Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print January 15, 2003, 10.1148/radiol.2263011892
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow Excluded Articles
Right arrow All Versions of this Article:
2263011892v1
226/3/837    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Oei, E. H. G.
Right arrow Articles by Myriam Hunink, M. G.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Oei, E. H. G.
Right arrow Articles by Myriam Hunink, M. G.
(Radiology 2003;226:837-848.)
© RSNA, 2003

Evidence-based Practice

MR Imaging of the Menisci and Cruciate Ligaments: A Systematic Review1

Edwin H. G. Oei, MSc, Jeroen J. Nikken, MD, Antonia C. M. Verstijnen, MSc, Abida Z. Ginai, MD, PhD and M. G. Myriam Hunink, MD, PhD

1 From the Program for the Assessment of Radiological Technology, Departments of Radiology (E.H.G.O., J.J.N., A.C.M.V., A.Z.G., M.G.M.H.) and Epidemiology and Biostatistics (E.H.G.O., J.J.N., A.C.M.V., M.G.M.H.), Erasmus University Medical Center Rotterdam, Dr Molewaterplein 50, Rm EE21-40a, 3015 GE Rotterdam, the Netherlands. From the 2001 RSNA scientific assembly. Received November 26, 2001; revision requested February 5, 2002; final revision received June 26; accepted July 16. Supported in part by the Foundation "Vereniging Trustfonds Erasmus Universiteit Rotterdam," the Foundation "Gerrit Jan Mulder Stichting," and the "Revolving Fund" of the University Hospital Rotterdam. Address correspondence to M.G.M.H. (e-mail: hunink@epib.fgg.eur.nl).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
PURPOSE: To systematically review and synthesize published data on the diagnostic performance of magnetic resonance (MR) imaging of the menisci and cruciate ligaments and to assess the effect of study design characteristics and magnetic field strength on diagnostic performance.

MATERIALS AND METHODS: Articles published between 1991 and 2000 were included if at least 30 patients were studied, arthroscopy was the reference standard, the magnetic field strength was reported, positivity criteria were defined, and the absolute numbers of true-positive, false-negative, true-negative, and false-positive results were available or derivable. Pooled weighted and summary receiver operating characteristic (ROC) analyses were performed for tears of both menisci and both cruciate ligaments separately and for the four lesions combined, by using random effects models. Differences were assessed according to lesion type.

RESULTS: Twenty-nine of 120 retrieved articles were included. Pooled weighted sensitivity was higher for medial meniscal tears than that for lateral meniscal tears. However, pooled weighted specificity for the medial meniscus was lower than that for the lateral meniscus. In summary ROC analyses performed per lesion, various study design characteristics were found to influence diagnostic performance. Higher magnetic field strength significantly improved discriminatory power only for anterior cruciate ligament tears. When all lesions were combined in one overall summary ROC analysis, magnetic field strength was a significant but modest predictor of diagnostic performance.

CONCLUSION: Diagnostic performance of MR imaging of the knee is different according to lesion type and is influenced by various study design characteristics. Higher magnetic field strength modestly improves diagnostic performance, but a significant effect was demonstrated only for anterior cruciate ligament tears.

Supplemental material: radiology.rsnajnls.org/cgi/content/full/2263011892/DC1.

© RSNA, 2003

Index terms: Knee, injuries, 452.485 • Knee, ligaments, menisci, and cartilage, 452.485 • Knee, MR, 452.121419 • Magnetic resonance (MR), technology


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
Since its introduction for clinical use in the mid-1980s, the role of magnetic resonance (MR) imaging in the diagnosis of knee lesions has been established. MR imaging has proved reliable and safe and offers advantages over diagnostic arthroscopy, which is currently regarded as the reference standard for the diagnosis of internal derangements of the knee. Arthroscopy is an invasive procedure with certain risks and discomfort for the patient and is preferably performed only for treatment purposes, provided that alternative noninvasive diagnostic modalities such as MR imaging are available (1,2).

Results of numerous diagnostic studies have been published in which MR imaging and arthroscopy of the knee were compared, and most have shown good diagnostic performance in detecting lesions of the menisci and cruciate ligaments. As a result, MR imaging has been increasingly used in the diagnostic work-up of knee lesions, but the high costs of purchasing and maintaining a high-field-strength MR imager and the often limited availability of such a unit has restricted its widespread use for this purpose. The use of middle- and low-field-strength MR imagers has created the possibility of using MR imaging more routinely in the diagnostic work-up of knee disorders at a lower cost. In addition, low-field-strength dedicated extremity MR imagers have been designed specifically for extremity imaging, and because of their compact size and low field strength, costs can be kept relatively low compared with those of middle- and high-field-strength imagers. The reported diagnostic performance of low-field-strength MR imaging is variable, however, which raises the question of whether low-field-strength MR imaging can reliably replace middle- and high-field-strength MR imaging for evaluation of knee lesions.

With the vast number of articles on MR imaging of knee lesions with a wide range in study design, imaging techniques, and results, it is difficult to get a good idea of the diagnostic performance of MR imaging and the factors that influence its accuracy. We found some limited review articles in which investigators attempted to summarize published results of MR imaging of the knee (2,3). To our knowledge, however, no systematic review with a meta-analysis of the diagnostic performance of MR imaging of the knee has been published to date.

The purpose of this study was to systematically review and synthesize the published data on the diagnostic performance of MR imaging of the knee, focusing on tears of the menisci and cruciate ligaments, and to assess the effect of study design characteristics and magnetic field strength on diagnostic performance. We followed published guidelines for conducting diagnostic meta-analyses (4).


