Investigating pulmonary embolism in the emergency department with lower limb plethysmography: the Manchester Investigation of Pulmonary Embolism Diagnosis (MIOPED) study

Abstract

Objectives

To measure the diagnostic accuracy of computerised strain gauge plethysmography in the diagnosis of pulmonary embolism (PE).

Methods

Two researchers prospectively recruited 425 patients with pleuritic chest pain presenting to the emergency department (ED). Lower limb computerised strain gauge plethysmography was performed in the ED. All patients underwent an independent reference standard diagnostic algorithm to establish the presence or absence of PE. A low modified Wells' clinical probability combined with a normal D‐dimer excluded PE. All others required diagnostic imaging with PIOPED interpreted ventilation perfusion scanning and/or computerised tomography (CT) pulmonary angiography. Patients with a nondiagnostic CT had digital subtraction pulmonary angiography. All patients were followed up clinically for 3 months.

Results

The sensitivity of computerised strain gauge plethysmography was 33.3% (95% confidence interval (CI) 16.3 to 56.2%) and specificity 64.1% (95% CI 59.0 to 68.8%). The negative likelihood ratio was 1.04 (95% CI 0.68 to 1.33) and positive likelihood ratio 0.93 (95% CI 0.45 to 1.60).

Conclusions

Lower limb computerised strain gauge plethysmography does not aid in the diagnosis of PE.

The Office of National Statistics recorded pulmonary embolism (PE) as the cause of death in 3016 patients in England and Wales during 2002, accounting for 0.6% of deaths. An analysis of the published literature¹ estimated the risk of death following diagnosis of PE at 1.5% (95% confidence interval (CI) 0.9 to 2.2%) and following recurrent PE as 26.4% (95% CI 16.7 to 38.1%).

Symptoms consistent with a diagnosis of pulmonary embolic disease are common in patients presenting to the emergency department (ED). Of presentations to Manchester Royal Infirmary, 4% involve chest pain,² and of these a quarter are pleuritic in nature (J Wright et al, unpublished observation, 2003). A minority of patients with pleuritic chest pain will have PE. Combining clinical probability and D‐dimer investigation is an effective method for screening patients suspected of PE,³,⁴ but may exclude PE in only 25–50% of patients.⁵,⁶ Diagnostic imaging is expensive and time consuming, and may subject patients to unnecessary radiation exposure and hospital admission.

Deep vein thrombosis (DVT) and PE are pathophysiologically related. In 1989, an analysis of 213 patients with angiographically proven pulmonary emboli⁷ revealed that 81.7% had a DVT on venography. A more recent diagnostic study⁸ showed that although ultrasound of the femoral and popliteal veins had a poorer sensitivity for PE, the specificity was high at 97% (95% CI 94 to 99%). Unlike venography, ultrasound has few side effects; however, it is only available during the working day.

Computerised strain gauge plethysmography (Amtec venometer; Amtec Medical Ltd, Belfast Road, Antrim, UK) is a bedside technique for diagnosing DVT. It can be performed rapidly by non‐specialist personnel, day or night. If this investigation could identify patients with PE, the emergency physician could commence anticoagulation. If the test could accurately exclude the disease, fewer patients would require diagnostic imaging. The aim of this study was to measure the diagnostic accuracy of computerised strain gauge plethysmography in the diagnosis of PE.

METHODS

The study cohort

Ethics approval for the study was granted by Central Manchester local research ethics committee (reference CEN/00/082). Between 11 February 2002 and 29 May 2003, two investigators prospectively recruited patients presenting to Manchester Royal Infirmary's ED with pleuritic chest pain, following their informed consent. To be included, the patient had to describe their pain as sharp, experienced between the neck and upper abdomen, and worse on deep inspiration and coughing. Patients with pneumothorax or ECG changes of cardiac ischaemia or pericarditis were excluded. Other exclusions were recent trauma, PaO₂ <7.5 kPa, contraindication to contrast investigations, weight >140 kg, pregnancy, age <18 years, inability to give consent, and refusal to participate. Throughout the study, all the ED notes were reviewed. Details were taken from the notes of any patient that described chest pain as sharp or pleuritic.

