In reply

Article Outline

• Copyright

We read your evidence-based medicine review with great interest, and thank you for the opportunity of responding to the letter. We are thankful for the thorough reanalysis of the data, and indeed we have to admit that the denominators were mistakenly switched for the endpoints of computed tomography and DPL-use, and laparotomy rates in the Boulanger study. Corrections that will be sent to the Cochrane Library also give us the opportunity of updating the review by data from the SOAP trial.

One might argue with the inclusion of the trial of Navarrete-Navarro et al, and we had clearly stressed in our review that the primary aim of this study was to evaluate the efficacy of computed tomography. However, with a 2-sided test, possible advantages of an ultrasonography-based algorithm should have been detected regardless of the direction of the study hypothesis. Let’s assume that this trial was beyond the scope of the review, and focus on the remaining studies.

In our review, we admitted that there was a marginal benefit with ultrasound-based pathways in reducing computed tomographic scans. Including the SOAP-trial increases the precision of the point estimate, now leading to a significant difference in favor of ultrasound-based pathways. So what? Given the low sensitivity of FAST to exclude intraabdominal injuries, is this really an advantage, or does it expose patients to additional harm by missing abdominal injuries?

From a European perspective, the reduction in DPL use is real, but clinically meaningless, since DPL has already been abandoned in most emergency departments.

We agree that laparotomy rates are dubious endpoints; however, they had been reported in the source trials, and may at least be regarded as patient-centered outcomes.

Of note, the SOAP trial is the only investigation suggesting a reduction in composite complication rates (0.27, 95% confidence interval 0.11 to 0.67). It must be stressed that in this trial only 217 of 525 patients screened for eligibility were included in the final analysis. Published data do not allow for cross-tabulation of mortality. However, the combined relative odds of death in the studies of Boulanger and Rose were estimated at 1.99 (95 % confidence interval 0.88 – 4.47) in favor of no-ultrasound-algorithms!

Altogether, ultrasound may reduce the number of computed tomographic scans and DPLs, which is a difficult to interpret surrogate measure. Although we thank the authors of the letter for pointing out the shift in denominators, we refuse their allegation of methodological flaws. After determining the diagnostic accuracy of FAST, we set out to identify all effectiveness studies on this popular screening method according to rigorous Cochrane Standards. Thus, the included studies still represent the best available evidence for health care decisions, and even the revised calculations neither affect the key messages of the original review, nor the evidence-based medicine review.

The statistically significant difference between study groups in the frequency of computed tomographic scans makes it even more difficult to determine the value of FAST. It is unclear whether the risk difference of 46% will translate to a number needed to treat or a number needed to harm of 2, since negative ultrasound-scans were not regularly verified by a reference test in the published effectiveness studies.

There is still no compelling evidence that the use of ultrasound has a benefit on patient-centered endpoints (eg, mortality). Unfortunately, with increasing utilization of contrast-enhanced, whole-body helical trauma computed tomography, it is unlikely that enough and sufficiently large randomized controlled trials on this topic will be conducted in the future to provide a conclusive answer.

Again, we thank you for the opportunity of correcting errors, and clarifying some open points.