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
Selection of Articles
We conducted a MEDLINE search of the English-language literature to identify original articles published between January 1991 and December 2000 on the diagnostic performance of MR imaging of knee lesions. Combinations of the following search terms were used: "magnetic resonance imaging," "knee," "meniscus," "cruciate ligament," and "arthroscopy." All articles that could not be excluded definitively on the basis of the title and abstract were retrieved in full text. We used the criteria for inclusion and exclusion listed in Table 1. The bibliographies of the original articles were screened to obtain additional references. For the articles that were excluded because absolute numbers or positivity criteria were lacking, we tried to contact the corresponding author to request additional information. Authors were also contacted if there were indications that in more than one article, they reported on overlapping patient populations. If the authors did not respond, we chose the article that provided the most unequivocal and clear data.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Criteria for Inclusion and Exclusion of Articles

 
Data Extraction
Two authors (E.H.G.O., A.C.M.V.) independently extracted results and characteristics of the study and MR imaging technique for each included article. For this purpose, a standardized data extraction sheet was used. Readers were not blinded to information about the authors, author affiliation, and journal name, since this has been shown to be unnecessary (5). The extracted study characteristics included publication year, country of origin, setting (academic or community hospital), patient characteristics, and aspects of study design, such as prospective versus retrospective design, inclusion of consecutive patients, and blinding. We also assessed the possibility of verification bias, which arises if patients are selected to undergo the reference test (arthroscopy) on the basis of the outcome of the test being evaluated (MR imaging), giving rise to an overestimation of sensitivity and an underestimation of specificity (6).

Three levels of likelihood were used to classify each study with respect to verification bias. We used "yes" if MR imaging was clearly used as a screening tool for arthroscopy. The category "possible" was assigned if, for unknown or unreported reasons, not all patients underwent both MR imaging and arthroscopy. We assumed that verification bias was absent if both MR imaging and arthroscopy were performed in all consecutive patients that were included in the study. Regarding the characteristics of MR imaging technique, we extracted the magnetic field strength of the MR imager and the number and type of MR imaging sequences.

From the articles in which investigators tabulated the results for different readers, we extracted the data of the first reader, unless the level of experience was explicitly stated to be different among the readers. In the latter case, we extracted the results of the most experienced reader. If results were tabulated for multiple observations per reader, we extracted the data of the first observation. Some articles reported separately the results obtained with multiple MR imagers, which we treated as individual studies if the MR imagers had different magnetic field strengths. If results were reported for multiple MR imaging sequences or techniques with the same MR imager, we extracted the data of the technique that was recommended by the authors.

All discrepancies between our two data extractors were recorded. To determine the level of agreement, {kappa} statistics were calculated for categorical variables, and Spearman correlation coefficients (r values) were calculated for continuous variables. To resolve discrepancies, a third data extractor (J.J.N.) assessed all discrepant items, and the majority opinion was used for analysis.

Data Analysis
Calculation of the natural logarithm of the diagnostic odds ratio.—For every lesion in each study, we calculated the natural logarithm of the diagnostic odds ratio (OR) (D), which represents a summary measure of the diagnostic performance or discriminatory power. This value is the measure of interest in summary receiver operating characteristic (ROC) analyses and is calculated as follows: D = ln{[(TP + 0.5)(TN + 0.5)]/[(FP + 0.5)(FN + 0.5)]}, where TP is the true-positive value, TN is the true-negative value, FP is the false-positive value, and FN is the false-negative value. Before calculating D, we first added 0.5 to each value to avoid undefined values of D and its variance resulting from zero values of each (7).

Assessment of publication bias.—We first evaluated the presence of publication bias, which could potentially arise if studies with positive results are more likely to be published than are those with negative results (8). Publication bias can be detected by constructing a funnel plot, in which the number of units measured (the number of knees) is plotted against the measure of interest (in our case, the natural logarithm of the diagnostic OR). In the absence of publication bias, the funnel plot shows a symmetric funnel-shaped distribution, whereas the distribution is asymmetric and skewed if publication bias is involved (9). Symmetry and shape of the funnel plots were judged by means of visual inspection.

Pooled weighted analyses.—To obtain a crude estimate of diagnostic performance of MR imaging in depicting tears of the medial and lateral menisci, anterior crutiate ligaments (ACLs), and posterior cruciate ligaments (PCLs), we performed a pooled weighted analysis per lesion, weighting with the reciprocal of the variance of each study. First, we tested for heterogeneity in effect size among studies, the result of which determines if a fixed or random effects model should be used (10,11). Pooled weighted sensitivity, specificity, and natural logarithm of the diagnostic OR (and 95% CIs) were then calculated by using a random effects model. We regarded two estimates as significantly different if their 95% CIs did not overlap.

Next, pooled weighted analyses for the different lesions were performed for various categories of magnetic field strength separately to obtain an idea of the influence of magnetic field strength on diagnostic performance. The range of field strengths was divided into either three categories (higher than 1.0 T, 0.5–1.0 T, or lower than 0.5 T) or two categories (1.0 T and higher vs lower than 1.0 T).

Summary ROC Analysis per Lesion Type
We subsequently performed summary ROC analysis for each of the lesions separately, which has advantages over pooled weighted analysis. First, the true- and false-positive rates for the different diagnostic studies can be summarized and synthesized, with adjustment for different positivity criteria among studies (12–14). Different positivity criteria exist among studies when institutions use different thresholds for labeling a test result as positive. For example, in one study, an area of high signal intensity in the course of the cruciate ligament may be considered indicative of a rupture, whereas in another study, investigators may require the absence of intact cruciate ligament margins as an additional finding before labeling the test result positive. Second, by using summary ROC analysis, one can identify and adjust for variables that have an influence on diagnostic performance. We applied a random effects model, which accounted for the residual interstudy heterogeneity, which may be present even after adjustment for characteristics such as population size, patient age and sex, positivity criteria, type of MR imager, or verification bias. A random effects regression model was developed in which the natural logarithm of the diagnostic OR of each study was the dependent variable. As the independent variable, a measure of the positivity criteria (S) of the study was calculated as follows: S = ln{[(TP + 0.5)(FP + 0.5)]/[(TN + 0.5)(FN + 0.5)]}, where TP is the true-positive value, FP is the false-positive value, TN is the true-negative value, and FN is the false-negative value. We added 0.5 to each value as before to prevent undefined values of the positivity criteria. By using this method, one adjusts for the variations in positivity criteria that are used implicitly or explicitly in different studies and that are assumed to influence diagnostic performance.

Other variables that influence diagnostic performance were identified by adding them to the regression model and by assessing their regression coefficient and influence on the model. The variable, depending on the type, can be included either directly as a continuous variable or as a dummy variable. In this bivariate summary ROC analysis, we considered variables as explanatory if they were statistically significant in the regression model (P < .05) or if the regression coefficient was at least 1.0 for dummy variables or 1.0 over the range of the variable values. Variables with a P value between .05 and .10 were retained in the model if their inclusion decreased the method-of-moments {tau}2 estimate by at least 10% compared with that in the univariate model. The method-of-moments {tau}2 calculation provides a measure of between-study variance and is higher if there is more heterogeneity among studies and zero if studies are homogeneous.

