Keywords
knee - magnetic resonance imaging - patellar instability - tomography, x-ray computed
Introduction
Patellar instability represents a common and significant health condition that affects
young subjects and can lead to early osteoarthritis,[1] with an incidence as high as 12.98/100,000 person-years in males between 15 and
19 years old.[2] Its etiology is multifactorial, requiring a precise diagnosis, since treatment options
range from conservative therapies to different surgical interventions.[3]
One of the most recognized risk factors for patellar instability is the increased
tibial tubercle-trochlear groove distance (TT-TG); when greater than 15 to 20 mm,
it is generally considered pathologic and has been proposed as a threshold for considering
a tibial tubercle osteotomy or distal realignment procedure.[4]
[5] The gold-standard imaging method for performing this measurement is computed tomography
(CT), that has proven to be reliable.[6] On the other hand, many of the patellar instability patients undergo magnetic resonance
imaging (MRI) of the knee to assess soft-tissue injuries such as ligament tears and
chondral defects, so it would be desirable to reliably assess the TT-TG distance in
the same imaging study and thus avoid the need of an extra CT study. Besides, many
patients with patellar instability are young and avoiding radiation exposure would
be considerably beneficial.
Many authors have found that CT and MRI TT-TG measurements are not equivalent, and
that MRI measurements are systematically underestimated,[3]
[4]
[7]
[8]
[9] which suggests it would be inaccurate to use the same threshold in MRI and CT in
diagnosis and surgical planning.
The TT-TG distance is highly sensitive to changes in knee positioning[4]
[10] and while CT is performed with the legs in full extension, the dedicated knee coil
in MRI surrounds the knee in a way it assumes variable grades of flexion (∼25°[7]) and varus deviation.[4] The literature is scarce and controversial about the influence of feet positioning
in knee measurements.
Another controversial topic in the literature is which landmarks to use for the measurement
of the TT-TG distance. Given the high soft-tissue contrast resolution of MRI, some
authors used soft-tissue parameters instead of bony parameters to measure the lateralization
of the tibial tubercle (TT): the nadir of the cartilaginous trochlear groove (CTG)
instead of the bony trochlear groove and the tibial insertion of the patellar tendon
(PT) instead of the TT.
To this date, no study has compared the TT-TG and PT-CTG measurement in MRI using
a body coil and CT to test interchangeability. Therefore, the aims of the present
study are to compare these measurement values between MRI using a body coil with CT
in asymptomatic volunteers, and to evaluate intermethods and interobserver agreement.
Our hypothesis is that warranting the same knee and feet positioning in MRI and CT,
the measurements would result similar.
Methods
Ethical committee approval was obtained (Plataforma Brasil number 3136833), as well
as informed consent of all participants. The sample size was calculated according
to Zou,[11] considering an effect size of 0.65, a 2-tailed significance level (α) of 5% and
a power (β) of 80%. This calculation resulted in a minimum of 30 knees.
Volunteers without any clinical knee symptoms were enrolled. The study group consisted
of 34 knees (17 subjects; 13 male and 4 female), with a mean age and standard deviation
(SD) of 38.6 ± 6.4 years, range between 29 and 50 years old. The inclusion criterium
was the absence of knee symptoms and the exclusion criteria were: previous knee surgery,
previous knee trauma, history of patellar instability or any other known knee pathologies.
A low-dose CT-scan and an axial T1-weighted MRI sequence of the knees were performed
in all subjects. Positioning was the same in both studies: the volunteers were scanned
in the supine position with full extension of the knees, using an acrylic supporting
device that kept the orientation of the medial faces of the feet parallel to each
other, with a distance of 3 to 5 cm between them ([Fig. 1]).
Fig. 1 (A and B) Positioning in the CT and MRI with body coil, both with the acrylic supporting
device.
Computed tomography studies were performed on a 64-detector Siemens CT scanner (SOMATOM
Definition Edge, Siemens Medical Solutions, Munich, Germany), and the CARE Dose control
system was selected to achieve radiation dose reduction. For the ethics committee
evaluation, we performed radiation dose calculation on standard phantoms and the effective
dose resulted in ∼ 0.01 mSv (half the dose of a posteroanterior chest X-ray).[12] The images were reformatted to 3 mm thickness using soft-tissue and bone windows.
Magnetic resonance imaging studies were performed on a GE/Optima 450w 1,5T MRI Scanner
(GE, Boston, MA, USA) with a body coil and consisted in an axial T1-weighted sequence
(TR: 375 ms/TE: 8,32 ms) of both knees, 5 mm thickness, 1 mm spacing, 320 × 256 matrix.
Also, both examinations included the femoral trochlea and the tibial tuberosity to
allow measurements.
