We analysed the combined data from two prospective multicentre cohort studies, which included 1721 consecutive outpatients with suspected pulmonary embolism.^{1, 12} These outcome studies were designed to evaluate diagnostic strategies for pulmonary embolism, combining clinical probability assessment, D-dimer measurement, lower limb venous compression ultrasonography, and helical computed tomography.

Briefly, all consecutive patients admitted to the emergency department of four general and teaching hospitals were included if they had a clinical suspicion of pulmonary embolism. The inclusion and exclusion criteria and the results of the two studies have been published previously.^{1, 12}

The first study, conducted at Geneva University Hospital, Geneva, and Centre Hospitalier Universitaire Vaudois, Lausanne, both in Switzerland, and at Angers University Hospital, Angers, France, between October 2000 and June 2002, comprised 965 patients.¹² The second study, conducted at Geneva University Hospital, Angers University Hospital, and H?pital Europ?en Georges Pompidou, Paris, France, between September 2002 and October 2003, comprised 756 patients.¹ Both studies were approved by the institutional review boards of each participating institution and written informed consent was obtained from all patients.

All patients underwent a sequential diagnostic investigation, including plasma D-dimer measurement by an enzyme linked immunosorbent assay (rapid ELISA assay, VIDAS D-Dimer Exclusion, Biom?rieux, Marcy-l?Etoile, France). For each patient, the Geneva score¹³ was assessed to assign the patient to a clinical probability category?with possible override by implicit assessment in case the result conflicted with the assessor?s clinical judgment.¹⁴ Variables included in the Wells clinical prediction rule for pulmonary embolism¹⁰ were also systematically and prospectively collected, allowing calculation of the Wells score.

Pulmonary embolism was ruled out by (a) a D-dimer concentration <500 ?g/l, except in patients with a high clinical probability, in the second study; (b) negative results from lower limb venous compression ultrasonography and from helical computed tomography in patients with a low or intermediate clinical probability of pulmonary embolism; or (c) by a normal ventilation-perfusion lung scan or a normal pulmonary angiogram in patients with a high clinical probability or with inconclusive helical computed tomogram. Pulmonary embolism was established by (a) finding a proximal deep vein thrombosis on lower limb ultrasonography; (b) a positive result from helical computed tomography; or (c) a high probability ventilation-perfusion lung scan or a positive pulmonary angiogram in high clinical probability patients with negative results from both compression ultrasonography and helical computed tomography, and in patients with inconclusive computed tomogram.

Patients were followed up by their family physicians and were interviewed by telephone by one of the study coordinators at the end of three months? follow-up. The outcome was an estimate of the three month thromboembolic risk in patients in whom pulmonary embolism was considered ruled out by the initial diagnostic investigations and who did not receive anticoagulants during follow-up. Confirmation of venous thromboembolic events during follow-up were established with the usual criteria.^{1, 2, 12}

For the first validation set, data from a third prospective multicentre cohort study were used.² This study evaluated the clinical effectiveness of a simplified algorithm using the dichotomised Wells rule, D-dimer testing, and computed tomography in patients with suspected pulmonary embolism. Briefly, all consecutive inpatients and outpatients with clinically suspected acute pulmonary embolism were eligible for the study, which was conducted between November 2002 and August 2004 in 12 hospitals in the Netherlands. The institutional review boards of all participating hospitals approved the study protocol. The results and the inclusion and exclusion criteria were published previously.² The study population comprised 3306 patients.

All patients underwent a sequential diagnostic investigation, consisting of clinical probability calculation, a D-dimer test (Tinaquant, Roche Diagnostica, Mannheim, Germany or Vidas D-Dimer Exclusion, Biomerieux) and computed tomography scanning. At admission, the clinical probability of pulmonary embolism was calculated by the treating physician using the Wells score. According to the protocol, a D-dimer test was performed only in patients with a Wells score of ?4. Pulmonary embolism was ruled out by (a) an unlikely clinical probability (Wells score ?4) combined with a D-dimer test ?500 ?g/l or (b) a negative helical computed tomogram in patients with a ?likely? clinical probability or an abnormal D-dimer test. Pulmonary embolism was established by a positive helical computed tomogram. The follow-up was performed the same way as in the derivation set studies.

For the second validation set, data from a fourth prospective multicentre study were used.³ This study investigated in a randomised non-inferiority trial whether the addition of venous ultrasonography to multi-detector computed tomography improved the detection of pulmonary embolism. Consecutive outpatients with clinically suspected acute pulmonary embolism were eligible for the study, which was conducted between January 2005 and August 2006 in six hospitals in France, Belgium, and Switzerland. The institutional review boards of all participating hospitals approved the study protocol. The results and inclusion and exclusion criteria were published previously.⁸ The study population comprised 1812 patients.