We assessed as potential predictors of diagnostic performance the effect of publication year (continuous variable), country of origin (in North America vs other), type of hospital (academic, community, or not reported), mean age (continuous variable, 35 years or older vs younger than 35 years), inclusion of consecutive patients (yes, no, or not reported), prospective versus retrospective study design, verification bias (yes, no, or possible), blinding of the arthroscopist to the MR imaging result (yes, no, or not reported), magnetic field strength of the MR imager (continuous variable, 1.5 T vs lower than 1.5 T), and number of MR sequences used (continuous variable). In five articles, mean age was not reported, and in the analyses in which mean age was examined, regression analysis was performed for the subset of studies with age available. In all studies, the radiologist was blinded to the arthroscopic findings because MR imaging was always performed prior to arthroscopy. Therefore, only the effect of blinding of the arthroscopist was assessed in this study.

Subsequently, we performed multivariate summary ROC analysis, in which the explanatory variables previously identified in the bivariate analysis were included one by one in a stepwise forward-selection regression model. We started with the variable that decreased the method-of-moments {tau}2 estimate the most and kept it in the model on the basis of the same criteria as those used in the bivariate summary ROC analysis. We always retained the measure of positivity criteria (S) in the model, since a difference in positivity criteria among studies is a key concept of summary ROC analysis. Even though the positivity criteria as defined in the articles seemed largely the same, interpretation of MR images of the knee may vary in subtle ways in different hospitals and among different radiologists.

This yielded the final model for each lesion from which multivariate summary ROC curves were plotted, adjusted for significant covariables that were set to the mean values or values indicating the ideal study design, as appropriate.

Overall Summary ROC Analysis of All Lesions
For the purpose of increasing statistical power and precision and comparing the results with those of the separate analyses per lesion, we also analyzed all four lesions in a single model. To compare diagnostic performance among lesion types, we created dummy variables to code if the lesion of interest was a meniscal versus cruciate ligament tear.

The effect of each significant predictor in the final models is reflected by its regression coefficient. A positive regression coefficient indicates better discriminatory power of MR imaging in studies with that predictor, compared with that in studies without the corresponding characteristic. A negative regression coefficient indicates reduced diagnostic performance in studies with that characteristic. Finally, we calculated relative diagnostic ORs of all predictors in the multivariate models by taking the antilogarithm of the regression coefficient. A relative diagnostic OR can be interpreted as the diagnostic performance of a test in studies with a certain characteristic, relative to its performance in studies without that corresponding feature. Thus, a relative diagnostic OR greater than 1.0 indicates that studies with that characteristic yield better diagnostic performance of MR imaging than that in studies without the corresponding feature, whereas a relative diagnostic OR less than 1.0 indicates reduced discriminatory power in studies with that characteristic. All analyses were performed by using STATA (versions 6.0 and 7.0; Stata, College Station, Tex) and SPSS for Windows (version 9.0.0; SPSS, Chicago, Ill) software.

Sensitivity Analyses
To assess the dependence of the five final multivariate models on the results of individual studies, we performed the jackknife type of sensitivity analyses for each model. By using this method, one can determine the contribution of the individual studies to the overall results by performing multiple summary ROC analyses with each article excluded in turn (the jackknife method).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 
Literature Search and Data Extraction
Our MEDLINE search resulted in 804 articles, of which 120 were retrieved after we evaluated the titles and abstracts. Eighty-nine articles were excluded because (a) the article was a review or descriptive article without original data on diagnostic performance (n = 20), (b) fewer than 30 patients were studied (n = 9), (c) magnetic field strength was not reported or was variable (n = 13), (d) absolute numbers were not available and could not be derived (n = 19), (e) no positivity criteria at MR imaging were reported (n = 3), (f) results were reported for the medial and lateral meniscus combined (n = 6), (g) only the value of specific indirect signs or features of knee lesions at MR imaging was assessed (n = 15), or (h) the patient population was suspected to overlap with that of another study (n = 4). Furthermore, we excluded one article in which patients and random control subjects were retrospectively selected on the basis of the presence or absence of a definite complete ACL tear (case-control design) without accounting for all patients in the study period (15). Another article was excluded in which the whole knee was used as the unit of analysis (16). All excluded articles, together with the reasons for exclusion, are listed in the table that is available as supplemental material on the Radiology website (Table E1, radiology.rsnajnls.org/cgi/content/full/2263011892/DC1).

In this table, each article is classified according to the reason why the article was excluded from our analysis, although multiple reasons may be applicable to one article. We attempted to contact 16 authors for more information, but this did not result in additional articles for inclusion. Thus, 29 articles were included in this meta-analysis, which comprised 27 studies on both menisci, 23 studies on ACL tears, and 12 studies on PCL tears. The articles, together with the corresponding study characteristics, are listed in Table 2. Results for the menisci and cruciate ligaments per study, as well as features of the MR imaging technique, are listed in Tables 3–5.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Characteristics of Included Studies

 

View this table:
[in this window]
[in a new window]

  		TABLE 3. MR Imaging Technique and Characteristics of Included Studies: Medial Meniscal Tears

 

View this table:
[in this window]
[in a new window]

  		TABLE 4. MR Imaging Technique and Characteristics of Included Studies: Lateral Meniscal Tears

 

View this table:
[in this window]
[in a new window]

  		TABLE 5. MR Imaging Technique and Characteristics of Included Studies: Cruciate Ligament Tears

 
We extracted 379 items from the articles; 46 discrepancies occurred between our two data extractors (Table 6), which were resolved by the third extractor. Depending on the variable, agreement between the two data extractors ranged from moderate to almost perfect (r value range, 0.80–1.00; {kappa} value range, 0.54–1.00).


View this table:
[in this window]
[in a new window]

  		TABLE 6. Analyzed Variables and Measures of Agreement between Two Data Extractors for 379 Items in 29 Articles

 
Publication Bias
At visual inspection, the funnel plots for the menisci showed an almost perfectly funnel-shaped distribution (Fig 1), suggesting that publication bias is unlikely. For both cruciate ligaments, the funnel plots were almost funnel shaped, although not perfectly so (Fig 2), but there were sufficient small studies with a low diagnostic performance to suggest the absence of substantial publication bias.