After a training session, two board-certified musculoskeletal radiologists (5 and
2 years of experience) evaluated the CT and MRI images independently and chose these
specific slices:
-
The most cranial slice that depicted complete cartilaginous coverage of the femoral
trochlea in MRI and CT (soft-tissue and bone window), allowing the determination of
the deepest point of the bony trochlea groove (TG) and the cartilaginous trochlea
groove (CTG);
-
The slice that showed the complete attachment of the patellar tendon at the tibial
tuberosity in MR and CT (soft-tissue window), and the midpoint of that enthesis was
defined as the patellar tendon (PT) landmark;
-
The most anterior point of the tibial tuberosity in MRI and CT (bone window), which
corresponded to the TT bony landmark.
After this first independent session, as our main interest was to study the relationship
between the knee position and the measurements, any differences in the slices chosen
were corrected by consensus. Then, they were superimposed and the TT-TG and PT-CTG
distances were measured independently in both methods ([Fig. 2]). The TT-TG distance was assessed between two parallel lines drawn through the bony
parameters: the most anterior point of the tibial tuberosity and the deepest point
of the bony TG, perpendicular to a line drawn tangent to the posterior aspect of the
femoral condyles.[13] The PT-CTG distance was measured similarly, but using the soft-tissue parameters:
the PT attachment to the tibia and the deepest point of the CTG.[14]
Fig. 2 The image shows the superimposed CT (left) and MRI (right) slices and depicts the
TT-TG measurement in the left knee of a 29-year-old female asymptomatic volunteer.
Statistical analysis was made using SPSS Statistics for Windows, Version 20.0 (IBM
Corp., Armonk, NY, USA), STATA 12 (Stata Software, College Station, TX, USA) and R
software (R Foundation, Vienna, Austria). Normality distribution was assessed by the
Kolmogorov-Smirnov test. The interrater reliability of slices chosen, TT-TG and PT-CTG
measurements on CT and MRI were evaluated for all measures studied using intraclass
correlation coefficient (ICC [2,1]) and Bland-Altman graphs. The type of ICC chosen
was based on the Koo et al. guidelines,[15] and the level of significance (α) = 0.05 was adopted.
Results
Normality assessed by the Kolmogorov-Smirnov test being the null hypothesis a normal
distribution resulted in a p > 0.05 for all variables. The ICC for all the slices chosen were excellent, except
for PT on MRI, which was good ([Table 1]). [Table 1] shows the percentage of knees in which the same or the next slice was chosen by
both observers. Good reliability and agreement was observed between CT and MRI measurements
for TT-TG and PT-CTG measurements with an ICC of 0.774 (0.659–0.854, p < 0.001) and 0.743 (0.615–0.833, p < 0.001), respectively. The distribution is shown in the Bland-Altman graphs ([Figs. 3] and [4]). The presence of < 6% of the observations outside the limits of agreement can be
observed (confidence interval [CI] of 95%). The TT-TG and PT-CTG measurements were
randomly scattered near the zero value of the difference and no systematic bias was
observed. The mean TT-TG on CT and MRI were 17.1 ± 4.2 mm and 16.2 ± 3.7 mm, respectively.
The mean PT-CTG distance were respectively 17.3 ± 4.2 mm and 16.5 ± 4.1 mm. The interrater
reliability was excellent for all measurements ([Table 2]).
Table 1
|
Slice
|
ICC(2,1) (95%CI)
|
Same slice
|
The same or the next slice
|
|
1 on CT
|
0.993 (0.985–0.997)
|
73.5%
|
97.1%
|
|
2 on CT
|
0.996 (0.989–0.998)
|
64.7%
|
100%
|
|
3 on CT
|
0.989 (0.928–0.996)
|
32.4%
|
76.5%
|
|
1 on MRI
|
0.967 (0.934–0.984)
|
70.6%
|
100%
|
|
2 on MRI
|
0.896 (0.796–0.947)
|
64.7%
|
94.1%
|
|
3 on MRI
|
0.961 (0.882–0.984)
|
52.9%
|
91.2%
|
Fig. 3 Bland-Altman graph shows the TT-TG measurements randomly scattered inside the CI.
Only 4,4% (3/68) of the cases are outside the limits of agreement.
Fig. 4 Bland-Altman graph shows the PT-CTG measurements randomly scattered inside the CI.
Only 5.9% (4/68) of the cases are outside the limits of agreement.
