See also pp. 43–51 and 115–116
In this issue of Emergency Medicine Australasia, two contributions appear that are likely to influence our current understanding of the value, utility and safety of whole-body computed tomography (CT) scan in major trauma, coined the ‘pan-scan’.
Asha and colleagues from the University of New South Wales report the results of a large cohort study comparing radiation exposure to trauma patients before and after the introduction of a pan-scan diagnostic algorithm employing a last-generation 64-slice scanner.1 Their findings are both intriguing and alarming. Patients undergoing a pan-scan, according to the calculated risk ratio, were 1.7 (95% confidence interval [CI] 1.3–2.2) times more likely to receive a radiation dose exceeding 20 mSv compared with a conventional CT work-up, but did not significantly benefit from the procedure in terms of the incidence of missed injuries (0.6% vs 0.9%).
The associated likelihood of being helped versus harmed (calculated as the ratio of the number needed to treat divided by the number needed to harm) was 333.3/12.8 = 26, which, irrespective of more sophisticated health-economic and utility analyses, suggests that the pan-scan might be 26 times as likely to harm patients in the long term as to help them in the acute setting. The results are even more concerning as they cast doubt on the common assumption that last-generation hardware and scanning protocols can minimize radiation exposure while increasing the diagnostic yield of the procedure.
Meanwhile, in their research letter, Cowan et al. stressed that many pan-scans may not be indicated (88/137, or 64% [95% CI 56–72]), and that so-called occult injuries requiring therapeutic action remain rare events (3/88, or 3% [95% CI 1–10]).2 Cowan et al.'s results are in accordance with reports from other authors who calculated a 0.3%3 and 0.6%4 incidence of missed relevant traumatic lesions, if the pan scan had been omitted for selective radiology.
Health-care professionals from various backgrounds, payers, patients and their relatives may ask why so many trauma surgeons and emergency physicians around the globe increasingly opt for an imaging modality (pan-scan) that apparently does more harm than good? A spontaneous but unsatisfying answer is: because there is no better diagnostic test available. Most trauma centres now provide a CT suite close to the ED. Transfer times are no longer a problem, nor are scanning times or the number of patients not responding to any attempt at damage-control resuscitation.
The underlying problem, however, is much more complex and requires a holistic approach. The lifetime risk attributable to medical imaging, specifically CT scans (apart from exposure to natural background radiation and other sources), remains theoretical, whereas a missed aortic dissection (although rare) may determine immediately between life and death. It may be worthy to step back to the hierarchical three levels of diagnostic test research to put the present research into context: (i) Is the test capable of distinguishing between a person with and without the condition of interest (efficacy or accuracy level)? (ii) Do the test results markedly influence clinical decision-making (effectiveness level)? (iii) Do patients benefit from test-based decision-making (efficiency level)?
There is little published information about the diagnostic accuracy of the pan-scan or its impact on decision-making in the setting of severe trauma. There is, however, evidence that established tests, such as ultrasound and plain X-rays, have too low a sensitivity to reliably exclude injuries of the abdomen, chest, spine and pelvis when negative. Interestingly, most validation studies used CT as the diagnostic reference standard.5,6 This leads to a bizarre situation: what is the ultimate reference standard to determine the accuracy of an accepted diagnostic gold standard, such as CT?
In philosophy, the inability to meet a certain demand or to answer a normative question is explained by the first bridge principle ‘Should implies Can’, as coined by Hans Albert.7 If something is required, it must be achievable. Many trauma patients present with haemorrhagic shock and a centralized circulation, and thus too early a scan may simply be unable to detect organ lacerations and contusions that demarcate only after minimal tissue perfusion has been established. A missed injury by a scan performed shortly after admission and resuscitation may not be classified as a false-negative finding. Also, should injuries depicted on scans that are missed or misinterpreted by radiological or surgeon readers be classified as false-negatives? Human factors play an important role in the health-technology assessment of imaging tests.
Lacking efficacy and conflicting effectiveness data seem to be in overwhelming contrast to the survival benefit demonstrated with use of the pan-scan use.8 It must be stressed that the reported survival benefit refers to the ratio of predicted versus observed trauma deaths, not to a favourable difference in raw event rates. Another explanation of this finding is stage migration, meaning that the pan-scan detects injuries that artificially increase injury severity, thereby decreasing the forecasted probability of survival without affecting management decisions.6
The pan-scan has quickly become a standard of care, before it has had any chance to be rigorously evaluated according to scientific standards. This is similar to the triumphal procession of surgical innovations, such as laparoscopic cholecystectomy, and vertebroplasty for osteoporotic vertebral compression fractures. The results of properly designed randomized trials trail off in the light of sensational stories in the mass media, which fuel patients' demands for new therapeutic options, and place pressure on institutions to offer those interventions irrespective of their proven value.
Compelling plausibility and convenience are knockout arguments that easily outrival concerns about the cost–benefit ratio of a medical novelty. According to Buxton's law, it is always too early for a rigorous evaluation of health technology until it is suddenly too late!9 The number of trauma centres in developed countries that have adopted the pan-scan algorithm may now exceed Gladwell's tipping point, and it will be impossible to turn back this trend.10
However, we may still be able to improve the therapeutic index of the trauma pan-scan by developing more accurate triage criteria. In Asha's study, less than one-fifth of their patients fulfilled the criteria of multiple trauma (e.g. Injury Severity Score ≥ 16). We must schedule only those patients to a pan-scan who deserve it (by means of the clinical pretest probability, supplemented for instance by results from imaging studies, such as thoraco-abdominal ultrasound, lab tests, such as S100B protein concentrations, and digital X-rays of the chest, spine and pelvis). The Advanced Trauma Life Support (ATLS; American College of Surgeons, Chicago, IL, USA) programme must in the future include decision criteria in favour of and against pan-scans. We should refrain from sending a patient for a pan-scan too early to avoid diagnostic gaps caused by insufficient arterial tissue perfusion. Finally, we must call for a second reading by experienced radiologists to fully capitalize on the diagnostic power of this method. Adhering to these simple prerequisites may considerably enhance the diagnostic gain, lead to better individualized care and reduce harms of the CT pan-scan to the general trauma patient.