View larger version (15K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Funnel plot for the menisci in which the number of knees is plotted against the discriminatory power of MR imaging (natural logarithm of the diagnostic OR). For both medial and lateral menisci, the distribution of data points appears to be fairly funnel shaped and symmetric, indicating that publication bias is unlikely. {blacksquare} = results for medial meniscus from individual studies, {circ} = results for lateral meniscus from individual studies.

 


View larger version (14K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Funnel plot for the cruciate ligaments in which the number of knees is plotted against the discriminatory power of MR imaging (natural logarithm of the diagnostic OR). For both ACL and PCL lesions, the data points show a slightly skewed distribution. There is a considerable number of small studies with low diagnostic performance, however, suggesting the absence of substantial publication bias. {blacksquare} = ACL results from individual studies, {circ} = PCL results from individual studies.

 
Pooled Weighted Analysis
The test for heterogeneity showed significant results for all lesions, with the exception of PCL tears. We therefore used a random effects model, which can accommodate both heterogeneous and homogeneous effect sizes among studies (10,11). For the diagnosis of meniscal tears (Table 7), pooled weighted sensitivity was higher for the medial meniscus (93.3% [95% CI: 91.7, 95.0]) than that for the lateral meniscus (79.3% [95% CI: 74.3, 84.2]), whereas pooled weighted specificity for medial meniscal tears (88.4% [95% CI: 85.4, 91.4]) was lower than that for lateral meniscal tears (95.7% [95% CI: 94.6, 96.8]). There was no statistically significant difference in sensitivity for complete ACL tears versus PCL tears (Table 7) (94.4% [95% CI: 92.3, 96.6] and 91.0% [95% CI: 83.2, 98.7], respectively), but the specificity for ACL tears was lower than that for PCL tears (94.3% [95% CI: 92.7, 95.9] and 99.4% [95% CI: 98.9, 99.9], respectively). No significant difference was demonstrated in the pooled natural logarithms of the diagnostic ORs for the different lesions.


View this table:
[in this window]
[in a new window]

  		TABLE 7. Pooled Weighted Results per Type of Lesion (random effects model)

 
The results of the separate pooled weighted analyses for various categories of magnetic field strengths (not tabulated) suggested a modest trend toward better diagnostic performance for higher-field-strength categories. None of the differences were found to approach statistical significance, however, and the CIs were all extremely wide because of a limited number of studies per category of magnetic field strength.

Summary ROC Analysis per Type of Lesion
In the multivariate summary ROC analysis for the medial meniscus, blinding of the arthroscopist was the only predictor of diagnostic performance. For the lateral meniscus, mean age was the only significant variable in the final model (Table 8). Publication year, mean age, and magnetic field strength were predictors in the multivariate model for ACL tears. The final model for PCL tears consisted of publication year, academic hospital setting, verification bias, and number of MR sequences.


View this table:
[in this window]
[in a new window]

  		TABLE 8. Multivariate Summary ROC Models per Type of Lesion and for All Lesions Combined (random effects model)

 
The summary ROC curves based on the final regression models, together with the results of the individual studies, are plotted in Figure 3. For the medial meniscus, we set the dummy variable for blinding to zero according to clinical practice, where arthroscopy is ideally performed on the basis of and with knowledge of the MR imaging findings. The summary ROC curve for lateral meniscal tears was plotted for a patient aged 30 years. For the ACL, the publication year was set to 1995, and the curve was plotted for a patient aged 30 years with use of a 1.0-T MR imager. The curve for PCL lesions was adjusted to publication in 1995, an academic hospital setting, use of three MR sequences, and the absence of verification bias. The adjusted summary ROC curves have different positions in ROC space (Fig 3), indicating that there are differences in diagnostic performance among the types of lesions. Whereas overall discriminatory power is not much different for the menisci, diagnostic performance is clearly better for the ACL compared with that for the PCL, since the curve for the ACL is located more toward the upper left corner.



View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Summary ROC curves for the four types of lesions separately and the results of the individual studies ({circ}) on the basis of the final models per lesion. Upper left: Summary ROC curve for medial meniscal tears. The dummy variable for blinding was set to zero, according to clinical practice, in which MR imaging is ideally performed before arthroscopy. Upper right: Summary ROC curve for lateral meniscal tears, adjusted to a patient aged 30 years. Lower left: Summary ROC curve for ACL tears, adjusted to publication in 1995, a patient aged 30 years, and a 1.0-T MR imager. Lower right: Summary ROC curve for PCL tears, adjusted for publication in 1995, academic hospital setting, three MR sequences in the imaging protocol, and absence of verification bias. The adjusted summary ROC curves demonstrate the differences in diagnostic performance among the lesion types. Whereas overall discriminatory power is not much different for the menisci, diagnostic performance is clearly better for the ACL compared with that for the PCL.

 
Overall Summary ROC Analysis of All Lesions
In the multivariate summary ROC analysis with all lesions combined, the meniscal tear versus cruciate ligament dummy variable (relative diagnostic OR, 0.52 [95% CI: 0.26, 1.01]), mean age (relative diagnostic OR, 1.13 [95% CI: 1.03, 1.23]), and magnetic field strength (relative diagnostic OR, 1.97 [95% CI: 1.13, 3.42]) were statistically significant predictors of diagnostic performance (Table 8). Although the dummy variable for meniscal versus cruciate ligament tears was of borderline significance, we kept it in the model because the nature of these lesions is totally different, and furthermore, it allowed comparison of the lesions. In Figure 4, summary ROC curves are shown for menisci and cruciate ligaments separately. The curves were plotted with the mean age covariable set at 30 years for both 0.2- and 1.5-T MR imagers. Compared with that for the menisci, the curve for cruciate ligaments is further up toward the upper left corner, suggesting better discriminatory power. Similarly, the curves for high-field-strength MR imaging show better diagnostic performance compared with those for low-field-strength MR imaging.



View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. Multivariate summary ROC curves based on the final regression model with all lesions combined. Curves are shown for both menisci and both cruciate ligaments combined. Curves were adjusted to patients aged 30 years and both 0.2- and 1.5-T MR imagers. The curves for cruciate ligaments are further up toward the upper left corner, as are the curves for high-field-strength MR imaging, suggesting better discriminatory power compared with that of the menisci and low-field-strength MR imaging.

