Imaging diagnosis for chronic lateral ankle ligament injury: a systemic review with meta-analysis

Background Various imaging techniques have been utilized for the diagnosis of chronic lateral ankle ligament injury. This systemic review will explore the effectiveness of different imaging techniques in diagnosing chronic lateral ankle ligament injury. Methods Relative studies were retrieved after searching 3 databases (MEDLINE, EMBASE, and Cochrane Central Register of Controlled Trails). Eligible studies were summarized. Data were extracted to calculate pooled sensitivity and specificity of magnetic resonance imaging (MRI), ultrasonography (US), stress radiography, and arthrography. Results Fifteen studies met our inclusion and exclusion criteria. A total of 695 participants were included. The pooled sensitivities in diagnosing chronic ATFL injury were 0.83 [0.78, 0.87] for MRI, 0.99 [0.96, 1.00] for US, and 0.81 [0.68, 0.90] for stress radiography. The pooled specificities in diagnosing chronic ATFL injury were 0.79 [0.69, 0.87] for MRI, 0.91 [0.82, 0.97] for US, and 0.92 [0.79, 0.98] for stress radiography. The pooled sensitivities in diagnosing chronic CFL injury were 0.56 [0.46, 0.66] for MRI, 0.94 [0.85, 0.98] for US, and 0.90 [0.73, 0.98] for arthrography. The pooled specificities in diagnosing chronic CFL injury were 0.88 [0.82, 0.93] for MRI, 0.91 [0.80, 0.97] for US, and 0.90 [0.77, 0.97] for arthrography. Conclusion This systematic review with meta-analysis investigated the accuracy of imaging for the diagnosis of chronic lateral ankle ligament injury. Ultrasound manifested high diagnostic accuracy in diagnosing chronic lateral ankle ligament injury. Clinicians should be aware of the limitations of MRI in detecting chronic CFL injuries.


Background
Ankle sprain is one of the most common sports injuries [1][2][3]. The injury mechanism, a combination of inversion and adduction in foot plantar flexion, can cause damage to the lateral ankle ligaments [4]. Lateral ankle ligaments comprise anterior talofibular ligament (ATFL), calcaneofibular ligament (CFL), and posterior talofibular ligament (PTFL) [5]. Chronic lateral ankle ligament injury is a long-term outcome in patients suffering acute ankle sprain. Some long-term cohort studies showed that 10 to 20% patients eventually developed persistent symptoms, including pain, swelling, perception of instability, and recurrent ankle sprain [6][7][8][9]. The precise cause of these symptoms is still in debate and probably multifactorial [5,[10][11][12]. But identifying chronic lateral ankle ligament injury is critical for locating primal symptoms, and thus, the decision making of surgical intervention [13].
Compared to acute ligament injury, the proper diagnosis of chronic ligament injury is difficult because of the uncertain association between initial inversion trauma history and ligament injury and diversified manifestation of chronic injuries on images [14,15]. The patients' history and clinical tests are important in diagnosis. A meta-analysis discussed the accuracy of clinical tests in diagnosing ankle ligament injury and claimed that clinicians cannot rule out ligamentous injury when physical tests are negative, and additional imaging is warranted [16]. Imaging is a helpful diagnostic method according to consensus of experts [17,18].
However, different from acute injury, chronic ligament injury could show as either stretching, rupture, scarring, or thickening of the ligament on images, which confuses the clinical practitioners [19]. Magnetic resonance imaging (MRI), ultrasonography (US), stress radiography, and arthrography are often utilized for diagnosing chronic lateral ankle ligament injury. Separate studies on various imaging diagnostic techniques have been reported; however, the diagnostic effectiveness of different imaging techniques was still controversial, partly due to different existing reference standards [20,21].
A previous systematic review assessed US for diagnosis of chronic ankle instability, but utilized variable gold standards, and did not provide pooled data [20]. Studies of MRI, US, stress radiography, and arthrography for diagnosis of chronic lateral ankle ligament injury were reported, but no previous systemic review provided synthesized data. This systemic review with meta-analysis is trying to analyze studies on diagnostic accuracy of different imaging techniques on chronic lateral ligament injury. Arthroscopic or surgical findings are set as the gold standards of ligament injuries [12,22].

