D-dimer assays - A help or hindrance in suspected pulmonary thromboembolism assessment?HK Bayes1, CA O'Dowd1, NJ Glassford2, A McKay1, S Davidson1
1 Department of Respiratory Medicine, Southern General Hospital, Glasgow, United Kingdom
2 Department of Medical Specialties, Southern General Hospital, Glasgow, United Kingdom
Background : Suspected pulmonary thromboembolism (PTE) is a common presentation to acute medical units and can cause diagnostic difficulty. National guidelines on PTE management highlight the need for clinical probability assessment and D-dimer assays to ensure appropriate use of diagnostic imaging. D-dimers are used widely in UK hospitals, yet concern exists regarding their misuse. Aims : In this study we aimed to assess the impact of the introduction of D-dimer assays, combined with clinical probability assessment, for evaluation of suspected PTE in our unit. Materials and Methods : This was a prospective audit of all patients presenting with suspected PTE over two 12-week periods, exactly 1 year apart. D-dimers were introduced into our unit between these two periods. We recorded the clinical probability score, potential causes of false-positive D-dimer assay, diagnostic imaging result, patient outcome, admission rates, and length of inpatient stay. Statistical Analysis : Categorical variables were compared using a 2 x 2 chi-square test or Fisher's exact test. Groups were compared utilizing the two-sample t-test or Mann-Whitney U test. Results : A total of 190 patients were included in the study; 65% were female. PTE was confirmed in 8.4%. Patients in both audit periods were comparable with regard to suitability for D-dimer measurement. Following D-dimer introduction, 40 out of 110 patients in period 2 could be discharged directly from the emergency department. Of those admitted to hospital, the median length of stay was significantly reduced in period 2 (3 days in period 1 vs 1 day in period 2; P=0.0007). Use of diagnostic imaging was significantly reduced following the introduction of D-dimers (90% in period 1 vs 40% in period 2; P<0.0001). The positive diagnostic yield for PTE on CT pulmonary angiogram (CTPA) increased significantly from 10% in period 1 to 23% in period 2 (P=0.039). Conclusion : D-dimers must be used judiciously in the assessment of suspected PTE. Appropriate use of D-dimers can provide many benefits, including reductions in diagnostic imaging (and thus radiation exposure), admission rates, and length of inpatient stay.
Keywords: D-dimer(s), investigation(s), pulmonary embolism, pulmonary thromboembolism, venous thromboembolism
Suspected pulmonary thromboembolism (PTE) is a common presentation in the emergency department (ED) and often necessitates admission of the patient to the acute medical unit (AMU) for further investigation. The annual incidence of PTE is 60-70 per 100000 of the population.  The overall mortality has been reported to be as high as 60% in untreated case,  which can be reduced to 2.5% with anticoagulation therapy.  PTE can however cause diagnostic difficulty as the presentation can be variable and symptoms may range widely in severity, with no specific physical findings being necessarily apparent. The perceived, and indeed real, consequences of a missed diagnosis of PTE may also influence a physician's assessment and propensity to further investigate such patients. 
Clinical judgment alone is not sufficient to make or exclude the diagnosis of PTE. On the other hand, performing diagnostic imaging - increasingly in the form of CT pulmonary angiogram (CTPA) - on all patients with suspected PTE is neither feasible nor cost-effective. British Thoracic Society (BTS) guidelines for the management of PTE outline an evidence-based standard of care,  which highlight the need for clinical probability assessment and the use of D-dimer assays to ensure appropriate use of diagnostic imaging. Judicious use of D-dimer assays, after consideration of both clinical probability and the possible alternative diagnoses, is necessary to ensure their use only in situations where the result will advance the diagnostic process.
There exist a number of clinical prediction scoring systems for assessing the probability of PTE. The BTS has recommend a simple scoring system for the assessment of suspected PTE.  However, the two most commonly recognized predictive scoring systems are the Wells rule  and the modified Geneva score. 
