2.1 Search strategy

In October 2023, A literature search to identify all studies evaluating the accuracy and reliability of plain radiograph, magnetic resonance imaging (MRI) and computed tomography (CT) in pre-operative meniscal sizing for meniscal allograft transplantation was conducted (Fig. 1). The search was performed in the following electronic databases: PubMed, Embase, Science direct and Web of Science. The keywords used for this search included “meniscus”,” sizing”,” measurement”, “comparison”, “magnetic resonance imaging”, “computed tomography”, “radiograph”, and “computed tomography” in various combinations using the Boolean operator “and”. Synonyms were adjusted for using the Boolean operator “or” for example ((“meniscus” OR “meniscal” OR “menisci”). The initial search yielded 184 articles. Citation tracking was then performed which yielded an additional 12 articles. These articles were screened by one author to exclude duplicate articles and those that were not relevant to the subject, resulting in 59 articles.

Close

Fig. 1 PRISMA flowchart showing the search strategy.

Export to PPT

2.2 Inclusion and exclusion

The inclusion criteria were the following: Studies that compared two or more imaging methods for meniscal sizing for allograft transplantation, Studies that provided quantitative data or measurements on meniscal dimensions obtained through specified imaging techniques, Studies that report on the accuracy, reliability or clinical outcomes of the imaging methods for meniscal sizing and Studies that involve in vivo or ex vivo data. The studies were limited to primary research articles published in peer-reviewed journals (level 1 to 4) and published in English.

The exclusion criteria were as follows: Animal studies, non-English language publications and studies without quantitative data. Fourteen studies were deemed suitable for inclusion and they are shown in Table 1.

The methodological quality and risk of bias of the included studies were assessed using the critical appraisal skills programme for cohort studies checklist²⁵ and no significant bias was demonstrated in the included studies. The findings of this systematic review are reported using PRISMA (preferred reporting items for systematic reviews and meta-analysis) guidelines.²⁶

3.1 Comparing radiographic methods to anatomic measurement

True, ex-vivo measurement of meniscus height, length and width (from either cadaveric knees or tibial plateau samples after total knee arthroplasty) was taken to be gold standard.

Radiographs are the most common method used for meniscal sizing – there are two methods employed. The first is the Pollard method²¹ which uses a surrogate measure of the tibial plateau on radiograph allowing for the proportion of the plateau that the meniscus occupies. On the original paper for the method, the average errors compared with direct measurement using 21 cadaveric knees were: Medial (Width 7.9 %, Length 7.4 %), Lateral (Width 8.4 %, Length 8.0 %) although the full data were not published.

Five other studies compared the Pollard method with true anatomic ex-vivo measurements and the pooled published data is summarised in Table 2. Three compared all four dimensions (both menisci, width and length)^23,27,28 with anatomic direct measurement. Two^22,29 compared lateral meniscus measurements only.

Interestingly, the studies differed widely as to the accuracy of the Pollard method compared to true measurement despite the relatively simple method. This is thought to be due to a combination of factors including beam angle, minor internal or external rotation and magnification factors increasing errors or making it difficult to identify bony landmarks. This may explain why different institutions with different hardware and radiographic acquisition protocols have obtained relatively different results. Prodromos et al. demonstrated a very poor concordance with medial meniscal width and length (width 23.1 %, length 15.1 %) whereas the other two studies and the original study all showed relatively good concordance with anatomical measurements of between 6.5 and 9.6 % of the true anatomical measurement.

For the lateral meniscus, nearly all studies showed similar concordance of between 4.0 and 13.3 % for width and length except the study by Ambra et al. which showed a low concordance of lateral meniscal width (20.3 % error) and was an outlier between the 5 studies.

There was no clear trend for under or overestimating measurements with 61 out of 148 (41 %) underestimating true dimensions across the four studies.

In 2011, Yoon et al.²² proposed a different method of measuring the lateral meniscus in which the beam is tilted caudally 10° and great care is taken to minimise rotation. In the index study the authors provided average error figures of 1.4 mm (average 4.8 %) for lateral meniscal width and 1.4 mm (average 4.3 %) for lateral meniscal length compared to 12.7 % for the Pollard method. The other study (Ambra et al.) looking at the Yoon method found 9.9 % error for the lateral meniscal length of a similar order to the Pollard method.

