| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Emergency Radiology |
1 From the Program for the Assessment of Radiological Technology (ART Program), Departments of Radiology (J.J.N., E.H.G.O., A.Z.G., G.P.K., M.G.M.H.), Epidemiology and Biostatistics (J.J.N., E.H.G.O., M.G.M.H.), Orthopaedic Surgery (J.A.N.V.), and Traumatologic Surgery (A.B.v.V.), Erasmus MC, University Medical Center Rotterdam, Dr Molewaterplein 40, 3015 GD Rotterdam, the Netherlands. Received July 7, 2003; revision requested September 23; revision received April 5, 2004; accepted May 19. Supported in part by the Revolving Fund from the Erasmus University Medical Center Rotterdam and by an unrestricted grant from Esaote, Genoa, Italy. Address correspondence to M.G.M.H. (e-mail: m.hunink@erasmusmc.nl).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Informed consent was obtained from all participating patients, and the institutional review board approved the randomized controlled trial and use of data to create prediction models. In a prospective controlled trial, 197 patients with recent ankle trauma (92 women, 105 men) were randomized into two groups: those who underwent radiography and those who underwent a combination of radiography and MR imaging. Data about side of injury, trauma mechanism, and results of radiography and MR imaging were collected. Additional treatment was necessary in 109 of 197 patients after their initial hospital visit. With univariable and multivariable regression analysis, four models were created for prediction of treatment.
RESULTS: In univariable analysis, age (odds ratio [OR], 1.02; 95% confidence interval: 1.00, 1.04), radiographic results (OR, 7.92; 95% confidence interval: 3.17, 19.8), and positive or uncertain results in patients who underwent MR imaging versus patients who did not (OR, 2.42; 95% confidence interval: 1.25, 4.70) were predictive of treatment. In the multivariable analysis, positive or uncertain MR imaging results (OR, 2.61; 95% confidence interval: 1.28, 5.30) contributed significantly to prediction of subsequent treatment. Negative MR imaging results did not contribute significantly (OR, 0.66; 95% confidence interval: 0.27, 1.61).
CONCLUSION: A limited MR imaging examination in initial evaluation of acute ankle injury with radiography has additional predictive value in identification of patients who need treatment but does not add significant information in identification of those who can be discharged without further follow-up. A limited MR imaging examination cannot replace radiography for prediction of need for additional treatment.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The potential value of MR imaging for examination in the acute phase after trauma is in the identification of patients who will not require any treatment and can thus be discharged without further follow-up. Furthermore, patients who have substantial soft-tissue injury are likely to receive a diagnosis at an earlier stage; thus, delay in treatment can be avoided, with potential cost savings to society because of diminished loss in productivity. To our knowledge, MR imaging has not been evaluated in this context before.
The purpose of this study was to assess the predictive value of a short MR imaging examination, in addition to or instead of radiography, performed in patients who present with acute ankle trauma to identify those who require additional treatment versus those who do not and can be discharged without further follow-up.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The included subjects were randomized to either the group of patients who underwent radiography only or the group of patients who underwent radiography combined with a short MR imaging examination. Randomization was performed by research staff and radiology technologists on service (including J.J.N. and E.H.G.O.). Patients were excluded if they had additional substantial injury of the head, the back, the thorax or the abdomen; if they had a compound ankle fracture; if they were in need of urgent treatment (eg, in case of ankle luxation); if they had preexisting symptoms of the same ankle; or if they were intoxicated with alcohol or had ingested nonmedicinal drugs. Patients were included from 8:00 AM to 11:00 PM, 7 days a week. Randomization was performed by research staff and radiology technologists on service. Although inclusion was intended to be consecutive, only about half of all eligible patients were randomized. A few patients refused to participate, but most patients were inadvertently missed and were thus not included.