Inclusion and exclusion criteria
The studies that met the following criteria were included: (1) cohort-type or cross-sectional studies; (2) evaluated MRI and/or US and/or stress radiography and/or arthrography for the diagnosis of chronic ATFL and/or CFL injury (regarding the classification of acute and chronic ligament injuries, we followed the decisions adopted by the authors of the studies included); (3) comparing imaging results with arthroscopic or surgical findings as reference standards, and (4) reported data that enabled the calculation of the number of true positive (TP), true negative (TN), false positive (FP), and false negative (FN).
The following criteria were used to exclude underqualified studies: (1) acute injury patients; (2) patients with confounding factors like ankle fracture, history of previous foot, and ankle surgeries; (3) without clearly described arthroscopic or surgical findings as their reference standards; (4) cadaveric studies or studies utilizing animal models; and (5) non-English articles.

Search strategy
A systematical literature search was conducted to include the following three databases: MEDLINE, EMBASE, and Cochrane Central Register of Controlled Trails (CEN-TRAL). The detailed search strategies (Table 1) were first developed in MEDLINE and were then adjusted and applied in the other two databases.
Retrieved articles from each database were at first screened for duplication. Then, after titles and abstracts screening, relevant studies for this systemic review underwent full-text screening. Eligible studies were included according to the aforementioned inclusion and exclusion criteria.

Data extraction and quality assessment
The extracted data include authors, publication years, demographic features of participants, study design, index  [14,19,23]. We eliminated this diversity by dichotomized imaging results to "injured" and "intact" in this study for better comparability among different included studies. Different kind of injuries such as "stretching," "rupture," "scarring," or "thickening" are all categorized as "injured." Two authors independently extracted these data and filled previously drafted forms for this review. Results of the two authors were cross-validated, and discrepancies were mediated by the third author. The quality of the included articles was assessed through revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool, a widely recommended scale for diagnostic test evaluation [24]. According to QUADAS-2 tool, risk of bias was assessed in terms of patient selection, index test, and reference standard.

Statistical analysis
Sensitivity and specificity of each index test in individual study were calculated in Meta-DiSc, version 1.4.0, using the extracted data of TP, FN, FP, and TN. Pooled sensitivity and specificity were calculated using the total number of TP, FN, FP, and TN subjects in all relevant studies. Likelihood ratio (LR) evaluates the discriminatory properties of the test results [25]. Positive and negative likelihood ratio evaluates the positive and negative test results respectively. Pooled positive and negative likelihood ratio was calculated using random effects model. The diagnostic odds ratio (DOR) is defined as (true positives × true negatives) / (false positives × false negatives), which evaluates the overall diagnostic test performance combining sensitivity, specificity, and likelihood ratio [26]. Pooled DORs were calculated using random effects model.
All final outcomes were presented with 95% confidential interval. Pooled sensitivity, specificity, likelihood ratio, and DOR are calculated concerning each subgroup. Heterogeneity testing was assessed using the I 2 test. A value of I 2 > 50% was considered to be significant heterogeneity among the pooled data.

Description of included studies
A total of 178 articles were retrieved from MEDLINE. One hundred eighty-eight articles were retrieved from EMBASE. Fifteen articles were retrieved from Cochrane Central Register of Controlled Trails. After deleting duplications, a total of 249 studies were identified in the primary search of three aforementioned databases. Then, these studies were screened for eligible studies as presented in the flow chart ( Fig. 1). Twenty-four studies underwent full-text screen, and 9 studies were excluded for the following reasons: inconsistent reference standards among subjects [21,[27][28][29], studies on other lesions associated with chronic ankle instability [30][31][32][33], and heterogeneous population with inadequate data for chronic injury group [34].

Methodological quality assessment
Methodological quality assessment was conducted for each study using QUADAS-2 tool (Fig. 2). All studies mentioned that participants were diagnosed of chronic ligament injury, of which 9 studies [36-38, 42, 43, 45, 46, 48, 49] mentioned the time from initial ankle inversion trauma to study commencement or duration when patients presented with chronic symptoms. Six of 15 studies [35,38,39,41,44,47] were categorized as high risk of bias due to patient selection. All studies mentioned arthroscopic or surgical findings as their reference standards, of which 3 studies [37,40,46] mentioned that the reference standard results were interpreted without knowledge of the results of the index tests. All studies mentioned that investigators were blind to arthroscopic or surgical findings when interpreting the diagnostic imaging.

