INTRODUCTION
Fractures account for 17% of injuries in Canadians ages 12 and older.Reference Billette and Janz 1 Long bone fractures (LBFs)–that is, fractures to the humerus, radius, ulna, femur, tibia or fibula–are most often caused by traumatic events such as motor vehicle accidents, sport injuries, falls, and assaults.Reference Cohen 2 Immobilization, with or without reduction, is the most common treatment strategy for fractures in the emergency department (ED).Reference Geiderman and Katz 3 Immobilization can be used as a temporizing measure until surgery is undergone or as a definitive treatment. In the latter case, adequate alignment is essential to optimize future functioning.
The diagnosis of LBFs is traditionally accomplished through plain radiographs taken at perpendicular angles to each other. Because of the resources required for around-the-clock radiographic capability, X-ray can be inaccessible in rural or remote areas, or in larger centres during off-hours.Reference Sippel, Muruganandan and Levine 4 In settings where a radiography technologist is not present on-site, a point-of-care ultrasound (PoCUS) examination can be performed earlier, and at the bedside, reducing diagnostic delays and time to initiation of management.Reference Grechenig, Clement and Schatz 5 Additional advantages of PoCUS include its simplicity, portability, repeatability, and its lack of ionizing radiation.
Fluoroscopy has been advocated as an adjunct to ED-based fracture reduction. However, this technique can be impractical because of cost, size of the equipment, the training necessary for the use of this technology, the need for lead protectors on health care providers, as well as the harmful effects to patients of ionizing radiation.Reference Brenner, Doll and Goodhead 6
The benefits of PoCUS may be even greater for the reduction of displaced fractures, where localized swelling and significant pain may lead to inadequate reduction and repeated attempts, which are known to increase the likelihood of developing compartment syndrome.Reference Chen, Kim and Moore 7 Moreover, repeated attempts may also put the patient through the risky process of sedation and/or analgesia multiple times, and could possibly lead to increased rates of operative management.Reference Ang, Lee and Lam 8 PoCUS allows for real-time assessment of reduction attempts, making it possible to re-manipulate and immobilize the fracture while the patient is still sedated. PoCUS-guided fracture reduction and immobilization produces minimal discomfort when liberal amounts of transmission gel are used.
There are multiple studies and reports on the use of PoCUS in the setting of LBFs, but to our knowledge this is the first comprehensive review that includes a description of the technique and its application to LBF reduction. The objective of this study is to summarize and, where applicable, pool the diagnostic accuracy of PoCUS compared to plain radiographs in children and adults for 1) the diagnosis of suspected LBFs and 2) the adequate reduction of known LBFs.
Description of the technique
For the diagnosis of LBFs with PoCUS, the high-frequency linear probe should be used in the longitudinal plane, with a thick layer of ultrasound gel to minimize pain and improve visualization (Figure 1). The probe should be gently moved along the axis of the bone while adjusting the depth and gain to optimize the view.Reference Majeed, Mukherjee and Paw 9 , Reference Cross 10 The deepest hyperechoic horizontal line on the PoCUS screen represents bony cortex, with its acoustic shadow preventing visualization of deeper structures. If a fracture is present, a discontinuity in this line will be seen (Figure 2, A). The degree of displacement and angulation of the distal fragment will be visualized as a step-off and as an angulation between the cortices, respectively.Reference Majeed, Mukherjee and Paw 9 Similar to plain radiography, the PoCUS examination should be obtained in two perpendicular planes for each bone of interest,Reference Saul, Ng and Lewiss 11 as it minimizes the false-negative examinations from fractures that may be visualized at certain angles only.Reference Majeed, Mukherjee and Paw 9
Simulated use of point-of-care ultrasound (PoCUS) for the diagnosis of distal radius fracture.
Diagnosis of A) displaced distal radius fracture and B) adequate reduction on the PoCUS screen. The patient’s distal radius is to the left of the screen.
