Background

The Cochrane Database of Systematic Reviews published a manuscript critical of the use of the FAST examination. The reference is Stengel D. Bauwens K. Sehouli J. Rademacher G. Mutze S. Ekkernkamp A. Porzsolt F. Emergency ultrasound-based algorithms for diagnosing blunt abdominal trauma. Cochrane Database of Systematic Reviews. (2):CD004446, 2005. UI: 15846717 [1]. The stated objective was the assessment of the “efficiency and effectiveness” of ultrasound-inclusive evaluative algorithms in patients with suspected blunt abdominal trauma (BAT). In the 4 trials with 1,037 patients reviewed, the primary outcome measures explored were mortality, CT use, and DPL and operative intervention rates. Little or no benefit was seen and the conclusion was that “there is insufficient evidence from randomized controlled trials to justify promotion” of FAST in patients with BAT.

While the review used the same rigorous methods employed in all Cochrane Reviews, it appears that several serious flaws plague the manuscript. First, basic power calculations using data from the National Trauma Data Bank of the American College of Surgeons [2] suggest that reliably measuring a mortality difference would require over 5,000 patients rendering the first outcome measure inappropriate given the combined sample size of all published studies. Second, published literature [3–6] included in the review) have shown significant reductions in CT and DPL use, when FAST is used. Finally, finding no difference in operative intervention rates in study and control patients in the RCTs (Boulanger, Rose, and Melniker), which was criticized in the Cochrane Review, in fact indicates the assignment groups were well match, is a strength of the reviewed studies, not a weakness.

The finest methodological rigor cannot yield usable results, if they are not applied to the right, clinically relevant question. Do we need FAST to visualize any amount of free fluid or any degree of organ injury? In a world of increasingly conservative management of BAT, the obvious answer is no. However, quantifying the ability of FAST to predict the need for immediate operative intervention (OR) is essential. Toward this end, two investigations were undertaken.

In a post hoc analysis of the SOAP-1 trial, of the 69 study arm patients with blunt torso trauma necessitating OR, the FAST correctly identified all patients needing OR and cleared all patients not requiring immediate thoraco-abdominal surgery. Of interest, there were no non-therapeutic laparotomies and CT identified no other patients with intra-abdominal or intra-thoracic injuries requiring operative intervention that had not been identified by FAST.

Next, a literature review was conducted using verbatim methodologies as used in the Cochrane Review with the exception of telephone contacts. This Cochrane-sk study is described below.

Methods

Criteria for considering studies for this review

Types of studies

Randomized and quasi-randomized controlled trials compared trauma algorithms with ultrasonography, alone or in combination with other established diagnostic tests (i.e., computed tomography [CT], diagnostic peritoneal lavage [DPL], clinical monitoring), to algorithms without the use of ultrasound. Reporting of FAST findings, other test findings, operative findings, and explanations of false-positive and -negative cases was required for inclusion. Trials were included irrespective of blinding, and number of patients randomized.

Types of participants

Hemodynamically stable or unstable patients with suspected torso injury after blunt trauma, as a single injury or an injury accompanying multiple trauma, were included. Studies investigating patients with only stab wounds and gunshot wounds were excluded.

Types of intervention

Diagnostic algorithms including ultrasonography to detect free intra-abdominal, intra-thoracic, and/or intra-pericardial fluid (focused assessment of sonography for trauma [FAST] or enhanced FAST [eFAST]), including ultrasound examinations performed by radiologists, non-radiologist clinicians, or ultrasound technicians, in combination with subsequent confirmatory tests (i.e., CT, DPL, OR reports, or clinical monitoring).

Objective

The objective is to study whether diagnostic algorithms using ultrasonography in the emergency department or trauma bay accurately predict which patients with blunt torso trauma require immediate OR.

The following hypotheses were tested:

  • That a positive FAST is predictive of a need for OR.

  • That a negative FAST is predictive of no need for OR.