All patients underwent a sequential diagnostic investigation, consisting of clinical probability calculation, a D-dimer test (Vidas D-Dimer Exclusion, Biomerieux) and then randomisation to either multi-detector computed tomography alone or compression ultrasonography of the legs followed by computed tomography. At admission, the clinical probability was calculated by the treating physician using the revised Geneva score.¹⁵ Pulmonary embolism was ruled out by (a) a non-high clinical probability (revised Geneva score <11) combined with a D-dimer test <500 ?g/l or (b) a negative helical computed tomography result in patients with a high revised Geneva score or an abnormal D-dimer test result. Pulmonary embolism was established by a positive result from helical computed tomography or compression ultrasonography. The follow-up was performed in the same way as the previous studies.

In the derivation set and validation set 1 the Wells clinical prediction rule for pulmonary embolism was calculated in all patients, and patients were classified according to the dichotomised Wells score as likely or unlikely to have pulmonary embolism. In validation set 2, the revised Geneva score was calculated in all patients and classified as a high or non-high (low or intermediate) score.

To derive a new D-dimer cut-off value, we divided patients aged >50 in the derivation set into 10 year age groups. We constructed receiver operating characteristics (ROC) curves of the D-dimer test for each age group to find the best cut-off value (with a sensitivity of 100% and the highest corresponding specificity). We plotted the D-dimer cut-off level against age group and performed linear regression analysis to obtain the regression coefficient, representing the increase in D-dimer cut-off value per decade. This coefficient was divided by 10 to find the coefficient per year. This coefficient was the multiplication factor for age in the new age-adjusted cut-off value.

We then calculated the proportion of patients with a negative D-dimer test result (as defined by the new cut-off point), the proportion in whom pulmonary embolism could be excluded (based on an ?unlikely? Wells score or ?non-high? revised Geneva score plus the negative D-dimer test result), and the false negative rates (that is, those patients who had pulmonary embolism in the diagnostic investigation or during follow-up). The number of patients needed to test by D-dimer to rule out one pulmonary embolism was computed as 1 divided by the proportion of patients with a negative D-dimer test result in each age group.¹⁶

In the validation sets, D-dimer test results were missing in a large proportion of patients with a ?likely? Wells score or high revised Geneva score. Therefore, the age adjusted D-dimer cut-off point was validated only in patients with an ?unlikely? or non-high clinical probability score in these two cohorts.

We calculated exact 95% confidence intervals using CIA software version 1.0 (Gardner et al. Confidence Interval Analysis (CIA), BMJ Books). All other analyses were performed with SPSS version 15.0 (SPSS, Chicago, IL).

Of the 1721 patients included in the derivation set, 416 (24.2%) had pulmonary embolism. The D-dimer test was not performed for nine patients from the second study with high clinical probability (all nine had pulmonary embolism diagnosed during initial investigation). Table 1tbl1 shows the clinical characteristics of the patients in the derivation cohort.

The figurefig1 shows the increase in optimal D-dimer cut-off value by patient age group, obtained from the ROC curves for each group. The cut-off point increased from 512 ?g/l in patients <50 years old to 934 ?g/l in patients >80 years old. The regression coefficient was 112 (SE 12.03) ?g/l increase per decade, or 11.2 ?g/l increase per year (r²=0.966). To be conservative and to facilitate clinical usefulness and practicality, we considered a 10 ?g/l increase per patient year to be an appropriate new D-dimer coefficient. Starting from the conventional cut-off point of 500 ?g/l until the age of 50, for older patients the patient age should be multiplied by 10 to calculate the age adjusted cut-off value.

With the conventional cut-off value, the VIDAS D-dimer test was normal (<500 ?g/l) in 512/1712 patients (29.9%), and none had pulmonary embolism during the initial investigation or the three month follow-up (0%, 95% CI 0% to 0.7%). The number of patients needed to test to find one normal D-dimer test result was 3.3. The Wells score could not be computed in 54 of the patients, mainly because information on the likelihood of an alternative diagnosis to that of pulmonary embolism was missing (n=42).

Using the age adjusted cut-off value ((age (years)?10) ?g/l), we found that D-dimer test results were negative in 615/1712 patients (46.2%, number needed to test 2.2). This resulted in a 20.1% (95% CI 16.9% to 23.8%) relative increase in the number of patients in whom D-dimer levels were considered normal. Of these 615 patients, five had pulmonary embolism during investigation or three month follow-up (0.8%, 0.4% to 1.9%).

