with low or intermediate clinical probability, the CPADD strategy would avoid doing a CTPA in an additional 298 patients (21% of those who have a D-dimer > 500 ng/mL and 17% of all patients) of which 5.0% had PE and the AADD strategy would avoid doing a CTPA in an additional 169 patients (11.9% of those who have a D-dimer > 500 ng/mL and 9.7% of all patients) of which 1.2% had PE. The mean age of the

8

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE additional patients who would avoid a CTPA with the CPADD strategy (298 patients) was 51.5 years and with the AADD strategy (169 patients) was 71.3 years.

T

Discussion

IP

Compared to the AADD strategy, use of the CPADD strategy would have excluded PE

CR

and avoided CTPA in an additional 7.4% (35.4% vs 28.0%) of all patients with a low or intermediate clinical probability of PE. However, this would have been at the cost of

US

reducing the negative predictive value of D-dimer testing by 1.9% (97.1% vs. 99.0%).

AN

There were 230 patients who were negative with the CPADD strategy but positive with the AADD strategy and 13 (5.7%) of these had PE. Conversely, there were 101 patients

M

who were negative with the AADD strategy but positive with the CPADD strategy and

ED

none of these had PE. It has been suggested that a diagnostic strategy for suspected PE should have a negative predictive value of at least 98% and, whereas the AADD

PT

strategy satisfied this requirement, the CPADD strategy did not.[12, 13] Therefore, our

CE

overall findings support using the AADD strategy for D-dimer interpretation in

AC

preference to the CPADD strategy.

A number of limitations of our study must be appreciated. First, our population was confined to patients who underwent CTPA and D-dimer testing so did not include patients who had PE excluded by clinical assessment and a D-dimer level less than 500 ng/ml. This study design likely selected for patients with low and intermediate probability for PE who have a higher risk of PE compared to unselected patients. Second, we used

9

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE the RGS system to assign clinical pre-test probability using retrospective chart review. We were unable to use the Wells score to calculate pretest probability as there was insufficient information in patient charts to retrospectively score the item ‘alternative diagnosis less likely than PE.’ Prospective scoring, or use of the Wells score,[14] may

T

have resulted in different clinical probability assessments and study results.[15] Third,

IP

we used a single D-dimer assay, so it is possible that other D-dimer assays may have

CR

yielded different results.[16] Fourth, this analysis compared the accuracy of the two Ddimer interpretation strategies against the findings of CT and did not test the safety of

US

using either strategy to manage patients with suspected PE. It is possible that a

AN

proportion of the patients who had negative D-dimer interpretations but PE on CT would have made a full recovery without anticoagulant therapy.[17] In particular, isolated

M

subsegmental PE may resolve without anticoagulant therapy, so the value of detecting

ED

such cases on imaging is uncertain. There were 3 cases in which PE was detected by CTPA which would not have been performed under both d-dimer strategies, and none

PT

were isolated subsegmental PE. Fifteen PE were found in patients who would not have

CE

had CTPA performed if the CPADD strategy was used, but would have undergone CTPA using the AADD strategy. Of these, 2 of 15 were isolated subsegmental PE (and

AC

both had visible associated pulmonary infarction. Therefore, it does not appear that the differential performance of the two strategies in this population was attributable to cases of isolated subsegmental PE. Finally, we included only out-patients with low or intermediate probability of PE which is expected to have reduced the overall prevalence of PE in our analysis.

10

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Our study also has strengths. First, the study was large and all 1745 cases underwent comprehensive standardized manual chart review, so we are confident that data abstraction was complete and accurate. Second, as we previously described, a sample of CTPA studies were reinterpreted by specialist chest radiologists and supported

T

accuracy of the CTPA interpretations.[8] Finally, we included all cases in which CTPA

IP

and D-dimer testing was performed for suspected PE, which makes our findings

CR

generalizable.

US

While our findings are consistent with other evaluations using an AADD interpretation in

AN

patients with suspected PE,[18] they differ from other evaluations of CPADD interpretation. A recent study found a 6% increase in the number of patients who had

M

VTE excluded using the CPADD strategy compared to the AADD strategy, with no loss

ED

in negative predictive value (99.7% vs. 99.6%).[15] That study differed from the present in that patients with suspected DVT (as well as suspected PE) were included, Wells

PT

scores were used to assess clinical probability, and clinical probability assessment

CE

occurred prospectively. Furthermore, when an AADD threshold was retrospectively analyzed in conjunction with a CPADD threshold in the YEARS study population, age-

AC

adjustment did not improve either specificity or negative predictive value.[19]

In summary, compared to AADD interpretation, CPADD interpretation improves the specificity of D-dimer testing in a selected population of patients with suspected PE with low or intermediate probability by the RGS, but was associated with an unacceptable reduction in negative predictive value in this analysis.