Patients
From August 1999 to May 2001, 202 patients, 107 male patients (mean age, 37.8 years; range, 11.9–84.7 years) and 95 female patients (mean age, 30.8 years; range, 13.7–79.6 years), were included. Male patients were significantly older than female patients (P = .001, t test). The number of patients randomized to the group of those who were undergoing radiography was 103, and the number randomized to the group of those who were undergoing radiography followed by MR imaging was 99. A flow diagram that shows the progress of subjects as they passed through the study is presented in Figure 1. One patient, randomized to the MR imaging group, did not undergo the MR imaging examination because the MR imaging system was unavailable. Five randomized patients were incorrectly included in the study: Three of them appeared to have preexisting symptoms of the ankle, and two had a ruptured Achilles tendon; in fact, there was no indication for radiography in these latter two patients, since they were not suspected of having another ankle injury. These five patients were excluded from the analysis, which left 197 patients in our study.
|
|
Outcome
We used the requirement of additional treatment after the initial presentation at our hospital as an outcome measure. This outcome measure was used to analyze whether initial MR imaging can help to add information in the discrimination between patients who do not need follow-up versus those who need additional treatment and therefore should have follow-up after their initial visit. Additional treatment could consist of antipronation taping, a soft cast, a plaster cast, physical therapy, or surgery. Since indications for physical therapy are generally not very well defined, one may question whether physical therapy should be considered an end point. Therefore, we performed a separate analysis in which physical therapy as such was not considered a specific treatment. The treatment decision could be made at the first visit of the patient, but it could also be made later during follow-up and not necessarily by the same physician. The number of patients who received additional treatment after their first visit to the emergency department was 109; 88 patients did not receive additional treatment.
Data Collection
Data were collected (J.J.N., E.H.G.O.) about age, sex, side of trauma, trauma mechanism (direct, indirect, or unknown), radiographic results, MR imaging results (traumatic injury visible, not visible, or uncertain), and therapeutic procedures. To evaluate the effect of the trauma mechanism on the outcome— since indirect trauma is associated with rupture of ligaments and osteochondral lesions, whereas this is seldom the case in direct trauma—the trauma mechanism was scored as direct trauma if patients suffered a direct hit on the joint (eg, by a direct hit with a foot or hockey stick) and as indirect trauma in case of an inversion, eversion, hyperflexion, or hyperextension injury or a combination thereof. Data were obtained by means of direct assessment and review of all information available from emergency department records, hospital records, outpatient clinic records, and the electronic hospital information system. Questionnaires were sent to all patients at 1, 6, and 12 weeks and at 6 months after the initial visit to obtain information about treatment received outside our hospital. When questionnaires were not returned, we interviewed the patients by telephone (J.J.N., E.H.G.O.), provided they could be reached within two attempts of calling them. The final diagnosis was obtained with review of all information from diagnostic imaging, follow-up, and questionnaires.
The follow-up period was 6 months. If the patients indicated on the questionnaire that they did not have daily symptoms anymore, however, the follow-up was terminated, since no additional effects of the injury were to be expected.
Image Interpretation
The radiographs and the MR images were interpreted as positive if signs of recent traumatic injury were visible. For the radiographs, these signs were fracture, luxation or subluxation, epiphysiolysis, osteochondral lesion, loose body with signs of recent traumatic origin (sharp edges), or widened distal tibiofibular syndesmosis. The MR image was interpreted as positive if one or more of the following lesions were present: fracture, luxation or subluxation, epiphysiolysis, osteochondral lesion, or ligament rupture. A fracture visible on the MR image but not on the radiograph was regarded as an occult fracture. A ligament was regarded as ruptured if discontinuity of the ligament was observed and the ligament was surrounded by edema or hematoma, which helped to distinguish it from an old rupture. A ligament was regarded as partially ruptured if the ligament was edematous but had intact continuity. If the result was uncertain, that result was interpreted as such.
Since even minor traumatic injuries of the joint often cause abnormalities visualized at MR imaging, we excluded findings that generally do not need specific treatment. For this reason, joint fluid, soft-tissue edema, bone marrow edema, and hematoma without other lesions were not considered positive findings, although these findings may indicate the presence of another lesion. It may be debated whether bone marrow edema should be regarded as not likely to contribute to the prediction of necessary additional therapy, since one may argue that bone marrow edema may mimic a trabecular fracture or it may be an initial sign of osteochondritis dissecans. Therefore, we reanalyzed the data, with bone marrow edema considered as a positive MR imaging finding.
