Ultrasonography in the diagnosis of pediatric distal forearm fracture: a systematic review

Distal forearm fractures are considered one of the commonest injuries in adults and children, due to falling on an outstretched hand⁽¹⁾. There are different types of fractures distinctively seen in children: torus (buckle), greenstick, complete, and epiphyseal plate fractures⁽²⁾. X-ray studies are the most common diagnostic modality for suspected fractures⁽³⁾. Ultrasound (US) has recently been used for the detection of fractures, with reports suggesting that it may be more sensitive than X-ray because bone acts as a natural obstacle against sound transmission at high frequencies. Furthermore, the US can analyze a region in multiple planes rather than the limited views offered by traditional radiography^(4,5).

Several important sonographic findings may be associated with bone fractures, both in the emergency setting and at follow-up. These include focal disruption of the hyperechoic cortical bone, hematoma with or without discontinuity of the periosteum (subperiosteal space), edema of the soft tissues surrounding the fracture, mechanical disruption or dissociation of the growth plate, and assessment of fracture healing and different stages of bone callus formation⁽⁶⁾. In addition, US can show abnormalities of the surrounding tissues⁽⁷⁾as well as bursitis and articular effusion in cases with intra-articular fractures⁽⁸⁾.

Currently, experience in bedside ultrasound is growing amongst emergency physicians^(9,10), with a relatively easy learning curve^(11,12). The role of ultrasound as a gold standard screening tool is currently being investigated^(13,14). An important feature in this debate is the actual diagnostic accuracy of ultrasound for detecting forearm fractures.

This systematic review was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy, while the reporting followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines⁽¹⁵⁾.

The results of the literature search, screening, and study selection are illustrated in the PRISMA 2020 flowchart (Fig. 1). The literature search found 105 results (93 from databases and 12 from other electronic resources). The results from other resources were all duplicates of those yielded by searching the Medline and Embase databases. After the removal of duplicates, the titles, and abstracts of 86 records were screened, and 65 records were excluded due to nonrelevance to the research question. The full text of the remaining 21 studies was retrieved and examined for eligibility; 14 studies did not fulfil the inclusion and exclusion criteria. Finally, seven studies were eligible to be included in this systematic review^(19–25).

Fig. 1.

The PRISMA flow chart diagram for the results of the literature search and study selection

Table 1 summarizes the basic characteristics of the included studies. One paper was a prospective cohort study⁽²⁴⁾, while the other six studies were prospective diagnostic test studies(^19–23,25). All studies were single-centered(^{19–22,24,25}), except for one study that was conducted in two centers⁽²³⁾. The studies were conducted in the United States (USA)^(20,21,24), Germany^(19,22,23), and Australia⁽²⁵⁾. The seven studies compared bedside US (index test) to standard plain X-ray (standard test).

The study by Patel et al.⁽²⁴⁾ prospectively enrolled 33 children aged two through 17 years with suspected radius, ulna, tibia, or fibula fractures, who presented from March 2006 through January 2007 to the Pediatric Emergency Department, New York. The exclusion criteria were open fractures, neurovascular compromise, and hemodynamic instability. Bedside US was performed before standard radiography by one of three pediatric emergency medicine physicians who had no previous formal training in ultrasonography. The three physicians completed – before initiating the study – a two-hour session of bedside US. The bedside US for upper extremity injuries showed a sensitivity of 1.00 (95% CI: 0.87–1.00) and a specificity of 0.91 (95% CI: 0.69–0.98). The study suggests that the US of the upper extremity may be equivalent to radiography for identifying fractures not involving the joint.

Ackermann et al.⁽¹⁹⁾ conducted their study from January 2007 to May 2008 in Germany on 93 patients aged 0–12 years with a suspected forearm fracture. This was defined by a mechanism of injury consistent with forearm injury, bony tenderness, and the absence of open wounds in the forearm area. The children were first subjected to US examination, followed by radiography. The US assessors in this study were residents and consultants without special training in osteosonography. The sensitivity of US diagnosis was 94%, and the specificity was 99%, compared with X-ray diagnosis as the gold standard.