D-dimers are produced when fibrin clots are degraded. D-dimer assays have been demonstrated to be highly accurate in excluding PTE in the context of a low or moderate clinical pretest probability,  but have insufficient sensitivity to exclude the diagnosis in cases with high pretest probability. D-dimers can also be elevated in a number of other common clinical situations (liver disease, pregnancy, recent trauma or surgery, cancer, and in the presence of elevated inflammatory markers, ,, and thus have a low specificity for the diagnosis of PTE. Unfortunately these factors are often not considered when ordering or interpreting these tests in the context of suspected PTE, and physicians often feel compelled to act on a positive D-dimer result, leading to unnecessary investigations. ,
Following the widespread implementation of D-dimer assays in EDs and AMUs, few institutes are in a position to prospectively study the impact of D-dimers on the management of suspected PTE. Although an investigation algorithm combining clinical probability assessment and D-dimer assays has been widely advocated, , there is limited published data demonstrating that this approach is appropriately applied in clinical practice. Indeed, concern exists that D-dimers are often used as a 'blanket screening tool' for PTE, with resultant increases in imaging procedures (and, therefore, inpatient stay and radiation exposure), decreased positive diagnostic yield, and potentially missed diagnoses in high-risk patients. ,
The purpose of this study was to prospectively audit the impact of the introduction of D-dimer assays for assessment of suspected PTE in our institute, with clinical probability assessment being a prerequisite for obtaining the test. We assessed the appropriateness of D-dimer use at our hospital and the impact it has had on diagnostic imaging, admission rates, and inpatient stay.
This prospective audit covered two 3-month periods (March to May) in 2008 and 2009 in a large city teaching hospital with a catchment population of 225000. A quantitative D-dimer assay was introduced for the assessment of PTE in the hospital following the first audit period, becoming routinely available 5 months prior to the second audit period. All patients with suspected PTE attending the ED and/or admitted to the AMU during the two periods were included. The audit intervention (i.e., D-dimer measurement) was being introduced in our unit in compliance with national guidelines;  there were no other changes made in patient care or management during the course of this audit. Confidentiality was maintained. Ethical review is not required for audit purposes in our institution, as also nationally.
A latex agglutination D-dimer assay was routinely available in our unit during the second audit period. As per our hospital's electronic hematology ordering system, it was mandatory that the requesting physician record the patient's modified Geneva clinical probability score  when ordering a D-dimer assay. A D-dimer value <500 ng/ml was considered normal.
In the first audit period, D-dimer assay was done for a few patients at some other hospital's hematology department on the request of a consultant physician.
The decision to undertake diagnostic imaging in patients with suspected PTE was made by the admitting medical consultant. Imaging was undertaken in the form of either a CTPA or a ventilation-perfusion (VQ) scan. In our unit a VQ scan was only performed in patients with a normal admission chest x-ray (CXR), in patients with renal failure, or in case of allergy to contrast dye. Perfusion-only (Q) scans were performed in pregnant patients with suspected PTE.
We recorded data regarding patient demographics, clinical presentation, risk factors for PTE, modified Geneva score, D-dimer result (if applicable), potential causes for false-positive D-dimer (i.e., raised inflammatory markers, liver disease, recent surgery, pregnancy, and active malignancy), diagnostic imaging (CTPA or VQ scan), and final diagnosis. For those patients who were admitted, the length of inpatient stay was recorded. In the case of patients who did not undergo diagnostic imaging for PTE during their admission at our hospital, we used a city-wide radiology database to find out whether any of them had been later admitted and investigated for PTE elsewhere over the following 3 months.
We used GraphPadTM software for analysis. Categorical variables were compared using contingency tables with either a 2 x 2 chi-square test or, where sample size was small, Fisher's exact test. Normality of the distribution of data sets was assessed using the Kolmogorov-Smirnov test. Where normality was present, groups compared using a two-sample t-test using IBM SPSS® Statistics software. Nonparametric data were compared with the two-tailed Mann-Whitney U test using SPSS® software. The data sets for both audit periods were complete.