Statistical Analysis
By using univariable logistic regression analysis, we determined the odds ratio (OR) and the 95% confidence interval of each variable used in the prediction of the outcome. For the regression analysis, we recoded the variable we called "MR imaging results," and it had three possible outcomes: positive or uncertain, negative, or MR imaging not performed. The MR imaging results were analyzed, with the outcome of "MR imaging not performed" as the reference, since we were interested in the additional value of MR imaging findings compared with the absence of MR imaging information in the initial phase. When the number of uncertain MR imaging or radiographic results was large enough to be analyzed as a separate dummy variable, this analysis was performed as such; otherwise, the uncertain results were grouped together with the positive results, since we wanted to avoid sending patients home without follow-up if the imaging results were uncertain.
Because the outcome of radiographic and MR imaging results in the same patient were correlated, we created an additional integrated imaging variable that contained both radiographic and MR imaging results. This variable had five possible outcomes: (a) negative radiographic results and MR imaging not performed, (b) positive radiographic results and MR imaging not performed, (c) negative results at both radiography and MR imaging, (d) discrepant results at both radiography and MR imaging, and (e) positive results at both radiography and MR imaging.
With this combined-imaging variable, we took into account the diagnostic interaction between radiography and MR imaging, since these variables were at least partly dependent: A positive radiographic result would likely be accompanied by a positive MR imaging result. The outcome with negative radiographic results and MR imaging not performed was considered the reference for analysis of the combined-imaging variable.
A variable was included in the multivariable logistic regression analysis if the P value was less than .1 (
2 test) in the univariable analysis. This lenient P value was chosen to avoid rejection of a variable that could potentially contribute in the prediction of the outcome (treatment) in the multivariable analysis. We combined the variables by using multivariable logistic regression analysis to create a prediction rule for the identification of those patients who needed further treatment after the initial visit versus those who did not. Several models, which all included age, were evaluated to explore the predictive value of the imaging variables: In model 1, radiographic results were added to age; in model 2, radiographic results were omitted and MR imaging results were used instead; in model 3, both radiographic and MR imaging results were included; and in model 4, the combined-imaging variable was used.
The likelihood ratio test was used to assess the significance of each variable in the model in a stepwise backward approach. The calibration of the models was assessed by using the Hosmer-Lemeshow goodness-of-fit test (15). With calibration, the degree of correspondence between the probabilities estimated by using the model and the actual treatment received by the patients is evaluated. If the P value of the test statistic is large, the model is well calibrated and the estimates fit the data well. The area under the receiver operating characteristic curve was used to evaluate the discriminative performance of the model and to compare predictive performance across models (16). The difference between the areas under the receiver operating characteristic curve of the different models was analyzed by using the z statistic and was corrected for the fact that the areas under the receiver operating characteristic curve were derived from the same cases. The correction factor was calculated by using the Pearson product moment correlation (17). The difference was considered significant if the P value was less than .05 for the z statistic.
The Akaike information criterion (AIC) was used to compare the predictive power of the models (18). The AIC is derived from the 2 statistic and takes both the accuracy and the complexity of the model into account by correcting for the number of degrees of freedom. In this way, direct comparison of models is possible. An AIC equal to or smaller than zero indicates that there is no predictive value of the model. The significance of a difference in AIC of nested models (eg, the model with radiographic results vs the model with radiographic and MR imaging results) was assessed by using the likelihood ratio test. In the case of nonnested models, a difference in AIC larger than 2.0 was considered significant. All analyses were performed by using software (SPSS for Windows, release 10.0.0; SPSS, Chicago, Ill).
