Page 76 - JSOM Winter 2017

Table 2 Accuracy of Ultrasound in Detecting Wooden Foreign Bodies, Stratified by Level of Training
Training Sensitivity (95% CI), % Specificity (95% CI), % LR+ (95% CI) LR− (95% CI)
Resident 47.0 (39.7–54.5) 68.6 (61.4–75.1) 1.5 (1.2–2.0) 0.8 (0.7–0.9)
Attending physician (no fellowship) 52.5 (36.3–68.2) 62.5 (45.8–76.8) 1.4 (0.9–2.3) 0.8 (0.5–1.1)
Physician assistant 46.7 (22.2–72.6) 66.7 (38.7–87.0) 1.4 (0.6–3.4) 0.8 (0.5–1.4)
Ultrasound fellow 60.0 (17.0–92.7) 60.0 (17.0–92.7) 1.5 (0.4–5.5) 0.7 (0.2–2.5)
Attending physician (ultrasound fellowship) 60.0 (17.0–92.7) 80.0 (29.9–99.0) 3.0 (.5–19.9) 0.5 (0.2–1.6)
CI, confidence interval; LR, likelihood ratio.

Figure 1 Ultrasound sensitivity to detect wooden foreign bodies ultrasound knowledge and expertise. Second, the Jacobson et
as a function of foreign body size. Attending physicians, n = 8; al. study used cadaver specimens, which may have created
20
ultrasound-trained physicians, n = 2; midlevel, n = 40.
an environment more amenable to ultrasound examinations
than our fresh-tissue models. Third, their study did not impose
a time limit for ultrasound examinations on subjects. Finally,
because our study does suggest a trend toward improved sen-
sitivity with increasing object size, it is possible that our study
lacked adequate power to demonstrate an association between
foreign body size and identification via ultrasound.

Limitations
The principal limitation of this study related to lack of gen-
eralizability to real-world clinical settings. There are several
reasons for this. First is the use of the tissue model. Although
the chicken-tissue model is fresh and not preserved, as is a
cadaveric model, the transmission of ultrasound through this
tissue may be different than that in living human models.
Moreover, subjects were not allowed to use physical examina-
tion maneuvers (e.g., point of maximal tenderness) to direct
their ultrasound examinations and increase suspicion for an
underlying foreign body.
10 mm. Furthermore, we did not identify any association be-
tween diagnostic accuracy and education level. The setting including a time limit imposed on our subjects
similarly may limit generalizability. Our intent was to simu-
The pooled sensitivity was 48.4% for the use of ultrasound to late the circumstances of a busy emergency department and
detect wooden foreign bodies. This aligns with the lower range consistently limit the resource of available time to complete
of values previously reported in the literature for detection of the procedure. However, had we offered subjects more time
various foreign bodies in tissue models (50% to 52.6% ). to complete their examinations, it is possible they would have
6
10
These values contrast with studies reporting significantly achieved higher diagnostic accuracy. On the other hand, to the
higher sensitivities for ultrasound in detecting foreign bodies point of simulating an emergency department environment,
ranging up to 93% to 100%. 12,13,19 There are several potential our subjects performed their examinations in a relatively
explanations why other studies reported significantly higher quiet, dark, and secluded room, which may have increased
sensitivities. First, some of these studies entailed didactic ses- their diagnostic accuracy and may not approximate the typi-
sions immediately before data collection, which may have cal clinical care setting in the emergency department.
improved subject diagnostic accuracy. Second, some used
13
differing tissue models, such as human cadaveric tissue, and it Another limitation is the sample size. Though our dataset
is possible that this medium is more amenable to sonographic comprised 500 ultrasound studies, only 50 studies were per-
penetration or creates more echogenic contrast between mod- formed for each foreign body size. This yielded a study pow-
eled tissue and foreign body than our fresh-tissue model. 12,19 ered to detect sensitivity differences of 20% or more. It may
Third, none of these prior studies reported an imposed time be that diagnostic accuracy, indeed, improves with increasing
limit on participants and may have enabled more time for foreign body size but in smaller increments, which could not
aspects of procedure completion, increasing diagnostic test be detected because of a small sample size.
characteristics. Finally, some of these studies aggregated ex-
aminations of tissue models with alternative substances (e.g., Future Research
metal, glass), the inclusion of which may have confounded ac- Our results highlight several potential areas for investigations.
curacy estimates. 13,17 An ideal investigation may be a clinical study of emergency de-
partment patients with suspected radiolucent foreign bodies in
Our findings contrast with at least one study that suggested im- penetrating wounds. Such a study would ideally compare the
proved sensitivity of ultrasound for detecting wooden foreign diagnostic accuracy of ultrasound as performed in the emer-
bodies with increasing size. That study, by Jacobson et al., gency department versus a gold standard such as magnetic res-
20
20
reported a sensitivity of 86.7% for foreign bodies measuring onance imaging or surgical exploration. Such a study would
2.5mm and 93.3% for foreign bodies measuring 5.0mm. How- likely be logistically challenging to accumulate sufficient num-
ever, the subjects differed from our emergency medicine pro- bers of patients for a well-powered analysis. In the interim,
vider population as musculoskeletal radiologists with extensive further insight may be gained from replication of our study

74 | JSOM Volume 17, Edition 4/Winter 2017