Computerised strain gauge plethysmography

Plethysmography was performed in the ED by one of two investigators, while the patient awaited their blood test results. Both investigators (the lead physician and research nurse) were trained to use the venometer by Amtec Medical Ltd prior to commencing the study. The patient lay supine with one foot elevated on a footrest. After a 4 minute relaxation period, a strain gauge was attached to the patient's calf. A thigh cuff then inflated to 55 mmHg, during which time the venometer generated a graph of strain gauge tension over time. The cuff automatically released after 2 minutes, and the computer gave a positive, negative, or invalid result. Both legs were tested. Following an invalid result, the test was repeated once, 20 minutes later. If the result remained invalid, the test was considered indeterminate. If either leg gave a positive reading, the test was considered positive, and if both legs gave a negative reading, the test was negative. Computerised strain gauge plethysmography was conducted while blinded to all reference standard results.

Reference standard

Patient investigation

PE was diagnosed with a high probability ventilation perfusion scan in a patient scoring moderate or high clinical probability, a positive computed tomography (CT) pulmonary angiogram, or a positive digital subtraction pulmonary angiogram. PE was excluded by a normal D‐dimer in a patient scoring low clinical probability, a normal ventilation perfusion scan, a low ventilation perfusion scan in a patient scoring low clinical probability, a negative CT pulmonary angiogram in a patient scoring low clinical probability, a CT showing another diagnosis, or a negative digital subtraction pulmonary angiogram. All patients were followed clinically for 3 months. The reference standard diagnostic algorithm is detailed in fig 1.

Figure 1 The MIOPED reference standard.

The study used the Manchester modified Wells' clinical probability score (table 1), which adds intravenous drug use (S. Jones et al, unpublished observation, 2001) to the Wells' score.⁹ The score was independently calculated by two physicians. Any discrepancies with the first score were resolved by a third independent physician. All patients had a latex agglutination laboratory based D‐dimer test (421 patients had the IL D‐dimer test and four patients the MDA D‐dimer test).

Table 1 Manchester modified Wells' score.

		Points
Clinical signs and symptoms of DVT: minimum of swelling and pain on palpation of the deep veins		3.0
Intravenous drug use		3.0
Pulmonary embolism is the most likely diagnosis in the opinion of the doctor, using all available blood results, ECG, and chest radiograph		3.0
Heart rate >100 beats/min		1.5
Immobilisation for a minimum of 3 days' bed rest or surgery within past 4 weeks		1.5
Previous DVT or PE		1.5
Haemoptysis		1.0
Malignancy with treatment or palliative care in the past 6 months		1.0

Low probability <2.0, moderate probability 2.0–6.0, high probability >6.0. DVT, deep vein thrombosis; PE, pulmonary embolism.

Ventilation perfusion scans were obtained with either ^81MKr krypton gas or ^99mTc technegas with ^99MTc macroaggregated albumin, and interpreted using PIOPED¹⁰ criteria. Two nuclear medicine physicians independently reported each scan. Any discrepancies were reviewed by a third blinded consultant. CT images were interpreted prospectively by one consultant chest radiologist on a computer. The criterion used to diagnose PE was direct visualisation of an endoluminal thrombus.¹¹ A random sample of 25% of the CT scans was independently reviewed by a second consultant chest radiologist. Digital subtraction pulmonary angiography images were interpreted independently by two consultant vascular radiologists. Any discrepancies were resolved by the opinion of a third. The reference standard was conducted while blinded to the plethysmography results.

Patient follow up

Each patient recruited into the study received written instructions to return or make contact if they experienced any further symptoms of chest pain, shortness of breath, or pain or swelling of the legs. In addition, all patients were contacted by telephone 3 months after recruitment. Patients were followed up for evidence of DVT or PE.

Adjudication committee

An independent adjudication committee reviewed all cases where a patient died during study follow up. The committee also reviewed any case where a patient had undergone further testing for PE during the clinical follow up period, without the knowledge of the study group. A consensus opinion was given on whether the patient's death or continued symptoms were “unlikely”, “possibly”, or “probably” caused by thromboembolic disease.