Diagnostic accuracy of imaging techniques
Considering the combination of different ligaments (ATFL and CFL) and imaging techniques (MRI, US, stress radiography, and arthrography), 8 subgroups were analyzed. Extracted data of 8 subgroups are listed in Fig. 3. Pooled sensitivity, pooled specificity, and their 95% confidential interval are listed in Fig. 4 in the form of zones of mostly bad imaging efficacy (ZOMBIE) plot as Richardson [50] described. Detailed pooled data are listed in Table 3

Discussion
Accurate diagnosis of chronic lateral ankle ligament injury is considered critical for surgical intervention of chronic ankle instability [13]. Imaging diagnosis is usually noninvasive and can be of value when physical tests are ambiguous [16]. Results for the four including imaging diagnostic techniques have been reported; however, the diagnostic accuracy of each imaging technique is variable in different studies, partly due to different reference standards applied [20,21]. Our study set arthroscopic or surgical findings as the gold standard to investigate four imaging techniques. This gold standard reduced the heterogeneity of different studies.
In our results, the pooled sensitivities and specificities of US in diagnosing chronic ATFL and CFL injury were all above 0.90. When diagnosing chronic ankle instability, another systemic review involving 6 articles reported sensitivity of US ranged from 84.6 to 100% and specificity of US ranged from 90.9 to 100% [20]. In agreement with this systemic review, our results demonstrated US is an effective imaging technique in diagnosing chronic lateral ankle ligament injury.
US can precisely discriminate different ligamentous conditions, such as torn, lax, or thickened ligaments [40]. Comparing US results of healthy people and patients with chronic ankle instability, Liu et al. reported the differences     [53]. The study of Cho et al., which was included in the current review [38], utilized stress US. Other studies included in the current review did not utilize stress US technique, but still presented high diagnostic accuracy. However, US is of limited value in assessing bone or cartilage and highly experiencereliant and may be much less efficacious in less experienced hands. Moreover, only 4 studies [37,38,40,45] included in this systemic review were related to the US examination of chronic lateral ankle ligament injuries, of them only one study [37] involved CFL injuries. Only two included studies [38,45] compared US with MRI findings and reported similar diagnostic effectiveness. US is an effective imaging technique in diagnosing chronic lateral ankle ligament injury, yet more studies are warranted to compare US with other imaging techniques. In our results, MRI exhibited the pooled sensitivities of 0.83 in diagnosing chronic ATFL injury and 0.56 in diagnosing chronic CFL injury. The pooled specificities were around 0.8 in diagnosing chronic ATFL and CFL injury. In clinical practice, MRI is highly recommended in diagnosing ligamentous injuries [54]. Also, MRI was reported to be effective in diagnosing intra-articular lesions of chronic ankle instability, including osteochondral lesions of talus, syndesmotic injuries, and impingement syndromes [55,56]. Using MRI, Tao et al. [57] reported more cartilage lesions in patients with combined injuries of the ATFL and CFL, compared to patients with only ATFL injury. A study showed that 86.7% of the experts recommended MRI before considering surgery in chronic ankle instability patients [18]. However, according to our results, MRI did not provide the highest sensitivity or specificity in diagnosing chronic ligament injuries. It presented different patterns in diagnosing chronic ATFL and CFL injuries. The sensitivity for diagnosing chronic ATFL injury (0.83 [0.78, 0.87]) was higher than that for diagnosing chronic CFL injury (0.56 [0.46, 0.66]). MRI is still irreplaceable in assessing chronic lateral ankle ligament injury because it is frequently performed to confirm or exclude the presence of concomitant lesions and influence the precise surgical technique for a certain patient. Two included studies [38,45] in the current systemic review compared US with MRI findings and did not report significantly different diagnostic effectiveness.
Arthroscopy was recommended as a complementary to MRI for definitive diagnosis [17,18,22]. An expert consensus from Guillo et al. recommended that an arthroscopy should be performed at the time of surgery unless intra-articular pathology has been excluded by MRI scan and there is no history of pain [17].