A total of 190 patients were admitted with suspected PTE during the periods covered by the audit (80 patients in period 1 and 110 patients in period 2). There was a female preponderance in both audit periods, with overall 65% of patients being female [Table 1]. Risk factors for PTE were similar between the two groups [Table 1]. Presentation with dyspnea, hypoxia, and symptoms of possible deep vein thrombosis (DVT) were more common in period 1, whereas chest pain was the commonest presenting symptom in period 2 [Table 2].
The patients in the two audit periods did not differ in their suitability for D-dimer measurement as part of their diagnostic assessment. The proportion of patients deemed suitable for D-dimer testing (i.e., having a low or intermediate clinical probability on the modified Geneva scoring system and no obvious cause for a false-positive D-dimer test) was 56% (45/80 patients) in period 1 vs 64.5% (71/110 patients) in period 2 (χ2 = 1.34, degree of freedom (df) = 1; P=.247).
Pulmonary embolism diagnosed in the study population
Overall, 8.4% (n=16) of patients were confirmed to have PTE on diagnostic imaging. CTPA confirmed the diagnosis in 14 patients, whereas high-probability VQ scan was taken as confirmation of PTE in 2 cases. There was no significant difference in the prevalence of PTE between the two audit periods: 8/80 patients (10%) with PTE in period 1 vs 8/110 patients (7.3%) with PTE in period 2 (P=.599).
Use of D-dimer assays
As can be expected, the use of D-dimers for assessment of suspected PTE increased significantly between audit periods, with D-dimer assays performed in 15/80 patients (19%) in period 1 vs 97/110 patients (88%) in period 2 (χ2 = 92.26, df = 1; P<0.0001). Inappropriate use of D-dimer measurement was deemed to have occurred when the patient's modified Geneva probability score was high or where potential causes of a false positive D-dimer were evident. There was no significant difference between the audit periods in the number of instances of inappropriate D-dimer use; 2/15 (13.3%) assays were deemed inappropriate in period 1 vs 27/97 (27.8%) in period 2 (P=0.218). However, given the limited number of D-dimer assays performed in audit period 1 and the fact that these were only performed under the auspices of a consultant physician, it is difficult to make a direct comparison between the two audit periods. The most common reason for considering use of D-dimer measurement as inappropriate was in patients who demonstrated raised inflammatory markers (i.e. total white cell count and C-reactive protein (CRP)). In the majority of patients with raised inflammatory markers, the admission CXR was also abnormal. In only one case was a D-dimer measurement performed in the presence of a high probability score; however, PTE was subsequently excluded in this patient.
Impact on diagnostic imaging
There was a significant reduction in the proportion of patients with suspected PTE undergoing diagnostic imaging in audit period 2; 72/80 patients (90%) underwent imaging in period 1 vs 44/110 patients (40%) in period 2 (χ2 = 48.695, df = 1; P<0.0001) [Table 3]. The positive diagnostic yield for PTE on CTPA also significantly increased from 10% in period 1 to 23% in period 2 (χ2 = 4.245, P=0.039).
A negative D-dimer result along with low/intermediate clinical probability by the modified Geneva score allowed avoidance of further investigation in 53% of patients in period 2 (58/110 patients) compared with 12.5% (10/80 patients) in period 1 (χ2 = 32.614, df =1; P<0.0001). None of the 68 patients who were discharged without being subjected to diagnostic imaging (on the basis of their negative D-dimer result and low/intermediate modified Geneva clinical probability score) were later investigated for PTE at any other local hospital in the 3 months following their presentation.
Effect on admission rates and length of stay
Following the introduction of D-dimer assays, 40 patients in period 2 could be discharged directly from the ED as they had negative results on D-dimer assay and low clinical probability scores. This form of direct discharge from the ED was not possible prior to the introduction of D-dimers. Among those admitted to hospital, the length of inpatient stay of patients with suspected PTE was significantly reduced in audit period 2: the median length of hospital stay in period 1 was 3 days (IQR: 2-9 days) vs 2 days (IQR: 1-4 days) in period 2 (P=0.0036).