11

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Authorship Details S Takach Lapner participated in study design, data interpretation and drafted the manuscript; SM Stevens participated in study design, data interpretation and drafted the manuscript; SC Woller participated in study design, data interpretation and drafted the

T

manuscript; G Snow participated in study design, performed statistical analysis,

IP

participated in data interpretation and drafted the manuscript; C Kearon participated in

CR

study design, data interpretation and drafted the manuscript.

AN

US

Declarations of interest: none

Funding source: There was no external funding for this study. The original study that

M

provided data to this analysis was supported by a grant from the Intermountain

AC

CE

PT

ED

Research & Medical Foundation (Project #691).

12

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE References

AC

CE

PT

ED

M

AN

US

CR

IP

T

[1] A. Torbicki, A. Perrier, S. Konstantinides, G. Agnelli, N. Galie, P. Pruszczyk, F. Bengel, A.J. Brady, D. Ferreira, U. Janssens, W. Klepetko, E. Mayer, M. Remy-Jardin, J.P. Bassand, E.S.C.C.f.P. Guidelines, Guidelines on the diagnosis and management of acute pulmonary embolism: the Task Force for the Diagnosis and Management of Acute Pulmonary Embolism of the European Society of Cardiology (ESC), Eur Heart J 29(18) (2008) 2276-315. [2] M. Righini, C. Goehring, H. Bounameaux, A. Perrier, Effects of age on the performance of common diagnostic tests for pulmonary embolism, Am J Med 109(5) (2000) 357-61. [3] F.R. Rosendaal, V.A.N.H.V. A, C.J. Doggen, Venous thrombosis in the elderly, J Thromb Haemost 5 Suppl 1 (2007) 310-7. [4] N. van Es, T. van der Hulle, J. van Es, P.L. den Exter, R.A. Douma, R.J. Goekoop, I.C. Mos, J. Galipienzo, P.W. Kamphuisen, M.V. Huisman, F.A. Klok, H.R. Buller, P.M. Bossuyt, Wells Rule and d-Dimer Testing to Rule Out Pulmonary Embolism: A Systematic Review and Individual-Patient Data Meta-analysis, Ann Intern Med 165(4) (2016) 253-261. [5] H.J. Schouten, G.J. Geersing, H.L. Koek, N.P. Zuithoff, K.J. Janssen, R.A. Douma, J.J. van Delden, K.G. Moons, J.B. Reitsma, Diagnostic accuracy of conventional or age adjusted D-dimer cut-off values in older patients with suspected venous thromboembolism: systematic review and meta-analysis, BMJ 346 (2013) f2492. [6] L.A. Linkins, S.M. Bates, J.S. Ginsberg, C. Kearon, Use of different D-dimer levels to exclude venous thromboembolism depending on clinical pretest probability, J Thromb Haemost 2(8) (2004) 1256-60. [7] S.C. Woller, S.M. Stevens, D.M. Adams, R.S. Evans, J.F. Lloyd, G.L. Snow, J.R. Bledsoe, D.Z. Gay, R.M. Patten, V.T. Aston, C.G. Elliott, Assessment of the safety and efficiency of using an age-adjusted D-dimer threshold to exclude suspected pulmonary embolism, Chest 146(6) (2014) 1444-51. [8] D.M. Adams, S.M. Stevens, S.C. Woller, R.S. Evans, J.F. Lloyd, G.L. Snow, T.L. Allen, J.R. Bledsoe, L.M. Brown, D.P. Blagev, T.D. Lovelace, T.L. Shill, K.E. Conner, V.T. Aston, C.G. Elliott, Adherence to PIOPED II investigators' recommendations for computed tomography pulmonary angiography, Am J Med 126(1) (2013) 36-42. [9] G. Le Gal, M. Righini, P.M. Roy, O. Sanchez, D. Aujesky, H. Bounameaux, A. Perrier, Prediction of pulmonary embolism in the emergency department: the revised Geneva score, Ann Intern Med 144(3) (2006) 165-71. [10] A. Agresti, Y. Min, Simple improved confidence intervals for comparing matched proportions, Stat Med 24(5) (2005) 729-40. [11] W. Leisenring, T. Alonzo, M.S. Pepe, Comparisons of predictive values of binary medical diagnostic tests for paired designs, Biometrics 56(2) (2000) 345-51. [12] M. Righini, H. Robert-Ebadi, G. Le Gal, Diagnosis of acute pulmonary embolism, Journal of Thrombosis and Haemostasis 15(7) (2017) 1251-1261. [13] C.E.A. Dronkers, Y.M. Ende-Verhaar, P.A. Kyrle, M. Righini, S.C. Cannegieter, M.V. Huisman, F.A. Klok, P. Subcommittee on, D. Diagnostic Variables in Thrombotic, Disease prevalence dependent failure rate in diagnostic management studies on 13