 
Sensitivity Analyses
The jackknife sensitivity analyses, in which articles were excluded one by one from the final models, did not demonstrate any disproportionate influences of individual studies. In the models for ACL tears and the overall model with all lesions combined, the effect of magnetic field strength remained relatively stable in the sensitivity analysis (relative diagnostic OR ranges, 2.63–4.41 and 1.80–2.16, respectively), and this effect was always statistically significant.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
 
In this systematic review and meta-analysis, attention was focused on tears of the menisci and cruciate ligaments, which are primarily caused by traumatic mechanisms. Our results confirm those of previous studies, which show MR imaging to be a highly accurate diagnostic tool for detecting tears of the menisci and cruciate ligaments. The results show that diagnostic performance is better for cruciate ligament tears than that for meniscal tears. With regard to the menisci, our results demonstrate that the sensitivity and specificity differ significantly for the medial and lateral meniscus.

Whereas MR imaging is more sensitive in the diagnosis of medial meniscal tears, the specificity is higher for lateral meniscal tears. However, the natural logarithm of the diagnostic OR, which is an overall measure of diagnostic performance that incorporates both sensitivity and specificity, is not significantly different for the two lesions, indicating that radiologists probably use different points along the same underlying ROC curve when evaluating the two lesions. The results confirm that findings of the two menisci are preferably considered separately in studies in which the diagnostic performance of MR imaging was assessed to avoid performance underestimation. Thus, our criterion to include only articles in which results for medial and lateral menisci were reported separately seems to be justified. Moreover, in clinical practice, it is usually a lesion of either the medial or lateral meniscus that is suspected, making diagnostic performance statistics for both menisci combined less meaningful. To increase statistical precision, a pooled weighted analysis with both menisci combined into one model is possible, provided that an appropriate technique (eg, summary ROC analysis) for combining lesions with different points on the same ROC curve is used, which is what we did. We acknowledge, however, that this type of pooled combined analysis may result in data for some patients being included more than once in the same regression model, which induced some dependence among observations.

With regard to the cruciate ligaments, we considered complete tears only, because these injuries are by far more serious than are partial ruptures. Whereas complete ruptures mostly necessitate intensive physical therapy or reconstructive surgery, partial tears usually do not require specific treatment.

Apart from the menisci and cruciate ligaments, there are more derangements involving the knee that might be visualized with MR imaging. Articular cartilage lesions may also have a traumatic origin and cause serious symptoms. Except for full-thickness lesions (Ficat grade IV or V), however, the role of MR imaging in detecting articular cartilage defects has not been well established. Especially for lesions that are limited to half of the cartilage thickness (Ficat grade I–III), the diagnostic accuracy of MR imaging seems poor (17–19). We performed a MEDLINE search of articles on MR imaging of knee cartilage, but among the articles that we found, we considered the patient population, grading system, definition of disease, and regions studied as too heterogeneous to justify a meta-analysis. For example, patient populations varied from patients who experienced trauma to patients with known osteoarthritis. Grading systems had a range of three to six categories, and positivity criteria varied from anything abnormal to full-thickness defects. Finally, in some studies, the knee cartilage was subdivided into regions, such as the lateral and medial femoral condyle or even the different facets of the patella, whereas in other articles, investigators considered the whole knee to be a unit of analysis.

To avoid missing important articles, we applied broad criteria for our MEDLINE search. This resulted in 804 retrieved references, most of which were clearly not suitable for inclusion in our systematic review and were excluded on the basis of either the title or the abstract. As an illustration, a substantial number of articles about other pathologic conditions in the knee or about postoperative MR imaging, cadaveric examinations, case reports, and even the meniscus in the temporomandibular joint were found among these 804 studies. After this initial selection, 120 articles were considered potentially eligible for inclusion, 20 of which were subsequently excluded because they did not present data on diagnostic performance. We applied criteria for inclusion and exclusion that are commonly used in evidence-based medicine, as well as criteria that were related to the aims of the present study and the methods that we intended to use. Studies with a sample size of fewer than 30 patients were excluded because small samples contribute little to the results of a meta-analysis. In addition, our purpose to assess the effect of magnetic field strength on diagnostic performance required that the magnetic field strength of the MR imager be reported in all included studies. Similarly, a summary ROC analysis is only possible if the absolute numbers of true-positive, false-negative, true-negative, and false-positive results are available or derivable.

Our literature search was limited to articles published between 1991 and 2000. We acknowledge that there have been relevant publications between 1985 and 1990, some of which would have fulfilled the inclusion criteria of this meta-analysis. However, the question arises whether the MR imaging techniques used in the earlier period can be compared with the improved techniques and sequences of more recent years.

Furthermore, we limited our literature search to articles published in the English language. It has been shown that inclusion of only English-language articles does not influence the results of a meta-analysis (20). Moreover, the decision as to which other languages to include will, in our view, always be based on highly arbitrary and geographically dependent criteria, namely, the ability of the data extractors to understand other languages.

We included only studies in which arthroscopy was regarded as the standard of reference. This procedure has always been the reference standard for the diagnosis of internal derangements of the knee, against which alternative diagnostic modalities should be compared. However, the use of arthroscopy alone as the reference standard has been criticized because some parts of the joint cannot be brought into view properly. The posterior horn of the medial meniscus is an especially difficult area to visualize, and the arthroscopic diagnosis of meniscal tears in this region is often assigned on the basis of probing rather than visualizing the meniscus. Quinn and Brown (21) retrospectively analyzed the arthroscopic videotapes of false-positive MR imaging results and found that the suspected area of the meniscus was never visualized in these cases. Therefore, false-negative findings at arthroscopy could potentially account for many false-positive MR imaging results. Likewise, the PCL is not usually visualized during arthroscopy if the ACL is intact, and in this case, physical examination is often performed with the patient anesthetized to demonstrate a rupture of the PCL. As a result, arthroscopy is ideally performed with knowledge of the findings from the preceding MR examination.