For PoCUS-guided reduction of LBFs, the fracture should be visualized with PoCUS prior to a reduction attempt, to provide a comparison for the post-reduction PoCUS images (Figure 2, B). If the post-reduction images do not appear satisfactory on PoCUS, based on step-off or angulation, additional attempts at reduction may be initiated in the same procedural sedation and/or analgesia episode.Reference Saul, Ng and Lewiss 11 When the best possible reduction is achieved, the health care provider may proceed with immobilization of the fracture and usual care strategies, including post-reduction plain radiography. All relevant PoCUS-generated images should be saved and archived for clinical and quality assurance purposes.
METHODS
Eligibility criteria
We included studies that prospectively enrolled patients with either suspected or known acute LBFs. Studies had to use PoCUS as the intervention, and compare it to plain radiography, the pragmatic standard of care. Studies had to report on the performance of PoCUS, either as a diagnostic or a management tool. Case reports and narrative review articles were excluded because they do not report diagnostic test characteristics. Other exclusion criteria included non-acute fractures, non-human models, or comparison done only with modalities other than plain radiography.
Information sources
MEDLINE and EMBASE databases were searched, from their earliest publication date to July 19, 2015. The search strategy was developed in conjunction with a health sciences librarian, and the strategy for the MEDLINE search is described in Supplementary Material Appendix 1 (available online).
Study selection
Two authors (LBC and LB) reviewed the title and abstract of the retrieved articles to determine potential eligibility for inclusion in the manuscript, and manually searched the reference list of review articles for additional relevant studies. They subsequently performed a full-text review to identify studies meeting inclusion and exclusion criteria. Disagreements were resolved by consensus.
Data collection process
A data extraction sheet (based on the Cochrane Consumers and Communication Review Group’s data extraction template)Reference Higgins 12 was pilot-tested for the first five studies, and refined according to the typical presentation of results. Information extracted from each study included type of study, study location, patient population, type of fracture and prevalence, number and specialty of the providers performing and interpreting PoCUS examinations, training that the providers received, the actual application of PoCUS (for initial diagnosis, reduction assessment, or both), and the performance characteristics of PoCUS accuracy, including sensitivity, specificity, and positive and negative likelihood ratios (LRs).
Risk of bias in individual studies
Methodological bias and applicability of the included studies were assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) guidelines.Reference Whiting, Rutjes and Westwood 13 As outlined in the 2011 guidelines, the authors agreed upon signaling questions that were pertinent to the review, encompassing five domains: patient selection, index test, reference standard, risk of bias, and concerns about applicability. Two of the authors (LBC and LB) independently performed the assessments, discrepancies were reconciled through discussion, and a consensus was reached.
Summary measures
The primary outcomes were the test characteristics of PoCUS compared to plain radiography for 1) the diagnosis of LBFs and 2) the successful reduction of LBFs. These included sensitivity, specificity, and positive and negative LRs. We also abstracted and reported additional measures of success in studies pertaining to the reduction of known LBFs, such as rates of repeat reduction and surgical management.
Meta-analysis
Where multiple studies reported on diagnostic accuracy under similar clinical and methodological conditions, a meta-analysis was performed. Forest plots were used to present sensitivity and specificity results from each study. In order to provide a quantitative measure of statistical heterogeneity, we used the model of diagnostic odds ratio used by DerSimonian and Laird.Reference DerSimonian and Laird 14 , Reference Doebler and Holling 15 We reported Cochran’s Q and Higgins’ I2.
Results were pooled using the bivariate random effects model proposed by Reitsma et al. for meta-analysis of data on diagnostic accuracy.Reference Reitsma, Glas and Rutjes 16 Results include estimates of the overall sensitivity, specificity, and the hierarchical summary receiver operating characteristic (HSROC) curve. In addition, Markov chain Monte Carlo sampling was used to produce summary estimates of LR(+) and LR(–) and their 95% confidence intervals (CIs).Reference Zwinderman and Bossuyt 17 All analyses were performed using the mada package version 0.5.4Reference Doebler 18 for R statistical software.Reference R 19
RESULTS
Study selection
The literature search yielded 1,034 articles, which were screened by title and abstract (Figure 3). Upon performing manual searches of the bibliographies of the most recent review articles (less than 5 years from publication), 16 additional studies were retrieved. Based on the screening of these 1,050 articles by title and abstract, 948 studies were excluded. Upon a full-text review of the remaining 102 studies, 72 articles were excluded; reasons for exclusion are listed in Figure 3. Thirty studies were selected for inclusion in the systematic review: 23 for the diagnosis of LBFs, 5 for the management of LBFs, and 2 for the assessment of both outcomes. There were no disagreements between the authors.