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE

AC

CE

PT

ED

M

AN

US

CR

IP

T

suspected deep vein thrombosis: communication from the SSC of the ISTH, J Thromb Haemost 15(11) (2017) 2270-2273. [14] P.S. Wells, D.R. Anderson, M. Rodger, J.S. Ginsberg, C. Kearon, M. Gent, A.G. Turpie, J. Bormanis, J. Weitz, M. Chamberlain, D. Bowie, D. Barnes, J. Hirsh, Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer, Thromb Haemost 83(3) (2000) 416-20. [15] S. Takach Lapner, J.A. Julian, L.A. Linkins, S. Bates, C. Kearon, Comparison of clinical probability-adjusted D-dimer and age-adjusted D-dimer interpretation to exclude venous thromboembolism, Thromb Haemost 117(10) (2017) 1937-1943. [16] M.C. Janssen, A.E. Heebels, M. de Metz, H. Verbruggen, H. Wollersheim, S. Janssen, M.M. Schuurmans, I.R. Novakova, Reliability of five rapid D-dimer assays compared to ELISA in the exclusion of deep venous thrombosis, Thromb Haemost 77(2) (1997) 262-6. [17] D.L. Sackett, R.B. Haynes, The architecture of diagnostic research, BMJ 324(7336) (2002) 539-41. [18] M. Righini, J. Van Es, P.L. Den Exter, P.M. Roy, F. Verschuren, A. Ghuysen, O.T. Rutschmann, O. Sanchez, M. Jaffrelot, A. Trinh-Duc, C. Le Gall, F. Moustafa, A. Principe, A.A. Van Houten, M. Ten Wolde, R.A. Douma, G. Hazelaar, P.M. Erkens, K.W. Van Kralingen, M.J. Grootenboers, M.F. Durian, Y.W. Cheung, G. Meyer, H. Bounameaux, M.V. Huisman, P.W. Kamphuisen, G. Le Gal, Age-adjusted D-dimer cutoff levels to rule out pulmonary embolism: the ADJUST-PE study, JAMA 311(11) (2014) 1117-24. [19] L.M. van der Pol, T. van der Hulle, Y.W. Cheung, A.T.A. Mairuhu, C.G. Schaar, L.M. Faber, M. Ten Wolde, H.M.A. Hofstee, M.M.C. Hovens, M. Nijkeuter, R.C.J. van Klink, M. Kruip, S. Middeldorp, M.V. Huisman, F.A. Klok, No added value of the ageadjusted D-dimer cut-off to the YEARS algorithm in patients with suspected pulmonary embolism, J Thromb Haemost 15(12) (2017) 2317-2324.

14

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Table 1: Revised Geneva Score Criterion

Points

Risk Factors 1

Previous DVT or PE

3

T

Age >65 years

IP

Surgery or lower limb fracture within 1 month

CR

Active malignancy within 1 year Symptoms

Hemoptysis

2

AN

Clinical Signs Heart rate 75-94 beats/min

3 5

M

Heart rate  95 beats/min

ED

Pain on palpation of lower limb deep venous system and/or unilateral edema Total

PT

Total Points 0-3

CE

Intermediate

4 0 to 25

Clinical Probability Category Low

2

3

US

Unilateral lower limb pain

2

4-10

AC

High DVT, deep vein thrombosis; PE, pulmonary embolism

15

11

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE

Patients ≤50

Patients >50

1745

822

923

Age, mean (SD), y

52 (19)

35 (9)

67 (11)

Female sex, n (%)

1128 (65%)

562 (68%)

T

566 (61%)

Low Probability, n (%)

646 (37%)

318 (39%)

328 (36%)

Prevalence of VTE

40 (6.2%)

18 (5.7%)

22 (6.7%)

1099 (63%)

504 (61%)

595 (64%)

Characteristic

CR

Number of patients n

IP

All patients

US

Table 2: Characteristics of all patients with low or intermediate probability of PE who had both a D-dimer test and a CTPA

Intermediate Probability, n (%)

105 (9.6%)

43 (8.5%)

62 (10.4%)

Previous VTE, n (%)

203 (12%)

85 (10%)

118 (13%)

139 (8%)

76 (9%)

63 (7%)

68 (3.9%)

22 (2.7%)

46 (5.0%)

65 (3.7%)

34(4%)

31(3%)

23 (1.3%)

15 (1.8%)

8 (0.9%)

AN

Prevalence of VTE

Surgery or Fracture, n (%)

M

Active malignancy, n (%)