Sample Size
In multivariable logistic regression analysis, at least 10 patients with an event and at least 10 patients without an event are necessary per variable analyzed (19). Because the data for this study were obtained from a randomized controlled trial in which the costs and effectiveness of early MR imaging of ankle trauma were addressed, the number of variables analyzed in the multivariable regression analysis was determined by the number of patients included in the randomized controlled trial. Since 109 of the included patients received additional therapy and 88 patients did not, we could analyze a maximum of nine (88 ÷ 10) variables in the multivariable regression analysis.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
|
|
The area under the receiver operating characteristic curve, which represents the discriminatory power of the model, was not significantly different between models 1 and 2, models 1 and 3, models 1 and 4, and models 3 and 4 (P = .21, .11, .59, and .15, respectively). Neither the replacement of radiographic results with MR imaging results (comparison of models 1 and 2) nor the addition of MR imaging results to radiographic results (comparison of models 1 and 3) changed the discriminatory power of the model significantly. The two models with both radiographic and MR imaging results (models 3 and 4), however, showed a significantly larger area under the receiver operating characteristic curve than did the model with only MR imaging results (model 2) (P = .002 and .04, respectively). This finding indicated the better discriminatory power of the models, with both radiographic and MR imaging results compared with the model with only MR imaging results.
Findings of repeat analysis with bone marrow edema defined as an abnormal MR imaging result showed only a marginal change of the ORs in the models containing MR imaging results. Reanalysis of the data without physical therapy as an end point showed that age was not significant in the univariable analysis anymore, but the ORs of the other variables and the model performance statistics changed only marginally.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
In the assessment of the value of a short dedicated extremity MR imaging examination, we did not aim at the demonstration of the accuracy of the short MR imaging examination for all possible lesions. This would not be feasible because of the variety of possible lesions. Moreover, for many lesions it would not be possible to verify the diagnosis, since only surgery would be an acceptable reference standard. Only a minority of patients were treated surgically, so for most patients a suitable reference standard was not available. Instead, we chose to perform an outcome study and assessed the predictive value of MR imaging, compared with the standard work-up, in the determination of the need for additional treatment after the initial visit.
Male patients were significantly older than female patients. Whether this difference was related to differences in activity in sports at different ages was unclear.
Although MR imaging has a high sensitivity in the detection of occult fractures, its sensitivity in the detection of avulsion fractures is low. Because of the limited field of view at MR imaging, two metatarsal fractures were missed.
We were not absolutely sure that bone marrow edema had to be regarded as a lesion that did not require treatment for the analysis. Bone marrow edema, as such, does not need treatment, but it may add to the prediction of the need of treatment. Findings of repeat analysis with bone marrow edema defined as an abnormal MR imaging result showed only a marginal change of the ORs in the models containing MR imaging results. One may argue that physical therapy should not be included as an end point, since the indication for physical therapy often is quite arbitrary. Findings of repeat analysis without physical therapy as an end point showed that age was not significant in the univariable analysis anymore, but the ORs of the other variables and the model performance statistics changed only marginally.
Among the limitations of this study was the fact that, although inclusion was intended to be consecutive, only about half of all eligible patients were randomized. This was mainly caused by missing eligible patients. The main reason for missing patients was that inclusion took place from 8:00 AM to 11:00 PM, 7 days a week, and thus many radiology technologists were on service, and they were supposed to recruit patients but forgot to do so regularly. Since it was often impossible to retrace whether these missed patients would have been eligible, exact numbers of missed eligible patients could not be obtained. Sometimes there was a language problem: In this case, patients did not understand the goal of the study, and informed consent could not be obtained. In two cases, the patient refused to participate. We do not expect that patients who were forgotten to be asked to participate or patients who were excluded because they did not understand the native language caused a selection bias, but it cannot be ruled out.
Another limitation of the study was that we assumed that the decision to treat a patient was always correct. It is, however, possible that in some cases treatment was unnecessary or that treatment was incorrectly withheld, although in the latter instance it is likely that most patients would return within the 6 months of follow-up because of persistent symptoms and would have been treated at that time. Furthermore, the patients were treated by different physicians. Since treatment of the various lesions was not standardized, the decision to treat and the choice of treatment varied across physicians. The influence of this variation on the study results, however, is likely to be very small, especially because differences in type of additional treatment would not influence the outcome.