Search methods for identification of studies

Trials indexed in MEDLINE and PUBMED between 1966 and May 2009 were identified by the following strategy:

  1. 1

    Abdominal injuries

  2. 2

    Thoracic injuries

  3. 3

    Wounds, nonpenetrating

  4. 4

    Multiple trauma OR polytrauma

  5. 5

    Retroperitoneum

  6. 6

    Rupture

  7. 7

    Shock, traumatic

  8. 8

    Hemoperitoneum OR haemoperitoneum OR free fluid OR intraperitoneal fluid

  9. 9

    Spleen OR splenic

  10. 10

    Liver OR hepatic

  11. 11

    Accidents

  12. 12

    Accidents, traffic

  13. 13

    Seat belts

  14. 14

    Bicycling

  15. 15

    Motorcycles

  16. 16

    Ultras* OR echotomogr* OR sonogr*

  17. 17

    Focused assessment of sonography for trauma OR FAST OR emergency ultras*

  18. 18

    (1 OR 2 OR 3 OR 4 OR 5 OR 6 OR 7 OR 8 OR 9 OR 10 OR 11 OR 12 OR 13 OR 14 OR 15) AND (16 OR 17)

  19. 19

    Randomised controlled trial OR randomized controlled trial

  20. 20

    Random allocation

  21. 21

    Double blind method

  22. 22

    Single blind method

  23. 23

    (19 OR 20 OR 21 OR 22)

  24. 24

    18 AND 23

Trials covered by EMBASE back to 1980 were tracked by

  1. 1

    ‘Intermethod comparison’/exp

  2. 2

    ‘Randomized controlled trial’/exp

  3. 3

    ‘Non invasive measurement’/exp

  4. 4

    1 OR 2 OR 3

  5. 5

    ‘Peritoneal fluid’/exp

  6. 6

    ‘Hemoperitoneum’/exp

  7. 7

    ‘Spleen rupture’/exp

  8. 8

    ‘Spleen injury’/exp

  9. 9

    ‘Liver injury’/exp

  10. 10

    ‘Multiple trauma’/exp

  11. 11

    ‘Abdominal blunt trauma’/exp

  12. 12

    ‘Abdominal bleeding’/exp

  13. 13

    ‘Thoracic injury’/exp

  14. 14

    ‘Thoracic bleeding’/exp

  15. 15

    5 OR 6 OR 7 OR 8 OR 9 OR 10 OR 11 OR 12 OR 13 OR 14

  16. 16

    ‘Peritoneum lavage’/exp

  17. 17

    ‘Clinical observation’/exp

  18. 18

    ‘Spiral computer assisted tomography’/exp

  19. 19

    ‘Diagnostic approach route’/exp

  20. 20

    16 OR 17 OR 18 OR 10

  21. 21

    ‘Echography’/exp

  22. 22

    ‘Ultrasound scanner’/exp

  23. 23

    ‘Ultrasound transducer’/exp

  24. 24

    21 OR 22 OR 23

  25. 25

    4 AND 15 AND 20 AND 24

Electronic databases

The Cochrane Injuries Group Specialized Register and CENTRAL (the Cochrane Central Register of Controlled Trials) were searched, as were the databases of the publishers SpringerLink (including the journal Abdominal Imaging, Emergency Radiology), Elsevier (including the journal Annals of Emergency Medicine,), Wiley (including the journal Academic Emergency Medicine, British Journal of Surgery), Lippincott Williams and Wilkins (including the Journal of Trauma, Annals of Surgery, Critical Care Medicine, Shock, Journal of Computer Assisted Tomography). Finally, searches on web-based resources including the Radiological Society of North America (RSNA, covering the journals Radiology and Radiographics as well as the RSNA Index to Imaging Literature), trials registers (such as Current Controlled Trials) and Google were run. The reference lists of all retrieved articles were reviewed for further trials.

Abstract searching

Abstracts presented to the following international scientific societies were searched: Society for Academic Emergency Medicine (1999–2008), the American College of Emergency Physicians (1999–2008), the American Association for the Surgery of Trauma (1999–2007), and the World Congress on Ultrasound in Emergency and Critical Care Medicine (2005–2008).

Methods of the review: trial identification and selection

The titles or abstracts of all studies identified were assessed by the initial search and excluded clearly non-relevant studies. Full text articles were obtained for potentially relevant studies and any studies with unclear methodology. All these studies were assessed as to whether they met the inclusion criteria for this review, their method of randomization or quasi-randomization, and their adequacy of allocation concealment.

Data extraction

The author extracted the results of each included paper on a data extraction sheet.

Assessment of methodological quality

Each included trial was read for the following aspects of internal and external validity.

  1. A.

    Was the assigned treatment adequately concealed prior to allocation?

    • 2 = method did not allow disclosure of assignment;

    • 1 = small but possible chance of disclosure of assignment or unclear;

    • 0 = quasi-randomized or open list/tables.