We first summarized and combined the data in a pooled weighted analysis performed separately for each lesion. This type of analysis has certain limitations, since it does not take into account the differences in positivity criteria that were used in the studies. For the lesions that we included in this meta-analysis, however, we found that articles were highly comparable with regard to the definition of disease at MR imaging. In all included articles, the meniscus was evaluated and graded by using the original or slightly modified criteria that were introduced previously by Reicher et al (22) or Crues et al (23). On the basis of these criteria, a meniscal tear is diagnosed if an intrameniscal area of high signal intensity extends to the articular surface, whereas an area of high signal intensity that does not reach the surface is considered degenerative. Since degenerative changes of the menisci are common and usually asymptomatic findings in patients after the 3rd decade (24), our interest was focused on meniscal tears only. With regard to the status of the cruciate ligaments, a complete tear was usually defined as nonvisualization of the ligament in its expected position or high signal intensity in the course of the ligament in the absence of intact margins. Even though positivity criteria were largely the same among studies, we considered a summary ROC analysis appropriate, since, in our view, there will always be implicit variations in interpretation in different settings and by different radiologists.

The results of our meta-analysis stress the importance of adequate study design when conducting a study on diagnostic performance. Especially, the presence of verification bias was found to dramatically increase the diagnostic performance for PCL tears. Also, blinding the performer of the reference test (arthroscopy) to the findings of the test under evaluation (MR imaging) was predictive in the final model for medial meniscal tears. In every included study, the radiologist who interpreted the MR images was blinded to the arthroscopy result. This reflects clinical practice, in which MR imaging is usually and ideally performed prior to arthroscopy. Therefore, the effect of blinding the MR image interpreter to the result of the reference test was not assessed in this meta-analysis. In addition, mean age of the patient population was a statistically significant predictor for various lesions. This finding implies that the age distribution of the patient population should be described sufficiently in diagnostic performance studies.

The effect of publication year on the diagnostic performance probably reflects the progress in MR imaging technique over the past 15 years with regard to the MR imagers and the sequences used. Whereas in the early years of MR imaging, only spin-echo and gradient-echo sequences were performed, nowadays the choice of sequences that can be incorporated in the imaging protocol has expanded, with various types of fat-suppression techniques and three-dimensional acquisitions that enable reconstructions in every plane. Moreover, the influence of publication year could represent a learning process over the years among interpreters of MR images of the knee.

Higher magnetic field strength of the MR imager improved the diagnostic performance for ACL tears, and a modest effect was demonstrated when all lesions were combined in one model. In the pooled weighted analyses per lesion for different magnetic field strength categories, a trend was observed toward better diagnostic performance for higher magnetic field strengths, but this effect was far from significant, probably because too few studies were available per category of magnetic field strength. We can only speculate whether the effect would have been statistically significant if more existing studies would have fulfilled our inclusion criteria. Except for the ACL, however, no effect was demonstrated in the summary ROC analysis per lesion, which is a more powerful technique for assessing predictors of diagnostic performance. Only for PCL tears was the number of sequences a predictor in the final multivariate model. Unexpectedly, this effect was negative with increasing numbers of sequences. The status of the PCL can usually be assessed well on the basis of a few images in the sagittal plane.

We conclude that MR imaging is highly accurate in the diagnosis of tears of the menisci and cruciate ligaments. MR imaging is an appropriate screening tool for therapeutic arthroscopy, making diagnostic arthroscopy unnecessary in most patients. Diagnostic performance of MR imaging differs significantly for menisci and cruciate ligaments and for the medial and lateral meniscus and is influenced by characteristics of study design, patient age, and magnetic field strength. The effect of magnetic field strength, however, was only statistically significant for ACL tears. The influence of study design characteristics and patient age should be taken into consideration whenever a diagnostic performance study on MR imaging of the knee is designed or reported.


   FOOTNOTES
 
Abbreviations: ACL = anterior cruciate ligament, OR = odds ratio, PCL = posterior cruciate ligament, ROC = receiver operating characteristic

Author contributions: Guarantor of integrity of entire study, M.G.M.H.; study concepts and design, all authors; literature research, E.H.G.O.; data acquisition, E.H.G.O., A.C.M.V.; data analysis/interpretation, all authors; statistical analysis, E.H.G.O.; manuscript preparation, E.H.G.O.; manuscript definition of intellectual content, all authors; manuscript editing, E.H.G.O.; manuscript revision/review and final version approval, all authors.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
DISCUSSION
 REFERENCES
 

