Value of structured clinical and scintigraphic protocols in acute pulmonary embolism
Tage Nilsson Department of Thoracic Radiology, Karolinska Hospital, S-171 76 Stockholm, Sweden (fax: +46 8 33 76 99; e-mail: email@example.com).
Abstract. Nilsson T, Måre K, Carlsson A (Karolinska Hospital, Linköping University Hospital; and Danderyd Hospital, Sweden). Value of structured clinical and scintigraphic protocols in acute pulmonary embolism. J Intern Med 2001; 250: 213–218.
Purpose. To study the use of a combination of a clinical and scintigraphic protocol in relation to the final outcome diagnosis in patients with clinical suspicion of acute pulmonary embolism (PE).
Material and methods. A total of 170 patients with clinical suspicion of acute PE were all examined with ECG, blood chemistry, chest X-ray, pulmonary scintigraphy and selective pulmonary arteriography. The scintigraphic and clinical probabilities of PE were estimated on visual analogue scales (VASs) by different readers unaware of each others’ results. The follow-up time was 6 months. In order to establish the final diagnosis a final outcome committee was created. They analysed in retrospect all the clinical and laboratory data and established whether the patient had had PE or not.
Results. The final outcome committee concluded that 53 patients had PE. When the scintigraphic and clinical probability judgements were congruent, a combined probability of 1–25% (i.e low probability) had a negative predictive value of 98%. When the combined probability was 26–75% (i.e. intermediate) half of the cases had PE. With a combined probability of 76–100% (i.e. high) the positive predictive value was 100%.
Conclusion. By applying a model of combined clinical and scintigraphic probabilities for PE, the diagnosis is ruled in when the combined probability is high, and ruled out when the combined probability is low. However, nearly half of the patients will still have an uncertain diagnosis for which further diagnostic procedures may be allocated.
Pulmonary embolism (PE) is a complex diagnosis and none of the existing imaging modalities have a 100% diagnostic specificity or sensitivity, particularly when it comes to subsegmental vessel emboli. The trend towards using spiral computed tomography (s-CT) directly visualizing the emboli is encouraged by several studies which show a specificity and sensitivity of over 90% for the diagnosis of central PE, i.e. main, lobar and segmental vessel location [1–3]. However, selective pulmonary angiography (PA) is considered the most specific test, but its availability varies and many clinicians hesitate to refer for PA because of an alleged high complication rate.
The diagnostic disadvantage of both PA and s-CT is the high frequency of interobserver disagreement when evaluating subsegmental vessels, but the clinical significance of untreated, undiagnosed subsegmental emboli, especially in patients with underlying cardiopulmonary disease, is not yet known.
However, ventilation/perfusion (V/Q) scintigraphy is still used as the first line imaging modality for PE in most hospitals worldwide. The purpose of this investigation was to evaluate if a structured clinical protocol assessing the likelihood of PE as estimated by a visual analogue scale (VAS), together with the scintigraphic probability also expressed by VAS, could improve the diagnostic accuracy and also denote when further examinations are called for.
Materials and methods
This study was performed at Danderyd Hospital, Stockholm, Sweden, which at the time served as a catchment area of 250 000 people. Between September 1991 and February 1994, there were 170 of 269 consecutive outpatients with a clinical suspicion of acute PE, who underwent chest X-ray, lungscan and PA (60 males, mean age 60 ± 15 years; 110 females, mean age 59 ± 17 years). Patients with known or suspected deep venous thrombosis were not included. Reasons for exclusion of 99 patients (33 males and 60 females, mean age 64 ± 17 years) were as follows; other disease than PE (n=15), refused to participate (n=23), too ill to participate (n=5), cath. lab had no capacity (n=41), contraindication to PA (n=4), and the investigation not being fulfilled according to protocol (n=11). All studies were carried out within 24 h from the patients’ arrival to the emergency ward. Informed consent was obtained from all patients and the study protocol was approved by the local ethics committee.
After admittance to the emergency ward all patients with a clinical suspicion of PE were screened for their probability of having PE by one of three physicians engaged in this project. The patients were evaluated according to a protocol (Table 1) including data about previous history, especially risk factors for thromboembolic disease, current symptoms and physical findings. The ECGs were scrutinized for signs of right ventricular load. It was up to the physician’s discretion to judge the clinical relevance of each positive item of the clinical protocol. The physician finally estimated the probability of PE by VAS. The VAS was a 100-mm long continuous line marked with 0% probability at its left margin and 100% at its right margin.
Clinical protocol for estimating the likelihood of pulmonary embolism in 170 patients
The clinical probability was subsequently divided in to three categories, i.e. low (1–25%), intermediate (26–75%) and high (76–100%).