  2. B.

    Were the outcomes of patients/participants who withdrew described and included in the analysis (intention to treat)?

    • 2 = withdrawals well described and accounted for in analysis;

    • 1 = withdrawals described and analysis not possible;

    • 0 = no mention, inadequate mention, or obvious differences and no adjustment.

  3. C.

    Were the outcome assessors blinded to the results of the index test (i.e., ultrasonography) and/or reference tests and/or patient outcome?

    • 2 = effective action taken to blind assessors;

    • 1 = small or moderate chance of unblinding of assessors;

    • 0 = not mentioned or not possible.

  4. D.

    Were the treatment and control group comparable at entry?

    • 2 = good comparability of groups, or confounding adjusted for in analysis;

    • 1 = confounding small or mentioned but not adjusted for;

    • 0 = large potential for confounding, or not discussed.

  5. E.

    Were care programs, other than the trial options, identical?

    • 2 = care programs clearly identical;

    • 1 = clear but trivial differences;

    • 0 = not mentioned, or clear and important differences in care programs.

  6. F.

    Were the inclusion and exclusion criteria clearly defined?

    • 2 = clearly defined;

    • 1 = inadequately defined;

    • 0 = not defined.

  7. G.

    Were the interventions clearly defined?

    • 2 = clearly defined interventions are applied with a standardized protocol;

    • 1 = clearly defined interventions are applied but the application protocol is not standardized;

    • 0 = intervention and/or application protocol are poor or not defined.

  8. H.

    Were the outcome measures used clearly defined (by outcome)?

    • 2 = clearly defined;

    • 1 = inadequately defined;

    • 0 = not defined.

  9. I.

    Was the surveillance active, and of clinically appropriate duration?

    • 2 = active surveillance and appropriate duration;

    • 1 = active surveillance, but inadequate duration;

    • 0 = surveillance not active or not defined.

Data analysis

Mean differences and 95% confidence intervals were calculated for continuous variables. For dichotomous outcomes, proportions with 95% confidence intervals were calculated. Due to the heterogeneity of the data, meta-analyses and mixed regression modeling were not conducted.

Description of studies

The search delivered 487 citations of studies investigating the use of ultrasound in torso trauma. Since ultrasound findings prompted different forms of further investigation, care programs varied between groups. Most studies examined the diagnostic accuracy of ultrasonography to detect free intra-peritoneal fluid, leaving 160 studies for further screening.

Identified were 49 studies [3–52] that compared the effectiveness and efficiency of ultrasound-based clinical pathways to algorithms that did not incorporate ultrasound examinations. Thirty-eight of these were retrospective, did not define an allocation schema, or did not describe the operative findings nor describe the “false-negative cases” and were excluded from further analysis.

The 11 remaining trials were included in the formal review (See Table 1).

Table 1 Assessment of methodological quality
Full size table

Methodological quality

The Melniker [3] study was a randomized clinical trial to primarily assess the effect of point-of-care, limited ultrasonography (PLUS) for trauma on the time to operative intervention; secondary outcomes included use of other diagnostics, hospital and ICU length of stay, and hospital charges. Regression models controlled for confounders and analyzed physician-to-physician variability. All analyses were conducted on an intention-to-treat basis. Results were presented as mean, first-quartile, median, and third-quartile with multiplicative change and 95% confidence intervals; or percentage with odds ratio and 95% confidence intervals. 444 patients with suspected torso trauma were eligible; 136 lacked consent and attendings refused enrollment of 46. 262 patients were enrolled: 135 PLUS and 127 controls; 45 patients were discharged from the ED or “Walked Out AMA”, leaving 111 PLUS and 106 Control patients in the final analysis. There were no baseline differences between groups. Time to OR was 64% (48, 76) less for PLUS compared to control patients. PLUS patients underwent fewer CT, Odds Ratio = 0.16 (0.07, 0.32), spent 27% (1, 46) fewer days in hospital, suffered fewer complications, Odds Ratio = 0.16 (0.07, 0.32), and charges were 35% (19, 48) less compared to control. The authors concluded that a PLUS-inclusive protocol significantly decreased time to OR in patients with suspected torso trauma, with improved resource utilization and lower charges.