Statistics

Assuming a prevalence of PE of 10% and a test sensitivity of 95%, a sample of 400 participants would give a lower 95% confidence limit of 83%. If the prevalence was 5%, the lower 95% confidence limit would be 76%. Confidence intervals for sensitivity, specificity, positive predictive value and negative predictive value were calculated using the Wilson “score” method¹²,¹³ Those for likelihood ratios were constructed using the likelihood based approach to binomial proportions by Gart and Nam.¹⁴

All prospectively accumulated information was stored anonymously in a database using SPSS software (SPSS Inc., Chicago, Illinois, USA), and confidence intervals were calculated using StatsDirect software (StatsDirect Ltd, Gresham Way, Sale, Cheshire, UK).

RESULTS

The study cohort

In total, 799 patients were assessed for inclusion into the study. Of these, 425 consented to participate, 97 declined, and 277 were excluded (171 patients were excluded because their pain did not fit the inclusion criteria). The mean age of the final MIOPED study cohort was 38.3 years, and 48.9% were men. Of the patients who were assessed, 64% were reviewed by a researcher between the hours of 0900 and 1700, 28% were assessed between 1700 and midnight, and 8.0% between midnight and 0900. During the same time period, an additional 633 patients attended the ED with chest pain described in the notes as either pleuritic or sharp. An analysis of these patients showed they did not differ clinically in age or sex distribution, clinical probability score, white cell count, or arterial blood gas analysis.

In total, 368 patients (86.6%) scored low clinical probability of PE, 43 (10.1%) moderate and 14 (3.3%) high. Table 2 shows the study patients' thromboembolic risk factors and examination findings.

Table 2 MIOPED study patient findings.

Risk factor		Prevalence (n = 425)
Family history of thromboembolism		50 (11.8%)
Previous thromboembolic event		51 (12%)
Anticoagulation		24 (5.6%)
Intravenous drug use		19 (4.5%)
Surgery in previous 4 weeks		6 (1.4%)
Immobilisation in past 4 weeks		30 (7.1%)
Limb immobilisation in past 4 weeks		11 (2.6%)
Cancer therapy in past 6 months		6 (1.4%)
Post‐partum		6 (1.4%)
Oral contraception or HRT		54 (12.7%)
Haemoptysis		43 (10.1%)
Heart rate >100 beats/min		21 (4.9%)
Oxygen saturations <95% on air		23 (5.4%)
PaO2 <10 kPa		96 (22.6%)
Temperature >37.2°C		49 (11.5%)
White cell count >11.0×109/l		86 (20.2%)
Abnormal ECG		60 (14.1%)
Abnormal chest radiograph		57 (13.4%)
Clinical signs of DVT		22 (5.2%)

HRT, hormone replacement therapy; DVT, deep vein thrombosis.

Reference standard outcomes

Fig 2 shows the reference standard and plethysmography results. During investigation, 20 patients were diagnosed with acute PE. In total, 98.8% of the patients completed follow up; 368 patients were contacted directly, 24 via their general practioner, 23 via next of kin, and 1 hospital consultant gave information on 1 patient. Four patients died. Two patients were determined by the adjudication committee to have pulmonary embolic disease after re‐presenting in follow up.

Figure 2 Flow diagram of results.

All patients were included in the analysis with the exception of eight. Five patients had protocol violations and were anticoagulated during follow up, One patient was diagnosed with pulmonary embolic disease on the basis of a high probability ventilation perfusion scan (initial report) and high clinical probability, but the ventilation perfusion mismatches were later demonstrated to be caused by bilateral apical emphysema. One patient had chronic PE demonstrated on CT scan and it was not possible to establish whether she also had acute emboli. One patient had a subpectoral abscess at presentation, which had caused her pleuritic chest pain. She developed a DVT 11 weeks later, but it was deemed unlikely that the thrombosis represented a missed PE at presentation.