In conclusion, the radiographic methods seem to vary significantly due to unclear factors probably including acquisition method and patient population. The average mean error for the Pollard method was between 7.4 and 12.6 % but with wide standard deviation. Overall across the 4 studies with raw data presented, 56 out of 148 meniscal measurements (38 %) fell outside the <10 % benchmark with the Pollard method and 5 out of 35 measurements (14 %) with the Yoon method.

3.2 Comparing MRI to anatomic measurement

Four studies directly compared MRI meniscal measurements with anatomic ex-vivo measurements of cadaveric knees^23,28–30 and the pooled published data is summarised in Table 3. However, the MRI imaging technique and the measurement technique differed between each study making the results difficult to synthesise. Prodromos et al. used a saline arthrography technique. 2D (4 mm thickness) and 3D spin echo images were measured from coronal and sagittal planes. Shaffer used 3D gradient echo images but measured from a reconstructed true axial plane, Ambra used 2D spin echo 4 mm thickness (2 mm thickness, 2 mm gap) but measured on axial plane and Haut used 4 mm 2D spin echo images (3 mm thickness, 1 mm gap) and measured on sagittal and coronal planes and the comparison in this study was to the contralateral, not ipsilateral knee.

The first three studies presented the absolute measurement data allowing direct comparison between the studies but Haut et al. only provided R² correlation coefficients.

The data showed near identical results for lateral meniscal measurements across all three studies with an average relative difference of 3.5–5.1 % for width and 5.5–6.8 % for length (corresponding to R² values of 0.83–0.9 and 0.63–0.70 respectively). This gives a total average relative difference of 4.3 % (R² = 0.88) for width and 6.3 % (R² = 0.66) for length. The Shaffer and Ambra studies showed a definite tendency for undermeasurement on MRI (27 out of 44 measurements with 13 out of 44 overmeasurements) not shown on the Prodromos study.

However, there was disagreement between the two studies looking at medial meniscal measurements with Prodromos et al. showing good concordance with anatomic measurements (Average relative difference (R²): Width 4.3 % (0.95), Length 3.6 % (0.93)) and Shaffer et al. showing only moderate concordance (Average relative difference (R²): Width 8.1 % (0.66), Length 8.6 % (0.43)). This gives a total average relative difference of 6.4 % (R² = 0.95) for width and 6.3 % (R² = 0.78) for length. Both the Prodromos and Shaffer data showed tendency for overmeasurement for the medial meniscus in both length and width (34 out of 44 measurements compared with 9 undermeasurements).

The Haut et al. study was conducted differently and looked at concordance of measurements between the cadaveric meniscus and the contralateral knee MRI. This study showed medial meniscal measurements equivalent to Width R² = 0.68, Length R² = 0.82 and lateral meniscal measurements equivalent to Width R² = 0.76, Length R² = 0.54. The fact that the contralateral knee was used would explain these lower correlation coefficients.

In conclusion, there is good correlation between MRI and true measurements of lateral meniscus (poorer for length). On average MRI is more accurate than radiographic methods for medial meniscus measurements and the differences may be due to the method of acquisition or measurement on MRI since the two more similar datasets (Ambra and Shaffer) both utilise axial measurements whereas the other study utilises sagittal and coronal plane measurements. Of the 108 knees compared using MRI providing raw data, 20 measurements fell outside the 10 % threshold (18.5 %)

3.3 Comparing CT to anatomic measurement

Only one study directly compared CT and cadaveric measurements³¹ although a previous short report³² did compare CT and other modalities to anatomic cadaveric measurements, but as this was a short report the data were not given in a way that was easily comparable to other studies. In the McConkey study, anatomic measurements were compared with XR (Pollard method) and CT to measure the dimensions of the tibial plateau as a substitute measure for the meniscus (as other authors had found in previous reports³³ that the two measurements are closely related). This report found good concordance of CT with anatomical measurements with errors between 3.8 % and 6.5 % of the true bony margins. However, if the further errors quoted by McDermott et al. in converting the plateau measurement to a meniscus measurement are taken into account, this increases the error in the CT method to between 9.0 and 11.3 %. It is not possible to tell how many measurements fell outside the 10 % threshold as this benchmark was not utilised in this study set.