One of the randomized trials [7] met some of the design standards. Patients were assigned by a computer-generated list, although it was not clear whether concealment was maintained. Sample size considerations called for 50 patients in each group to detect a 20% difference in CT scan use between groups. A secondary outcome (30-min difference in time to operative intervention) mandating inclusion of 420 patients was mentioned in the methods section of the original paper. However, no data were provided on this endpoint. A flowchart sketched the study profile according to the CONSORT recommendations.

Two other studies enrolled patients in a quasi-randomized fashion. The suitable algorithm was defined by ultrasound availability: ultrasound on weekdays from 8 a.m. to 5 p.m; no ultrasound on weekdays from 5 p.m. to 8 a.m. and on weekends [4] or the presence of one of the investigators [5]. Since no patient had the opportunity to influence the date of injury, these methods were considered proper random allocation.

In general, details of the study populations in the remaining 7 papers were sparse or missing.

Results

In the 11 studies included in the review, 2,755 patients were prospectively evaluated with Focused Assessment with Sonography in Trauma (FAST) and 448 (16%) went to operative intervention. The data demonstrated that for the detection of any amount of free fluid by ultrasound the sensitivity was 90.6% (95% CI) and the specificity was 98.6% (95% CI). As a screening tool to assess the “Need for OR”, the sensitivity was 94.2% (95% CI) and the specificity was 98.1% (95% CI).

The published reports indicated that 26 (5.8%) patients had false-negative FAST, but upon further review, 21 of these cases did not undergo operative intervention or the FAST was not done contemporaneously with the decision to operate, e.g., negative FAST on presentation and on Hospital Day-2 instability developed necessitating operation, which revealed hemoperitoneum. Two cases were associated with rare blunt trauma-induced small bowel perforation without hemoperitoneum, an injury type that CT is insensitive to identifying; and 3 FAST exams were technically poor and, therefore, uninterpretable resulting in a legitimate false-negative rate of 1.1 or 98.9% sensitivity for the “Need for OR”, when an adequate FAST exam was completed.

The debate at WCU2

Dr. Stengel, primary author of the Cochrane Review on FAST, was contacted in August 2005 and invited to attend the 2nd World Congress on Ultrasound in Emergency and Critical Care Medicine (WCU2) in June 2006 in New York City to debate the ““Emergency ultrasound-based algorithms for diagnosing blunt abdominal trauma (Review), from the Cochrane Collaboration.” The Congress organizers funded the air travel and accommodations in New York for Drs. Stengel, Bauwens, and Sehouli. Participants in the debate were Drs. Melniker (Chair, WCU2 Organizing Committee and Principal Investigator, SOAP-1 Trial), Dulchavsky (representing the American College of Surgeons), and Kirkpatrick (enhanced FAST investigator). On June 12, 2006, the Opening Plenary Session of WCU2 featured the debate of the Cochrane Review on FAST, which was digitally recorded.

Dr. Stengel and his colleagues presented the review and each speaker and several delegates to the Congress offered comments and posed questions. Key points from the session were that the review, while methodologically strong, did not evaluate all critical endpoints needed to judge the effectiveness of FAST. Dr. Stengel agreed to include the commentary from the debate and any new published data in all future updates of the Cochrane Review. Furthermore, it was agreed that more investigation was needed and, toward that end, it was desirable to establish a “FAST Registry” or build on existing registries, e.g., the National Trauma Data Bank of the American College of Surgeons, to include more FAST-related data points.

Cochrane Review of FAST: 2008 update

In January of 2008, the Cochrane Review of FAST was updated by Stengel et al. [53]. The update neither includes any methodological changes nor any new endpoints analyzed. There was no mention of the debate at WCU2 in New York or the need for a FAST Registry. Some limited aspects of the Melniker et al. [3] findings were presented, but it was erroneously stated that the data were analyzed in a manner “contradicting the intention-to-treat principle” and included the following statement: “We did not receive a response to our letter to the research team.”

The 2008 Update concluded: “There is currently insufficient evidence from RCTs to justify promotion of ultrasound-based clinical pathways in diagnosing patients with suspected blunt abdominal trauma.” The authors recommended widespread use of CT of the chest, abdomen, and pelvis for patients with blunt torso trauma.

The Cochrane Reviewers

The authors of the Cochrane Review on FAST are well published with over 100 citations, individually and collaboratively, but only 3 references for ultrasound-related investigations. None of the 3 ultrasound studies involved prospective investigations or any use of ultrasound by the authors; all are literature reviews with data-pooling and meta-analyses and each speak to a lack of methodological rigor in most studies of clinician-performed ultrasound for trauma.

