Statistical Methods of the Diagnostic Section (I–III)

Arthroscopic findings served as the gold standard when the diagnostic validity of MRI was analysed. Sensitivity, specificity, diagnostic accuracy, as well as positive and negative predictive values, were calculated with single table analysis. When the results for the different grades of chondral lesions were analysed, only sensitivity and accuracy were calculated because the number of true-negative and false-positive results in this setting is always 0. (Specificity = True-negative results/[False-negative results + False-positive results]). For the same reason, diagnostic accuracy was identical with the sensitivity (Sensitivity = True-positives/[True Positives + False-negatives] and accuracy = [True-positives + True Negatives]/[True Positives + True negatives + [True-positives + False-negatives]). The Fisher’s exact test was used in two-way tables and 95% CI was analysed with the Wilson score method (Agresti & Coull, 1998).

8.6.1 Data of Fresh Chondral Lesions (I)

To analyse fresh traumatic chondral lesions found at MRI and arthroscopy, the data were classified according to the numeric grading system developed by Tyrrell et al (Tyrrell, et al., 1988) (Table 5). This system is similar to ICRS classification in that chondral lesions are graded according to their depth. Articular surfaces of the patella, femoral sulcus, medial and lateral femoral condyles, and medial and lateral tibial plateaus were included

in the analysis. Possible bone bruises seen in MRI were also taken into account as they may indicate damage to the overlying articular cartilage (Johnson, et al., 1998).

These data were collected in connection with normal clinical practice, and thus the orthopaedic surgeon received the original MRI reports before the arthroscopy. The exact grading of the chondral lesions was not included in the data. For the purposes of this study, the MRI data were re-evaluated by a musculoskeletal radiologist blinded to the prior results of arthroscopy or MRI. Sensitivity, specificity, and diagnostic accuracy of MRI for fresh traumatic chondral lesions were calculated. Bone bruises in MRI were primarily interpreted as a negative result for chondral lesions. As the underlying cartilage may be injured, however, the sensitivity, specificity, and diagnostic accuracy were also calculated to allow for such an interpretation.

Table 5. Depth of articular cartilage lesions according to grading scale developed by Tyrrell et al. (Tyrrell, et al., 1988).

Grade Description

0 Normal

1 Moderate irregularity

2 Severe irregularity but not full thickness 3 Full-thickness loss

Reprinted from original publication I (Kuikka, et al., 2006) with permission from Elsevier.

8.6.2 Data of Meniscal Tears (II)

Generally used MRI criteria for meniscal tears are an internal tear line extending to the upper, lower, or apical surface of the meniscus, or a tear line splitting the meniscus into two or more pieces and showing partial dislocation of the meniscus (Crues, et al., 1987).

These criteria were also used for samples with fresh and old meniscal tears. Due to the retrospective nature of the study, the arthroscopist was aware of the MRI report before surgery. The original MRI findings were reviewed and compared with the arthroscopic findings. Lateral and medial menisci of all the knees were evaluated and interpreted independently. The sensitivity, specificity, and diagnostic accuracy of MRI for both acute and chronic meniscal tears were calculated. To calculate the interobserver correlations and intraobserver agreement, the original MRI and arthroscopy records were reviewed twice by another radiologist with musculoskeletal training who was blinded to previous MRI and arthroscopy findings. Agreement was almost perfect when the kappa-value was 0.81 to 1.00, substantial at 0.61 to 0.80, moderate at 0.41 to 0.60, fair at 0.21 to 0.40, and poor at a kappa-value of less than 0.20 (Landis & Koch, 1977).

8.6.3 Data of Anterior Knee Pain and Patellar Chondral Lesions (III) The MRI and arthroscopic findings of patellar chondral lesions in a prospective sample of nontraumatic AKP patients were recorded as showing either normal cartilage thickness or cartilage lesions of different depths or severity. Grading was based on the arthroscopic grading system by Shahriaree (Shahriaree, 1985), with 0 indicating normal; I, softening or a blister; II, fissuring; III, fragmentation and fissuring; and IV, full-thickness fissuring and exposed bone (Table 6). This grading system is a modification of the Outerbridge classification in which grade I represents softening and swelling of the cartilage, grade II/

III represent fissure and fragmentation in an area less/more than 1.3 cm in diameter and grade IV represents erosion of cartilage down to bone (Outerbridge, 1961).

The images were initially evaluated according to a standard protocol for interpreting MRI findings by a musculoskeletal radiologist on duty. In addition, another musculoskeletal radiologist who was blinded to the previous MRI and arthroscopy findings then re-evaluated the images. These re-evaluations differed from the original interpretation in three cases, and thus a final decision was reached by consensus.

Arthroscopy was performed by an experienced orthopaedic surgeon who evaluated the articular surfaces being blinded to the MRI findings.

When calculating the results for grade-I patellar chondral lesions, only arthroscopically detected grade-I lesions were taken into account as positive results (the gold standard) for patellar chondral lesions, and they were compared with the MRI results (with grade 0 on the images considered negative and grades I through IV considered positive). The results for grades II, III, and IV were calculated with arthroscopically detected grade-II, III, and IV lesions considered positive and compared with the MRI results (with grade 0 on the images considered negative and grades I through IV considered positive). Other internal derangements of the knee revealed by arthroscopy were also recorded and taken into account.

Possible PFM was evaluated on patellofemoral radiographs by measuring the lateral patellofemoral angle by drawing a line connecting the anterior aspects of the femoral condyles and a second line along the lateral facet of the patella (Figure 1) (Laurin, et al., 1978). The sulcus angle was measured between the lines extending from the deepest point of the intercondylar sulcus, both medially and laterally, to the tops of the femoral condyles. A sulcus angle >138˚ was considered to be shallow (Merchant, et al., 1974).

A line bisecting the sulcus angle was then compared with a line drawn from the apex of the sulcus angle through the lowest point of the articular ridge of the patella (the congruence angle) to detect possible patellar lateralization.

Table 6. Classification of patellar chondral lesions (CMP) according to the system described by Shahriaree (Shahriaree, 1985).

Grade

I Fibrillation II Fissure formation III Fragmentation

IV Crater formation and eburnation

Slightly reformatted from Table I in original publication III (Pihlajamäki, et al., 2010).