All studies included a chest radiograph performed <24 h before the perfusion lung scintigraphy and PA, using standing posterior–anterior and lateral projections or when necessary, a portable radiogram.
Perfusion and ventilation scintigraphy were performed with a parallel-hole, low energy, all-purpose collimator on a γ camera (General Electric 400 AT, Waukesha, USA).
Perfusion lung scintigraphy was performed after intravenous injection of 75 MBq 99mTc macro-aggregated albumin (Solco Nuclear, Basel, Switzerland). Registrations were made with the patient in the supine position in eight standard projections (anterior, posterior, left anterior oblique, right anterior oblique, left posterior oblique, right posterior oblique and two laterals). Anterior and posterior projections were registered with 800 000 counts per view, and the remaining projections were registered with 400 000 counts per view.
Ventilation scintigraphy was performed only when the perfusion scan was considered abnormal. This decision was left to one of the two nuclear medicine physicians in charge. Of 170 patients, the ventilation was not performed in 35 cases. For those who underwent ventilation scintigraphy, 99mTc technegas (Tetley Technologies, Sydney, Australia) was used, and registrations were made in selected projections according to the findings of perfusion imaging. If performed on the same day as perfusion scintigraphy, the count rate at ventilation scintigraphy had to reach three times or more the activity of the remaining 99mTc macro-aggregated albumin. Most ventilation scintigrams were however, performed the day after the perfusion scintigraphy.
The scintigrams were evaluated by a physician experienced in nuclear medicine who also had access to the chest X-rays, but was blinded to the clinical information and the results of angiography.
Immediately after scrutinizing the scintigraphy according to the modified prospective investigation of pulmonary embolism diagnosis (PIOPED) criteria , the reader made a subjective personal probability estimate as a mark on a VAS. Later this mark was measured by means of a ruler and expressed as percent probability. The estimate was based solely on the personal experience and opinion of the reader. The result of the estimation was subsequently divided into four groups, 0, 1–25% (low), 26–75% (intermediate) and 76–100% (high) probability. This categorization only is used in the following.
All studied patients underwent PA, regardless of the scintigraphic findings of which the angiographer was blinded. PA was performed with cine-technique and standard Seldinger technique using the femoral approach. At least two oblique projections of each lung together with an additional PA-projection were performed. Only direct criteria of PE were used, i.e. an intraluminal filling defect or an occlusion with a concave border of the end of the contrast medium column, indicating a trailing edge of an embolus. The results and methods used are described in detail elsewhere . In short, concensus reading by three thoracic radiologists revealed 51 patients with PE.
In order to establish a ‘final outcome diagnosis’, a committee of experts including two radiologists, one nuclear physician and three internists analysed all available data during one session (i.e. patient records including follow-up data at 6 months, lungscan results, radiographic examinations including PA, clinical, laboratory, and autopsy findings) and decided whether the patient did or did not have PE at the time of the hospital episode.
In order to evaluate hypotheses of variables in contingency tables, the χ2 test was used or, in the case of small expected frequencies, Fisher’s exact test. In addition to that, descriptive statistics were used to characterize the data. All analyses were carried out by use of the SAS system, version 6.12 (Statistical Analysis System, Cary, NC, SAS Institute) and the 5, 1 and 0.1% levels of significance were considered. In the case of a statistically significant result the probability value (P-value) has been given.
The final outcome committee decided that 53 of 170 patients had PE. One case of subsegmental PE on PA was overruled and three cases with a negative PA were considered as having PE. Thus the prevalence of PE in this study was 31%.
Three patients had the diagnosis of PE reversed from negative to positive by the outcome committee. These three patients did not receive anticoagulants. One of these patients died 4 months after the pulmonary angiogram and was found to have recurrent PE at autopsy. The other two patients were uneventful at follow-up. The three readers in the angio consensus group had initially different opinions whether there were small subsegmental emboli or not. Finally they decided that the angiograms were negative. However, the clinical suspicion were reported as high in these patients. The scintigraphic probabilities were judged as low, intermediate and high, respectively.
One patient, in whom the angio consensus group judged the angiogram as positive, based on a possible subsegmental embolus, had the diagnosis of PE reversed from positive to negative by the outcome committee. The clinical suspicion was reported as low and so was the scintigraphic assessment.
Nine patients died during follow-up. Four of these patients had PE at the time of the hospital episode. One patient was found to have recurrent pulmonary embolism at autopsy and the outcome committee reversed this patient from negative to positive. The other three patients were treated for pulmonary embolism initially. Five patients with negative PAs, died during follow-up for other reasons, i.e. cardiac failure, cancer and respiratory insufficiency.