Overall this one study shows CT without arthography is likely to have an accuracy of between MRI and XR. No studies to data have compared CT arthrography with anatomic benchmark.

3.4 Comparing left and right knee meniscus measurements

Symmetry of meniscal dimensions is important, as MRI measurements of meniscal size rely on measurement of the contralateral, normal knee (due to distortion and/or partial resection of the meniscus in the recipient compartment). Five studies^34–38 compared MRI measurements and three studies^23,28,33 compared direct measurements of left and right paired knees. In the largest study on cadaveric knees (Prodromos et al., n = 428), of the lateral compartments considered, 98.0 % of measurements were within 3 mm of the anatomic measurements. In the medial compartments, 96.8 % of measurements were within 3 mm, but there were significant outliers with some up to 8 or 10 mm different. Using the mean of the measured menisci from the second arm of the same study these would correspond to between 3.1 and 3.3 % relative difference between sides.

McDermott et al. measured 22 pairs of knees and found differences between sides of between 4.5 and 9.8 % for different dimensions.

Shaffer et al. also did left and right comparisons of their 10 knees (which were paired) and found wide differences between sides with r values between 0.75 and 0.79 for most measurements, but an outlier r value of 0.17 for medial meniscal length.

Kaleka et al. measured 22 pairs of knees on MRI and found r values of >0.98 for all measurements of meniscal width and length between knees in the same patient. Netto et al. did a similar study on 25 pairs in 2018 which also showed similar interclass coefficients of between 0.84 and 0.93.

In a 2014 study by Yoon et al., 60 pairs of knees were analysed by MRI showed no significant difference between the meniscal length and width measurements between pairs of knees with average difference from the mean being 1.5 % and 0.6 % for medial meniscus width and length and 2.7 % and 4.4 % for the lateral meniscal width and length respectively and similar data was shown by Hamdan in 2020.

A different study utilising 3D modelling of 40 pairs of knees (Beeler, 2020) and 2D and 3D measurements showed good concordance of 3D coordinate points between the two knees and of conventional 2-dimensional measurements with average differences from the mean of 6.8 % and 3.4 % for medial meniscus and 5.3 % and 4.6 % for lateral meniscus width and length respectively.

Overall, the current evidence favours that knees are relatively symmetrical although there are a few outliers. However, one has to consider that this 0.6–6.8 % error needs to be combined with the of 4.3–6.6 % error of MRI compared with cadaver. In the worst case scenario using Beeler et al.’s figures, this would increase the error between MRI and recipient true measurement to between 9.6 and 13.2 %. This is supported by the findings by Haut et al. that the correlation between MRI and direct measurement was lower than the figures obtained by other studies of the ipsilateral knee and were comparable to that of radiographs except for lateral meniscus width (r² = 0.25 for Pollard, 0.76 for MRI).

3.5 Comparison between modalities

Two small studies compared between modalities without correlation with true cadaveric measurements. The study by Kaleka et al.³⁶ was a prospective study of 22 patients (44 knees) and showed good concordance between radiographic methods and MRI for all dimensions except for lateral meniscal width. Average mean difference compared to MRI was 0.7 % and 2.6 % for medial meniscus and 10.2 % and 2.4 % for lateral meniscus width and length respectively, length using the Yoon method.

The second study by Haen et al.²⁴ was a retrospective study and was of poorer quality as radiographic and MRI and CT methods varied due to being done at different institutions. This showed best correlation between radiographs and MRI for medial meniscal length, but the mean differences between radiograph and MRI were very large – 11.3 % difference for medial meniscal length and between 22.3 and 47.4 % difference for the remaining dimensions. Given the results from all of the other studies in this review, this case series could be taken as an outlier.