PRISMA flow diagram, including reasons for exclusions of the full-text articles reviewed.
Study characteristics
The 30 included studies recruited patients in unscheduled care environments, whether in the emergency or radiology departments, and almost all were in university-affiliated hospitals (Table 1).Reference Chen, Kim and Moore 7 , Reference Ang, Lee and Lam 8 , Reference Herren, Sobottke and Ringe 20 - Reference Chern, Jou and Lai 47 Two studies used a second reference standard for some of the patients, either computed tomography when the plain radiographs were not definitive or a daily clinical evaluation for 72 hours with delayed imaging or return to full duty for military personnel.Reference McNeil, McManus and Mehta 34 , Reference Marshburn, Legome and Sargsyan 42
Characteristics of studies included in the systematic review
UAH=university-affiliated hospital; ED=emergency department; PED=pediatric emergency department; Dept=department; EM=emergency medicine; NR=not reported; N/A=not available; EP=emergency physician; PEP=pediatric emergency physician.
Both the medical background and PoCUS training of the operators varied greatly across studies. Sixteen studies (53.3%) had emergency physicians performing and interpreting PoCUS examinations, four studies (13.3%) had surgeons (pediatric, hand, and/or orthopedic), five studies (16.7%) had radiologists or ultrasound fellows or technicians, and four studies (13.3%) included medical students or residents. The musculoskeletal PoCUS training received by the various operators ranged from a 3-minute tutorial followed by hands-on training,Reference Chinnock, Khaletskiy and Kuo 40 to a 1-month musculoskeletal ultrasound course.Reference Bolandparvaz, Moharamzadeh and Jamali 33 On average, most providers received between 1 and 4 hours of didactic and practical training on basic PoCUS skills, which often focused on musculoskeletal indications.
Diagnostic accuracy of PoCUS
Table 2 shows the results of the systematic review for the studies pertaining to the diagnosis of LBFs with PoCUS. All studies reported both the sensitivity and specificity of PoCUS, and LRs were calculated from these values when the authors did not include them in their manuscripts. Twenty-five studiesReference Chen, Kim and Moore 7 , Reference Herren, Sobottke and Ringe 20 – Reference Hübner, Schlicht and Outzen 39 , Reference Waterbrook, Adhikari and Stolz 41 – Reference Beltrame, Stramare and Rebellato 44 were included, ranging from 26 to 653 patients. Collectively, these studies included 2,982 patients and identified 1,200 fractures, with fracture prevalence ranging from 7% to 71% (aggregate prevalence of 40.2%). Twelve of the 25 (48%) studies focused on one body area only, either the forearmReference Chen, Kim and Moore 7 , Reference Herren, Sobottke and Ringe 20 – Reference Kozaci, Ay and Akcimen 26 or the ankle,Reference Atilla, Yesilaras and Kilic 27 – Reference Canagasabey, Callaghan and Carley 30 and 13 of 25 (52%) studies recruited exclusively pediatric patients.Reference Chen, Kim and Moore 7 , Reference Herren, Sobottke and Ringe 20 – Reference Kozaci, Ay and Akcimen 26 , Reference Patel, Blumberg and Crain 31 , Reference Barata, Spencer and Suppiah 36 – Reference Hübner, Schlicht and Outzen 39
Diagnostic accuracy of point-of-care ultrasound compared to plain radiography in long bone fractures
* Data from personal correspondence with the authors.
Studies are listed by bone type studied, then ped/adult, then reverse chronological publication date.
Ped=pediatric; UE=upper extremity; LE=lower extremity; LB=long bone; All bones=long bones and non-long bones.
