| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
Department of Radiology, Chelsea & Westminster Hospital, 369 Fulham Road, London SW10 9NH, UK
Certain truisms preface many articles about pulmonary embolism (PE). It is correctly emphasized that PE is a difficult diagnosis to make. In one much quoted post-mortem series, published nearly 30 years ago, two-thirds of patients with an in vivo diagnosis of PE had no evidence of embolism at post-mortem [1]. In the same study, pulmonary emboli were correctly diagnosed in only one-third of patients in whom they where subsequently found at post-mortem. The false negative and false positive rates for the diagnosis of PE were therefore nearly the same and were spectacularly poor.
Most pulmonary emboli arise in the pelvic or lower limb vessels. Clinical PE will occur in up to half of patients when acute above-knee deep vein thrombosis (DVT) is left untreated [2]. A significant number of asymptomatic patients will undergo subclinical PE [3].
The dangers of leaving PE untreated are obvious, if not based on hard evidence. Emboli tend to recur and, although the condition may be self-limiting, it may also result in sudden death or prolonged morbidity. The mortality rate from untreated PE is usually quoted as 30%. The fact that this figure is based upon a randomized trial in which the non-treatment arm included only 19 patients is less well recognized [2]. However, the dangers of known but untreated PE are now accepted as being so high that the 30% figure is unlikely to be redefined. What is also generally accepted is that institution of anticoagulation, nowadays with low molecular weight heparin, greatly reduces the subsequent morbidity and mortality in patients known to have DVT or PE [4].
The imaging investigation of any patient with a possible acute PE will commence, almost without exception, with a chest radiograph. Whatever the result, further investigation will almost invariably follow once the possibility of a pulmonary embolus has been raised in the mind of the referring clinician. Although the chest radiograph is unable to confirm the diagnosis of an embolism, it is still of considerable value in defining the subsequent diagnostic pathway.
In the UK, by far the next most common investigation is a perfusion scintigram [5]. This is a well established investigation, perceived as being safe and widely available, and may be extended to include a ventilation component when necessary. As it is non-invasive and there is a familiarity with the results, this technique is usually requested with equanimity. However, regular users will recognize that it suffers from a number of shortcomings. Perfusion scintigraphy can provide only indirect evidence of the presence of an embolus, and there have been a number of attempts to clarify and improve reporting methods. Many of these refinements have been instituted since the PIOPED study [6], but even the most widely accepted of them, such as those proposed by thePISA-PED study [7], leave a large cohort of patients without a definitive diagnosis. Indeed, depending on the study in question, between 30% and 70% of patients will still be left drifting in a diagnostic middle ground of uncertainty [8–10]. Part of this problem stems from the fact that perfusion scintigraphy is more likely to fail to be definitive when the chest radiograph is abnormal. For example, the relative shortcomings of scintigraphy in the presence of underlying chronic obstructive pulmonary disease (COPD) were highlighted by the study of Hartmann et al [11]. They demonstrated that whilst COPD significantly diminishes the diagnostic performance of scintigraphy, it does not affect the performance of clinical probability estimation, D-dimer testing, CT pulmonary angiography (CTPA) or pulmonary angiography.
It must be emphasized that there remains a sizeable group of patients in whom perfusion scintigraphy is of great utility, namely those patients in whom the clinical assessment and ventilation–perfusion (V/Q) scan results are concordant and give either a normal or high probability result. In the PIOPED study this was a minority of patients, and it has varied from study to study since 1990, but realistically is between 30% and 40% [9]. The fact remains that whilst the PIOPED study demonstrated that nearly all patients with a PE had an abnormal scintigram (sensitivity 98%), so did most patients without a PE (specificity 10%).
Logic (and nowadays clinical governance) would suggest that the one-half to two-thirds of patients left without a diagnosis should be further investigated. Although the debate about how this should be undertaken may continue on paper, in practice the debate is largely over. Pulmonary angiography is performed infrequently in this country. In a recent postal survey, only 1 pulmonary angiogram and 5 CTPA examinations were undertaken for every 100 perfusion or V/Q scans performed for suspected acute PE [5]. It is likely that the number of CTPA examinations for PE has increased significantly in the 3 years since this study. However, it remains likely that the number of patients not investigated further still far exceeds the abilities of perfusion scintigraphy, whether or not combined with clinical assessment, to satisfactorily resolve the diagnosis of possible acute embolism. Those radiologists who undertake pulmonary angiography regularly and often, and who continue to find it a safe and robust technique, would find themselves rather busier if they were asked to investigate the 60–70% or so of patients who have an equivocal V/Q scintigram result. It therefore seems an inescapable conclusion that many patients who undergo V/Q scintigraphy are treated and discharged without further investigation, even though a firm diagnosis is never made. It also seems likely that the downward trend in pulmonary angiography uptake is unlikely to reverse, even if it ceases to decline.