Stengel et al. are qualified researchers, but lack professional investment in point-of-care testing, specifically, bedside ultrasound. They have demonstrated a predilection for the use of CT in their studies and recommendations.

Discussion

Many torso injuries do not require repair; therefore, operative intervention rather than the presence of free fluid or organ injury is the optimal endpoint for study. Although testing is normally thought of as identifying the presence of injury, not “need for treatment”, the unstated assumption is that all “injuries” require “treatment”, and it is clear that with torso injury, especially in the pediatric population, this is not the case. Likewise, after head CT, not all patients with subdural hemorrhage are taken to neurosurgery, but for the purpose of defining the effectiveness of CT, it is reasonable to ask whether all patients ultimately requiring neurosurgery are identified by a positive CT. What is essential to know is which patients require immediate operative intervention, the direct and not a proxy endpoint is preferred.

Need for operative intervention, as opposed to the decision to operate, can be objectively defined and determined prospectively or retrospectively. This allows determination of whether a positive FAST exam result accurately predicts those patients who require immediate OR from those who do not. The ways in which this result is integrated into the surgical decision-making process are, of course, subjective, because not all patients with positive FAST scans are taken to OR.

Indeed, a part of the impetus to bring ultrasound machines into the trauma bay in the first place was the high reported rates of non-therapeutic laparotomies caused by the oversensitivity of the diagnostic peritoneal lavage (DPL), in the range of 20–25%. This is probably a result of growing recognition of the ability of many intra-abdominal injuries to heal without surgical repair. It should be noted that no clear and consistent definition of non-therapeutic laparotomy has been reported. It is also apparent that there is a clinically relevant, albeit small, incidence of complications with DPL, many of which require surgical repair. FAST allows the grading of hemoperitoneum, a great improvement over the ‘Yes/No’ binary response from DPL.

The primary measure of accuracy of a screening test is its ability to “rule out” a disease process, characterized by sensitivity: a measure of false negativity. The results of this review demonstrate a sensitivity of 94.2%, with 422 of 448 ultrasound-screened patients who needed operative intervention having positive findings on their FAST scan. Twenty-one of the 26 patients without a positive scan had either delayed onset of internal bleeding, which is not a deficiency of FAST—it is designed to demonstrate pooling of blood in body cavities, not to predict future bleeding. Excluding “pseudo” false negatives in the sensitivity analysis enhances the rate of detection of “Need for OR” to 422 of 427 patients, or 98.9%. This figure is more representative of the true value of adequately completed FAST exams. Five patients who needed operative intervention were not identified on their FAST exams, which upon review 3 exams were found to be technically inadequate. The other 2 false-negative findings were in patients with rare blunt trauma-induced small bowel perforations without hemoperitoneum whose injuries were not of such severity as to make the delay of clinical significance and, interestingly, CT is also insensitive to detect these injuries. Although proponents of CT may argue that current results show better accuracy, the logistic difficulties and dangers of transport during the first few minutes after a trauma patient’s arrival are such that it is often not feasible as a screening test. The sensitivity of 94.2–98.9% found in this review compares favorably with CT results.

The specificity was 98.1% (2263 negative studies among 2,307 ultrasound-screened patients who did not require operative intervention). Thirty of the 44 patients whose FAST exam was judged clearly positive were found not to require operative intervention. Only 14 (3.0%) patients with positive FAST exams resulted in non-therapeutic laparotomies, and 7 of them had significant hemoperitoneum. This is notably better than the rates reported for DPL.

Trauma to the torso is a dynamic process; occult injuries may evolve. FAST allows monitoring of deterioration due to the ability to conduct serial exams at the point-of-care, in the emergency department, operative suite, and in the hospital. Hemoperitoneum Scoring Systems such as the University of Miami/McKenney Score have been developed and validated. DPL is more difficult to repeat and is not reliably quantifiable, but may have lingering utility when small bowel perforation is suspected. Repeat CT scans are logistically difficult and somewhat dangerous in view of the need to move the patient out of the resuscitation suite.

The FAST examination is the logical choice for screening for the “Need for OR” in patients with possible torso injury due to blunt trauma. Finally, while in developed countries CT is generally available, in most of the world this kind of advanced imaging is virtually non-existent. Therefore, on a global basis, the expansion of portable ultrasound use, clinician-performed at the point-of-care for trauma victims, represents a low-cost, high-technology solution.