Sensitivity for the detection of LBFs with PoCUS ranged from 64.7% to 100%, with 20 of 25 (80%) studies achieving rates of >90%. Specificity ranged from 79.2% to 100%, with 21 of 25 (84%) studies achieving rates of >90%. PoCUS was found to have an LR(+) ranging from 3.11 to infinity, with 19 of 25 (76%) studies having an LR(+) above 10. PoCUS was found to have an LR(–) ranging from zero to 0.45, with 19 of 25 (76%) studies having an LR(–) at or below 0.1.
Reduction success with PoCUS
Table 3 shows the results of the systematic review for the studies pertaining to the management of LBFs with PoCUS. Seven studies were included,Reference Chen, Kim and Moore 7 , Reference Ang, Lee and Lam 8 , Reference Patel, Blumberg and Crain 31 , Reference Chinnock, Khaletskiy and Kuo 40 , Reference Wong, Ang and Ng 45 - Reference Chern, Jou and Lai 47 ranging from 27 to 164 patients. Collectively, these studies recruited 524 patients with 484 known fractures. Three studies recruited exclusively pediatric patients.Reference Chen, Kim and Moore 7 , Reference Patel, Blumberg and Crain 31 , Reference Wong, Ang and Ng 45 Only one study did not specifically target patients with forearm fractures,Reference Patel, Blumberg and Crain 31 and four focused specifically on distal radius fractures only.Reference Ang, Lee and Lam 8 , Reference Chinnock, Khaletskiy and Kuo 40 , Reference Kodama, Takemura and Ueba 46 , Reference Chern, Jou and Lai 47 Despite these clinical similarities, there was significant methodological heterogeneity between included studies, thus precluding performance of a meta-analysis.
Reduction success of point-of-care ultrasound compared to plain radiography in long bone fractures
PoCUS=point-of-care ultrasound; NR=not reported.
The sensitivity and the specificity for the adequate reduction of LBFs with PoCUS ranged from 94% to 100% and from 56% to 100%, respectively. One study found a decrease in the rate of repeat manipulation and reductions, and showed better post-reduction volar tilt measurements and lower rates of operative management in patients with PoCUS-guided reduction compared to blind reduction.Reference Ang, Lee and Lam 8 One other study found that PoCUS-guided reduction was successful in 95% of cases, as opposed to 68% with blind technique.Reference Kodama, Takemura and Ueba 46
Meta-analysis of selected studies
Only studies of two subpopulations of patients had sufficient clinical and methodological homogeneity to warrant performance of a meta-analysis: pediatric forearm and adult ankle fractures. Many studies that evaluated multiple bones amalgamated their results and had incomplete information on the relative prevalence of each type of fracture, thus preventing inclusion in the meta-analysis. We attempted to contact study lead authors but were successful for only two of these (Drs. Ekinci and Canagasabey; personal correspondence).
Ten studies were included in the meta-analysis of the PoCUS diagnosis of forearm fractures in pediatric patients.Reference Chen, Kim and Moore 7 , Reference Herren, Sobottke and Ringe 20 – Reference Williamson, Watura and Cobby 25 , Reference Hübner, Schlicht and Outzen 39 , Reference Weinberg, Tunik and Tsung 43 , Reference Beltrame, Stramare and Rebellato 44 Forest plots of the included studies’ sensitivity and specificity are shown in Figure 4. The studies’ pooled test characteristics were as follows: sensitivity 93.1% (95% CI, 87.2%–96.4%), specificity 92.9% (95% CI, 86.6%–96.4%), LR(+) 14.1 (95% CI, 6.71–25), and LR(–) 0.08 (95% CI, 0.04–0.15). Supplementary Material Appendix 2 (available online) shows the HSROC curve, demonstrating very good test operating characteristics. The calculated Cochran’s Q was 8.382 (9 df, p=0.496), and Higgins’ I2 was 0%, demonstrating statistical homogeneity between the included studies.
Forest plots of the sensitivity and specificity of studies included in the meta-analysis of pediatric forearm fractures. Point estimates are presented with their 95% confidence intervals.