If asked to design the ideal test for PE, most would suggest that it should be non-invasive and yet be able to provide direct visualization of thrombus within pulmonary arteries. CTPA appears to fulfil these criteria. In common with the journey from introduction to acceptance of many other tests, the initial enthusiasm for CTPA has now been tempered with a little realism. Nevertheless, the sensitivities and specificities for pulmonary emboli detection by CTPA at main, lobar and segmental levels are almost always reported as greater than 90% [12–18]. This impressive track record has led some commentators [19] to suggest the abandonment of most other investigations in favour of CTPA, although in this country most would agree with the opinion that this is not yet appropriate [20]. Many of these studies of CTPA have also emphasized the improved interobserver agreement of CTPA (83–95%) compared with V/Q scintigraphy for intermediate or low probability studies (70%), the ability to make alternative diagnoses in that majority of patients without a PE, and the low incidence of non-diagnostic studies. There is also evidence that the CTPA technique may be extended to include assessment of the pelvic and lower limb veins. Cham et al [21] looked at the added value of indirect CT venography and CTPA over CTPA alone in the investigation of suspected PE. 93 of 541 patients examined had pulmonary emboli, and an additional 16 patients had DVT but no pulmonary emboli. Very few of these patients had thrombus that would not have been detected by lower limb ultrasound.
The optimal technique required for CTPA has been refined since 1992 and will change further as multislice CT systems become more commonplace. There has been a recent trend towards narrower collimation, enabling more reliable depiction of segmental and subsegmental arteries [22]. There has also been some emphasis on the use of work station-based image analysis rather than traditional hard copy images, an approach that resulted in the detection of up to 25% more emboli in one study [23]. The CT signs of PE are now well defined [8] and, while familiarity with the technique is required for accurate and confident interpretation, the cardinal sign of an intraluminal filling defect remains the central diagnostic criterion for an embolus.
Most of the above studies were confined to the detection of emboli in the central lobar and segmental vessels. The prevalence and significance of subsegmental emboli remains open to question, and varies from as low as 5% [6, 12] to as high as 36% [24], although it has been estimated that approximately 10% of patients in an unselected population will have isolated subsegmental emboli. While these emboli may be difficult to detect, it should be borne in mind that interobserver agreement for two observers for these emboli on angiography was only 66% and fell to 13% for three observers [25]. Baile et al [26] recently undertook a meticulous study designed to address the question of the relative accuracies of CTPA and conventional angiography for the detection of subsegmental sized thrombi. They embolised methacrylate beads into the pulmonary circulation of pigs, which was followed by CTPA and conventional pulmonary angiography. These studies were interpreted separately and the number and location of the artificial emboli were determined and compared with casts made of the pulmonary circulation to show the true location of the methacrylate beads. No difference in the abilities of CTPA and pulmonary angiography was found in this setting.
The clinical significance of subsegmental emboli remains unclear. While they may herald the arrival of a subsequent significant embolus, follow-up studies of patients with negative CT studies who are left untreated suggest a re-presentation rate following a normal CT of less than 1% [27, 28], similar to that found after a normal V/Q scan [6] or angiography [29].
When the price of CTPA is compared with the unit price of a perfusion scintigram there seems no doubt that perfusion scintigraphy is cheaper. However, a number of studies suggest that, although cheaper, it is not a more cost effective means of reaching a diagnosis [30], particularly if access is limited during the working week. The hotel costs for prolonging patient admission prior to investigation for PE amount to around £350 per day. If this is coupled with the fact that many, if not the majority, of patients undergoing perfusion scintigraphy will be left with a significant degree of diagnostic uncertainty requiring further investigation, then the discrepancy becomes even more apparent.
The answer to the question "should all patients with suspected PE undergo CTPA?" must remain an emphatic "no" for a number of reasons [20]. First, the current provision of CT services in the UK, with many competing pressures for table time, renders widespread CTPA an impractical solution to a difficult but common problem. Second, CT already contributes 20% of the medical radiation dose to the general population, even though it accounts for only 2% of the total number of investigations. Wider application of CT to solve acommon clinical question must therefore be coupled with a locally agreed method of patient selection to ensure only those patients with a high pre-test probability undergo CTPA. The choice of assessment method is varied, but the approach suggested by Wells et al [31] seems simple and robust. Use of a simple risk stratification score coupled with a D-dimer level allowed correct assignment of risk in patients presenting with suspected PE. In patients defined as PE unlikely, PE was absent on further investigation in 98%.
An approach utilizing clinical risk assessment and D-dimer determination has been estimated to have the potential to reduce overall demand for V/Q and CT examinations by at least 40% [30]. If PE is still felt to be likely after this assessment then the subsequent diagnostic pathway could be reasonably determined by the initial chest radiograph and the absence of clinical evidence of DVT (which should prompt directed investigation). Patients with normal chest radiographs and without pre-existing cardiopulmonary disease, including previous PE, are much more likely to undergo a diagnostic perfusion scintigram and should therefore be investigated accordingly. Patients with abnormal chest radiographs should undergo CTPA as the initial radiological test of choice.
In conclusion, lung scintigraphy continues to play an important role in a significant proportion of patients presenting with possible PE. In a recent extensive review [32], it was suggested that about 25% of patients with suspected PE will have the diagnosis refuted by normal scintigraphy, and anticoagulants may be safely withheld. Another 25% with suspected PE will have a high probability study and anticoagulant therapy may be commenced. The remainder require—although may not undergo—further investigation. There is an ever-growing body of evidence to suggest that CT is capable of replacing angiography as the definitive first or second line test for suspected PE and, with further refinements in technology and technique, it may eventually prove to be superior to angiography in many regards.
Received for publication April 30, 2001. Revision received August 23, 2001. Accepted for publication September 18, 2001.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