Chaar-Alvarez et al.⁽²¹⁾ conducted their study in Palmer Children’s Hospital, Orlando, from October 2007 to March 2009. The study enrolled 101 children between the ages of one and 17 years who presented to the ED with non-angulated distal forearm injuries. Children were excluded if they had a clinical forearm deformity, open wounds, or neurovascular injuries. The US was performed by pediatric emergency doctors who had completed training in emergency ultrasonography. Standard radiography was performed afterwards. The overall diagnostic accuracy of the blinded reviewer’s US interpretation was 94% (95% CI: 88–99%). Sensitivity and specificity were 96% (95% CI: 85–99%) and 93% (95% CI: 82–98%), respectively. At the observed prevalence, the PPV and NPV were 92% and 96%, respectively. The overall diagnostic accuracy of the bedside interpretation of the US was 79% (95% CI: 70–86%), with a sensitivity of 85% (95% CI: 72–94%), a specificity of 73% (95% CI: 60–84%), a PPV of 73%, and an NPV of 85%. The inter-rater reliability (J) was 0.57 (95% CI: 0.41–0.73). The setting of the study was a large pediatric ED covered by Pediatric Emergency Medicine physicians. Therefore, the findings may not apply to other types of hospitals.

Barata et al.⁽²⁰⁾ enrolled 53 pediatric patients (under 18 years of age) who presented to the ED of a university-affiliated, level I trauma center, New York, between March 2008 and January 2009, with suspected long-bone fracture. Suspected fractures were characterized by swelling, erythema, and localized pain. Patients who had a history of fracture at the suspected site, extremity deformity, or an open fracture were excluded from this study. True blinding was applied. The sensitivity and specificity of US were 95.3% (95% CI: 82.9–99.2%) and 85.5% (95% CI: 72.8–93.1%), respectively. The PPV and NPV were 83.7% (95% CI: 68.8–92.2%) and 96% (95% CI: 84.9–99.3%), respectively.

The study by Eckert et al.⁽²²⁾ included 76 patients aged between one and 14 years, and presenting with suspected distal forearm fractures (defined by adequate trauma and appropriate clinical symptoms). After US examination, standard X-rays of the wrist from the anteroposterior and lateral view were taken. The training of the US assessors was not reported. All radiologically diagnosed radius fractures and all patients with no fractures were correctly diagnosed by ultrasound. Compared to X-ray, the sensitivity of the ultrasound method was 96.1%, and the specificity was 97%, with a PPV of 94.3% and an NPV of 97.9%.

Herren et al.⁽²³⁾ carried out a prospective study including 201 patients between four and 11 years of age who presented at two trauma surgery clinics in Germany with a presumptive diagnosis of the distal radius or forearm fracture between January and December 2012. First, US imaging of the distal forearm was carried out on six standardized planes by physicians who had undergone a short training in US-guided fracture diagnosis. Afterwards, radiographs of the wrist were taken and interpreted by attending experts in radiography. The specificity and sensitivity of ultrasound diagnosis were both 99.5%.

The study by Rowlands et al.⁽²⁵⁾ was conducted at Princess Margaret Hospital for Children, the sole tertiary pediatric hospital in Western Australia. The study included 409 children under 16 years who had a suspected fracture of the forearm. Patients with evidence of an open fracture were excluded, as well as patients who had imaging performed before arrival. The study was double-blinded to avoid bias. The results showed that physicians could diagnose forearm fractures in children with a sensitivity of 91.5% and a specificity of 87.6%.

Table 2 shows the assessment of the risk of bias using the QUADAS II and three of the STARD criteria. All studies had a QUADAS II score above 10, indicating high quality of the studies. In all the studies, the reference standard was plain X-ray, which is considered the gold standard for diagnosing fractures. The performance and interpretation of US and radiography were independent of each other. All the studies clearly described their criteria for selecting patients. The sampling process was detailed in three studies only^(20,21,25). The confidence intervals for sensitivity and specificity were mentioned by three studies only^(20,21,24).

The study by Patel et al.⁽²⁴⁾ was well-powered and has many strong points. The sample size was calculated, as well as the kappa value, and the authors ensured blinding the radiologist to US results to avoid bias. Unfortunately, there was no STARD Flow Diagram to visualize the patient cohort and follow-up (QUADAS 12, level 2b).

Several limitations were also observed in the study by Ackermann et al.⁽¹⁹⁾. The method of patient recruiting was not clear, and the sample size was not calculated. The study carried a high risk of bias due to a lack of sample blinding. The lack of calculation of the kappa value and confidence intervals adds more limitations to the study (QUADAS 11, level 2b).

As for the study by Chaar-Alvarez et al.⁽²¹⁾, the limitations included the lack of a power calculation and the fact that the sample was convenient, which is a potential source of bias (QUADAS 11.5, level 2b).

In the study by Barata et al.⁽²⁰⁾, the inclusion of confidence intervals in the results and blinding increased the internal validity of the study. However, the small sample size with no power calculations and a single site may have affected its internal as well as external validity. Furthermore, the lack of a STARD diagram in the study makes it difficult to interpret the findings (QUADAS 11, level 2b).