Four studies were included in the meta-analysis of the PoCUS diagnosis of ankle (distal tibia and fibula) fractures in adult patients.Reference Atilla, Yesilaras and Kilic 27 – Reference Canagasabey, Callaghan and Carley 30 Because they are clinically different, we did not include studies of pediatric or mixed age patients, or studies of patients with mid-shaft lower leg fractures.Reference Dulchavsky, Henry and Moed 35 , Reference Hübner, Schlicht and Outzen 39 , Reference Weinberg, Tunik and Tsung 43 , Reference Beltrame, Stramare and Rebellato 44 Forest plots of the included studies’ sensitivity and specificity are shown in Figure 5. The studies’ pooled test characteristics were as follows: sensitivity 89.5% (95% CI, 77.0%–95.6%), specificity 94.2% (95% CI, 86.1%–97.7%), LR(+) 16.4 (95% CI, 6.57–33.5), and LR(–) 0.12 (95% CI, 0.05–0.24). Supplementary Material Appendix 3 (available online) shows the HSROC curve, demonstrating good test characteristics. The calculated Cochran’s Q was 3.834 (3 df, p=0.28), and Higgins’ I2 was 21.75%, demonstrating statistical homogeneity between the included studies.
Forest plots of the sensitivity and specificity of studies included in the meta-analysis of adult ankle fractures. Point estimates are presented with their 95% confidence intervals.
Risk of bias
Supplementary Material Appendix 4 (available online) shows the assessment of methodological bias and applicability of the included studies, using the 2001 QUADAS-2 guidelines. All studies used convenience as opposed to consecutive sampling, mostly due to the small number of PoCUS-trained providers available. All PoCUS examinations pertaining to the diagnosis of LBFs occurred before radiographs were taken. Management of patients was always based on the results of the radiographs. Only 3 of 30 (10%) studies reported on the inter-rater reliability of PoCUS: 78% agreement in one study, and kappa values of 0.57 and 0.92 in two others.Reference Chaar-Alvarez, Warkentine and Cross 23 , Reference Chinnock, Khaletskiy and Kuo 40 , Reference Waterbrook, Adhikari and Stolz 41 Only one study explicitly reported on a quality assurance process, with one senior sonologist reviewing all enrolled patients’ scans.Reference Weinberg, Tunik and Tsung 43 Twenty-nine studies used plain radiographs as the only reference test for LBFs, with one using computed tomography as a secondary confirmatory tool only if necessary.Reference Marshburn, Legome and Sargsyan 42 Twenty-five studies reported on the presence of PoCUS training prior to initiating the study. Overall, all studies were of good methodological quality and were felt to be appropriate for inclusion in the analysis of this review.
DISCUSSION
Our review of the literature on the accuracy of PoCUS for the diagnosis of LBFs included 25 studies and 2,982 patients, of which 20 reported sensitivity rates of >90%, and 21 specificity rates of >90%. Pooled results of the meta-analyses demonstrated good performance characteristics, with a sensitivity of 93.1% and specificity of 92.9% for pediatric forearm fractures, and a sensitivity of 89.5% and specificity of 94.2% for adult ankle fractures. Seven studies of PoCUS for the reduction of LBFs met inclusion criteria, comprising 524 patients, and all demonstrated success rates of above 90%. As practitioners become more proficient with PoCUS over time, its accuracy may improve further. However, although the current evidence is promising, large randomized controlled trials (RCTs) are needed to confirm the benefits seen in smaller, non-randomized trials.
PoCUS, a relatively new tool in the emergency physician’s armamentarium, is a powerful and useful adjunct to the clinical assessment and plain radiographs. In some settings, PoCUS may be preferable to plain radiographs, because of its ease of use and portability. Nonetheless, the need for medicolegal documentation, as well as baseline images with which to compare follow-up radiographs, justify the continued use of traditional diagnostic imaging in settings where it is available. The studies that reported lower PoCUS accuracy tended to be those that recruited adult (versus pediatric) patients, and that assessed multiple bones (versus focusing on one specific body area only). This may indicate that the PoCUS examination is easier to accomplish and/or to interpret in pediatric patients, and that there is a learning curve associated with each specific bone being assessed.