  1. Rangger C, Klestil T, Kathrein A, Inderster A, Hamid L. Influence of magnetic resonance imaging on indications for arthroscopy of the knee. Clin Orthop 1996; 330:133-142.
  2. Rappeport ED, Mehta S, Wieslander SB, Lausten GS, Thomsen HS. MR imaging before arthroscopy in knee joint disorders? Acta Radiol 1996; 37:602-609.[Medline]
  3. Mackenzie R, Palmer CR, Lomas DJ, Dixon AK. Magnetic resonance imaging of the knee: diagnostic performance studies. Clin Radiol 1996; 51:251-257.[CrossRef][Medline]
  4. Irwig L, Tosteson AN, Gatsonis C, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994; 120:667-676.[Abstract/Free Full Text]
  5. Berlin JA. Does blinding of readers affect the results of meta-analyses?University of Pennsylvania Meta-analysis Blinding Study Group. Lancet 1997; 350:185-186.[Medline]
  6. Begg CB, Greenes RA. Assessment of diagnostic tests when disease verification is subject to selection bias. Biometrics 1983; 39:207-215.[CrossRef][Medline]
  7. Littenberg B, Moses LE. Estimating diagnostic accuracy from multiple conflicting reports: a new meta-analytic method. Med Decis Making 1993; 13:313-321.
  8. Begg CB, Berlin JA. Publication bias and dissemination of clinical research. J Natl Cancer Inst 1989; 81:107-115.[Abstract/Free Full Text]
  9. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315:629-634.[Abstract/Free Full Text]
  10. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177-188.[CrossRef][Medline]
  11. Fleiss JL. The statistical basis of meta-analysis. Stat Methods Med Res 1993; 2:121-145.[Medline]
  12. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12:1293-1316.[Medline]
  13. De Vries SO, Hunink MG, Polak JF. Summary receiver operating characteristic curves as a technique for meta-analysis of the diagnostic performance of duplex ultrasonography in peripheral arterial disease. Acad Radiol 1996; 3:361-369.[CrossRef][Medline]
  14. Irwig L, Macaskill P, Glasziou P, Fahey M. Meta-analytic methods for diagnostic test accuracy. J Clin Epidemiol 1995; 48:119-130.[CrossRef][Medline]
  15. Vahey TN, Broome DR, Kayes KJ, Shelbourne KD. Acute and chronic tears of the anterior cruciate ligament: differential features at MR imaging. Radiology 1991; 181:251-253.[Abstract/Free Full Text]
  16. Warwick DJ, Cavanagh P, Bell M, Marsh CH. Influence of magnetic resonance imaging on a knee arthroscopy waiting list. Injury 1993; 24:380-382.[CrossRef][Medline]
  17. Munk B, Madsen F, Lundorf E, et al. Clinical magnetic resonance imaging and arthroscopic findings in knees: a comparative prospective study of meniscus anterior cruciate ligament and cartilage lesions. Arthroscopy 1998; 14:171-175.[Medline]
  18. Kawahara Y, Uetani M, Nakahara N, et al. Fast spin-echo MR of the articular cartilage in the osteoarthrotic knee: correlation of MR and arthroscopic findings. Acta Radiol 1998; 39:120-125.[Medline]
  19. Mori R, Ochi M, Sakai Y, Adachi N, Uchio Y. Clinical significance of magnetic resonance imaging (MRI) for focal chondral lesions. Magn Reson Imaging 1999; 17:1135-1140.[CrossRef][Medline]
  20. Moher D, Pham B, Klassen TP, et al. What contributions do languages other than English make on the results of meta-analyses? J Clin Epidemiol 2000; 53:964-972.[CrossRef][Medline]
  21. Quinn SF, Brown TF. Meniscal tears diagnosed with MR imaging versus arthroscopy: how reliable a standard is arthroscopy? Radiology 1991; 181:843-847.[Abstract/Free Full Text]
  22. Reicher MA, Hartzman S, Duckwiler GR, Bassett LW, Anderson LJ, Gold RH. Meniscal injuries: detection using MR imaging. Radiology 1986; 159:753-757.[Abstract/Free Full Text]
  23. Crues JV, III, Mink J, Levy TL, Lotysch M, Stoller DW. Meniscal tears of the knee: accuracy of MR imaging. Radiology 1987; 164:445-448.[Abstract/Free Full Text]
  24. Stoller DW, Martin C, Crues JV, III, Kaplan L, Mink JH. Meniscal tears: pathologic correlation with MR imaging. Radiology 1987; 163:731-735.[Abstract/Free Full Text]
  25. Boeree NR, Watkinson AF, Ackroyd CE, Johnson C. Magnetic resonance imaging of meniscal and cruciate injuries of the knee. J Bone Joint Surg Br 1991; 73:452-457.
  26. Fischer SP, Fox JM, Del Pizzo W, Friedman MJ, Snyder SJ, Ferkel RD. Accuracy of diagnoses from magnetic resonance imaging of the knee: a multi-center analysis of one thousand and fourteen patients. J Bone Joint Surg Am 1991; 73:2-10.[Abstract/Free Full Text]
  27. Kelly MA, Flock TJ, Kimmel JA, et al. MR imaging of the knee: clarification of its role. Arthroscopy 1991; 7:78-85.[Medline]
  28. Niitsu M, Anno I, Fukubayashi T, Shimojo H, Kuno S, Akisada M. Tears of cruciate ligaments and menisci: evaluation with cine MR imaging. Radiology 1991; 178:859-864.[Abstract/Free Full Text]
  29. Araki Y, Ootani F, Tsukaguchi I, et al. MR diagnosis of meniscal tears of the knee: value of axial three-dimensional Fourier transformation GRASS images. AJR Am J Roentgenol 1992; 158:587-590.[Abstract/Free Full Text]
  30. Gluckert K, Kladny B, Blank-Schal A, Hofmann G. MRI of the knee joint with a 3-D gradient echo sequence: equivalent to diagnostic arthroscopy? Arch Orthop Trauma Surg 1992; 112:5-14.
  31. Grevitt MP, Pool CJ, Bodley RN, Savage PE. Magnetic resonance imaging of the knee: initial experience in a district general hospital. Injury 1992; 23:410-412.[CrossRef][Medline]
  32. Gross ML, Grover JS, Bassett LW, Seeger LL, Finerman GA. Magnetic resonance imaging of the posterior cruciate ligament: clinical use to improve diagnostic accuracy. Am J Sports Med 1992; 20:732-737.[Abstract/Free Full Text]
  33. Heron CW, Calvert PT. Three-dimensional gradient-echo MR imaging of the knee: comparison with arthroscopy in 100 patients. Radiology 1992; 183:839-844.[Abstract/Free Full Text]
  34. Barnett MJ. MR diagnosis of internal derangements of the knee: effect of field strength on efficacy. AJR Am J Roentgenol 1993; 161:115-118.[Abstract/Free Full Text]
  35. De Smet AA, Norris MA, Yandow DR, Quintana FA, Graf BK, Keene JS. MR diagnosis of meniscal tears of the knee: importance of high signal in the meniscus that extends to the surface. AJR Am J Roentgenol 1993; 161:101-107.[Abstract/Free Full Text]
  36. Chan WP, Peterfy C, Fritz RC, Genant HK. MR diagnosis of complete tears of the anterior cruciate ligament of the knee: importance of anterior subluxation of the tibia. AJR Am J Roentgenol 1994; 162:355-360.[Abstract/Free Full Text]
  37. Gentili A, Seeger LL, Yao L, Do HM. Anterior cruciate ligament tear: indirect signs at MR imaging. Radiology 1994; 193:835-840.[Abstract/Free Full Text]
  38. Kinnunen J, Bondestam S, Kivioja A, et al. Diagnostic performance of low field MRI in acute knee injuries. Magn Reson Imaging 1994; 12:1155-1160.[CrossRef][Medline]
  39. LaPrade RF, Burnett QM, Veenstra MA, Hodgman CG. The prevalence of abnormal magnetic resonance imaging findings in asymptomatic knees: with correlation of magnetic resonance imaging to arthroscopic findings in symptomatic knees. Am J Sports Med 1994; 22:739-745.[Abstract/Free Full Text]
  40. Chen MC, Shih TT, Jiang CC, Su CT, Huang KM. MRI of meniscus and cruciate ligament tears correlated with arthroscopy. J Formos Med Assoc 1995; 94:605-611.[Medline]
  41. Justice WW, Quinn SF. Error patterns in the MR imaging evaluation of menisci of the knee. Radiology 1995; 196:617-621.[Abstract/Free Full Text]
  42. Lerman JE, Gray DS, Schweitzer ME, Bartolozzi A. MR evaluation of the anterior cruciate ligament: value of axial images. J Comput Assist Tomogr 1995; 19:604-607.[Medline]
  43. Barry KP, Mesgarzadeh M, Triolo J, Moyer R, Tehranzadeh J, Bonakdarpour A. Accuracy of MRI patterns in evaluating anterior cruciate ligament tears. Skeletal Radiol 1996; 25:365-370.[CrossRef][Medline]
  44. Lundberg M, Odensten M, Thuomas KA, Messner K. The diagnostic validity of magnetic resonance imaging in acute knee injuries with hemarthrosis: a single-blinded evaluation in 69 patients using high-field MRI before arthroscopy. Int J Sports Med 1996; 17:218-222.[Medline]
  45. Riel KA, Reinisch M, Kersting-Sommerhoff B, Hof N, Merl T. 0.2-T magnetic resonance imaging of internal lesions of the knee joint: a prospective arthroscopically controlled clinical study. Knee Surg Sports Traumatol Arthrosc 1999; 7:37-41.[CrossRef][Medline]
  46. Bui-Mansfield LT, Youngberg RA, Warme W, Pitcher JD, Nguyen PL. Potential cost savings of MR imaging obtained before arthroscopy of the knee: evaluation of 50 consecutive patients. AJR Am J Roentgenol 1997; 168:913-918.[Abstract/Free Full Text]
  47. Cheung LP, Li KC, Hollett MD, Bergman AG, Herfkens RJ. Meniscal tears of the knee: accuracy of detection with fast spin-echo MR imaging and arthroscopic correlation in 293 patients. Radiology 1997; 203:508-512.[Abstract/Free Full Text]
  48. Franklin PD, Lemon RA, Barden HS. Accuracy of imaging the menisci on an in-office, dedicated, magnetic resonance imaging extremity system. Am J Sports Med 1997; 25:382-388.[Abstract/Free Full Text]
  49. Rappeport ED, Wieslander SB, Stephensen S, Lausten GS, Thomsen HS. MRI preferable to diagnostic arthroscopy in knee joint injuries: a double-blind comparison of 47 patients. Acta Orthop Scand 1997; 68:277-281.[Medline]
  50. Weinstabl R, Muellner T, Vecsei V, Kainberger F, Kramer M. Economic considerations for the diagnosis and therapy of meniscal lesions: can magnetic resonance imaging help reduce the expense? World J Surg 1997; 21:363-368.[CrossRef][Medline]
  51. Ha TP, Li KC, Beaulieu CF, et al. Anterior cruciate ligament injury: fast spin-echo MR imaging with arthroscopic correlation in 217 examinations. AJR Am J Roentgenol 1998; 170:1215-1219.[Abstract/Free Full Text]
  52. Cotten A, Delfaut E, Demondion X, et al. MR imaging of the knee at 0.2 and 1.5 T: correlation with surgery. AJR Am J Roentgenol 2000; 174:1093-1097.[Abstract/Free Full Text]
  53. Elvenes J, Jerome CP, Reikeras O, Johansen O. Magnetic resonance imaging as a screening procedure to avoid arthroscopy for meniscal tears. Arch Orthop Trauma Surg 2000; 120:14-16.