Table 2tbl2 shows the proportion of the 1331 patients with an unlikely clinical probability in whom pulmonary embolism could be excluded based on the conventional and the age adjusted D-dimer cut-off values. There was a 17.4% (95% CI 14.3% to 21.1%) increase in the number of patients with a negative D-dimer test result when the age adjusted cut-off value was used. The false negative rate was 0 (0%, 0% to 0.8%) for the conventional cut-off value compared with 1 (0.2%, 0% to 1.0%) for the age adjusted cut-off value. Table 2 also shows the increase in the proportion of patients with an unlikely clinical probability in whom D-dimer levels would be considered normal for specific age groups by using the age adjusted cut-off value: this increase was highest among the oldest patient groups (>70 years), with an absolute increase of 14% compared with the conventional cut-off point.

The concept of using an increasing cut-off value for the D-dimer test according to age was validated in two independent cohorts of patients with suspected pulmonary embolism. The clinical characteristics of the patients in these cohorts were similar to those of the patients in the derivation set (table 1tbl1).

A recent cost effectiveness analysis showed that D-dimer measurement as part of the diagnostic investigation of patients with suspected pulmonary embolism was cost saving until the age of 79 years.¹⁷ After 80 years, the test?s clinical utility was too low to be cost effective, and the costs of strategies with or without D-dimer testing were similar. This analysis was based on the conventional cut-off point of 500 ?g/l for ELISA based assays. It can be expected, however, that the test?s cost effectiveness in older patients would increase with the new cut-off value, as the number needed to test was lower with the age adjusted cut-off value compared with the conventional cut-off value in patients >80 years old (3.5 versus 6.6). For a given clinical setting, this means that for every 35 patients aged >80 with a low/intermediate or ?unlikely? clinical probability of pulmonary embolism, imaging tests can be avoided in five patients when the conventional cut-off is used compared with 10 patients when the age adjusted cut-off value is used. Avoiding imaging tests (that is, ventilation-perfusion (V/Q) scintigraphy or computed tomography) would be of particular benefit for older patients because of the high frequency of non-conclusive scintigraphy results, the risk from injection with iodine contrast agent for computed tomography scanning, and the length of hospital stay when ordering imaging tests in this patient population.

Some may argue that having to calculate a D-dimer cut-off value is impractical in a clinical setting, and that the new value is really a series of multiple cut-off points. However, the treating physician needs to remember only the coefficient of 10 in order to calculate the new cut-off value, which is an easy multiplication.

The calculation of the age adjusted cut-off value was facilitated by the large size of the study population, which is a major strength of this analysis. However, some aspects of our study warrant comment. Firstly, two different D-dimer assays were used in the validation cohorts. Although there was no significant difference between the two tests, and the new cut-off value performed equally well with both assays, the study may not have been sufficiently powered to detect a difference between the two. It is unknown how the new cut-off value will perform in other D-dimer assays. Secondly, although D-dimer tests and the (variables for the) clinical decision rule were collected prospectively, this study was a retrospective analysis. After derivation and independent validation in a completely distinct cohort of patients, the next step would be to validate this cut-off value prospectively in a formal outcome study with patient follow-up.

In conclusion, a cut-off value adjusted to age combined with clinical probability greatly increased the utility of the D-dimer test for the exclusion of pulmonary embolism among older patients without reducing safety. This new cut-off is therefore clinically relevant and has sustained external validation. The next step would be to validate this new D-dimer cut-off value prospectively before implementation in daily practice.

Contributors: RAD, GlG, PWK, and PMR conceived and designed the study and performed analysis and interpretation of data. MS, MR, AP, and MJHAK collected and interpreted data. RAD drafted the manuscript. All authors were involved in revision and final approval of the manuscript. All authors had full access to the data in the study.

Funding: None

Competing interests: All authors have completed the unified competing interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare (1) no financial support for the submitted work from anyone other than their employer; (2) no financial relationships with commercial entities that might have an interest in the submitted work; (3) no spouses, partners, or children with relationships with commercial entities that might have an interest in the submitted work; and (4) no non-financial interests that may be relevant to the submitted work.

Ethical approval: Not required.

Data sharing: No additional data available.

IQR=interquartile range

*Based on Wells clinical decision rule.

?Conventional cut-off value for D-dimer test=500 ?g/l, age adjusted cut-off value=(age?10) ?g/l (if age >50).

?Number needed to test to find one normal D-dimer test result

IQR=interquartile range

*Based on Wells clinical decision rule.

?Conventional cut-off value for D-dimer test=500 ?g/l, age adjusted cut-off value=(age?10) ?g/l (if age >50).

?Number needed to test to find one normal D-dimer test result

*Based on Wells clinical decision rule.

?Conventional cut-off value for D-dimer test=500 ?g/l, age adjusted cut-off value=(age?10) ?g/l (if age >50).

IQR=interquartile range

*Based on revised Geneva score <11.

?Conventional cut-off value for D-dimer test=500 ?g/l, age adjusted cut-off value=(age?10) ?g/l (if age >50).

?Number needed to test to find one normal D-dimer test result