In 2013, Joshi et al. published a review of the diagnostic accuracy of history, physical examination, and bedside ultrasound for the diagnosis of extremity fractures in the ED.Reference Joshi, Lira and Mehta 48 In addition to eight articles that had not been published at that time,Reference Herren, Sobottke and Ringe 20 , Reference Kozaci, Ay and Akcimen 26 – Reference Hedelin, Goksör and Karlsson 28 , Reference Dallaudière, Larbi and Lefere 32 , Reference Bolandparvaz, Moharamzadeh and Jamali 33 , Reference Waterbrook, Adhikari and Stolz 41 , Reference Kodama, Takemura and Ueba 46 our study includes thirteen other articles not included in that review.Reference Eckert, Ackermann and Schweiger 21 , Reference Eckert, Ackermann and Schweige 22 , Reference Ackermann, Liedgens and Eckert 24 , Reference Williamson, Watura and Cobby 25 , Reference Ekinci, Polat and Günalp 29 , Reference Canagasabey, Callaghan and Carley 30 , Reference McNeil, McManus and Mehta 34 – Reference Barata, Spencer and Suppiah 36 , Reference Moritz, Berthold and Soenksen 38 , Reference Waterbrook, Adhikari and Stolz 41 , Reference Weinberg, Tunik and Tsung 43 , Reference Beltrame, Stramare and Rebellato 44 Additional studies were possibly found as a result of a more comprehensive search strategy. Our review reaches similar conclusions, that PoCUS is a useful tool to supplement radiographs for the detection of fractures. We have provided an inclusive and updated review of the literature regarding the use of PoCUS in LBFs. Given the increasing proportion of emergency medicine practitioners who have access to and use PoCUS regularly, we hope that this article can serve as a reference to inform their practice with regards to LBFs.
As with any newer technology or procedure, training remains the most important challenge. Without proper training, practitioners run the risk of having a large number of false-positive or negative examinations, which may lead to worse patient outcomes. In the three studies where it was reported, the inter-reliability ranged from moderate to very good. However, training undergone by clinicians in these studies was minimal (3-minute tutorial, 15-minute bedside teaching, and knowledge from prior completion of a general bedside ultrasound course), and further training may therefore have increased their accuracy.
LIMITATIONS
Many of the included studies had small samples sizes, and none were RCTs. Overall, 40% of the recruited patients with possible LBF were diagnosed with a fracture with the reference test. This is likely higher than the true proportion of LBFs in undifferentiated patients presenting to most EDs with traumatic pain, swelling, or deformity. This may have been affected by the convenience sample selected by all authors, which introduces bias in the selected population. This recruitment strategy, by a small number of motivated PoCUS practitioners, may also have affected the results and limits the generalizability of our study findings. The lack of inter- and intra-rater reliability in most studies is also a limitation. However, the fact that in many studies the providers had no prior PoCUS experience and only limited training is reassuring for clinicians beginning to use musculoskeletal PoCUS.
CONCLUSION
This systematic review and meta-analysis summarizes current knowledge regarding the use of PoCUS in LBFs. The 30 prospective studies, including 2,982 individual patients, showed good test characteristics for the diagnosis and reduction of LBFs. Although results were varied, most studies reported sensitivities and specificities of 90% and above for the diagnosis of LBF with PoCUS, including in the two meta-analyses targeting pediatric forearm and adult ankle fractures. Although PoCUS should not completely replace the use of plain radiography at the current time, this review demonstrates that it is a powerful adjunct in all cases and could be used as a reasonable alternative in certain settings.
RCTs are needed to further the current knowledge on the use of PoCUS for the diagnosis and management of LBFs. Specifically, future studies should focus on determining standards for appropriate training and credentialing, on patient-centered outcomes such as satisfaction and morbidity, and on the cost-effectiveness of the use of PoCUS in the ED. In low-resource settings where access to plain radiographs may be challenging or minimal, studies should focus on the feasibility and safety of using PoCUS as an alternative to plain radiographs.
Acknowledgements
The authors wish to acknowledge the support of Henry Lam, librarian and information specialist, for his help defining the systematic review search strategy.
Competing interests: None declared.
SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit https://doi.org/10.1017/cem.2016.397