In our results, stress radiography demonstrated a sensitivity of 0.81 and a specificity of 0.92 in diagnosing chronic ATFL injury. Tourné et al. suggested that stress radiography presented high specificity (up to 100%) but low sensitivity (57%), suggesting dynamic radiographs only have diagnostic value if they are positive [58]. According to our results, stress radiography showed similar high specificity as US. When diagnosing chronic ATFL injury, the sensitivity of stress radiography is still above 0.80, and the pooled LR− is 0. 24 [0.11-0.54], similar as that of MRI. Negative stress radiographic findings decreases the post-test probability of chronic ATFL and CFL injury in patients with ankle inversion trauma history. However, ligamentous laxity detected through stress radiography is not synonymous with chronic ankle instability [59]. Large variability in talar tilt and anterior drawer stress radiography precludes their routine use in diagnosing chronic ankle instability [60].
The role of anesthesia in stress radiography was reported by McCaskie et al. [61], suggesting larger discriminative capability under anesthesia. Amongst studies included in the current review, only one study reported stress radiography under anesthesia [43]. The large heterogeneity between stress radiography results for chronic CFL injury may partly explained by anesthetic condition. Though stress radiography is helpful in the diagnosis, it is hardly a useful tool in predicting surgical outcomes. Jeong et al. reported stress radiography-positive patients were not statistically significantly different from stress radiography-negative patients in prognostic features [62]. Arthrography is indicated for staging and detecting intra-articular lesions of chronic lateral ankle instability [63]. Samoto et al. reported that patients with combined injuries of the ATFL and CFL diagnosed through arthrography had worse prognosis than patients with only ATFL injury [64]. Arthrography is an invasive imaging technique. With the spread of arthroscopy, which is also invasive but much more accurate, this technique is now losing its popularity. Only 3 studies before year 2010 discussed arthrography [36,39,48].
There are several limitations in the current systemic review. First, 6 of the 15 included studies were graded as high risk of bias due to patient selection. Unlike metaanalysis of clinical intervention, in meta-analysis of diagnostic tests, it is common to include case-control studies considered as high risk of bias [65]. Case-control studies create a preselected patient population and should be interpreted with caution. Second, associated lesions of chronic lateral ankle ligament injury were not discussed in the current review; however, these associated lesions spotted on images would certainly affect the judgement of clinicians. Third, some studies compared the diagnostic accuracy of identical imaging technique with different parameters and/or configurations on diagnosing chronic lateral ankle ligament injury. Strength of the MRI machines varied among different studies. This diversity in configuration may cause the heterogeneity within each subgroup in this study. Moreover, the size of the included studies was relatively small. Of the 15 included studies, only a total of 695 participants were   . This is based on the receiver operating characteristic (ROC) plot. The colored dot stands for pooled sensitivity and specificity of each imaging technique, and the colored oval stands for 95% confidence interval of each imaging technique (green, MR; blue, US; red, arthrography; yellow, stress radiograph). Upper-left boomerang-shaped area formed by white and light grey zones defines zone of acceptable efficacy. Upper arm of the boomerang-shaped area stands for negative likelihood ratio below 0.33. Left arm of the boomerang-shaped area stands for positive likelihood ratio above 3

Conclusion
This systematic review with meta-analysis investigated the accuracy of imaging for the diagnosis of chronic lateral ankle ligament injury. Our results demonstrated that ultrasound manifested high diagnostic accuracy in diagnosing chronic lateral ankle ligament injury. MRI presented the diagnostic sensitivity and specificity around 0.8 for chronic ATFL injuries, but much lower sensitivity for chronic CFL injuries. Clinicians should be aware of the limitations of MRI in detecting chronic CFL injuries. Stress radiography showed similar high specificity as US in diagnosing chronic lateral ankle ligament injury and showed diagnostic value of both positive and negative results. Arthrography demonstrated sensitivity and specificity around 0.9 in diagnosing chronic CFL injury, but limited literature recommended arthrography due to invasiveness, especially in recent years. When the diagnosis of chronic lateral ankle ligament injury is uncertain after careful inquiry of the patient's history and physical tests, US may be a helpful imaging technique in experienced hands.

Availability of data and materials
All data generated or analyzed during this study are included in this published article.