Many limitations were noted in the study by Ecker et al.⁽²²⁾. The method of subject recruitment was not clear, and the sample size was not calculated. It was not clear from the study who performed the US and where it was done. Moreover, no tables or even STARD diagrams were prepared to add value to the study. Finally, the results were limited to sensitivity and specificity, without confidence intervals (QUADAS 11, level 2b).

In the study by Herren et al.⁽²³⁾, the residents who evaluated the US were not blinded to any information about the child. The attending experts in radiography who interpreted the X-rays were blinded to the US results. The physicians who performed US diagnoses were not experts in the US, but they had undergone a short training in US-guided fracture diagnosis. However, inter-observer variability was not determined. In addition, the author did not provide a true double-blind method for the analysis of both procedures, which put him at risk of diagnostic and test review bias (QUADAS 12.5, level 2b).

The study by Rowlands et al.⁽²⁵⁾ shows potential for bias in the selection of patients, given that the recruitment used prospective convenience sampling. Furthermore, the results of the primary outcome were represented as sensitivity and specificity, without confidence intervals or any other values such as likelihood ratios (QUADAS 11.5, level 2b).

The diagnostic performance of the US in the included studies is summarized in Tab. 3 and Fig. 2 and Fig. 3. The overall accuracy of the US in diagnosing fractures ranged from 99.5% to 78.6%. The sensitivity and specificity ranged from 85% to 100%, and from 73% to 100%, respectively. These findings suggest that the US can be used to diagnose forearm fractures in children with sufficient accuracy. The HSROC curve (Fig. 3) shows that the AUC for the US ranged from 0.79 to 1.00, with the study by Chaar-Alvarez et al.⁽²¹⁾ showing a lower AUC than the other six studies.

Fig. 2.

Forest plot showing the sensitivity and specificity of the included studies. CI – confidence interval; FN – false negative; FP – false positive; TN – true negative; TP – true positive

Fig. 3.

Hierarchical summary receiver operating characteristics (HSROC) curve for the diagnostic performance of bedside ultrasonography in the included studies

Ultrasound imaging has several advantages, including the absence of ionizing radiation besides the ability to carry out comparative scanning between the pathological and healthy sides in doubtful cases and to perform ultrasound palpation of the painful site with the transducer. In addition, it is cost-effective and can be repeated several times during the first hours after trauma to assess possible local complications, such as bleeding around the fracture site. All these advantages, together with its availability in primary care and in low-resource rural areas, make it an invaluable tool for the assessment and follow-up of bone fractures⁽⁶⁾.

Therefore, this review was conducted to summarize the evidence regarding the diagnostic accuracy of bedside US for identifying distal forearm fractures in pediatric patients.

All studies are prospective cohorts comparing a new diagnostic test (the US scan for fracture detection) against a gold standard of X-ray, with only two of them^(21,25) comparing pain associated with clinical examination, US, and X-ray. Appropriately, all the authors excluded open fractures and obvious deformity when ultrasound would not be appropriate. The quality of most studies was average to high (QUADAS 11–12.5). Despite the heterogeneity of the studies, they all show the sensitivity of the ultrasound to be high, varying from 91.5% to 100%, and specificity from 87.5% to 99.5%, and the majority presented the values with the confidence interval.

The seven studies showed some limitations which may introduce bias into their results. The populations were generally recruited as convenience samples, allowing for the introduction of selection bias. In two studies^(24,25), the sample size was calculated to detect high sensitivity and specificity with a power of 80% and p-value <0.05. Although there is a wide variation in the ultrasound experience of the doctors carrying out the scans in each study, ranging from a consultant to surgeons with vast previous scanning experience, to ED doctors with 1–2 hours of US teaching^(23–25), inter-rater variability (kappa) was calculated only in three studies^(21,24,25) and the value was between 0.57–0.79. None of the studies documented the number of patients who withdrew or declined study inclusion.

Previous systematic reviews assessed the US as a diagnostic test for pediatric fractures of the upper limb^(26–28). However, these systematic reviews showed some limitations. Other types of reference tests, such as computed tomography or magnetic resonance imaging, were used in the review by Joshi et al.⁽²⁷⁾. The review by Katzer et al.⁽²⁸⁾ included studies enrolling patients older than 18 years old. The eligibility criteria for selecting studies were not clearly stated by Douma-den Hamer et al.⁽²⁶⁾. The present review attempted to avoid the limitations of earlier systematic reviews and to summarize the updated evidence, as some new studies were published after the previous reviews.