The median length of stay in those with confirmed PTE was longer prior to introduction of D-dimer assays but the difference failed to reach statistical significance: median length of stay 9.5 days (IQR: 5-21 days) in period 1 vs 5 days (IQR: 5-7 days) in period 2 (P=.201). However, there was a significant reduction in the length of inpatient stay in those patients who were admitted to the AMU, with PTE subsequently excluded by either radiological investigation or a negative D-dimer plus low clinical probability: median of 3 days (IQR: 2-7 days) inpatient stay in period 1 vs 1 day (IQR: 1-4 days) in period 2 (P=0.0007).
Pulmonary embolism is a major cause of cardiovascular mortality, and because early anticoagulation can significantly reduce mortality, prompt and accurate diagnosis is crucial. , The investigation of suspected acute PTE relies on the accurate assessment of clinical probability and appropriate use of D-dimer assay, followed by diagnostic imaging if required.  D-dimer assays are now widely used in UK hospitals for the assessment of suspected PTE. However, concern exists regarding inappropriate use and the consequent implications on patient management. , Our unit is one of the few where D-dimers were not in routine use until recently and thus we are in a position to prospectively audit the impact that the introduction of D-dimers has had on the management of suspected PTE. D-dimers were introduced into our practice with the mandatory requirement that a clinical probability assessment be made prior to proceeding with the test. Ensuring compliance with this requirement was aided by the use of a computerized hematology ordering system, where the clinical probability score parameters had to be entered and calculated prior to ordering D-dimer assay.
Our audit clearly demonstrates that appropriate use of D-dimers, in combination with the prerequisite of assessment of clinical probability, has benefits in the management of suspected PTE. The most compelling finding was a significant reduction in the use of diagnostic imaging (and therefore patients' exposure to radiation). The BTS PTE guidelines state that use of D-dimers in combination with clinical probability assessment to evaluate patients with suspected PTE will result in radiological confirmation of PTE in 25% of cases who undergo subsequent diagnostic imaging.  Our audit demonstrates a significant improvement in diagnostic imaging yield following the introduction of D-dimers, with positive imaging findings in 23% of patients in audit period 2; this is close to the figure mentioned in the BTS guidelines.
Use of D-dimers allowed 21% of our cohort to be discharged directly from the ED, avoiding unnecessary admission to hospital. Additionally, in those who were admitted for further investigations, there was a significant reduction in the duration of inpatient stay.
Despite the rule in our unit requiring the requesting physician to record a modified Geneva clinical probability score  prior to obtaining a D-dimer assay, our study shows that we continued to use D-dimers 'inappropriately' in a quarter of cases. This is, in part, because of our strict definition of 'inappropriate' D-dimer use, including both an absence of potential causes of a false-positive D-dimer and a low/intermediate clinical probability; in contrast current guidance is to proceed to D-dimer measurement only when a suitable clinical probability is obtained and consideration has been given to alternative diagnoses.  Although D-dimer is useful for exclusion of PTE in patients with low pretest probability,  false-positive results are possible as D-dimers may be elevated in a number of other clinical situations. ,, Often, a clear alternative diagnosis and the likely cause of a false-positive D-dimer is evident at presentation but is missed or ignored by the clinician.  Thus further improvements in the use of D-dimers may be possible by educating the medical staff on the need to consider alternative diagnoses before ordering a D-dimer assay and by improving awareness regarding clinical situations where a D-dimer assay may be falsely positive. The measures should certainly involve review of the CXR by a middle-grade doctor before D-dimer assay is ordered but could also include awaiting the results of baseline blood tests (e.g., inflammatory markers and liver function tests) before processing or requesting a D-dimer assay. However, given that performing blood tests in the ED is reported to increase the length of stay by about 2 hours,  and with many countries now requiring patients to be processed through the ED within a limited time frame,  waiting on routine blood tests before proceeding to D-dimer assessment is unlikely to be feasible.