This article has been cited by other articles:


Home page
 Med Decis MakingHome page
L.R. Arends, T.H. Hamza, J.C. van Houwelingen, M.H. Heijenbrok-Kal, M.G.M. Hunink, and T. Stijnen
Bivariate Random Effects Meta-Analysis of ROC Curves
Med Decis Making, September 1, 2008; 28(5): 621 - 638.
[Abstract] [PDF]

Home page
 Med Decis MakingHome page
T. H. Hamza, J. B. Reitsma, and T. Stijnen
Meta-Analysis of Diagnostic Studies: A Comparison of Random Intercept, Normal-Normal, and Binomial-Normal Bivariate Summary ROC Approaches
Med Decis Making, September 1, 2008; 28(5): 639 - 649.
[Abstract] [PDF]

Home page
 Am. J. Roentgenol.Home page
A. A. De Smet, D. H. Nathan, B. K. Graf, B. A. Haaland, and J. P. Fine
Clinical and MRI Findings Associated with False-Positive Knee MR Diagnoses of Medial Meniscal Tears
Am. J. Roentgenol., July 1, 2008; 191(1): 93 - 99.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
S. R. Duc, C. W. A. Pfirrmann, P. P. Koch, M. Zanetti, and J. Hodler
Internal Knee Derangement Assessed with 3-minute Three-dimensional Isovoxel True FISP MR Sequence: Preliminary Study
Radiology, February 1, 2008; 246(2): 526 - 535.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
A. A. De Smet and R. Mukherjee
Clinical, MRI, and Arthroscopic Findings Associated with Failure to Diagnose a Lateral Meniscal Tear on Knee MRI
Am. J. Roentgenol., January 1, 2008; 190(1): 22 - 26.
[Abstract] [Full Text] [PDF]

Home page
 Br Med BullHome page
R. Crawford, G. Walley, S. Bridgman, and N. Maffulli
Magnetic resonance imaging versus arthroscopy in the diagnosis of knee pathology, concentrating on meniscal lesions and ACL tears: a systematic review
Br. Med. Bull., December 1, 2007; 84(1): 5 - 23.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page
P. W. J. Vincken, A. P. M. ter Braak, A. R. van Erkel, E. G. Coerkamp, T. P. W. de Rooy, S. de Lange, W. M. C. Mallens, L. N. J. E. M. Coene, R. M. Bloem, P. A. van Luijt, et al.
MR Imaging: Effectiveness and Costs at Triage of Patients with Nonacute Knee Symptoms
Radiology, December 1, 2006; 242(1): 85 - 93.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Roentgenol.Home page
J. E. Roos, B. Chilla, M. Zanetti, M. Schmid, P. Koch, C. W. A. Pfirrmann, and J. Hodler
MRI of Meniscal Lesions: Soft-Copy (PACS) and Hard-Copy Evaluation Versus Reviewer Experience.
Am. J. Roentgenol., March 1, 2006; 186(3): 786 - 790.
[Abstract] [Full Text] [PDF]

Home page
 RadiologyHome page