Using the clinical probability scores alone for the likelihood of PE gave the following results; Low; seven PE in 74 (9%) patients; intermediate, 25 of 70 (36%) and with high probability 21 of 26 (81%) patients had PE (Table 2).
Clinical probability of pulmonary embolism (PE) as estimated by a visual analogue scale in 170 patients with suspected PE
The scintigraphic assessment as expressed by VAS yielded a high diagnostic accuracy with 100% negative predictive value when the scan was read normal, 90% accuracy with low probability and a 94% positive predictive value when there was a high probability. In 45 of 170 (26%) patients the scans were classified as intermediate.
The lungscan score was considered normal or near normal (0 + low probability) in 94 patients. Amongst these, 27 patients had normal scans of whom none had PE. In the remaining 67 patients with low probability scans there were seven (10%) with PE. Intermediate probability scan score was considered in 45 patients and of these 17 (38%) had PE. In 31 patients with high probability lungscan score, 29 (94%) had PE (Table 3).
Scintigraphic probability of PE as estimated by a visual analogue scale in 170 patients with suspected PE
The sensitivity and the specificity of high probability lungscans was 55 and 98%, respectively. The positive predictive value of high probability lungscan was 94%, i.e. 29 of 31 with a high probability lungscan had PE according to the final outcome committee. Combining high clinical with high scan probability yielded 100% diagnostic accuracy.
Only one of 48 patients with combined low clinical probability and normal or low probability lung scan had PE, resulting in 98% negative predictive value.
In the group of both intermediate clinical and lungscan probability (26–75%) 10 of 20 patients had PE resulting in a specificity of 50%.
All 27 patients with intermediate or high clinical probability and high lungscan probability (76–100%) had PE. Also the combination of high clinical probability and intermediate lungscan probability yielded a 100% specificity, although the numbers were small (3/3) (Table 4).
Combined clinical and scintigraphic probabilities of PE as estimated by visual analogue scales in 170 patients with suspected PE
The present study was designed to study the diagnostic accuracy of combining a structured clinical probability judgement of PE with that of pulmonary scintigraphy. The main findings in our study were that a combination of high scintigraphic probability with intermediate to high clinical probability or intermediate scintigraphic probability and high clinical probability resulted in a reliable diagnosis. A normal lungscan ruled out PE. Combining a normal or low probability lungscan with low clinical probability yielded a negative predictive value of 98%.
The positive predictive value of high probability lungscans in our study was 94% compared with PIOPED  with 88% and prospective investigative study of acute PE diagnosis (PISAPED)  with 93%. It is noteworthy that adding the information of intermediate or high clinical probability yielded a 100% positive predictive value. The prevalence of PE was 31% in our study, 27% in PIOPED and 39% in PISAPED.
The clinical judgement alone revealed high diagnostic accuracy in the groups with low or high probability, i.e. 91 and 81%, respectively, thus including 100 of the 170 patients. The disadvantage with clinical judgement alone is that only experienced physicians reach high diagnostic accuracy. The physicians’ judgement may also be affected by variables that is not in a protocol, i.e. if the patient has visited the emergency ward earlier with a suspicion of PE, other colleagues’ influence and other reasons.
In the PIOPED study, 87% of the patients with abnormal and only 44% of those with normal or near normal lungscan had an angiography performed. In the PISAPED study only 62% of patients with abnormal scans underwent PA and patients with a normal lungscan were omitted for PA.
The strength in our study was that all of the 170 patients underwent PA, lung perfusion or V/Q scanning and chest X-ray within 24 h, thus avoiding a selection bias.
The V/Q lung scintigraphy results in large number of inconclusive results because: (i) only a minority of patients with suspected PE present with a high probability V/Q scan (31/170 or 18% in the present study); (ii) the high probability V/Q scan is very specific for PE (98% in the present study) but lacks sensitivity (55% in the present study); (iii) consequently, a large proportion of patients with proven PE (45% in the present study) do have V/Q abnormalities other than those of high probability V/Q scan category.
Several authors have proposed different diagnostic strategies including lower extremity ultrasound, s-CT, and PA or perfusion scintigraphy as the first-line imaging study for the diagnosis of PE [8–12]. The strength of s-CT and PA is that these modalities minimize the group of inconclusive patients because of direct visualization of the thrombus thus resulting in a correct diagnosis and treatment in a higher number of patients. Combining the clinical probability with findings on s-CT or PA might enhance the diagnostic value even more.
We have shown that by applying a model of combined clinical and scintigraphic probabilities for PE on a consecutive population, the diagnosis is ruled in when the combined probability is high, and ruled out when the combined probability is low, both in a highly reliable way. However, nearly half of the patients in such a population will still have an uncertain diagnosis for which further diagnostic procedures may be allocated.