Computerised strain gauge plethysmography outcomes

Computerised strain gauge plethysmography was performed on 419 of the 425 patients, of whom 23 had an indeterminate plethysmography result.

Table 3 shows the diagnostic accuracy of computerised strain gauge plethysmography. The sensitivity is 33.3% (95% CI 16.3 to 56.2%) and the specificity, 64.1% (95% CI 59.0 to 68.8%). The negative predictive value is 95.2% (95% CI 91.7 to 97.5%) and the positive predictive value is 4.3% (95% CI 1.6 to 9.2%). The test results do not change the probability of a patient having PE. The negative likelihood ratio is 1.04 (95% CI 0.68 to 1.33) and positive likelihood ratio 0.93 (95% CI 0.45 to 1.60).

Table 3 Diagnostic accuracy of plethysmography.

		PE+		PE−
CSGP+		6		133
CSGP−		12		237

Six patients diagnosed with pulmonary embolic disease complained of leg pain. Of these, four patients had a positive plethysmography test and two had a negative test (sensitivity 66.7%, 95% CI 30.0 to 90.3%). Only two patients with objective signs of a DVT were diagnosed with a PE. One had a positive test, the other a negative test (sensitivity 50.0%, 95% CI 9.5 to 90.5%).

DISCUSSION

Computerised strain gauge plethysmography has poor sensitivity and specificity for PE. The MIOPED patient cohort had a 5.3% prevalence of PE. Neither a positive nor a negative plethysmography result had an impact on post‐test probability, and 5.5% of plethysmography tests were indeterminate.

This is the first study to investigate the usefulness of computerised strain gauge plethysmography in the investigation of PE. Several studies have assessed the test in patients with signs of DVT, but these have provided conflicting results. One study¹⁵ reported a sensitivity of 93.5% (95% CI 82.5 to 97.7%) and specificity of 80.3% (95% CI 72.6 to 86.3%) compared with above knee ultrasonography. Another comparing plethysmography with venography¹⁶ demonstrated a sensitivity of 84.6% (95% CI 74.0 to 91.4%) and specificity of 83.9% (95% CI 77.8 to 88.6%). This study considered all distal thrombi as a negative reference standard, thus if distal thrombi had been considered positive, the sensitivity would have been even lower. Maskell et al.¹⁷ calculated the sensitivity for below knee DVT as 66%.

One study has assessed the venometer as a screening tool for DVT following hip replacement surgery.¹⁸ Like the MIOPED study, many of the patients had no signs of lower limb thrombosis. Interestingly, the study found that of nine patients with thrombosis demonstrated on venogram, only seven had positive plethysmography results (sensitivity 77.7%, 95% CI 45.3 to 96.7%). The same study estimated specificity at 85.9% (CI 77.6 to 91.4%) compared with the MIOPED results of 64.1% (95% CI 59.0 to 68.8%). This postoperative study excluded patients with a history of DVT. There were 51 MIOPED patients with a history of thromboembolic disease. Ten MIOPED patients had recently been diagnosed with DVT, none of whom were subsequently diagnosed with a pulmonary embolus. In addition, only two of the 19 intravenous drug users who had ulcerated swollen legs were diagnosed with PE. This could explain the drop in specificity.

Patients recruited to the MIOPED study had pleuritic chest pain, which is characteristic of small, peripheral emboli. The mean (SD) arterial PaO₂ was 11.8 (2.2) kPa. No patient was diagnosed with massive PE. The original venous thrombi may have been small or distal in the limb venous system, thus by the time the clot embolised, it is possible that very little remained of the original leg clot, giving a negative plethysmography test.

Performing lower limb computerised strain gauge plethysmography at the ED bedside does not aid in the diagnosis of PE in patients with pleuritic chest pain. Our results are so poor that further research involving other patients suspected of pulmonary embolic disease is unlikely to yield high sensitivity or specificity. We would not recommend the use of computerised strain gauge plethysmography as an aid in the diagnosis of PE in this patient group.

Abbreviations

CT - computerised tomography

DVT - deep vein thrombosis

ED - emergency department

PE - pulmonary embolism