An alternative means of improving diagnostic yield would be to consider our method of assessing clinical probability. The results of prospective studies lend support to the concept that clinical probability assessment is a fundamental step in the diagnosis of PTE. ,, A number of structured clinical prediction models for PTE have been developed with the purpose of improving and easing the diagnostic approach. Our unit utilizes the modified Geneva score,  which is entirely standardized. Alternative models are the BTS recommended score  or Wells rule;  however, both of these include a subjective element, with the assessor having to decide if an alternative diagnosis is less probable than that of PTE. The Wells and Geneva scoring systems have previously been found to have comparable predictive accuracy.  Recently a more complex prediction model has been proposed that is claimed to out-perform both the Wells and the modified Geneva models. 
Knowledge of clinical probability scores for PTE and their interpretation remains poor across all grades of clinicians.  In this regard, the BTS recommended score offers the advantage of simplicity, which may be beneficial in the acute setting.  In addition easy-to-use software has been developed that can be installed on palm computers and mobile phones to facilitate the use of more complex prediction rules in the clinical setting.  Use of such computer software may aid both scoring of patients and the interpretation of pretest scores in combination with D-dimer results, thus further improving assessment of suspected PTE..
A potential criticism of conclusions drawn from our comparison of the two audit periods is that differences exist in the patients' age range and presentation. Given that this was a prospective audit, including all consecutive admissions with suspected PTE, controlling for such factors was not possible. In audit period 2, the patient population was significantly younger and chest pain was the most frequent presenting complaint. Arguably, this may be interpreted as meaning that D-dimers were being used in younger patients with musculoskeletal chest pain who, prior to the introduction of D-dimers, would have normally been discharged from the ED on the basis of clinical judgment alone. Such a possibility is difficult to exclude on the basis of the data collected in this audit. However, the patient populations were similar in terms of risk factors for PTE, clinical probability scoring, and overall prevalence of PTE, suggesting that direct comparison is valid. In addition, we observed an absolute reduction in CTPA use and increase in positive diagnostic yield on imaging, thus confirming a benefit from D-dimer use which would not be influenced by the inclusion of these additional patients.
We observed a significant reduction in the length of inpatient stay between audit periods 1 and 2. Arguably, performing a D-dimer early in patients with suspected PTE admitted to our acute medical ward would have allowed earlier discharge as there would be no time delay caused by having to wait for exclusion of PTE via radiological investigations. With an absolute reduction in number of CTPAs performed, there may also have been reduced pressure on CT scanning slots and therefore reduced time delay to radiological investigation of suspected PTE in period 2. Both of these factors may have resulted in the observed reduction in length of inpatient stay following D-dimer introduction. However, in audit period 1, the patients were, on average, 8 years older, and presentation with dyspnea and hypoxia was more common. Such patients are less likely to be discharged directly from the ED and this may have contributed to the longer inpatient stay. In both audit populations, additional factors that may have influenced the period of inpatient care were not assessed, such as time to therapeutic anticoagulation, comorbidities, and social circumstances.
A strength of our study is that it is a prospective assessment of a number of important outcome measures that were influenced by the introduction of D-dimer assays. Although, via use of a city-wide electronic radiology database, we observed no readmission and investigation of patients who had had PTE excluded on the basis of a negative D-dimer result, this study was not designed to assess the safety of PTE exclusion on the basic of low/intermediate clinical probability and a negative D-dimer. This requires thorough follow-up in a large patient population and this has been done elsewhere. ,
D-dimers must be used judiciously in the assessment of suspected PTE, taking into consideration both potential causes for a false-positive result and pretest clinical probability. However, this study confirms that the appropriate use of D-dimers in clinical practice can provide benefits such as facilitating direct discharge from the ED and reducing use of diagnostic imaging.
HKB and CAD contributed equally to this manuscript.
[Table 1], [Table 2], [Table 3]