ED

Leg pain, n (%) Hemoptysis

AC

CE

PT

VTE, venous thromboembolism

16

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE

Table 3: Accuracy of D-dimer interpretation strategies for VTE Reference

Accuracy

Clinical probability-

Age-adjusted

Parameter

adjusted Strategy

Strategy

127/145

140/145

87.5%

96.6%

(81.6%-92.3%)

(92.7%-98.8%)

600/1600

484/1600

37.5%

30.3%

-7.2%

19.8%

(35.2%-39.9%)

(28.0%-32.5%)

(-9.4% - -5.1%)

(17.9% - 21.8%)

Difference

500 ng/mL)

n/N

97.1%

% (95% CI)

PT

(95.6%-98.2%)

Negative results n/N

CE

618/1745

97.9% (94.6%-99.5%)

317/320

99.0%

1.9%

99.1%

(97.8%-99.6%)

(0.7%-3.1%)

(97.5%-99.8%)

489/1745

320/1745

35.4%

28.0%

-7.4%

18.3%

(33.2% - 37.7%)

(25.9%-30.2%)

(-9.4% - -5.4%)

(16.6% - 20.2%)

AC

% (95% CI)

484/489

ED

600/618

142/145

317/1600

M

Negative predictive value

IP

AN

% (95% CI)

(4.3%-14.0%)

US

Specificity n/N

8.9

CR

% (95% CI)

T

Sensitivity n/N

(threshold of

17

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE

Table 4: Comparison of the clinical probability-adjusted strategy and the ageadjusted strategy regarding the prevalence of PE according to agreement A: All patients Age-adjusted Strategy D-dimer Positive

IP

T

D-dimer Negative

Total

D-dimer

388

230

probability-

Negative

5 PE (1.3%)

13 PE (5.7%)

18 PE (2.9%)

adjusted

D-dimer

101

1026

1127

Strategy

Positive

0 PE (0%)

US

AN

489 Total

CR

Clinical

127 PE (12.4%)

127 PE (11.3%)

1256

1745

140 PE (11.1%)

145 PE (8.3%)

ED

M

5 PE (1.0%)

618

PT

B: Patients ≤50 years

Age-adjusted Strategy Total

probability-

CE

D-dimer Negative D-dimer Positive

Negative

3 PE (1.4%)

7 PE (5.0%)

10 PE (2.8%)

adjusted

D-dimer

0

466

466

Strategy

Positive

0 PE (0%)

51 PE (10.9%)

51 PE (10.9%)

216

606

822

3 PE (1.4%)

58 PE (9.6%)

61 PE (7.4%)

D-dimer

216

140

356

AC

Clinical

Total

18

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE C: Patients >50 years Age-adjusted Strategy D-dimer Negative D-dimer Positive

Total

D-dimer

172

90

262

probability-

Negative

2 PE (1.2%)

6 PE (6.7%)

8 PE (3.1%)

adjusted

D-dimer

101

560

Strategy

Positive

0 PE (0.0%)

76 PE (13.6%)

76 PE (11.5%)

273

650

923

IP

CR

82 PE (12.6%)

661

84 PE (9.1%)

AN

2 PE (0.7%)

US

Total

T

Clinical

D: Patients with low clinical probability (RGS 0-3)

M

Age-adjusted Strategy

D-dimer

probabilityadjusted

230

438

Negative

3 PE (1.4%)

13 PE (5.7%)

16 PE (3.7%)

D-dimer

0

208

208

Positive

0 PE (0%)

24 PE (11.5%)

24 PE (11.5%)

208

238

646

3 PE (1.4%)

37 PE (15.5%)

40 PE (6.2%)

CE

AC

Strategy

Total

208

PT

Clinical

ED

D-dimer Negative D-dimer Positive

Total

19

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE E: Patients with intermediate clinical probability (RGS 4-10) Age-adjusted Strategy D-dimer Negative D-dimer Positive

Total

D-dimer

180

0

180

Negative

2 PE (1.1%)

0 PE (0%)

2 PE (1.1%)

D-dimer

101

T

Clinical

818 adjusted

IP

probability-

919

Strategy

Positive

CR

103 PE

0 PE (0%)

103 PE (11.2%)

281 2 PE (0.7%)

AC

CE

PT

ED

M

AN

Total

US

(12.6%)

20

818

1099

103 PE

105 PE (9.6%)

(12.6%)

ACCEPTED MANUSCRIPT Ms. No. TR-D-18-00237: Takach Lapner et al. Age or probability adjusted D-dimer to exclude PE Highlights D-dimer is a sensitive test to rule out pulmonary embolism, but is non-specific



A higher D-dimer threshold in low probability patients improves specificity



A higher D-dimer threshold in elderly patients also improves specificity



More pulmonary embolism ruled out if threshold increased with probability versus

T



CR

More pulmonary embolism missed if threshold increased with probability versus

CE

PT

ED

M

AN

US

age

AC



IP

age

21