The follow-up period in our study was 6 months. Although it is likely that treatment takes place within 6 months after the injury, it is possible that some patients underwent surgery after the follow-up period (eg, for ankle instability). However, most of these patients would have received physical therapy for proprioceptive training and improvement of muscle strength in the period preceding surgery and would have been considered as treated anyway.
The radiographic result as a single predictor of the need for therapy appeared to be of higher predictive value than was the MR imaging result. This can be explained by the fact that the majority of ankle traumas are inversion injuries. At our hospital, initial treatment of these injuries consists of bandaging, followed by taping or a soft cast, depending on the amount of swelling and pain, and does not depend on demonstration of a ruptured ligament. Ruptured ligaments are sometimes surgically repaired, but only in case of chronic ankle instability. At our hospital, even if an MR imaging examination reveals one or more ruptured ligaments, the treatment is the same as that for an ankle sprain without a ruptured ligament, and the MR imaging examination will have limited additional value in the prediction of need for treatment. This is one of the reasons why we did not find a significant predictive value for negative imaging results. Another reason is that many patients will undergo physical therapy if pain persists, irrespective of demonstrated lesions.
The short MR imaging protocol used in this study is associated with some loss of image quality, but in our view, the quality of the images we obtained was acceptable. Compared with a state-of-the-art protocol, however, it is possible that our accuracy was low, which may in part be responsible for the lack of discriminatory power that we found. The main reason, however, for lack of discriminatory power is the fact that, besides ligament rupture, very few other lesions that can be demonstrated with MR imaging and not with radiography were found in our patient group: We found no osteochondral lesions, no osteochondritis dissecans, only two occult fractures, and one case of epiphysiolysis of the distal fibula that was not recognized on the radiograph. By far, the majority of traumatic ankle injuries consisted of fractures and ankle sprains, with or without ligament rupture.
If the treatment policy for an ankle sprain, with or without ligament rupture, consists of a pressure bandage followed by taping or a soft cast, the consequences of a false-positive treatment are limited. This would be different if ruptured ankle ligaments are surgically repaired initially, in which case early information about the loss of integrity of the lateral ankle ligament is essential. Whether lateral ankle ligament ruptures should initially be treated conservatively or surgically is a matter of debate (20–23). Researchers in a meta-analysis study (24) concluded that surgical treatment leads to a better result than does functional treatment. If the treatment policy for ruptured ankle ligaments were to shift in the direction of initial surgery, there would be a role for an initial MR imaging examination in acute ankle injury. Since this is not the case at our hospital, we were not able to evaluate the added value of early MR imaging in this respect.
Whether a short MR imaging examination should be routinely used in the evaluation of patients with acute ankle injury will mainly depend on the cost-effectiveness of this strategy, which we intend to evaluate in future work.
In conclusion, a limited MR imaging examination in the initial evaluation of acute ankle injury has additional predictive value over and above radiography in the identification of patients who need treatment, but it cannot help in the identification of those who can be discharged without further follow-up. A limited MR imaging examination cannot replace radiography in the prediction of the need for additional treatment. MR imaging in the initial evaluation may, however, be valuable in a setting where ruptured ankle ligaments are immediately repaired surgically.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
See also the articles by Nikken et al and Oei et al in this issue.
Author contributions: Guarantor of integrity of entire study, M.G.M.H.; study concepts and design, all authors; literature research, J.J.N.; clinical studies, J.J.N., E.H.G.O.; data acquisition and analysis/interpretation, all authors; statistical analysis, J.J.N.; manuscript preparation, definition of intellectual content, revision/review, and final version approval, all authors; manuscript editing, J.J.N.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Related Articles
This article has been cited by other articles:
|
|
 |
|
  J. J. Nikken, E. H. G. Oei, A. Z. Ginai, G. P. Krestin, J. A. N. Verhaar, A. B. van Vugt, and M. G. M. Hunink Acute Peripheral Joint Injury: Cost and Effectiveness of Low-Field-Strength MR Imaging--Results of Randomized Controlled Trial Radiology, September 1, 2005; 236(3): 958 - 967. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