Table 2
|
Intraclass correlation coefficient
|
|
(95%CI)
|
p-value
|
|
CT
|
|
|
|
TT-TG
|
0.872 (0.760–0.934)
|
< 0.001
|
|
PT-CTG
|
0.918 (0.844–0.958)
|
< 0.001
|
|
MR
|
|
|
|
TT-TG
|
0.833 (0.693–0.912)
|
< 0.001
|
|
PT-CTG
|
0.907 (0.824–0.952)
|
< 0.001
|
Discussion
Our most important finding was the good reliability and agreement of the TT-TG and
PT-CTG measurements between MRI (using a body coil) and CT. Since the grade of knee
flexion influences the tibiofemoral rotation and hence the distances,[16] the rigorous standardization in the positioning of the knees was essential to achieve
that result. When TT-TG measurement is necessary, the CT study can be substituted
by an axial T1-weighted sequence of the knees using the body coil, removing the need
of unnecessary radiation exposure in this setting and, most importantly, allowing
the use of the same threshold (15–20 mm) classically used in CT. Our study also confirms
the excellent interrater reliability of MRI measurements, which had already been shown
in previous studies.[14]
[17]
[18]
Schoettle et al.[19] compared knee CT and MRI (with a routine knee protocol) and found an excellent intermethods
reliability, stating that additional CT scans were not necessary. However, many later
studies have not been able to reproduce these results, concluding that CT and MRI
TT-TG measurements are not equivalent, and that MRI measurements are systematically
underestimated,[3]
[4]
[7]
[8]
[9] which suggests it would be inaccurate to use the same threshold in MRI and CT in
diagnosis and surgical planning.
The TT-TG distance is highly sensitive to changes in knee positioning,[4]
[10] and while CT is performed with the legs in full extension, the dedicated knee coil
in MRI surrounds the knee in a way it assumes variable grades of flexion (∼ 25°[7]) and varus deviation.[4] A partially flexed position of the knee reduces the TT-TG measurements[7] due to the progressive internal rotation of the tibia in relation to the femur during
flexion. Seitlinger et al.[16] studied the TT-TG distance in extension and in different grades of flexion in MRI
and found that the TT-TG distance decreased significantly during flexion in knees
with patellofemoral instability and in healthy volunteers. Aarvold et al.[7] compared the TT-TG distance in symptomatic patients measured in MRI studies using
a body coil to guarantee full extension of the knees and in MRI using a dedicated
knee coil, finding that the latter underestimates the measurements (mean difference:
8.6 mm).
In none of these studies, the positioning of the feet was mentioned. Galland et al.[20] performed the CT studies using a plantar support device to avoid quadriceps contraction,
and although they mention a recommendation of placing the feet in the angle of step,
they considered feet positioning would not affect patellofemoral measurements (but
unfortunately did not present data to support it). We decided to standardize the positioning
of the feet for two theoretical reasons. One is the possibility of an undesired oblique
alignment of the examined lower extremity in relation to the longitudinal axis of
the machine, and the second is that the gravity acting on the feet of a lying supine
patient could produce a torque on the knee and rotation of the tibia in relation to
the femur.
The slice selection might be a source of a disagreement of the final TT-TG or PT-CTG
on both methods. Even though the ICC was excellent or good for all the slices, the
agreement over the same slice may be considered poor for slice 3 on CT (32.4%) and
MRI (52,9%). We believe the long TT cranio-caudal diameter may cause trouble to decide
which slice to choose. Regarding the use of bony or soft-tissue parameters, although
both were reliable between CT and MRI, there was a tendency for higher correlation
coefficients when using soft-tissue parameters, in accordance to what was observed
in MRI measurements by Wilcox et al.[14] These findings point us to recommend the use of the PT as the distal landmark instead
of the TT.
The only systematic review and meta-analysis on the topic[5] suggests the use of different thresholds for CT and MRI (15.5 ± 1.5 mm for TT–TG
distance measured on CT and 12.5 ± 2 mm for MRI), with the limitation that there was
no standardization of the positioning and flexion of the knees and the landmarks used.
Ho et al.[4] concluded that establishing a controlled, reproducible positioning of the patient
would be vital to allow the interchangeability of the use of CT and MRI in measuring
the TT-TG distance, and that was the main goal of our study.
Precluding the CT use in this setting would avoid radiation exposure in a mostly young
population, thus reducing its potential risks throughout life, and also reduce overall
costs, though adding a sequence to the knee MRI would increase the MRI study time.
The main limitations of our study include a small sample and the exclusive evaluation
of asymptomatic volunteers. Future research should assess the interchangeability in
patients with patellar instability.
Another limitation would be that we could not assess the isolated importance of the
feet positioning, given that we chose to rigorously standardize positioning of both
the knees and feet and did not test different positioning of the feet.
In conclusion, this was the first study that compared MRI using a body coil with the
gold-standard CT in measuring the TT-TG and PT-CTG distances, with a good agreement
between those methods and an excellent interrater reliability, with the potential
clinical implication that the knee CT could be substituted by MRI using the body coil
in this clinical setting.
