Purpose: The aims of this study were to: i) describe the association between MRI, macroscopic and histological assessments of synovitis in end-stage knee osteoarthritis (KOA) using correlation analyses and ii) determine if static and dynamic MRI variables can predict histological synovitis in KOA. We hypothesized that MRI based estimates of synovitis are highly and positively correlated (r ≥ 0.70) with histopathological findings consistent with synovitis in end-stage KOA.
Methods: Non contrast-enhanced (CE), CE- and dynamic CE-MRI of end-stage osteoarthritic knees performed before total knee replacement were analysed and correlated with microscopic and macroscopic assessments of synovitis obtained intraoperatively.
Synovitis was assessed with: i) MOAKS (MRI in OA Knee Score) on non-CE-MRI, ii) BLOKS (Boston-Leeds OA Knee Score) and the whole-knee synovitis score according to Guermazi et al. on CE-MRI and iii) both pharmacokinetic and heuristic perfusion variables using Dynamika® (www.imageanalysis.uk.org) on DCE-MRI. Microscopic (6 locations) and macroscopic assessments were performed according to Krenn et al. and af Klint et al. respectively.
To compensate for the issue of multiple testing, only Spearman's rank correlations coefficients ≥ 0.70 were regarded as statistically significant.
This was followed by multiple regression analyses with the basic characteristics and MRI variables as predictors and the histology score as outcome variable. As contrast-enhanced MRI is not routinely performed in KOA, three multiple regression analyses with different sets of predictors were performed in order to increase the feasibility and clinical applicability:
- -Model 1: basic characteristics and static, non-CE MRI variables
- -Model 2: the aforementioned variables from model 1 and the static, CE-MRI variables
- -Model 3: the variables from the two previous models and all DCE-MRI variables
In all three cases, multiple regression analyses were performed with the intention to find the subset of independent variables (MRI-variables) that best predict the dependent variable (histology) by linear regression in terms of the largest adjusted R2. The regression equation of the final model was used as a surrogate marker of synovial inflammation.
Results: 39 patients (56.4% women) had complete MRI and histological assessments. The mean age was 68 years with a median Kellgren-Lawrence grade of 4. Only the DCE-MRI time intensity variable MExNvoxel (the maximum enhancement multiplied by the number of highest perfused voxels, i.e. a surrogate of the degree and volume of inflammation) and the macroscopic score (Macro-Total) showed correlations with histological inflammation (Micro-Total) above the pre-specified threshold for acceptance (r ≥ 0.70, p<0.05) (Table 1). MExNvoxel was also correlated with the macroscopic score (r = 0.72), MOAKS-Synovitis (r = 0.71), BLOKS-Effusion (r = 0.71) and CE-Synovitis (r = 0.73). None of the static MRI-variables (neither non-CE nor CE-MRI) were correlated above the threshold with either the microscopic or the macroscopic scores (Table 1).
The maximum R2-value obtained in Model 1 (basic characteristics and non-CE-MRI) regression analysis was R2 = 0.39 and included the log transformed MOAKS-Synovitis and KL-grade as predictors. In Model 2, where the CE-MRI variables were added to the first model, the highest R2 = 0.52 and was obtained by including the square root of CE-Synovitis, log BLOKS-Effusion and Sex. In Model 3 all dynamic and static MRI-variables were included in addition to the basic characteristic. A 4-variable model consisting of log IRExNvoxel, √iAUGC60, √CE-Synovitis and Sex created the highest R2 = 0.71 (Micro-Total = 0.78 + 1.0(logIRExNvoxel) + 0.7(√CE-Synovitis) + 0.7(Sex) - 5.9(√iAUGC60)).
Conclusions: The present study showed that DCE-MRI is strongly correlated with histological synovitis in end-stage KOA and the combination of CE and DCE-MRI could explain 71% of the variance of histological synovitis, thus potentially contribute to the development of a MRI-based “virtual histology” modality.
Table 1Spearmans correlation matrix of microscopic, macroscopic and MRI variables.
| Histology | Macroscopic | non-CE-MRI | CE-MRI | DCE-MRI (heuristic) | DCE-MRI (pharmacokinetic) | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Micro-Total | Macro-Total | MOAKS-Synovitis | BLOKS-Effusion | CE-Synovitis | Nvoxel | MExNvoxel | IRExNvoxel | IRExME | Ktrans | iAUGC | |
| Micro-Total | 1.00 | 0.72*(p<0.0001) | 0.54 (p = 0.0004) | 0.47 (p = 0.0025) | 0.68 (p<0.0001) | 0.66 (p<0.0001) | 0.70* (p<0.0001) | 0.59 (p<0.0001) | 0.39 (p = 0.0142) | 0.38 (p = 0.0163) | 0.30 (p = 0.0615) |
| Macro-Total | 1.00 | 0.54 (p = 0.0004) | 0.40 (p = 0.0117) | 0.63 (p<0.0001) | 0.67 (p<0.0001) | 0.72* (p<0.0001) | 0.68 (p<0.0001) | 0.55 (p = 0.0003) | 0.38 (p = 0.0170) | 0.42 (p = 0.0075) | |
| MOAKS-Synovitis | 1.00 | 0.78* (p<0.0001) | 0.51 (p = 0.0008) | 0.75* (p<0.0001) | 0.71* (p<0.0001) | 0.59 (p<0.0001) | 0.13 (p = 0.4226) | 0.18 (p = 0.2851) | 0.22 (p = 0.1854) | ||
| BLOKS-Effusion | 1.00 | 0.41 (p = 0.0102) | 0.75* (p<0.0001) | 0.71* (p<0.0001) | 0.59 (p = 0.0002) | 0.06 (p = 0.7353) | 0.13 (p = 0.4282) | 0.24 (p = 0.1409) | |||
| CE-Synovitis | 1.00 | 0.69 (p<0.0001) | 0.73* (p<0.0001) | 0.72* (p<0.0001) | 0.60 (p<0.0001) | 0.50 (p = 0.0011) | 0.59 (p<0.0001) | ||||
*Correlations are stronger than the pre-specified threshold (r ≥ 0.70; p < 0.05); (D)CE-MRI: (dynamic) contrast-enhanced magnetic resonance imaging; Micro-Total: histology score according to Krenn et al.; Macro-Total: macroscopic score according to af Klint et al.; MOAKS-Synovitis: sum of the effusion-synovitis and Hoffa-synovitis subscales of the MRI in OA Knee score; BLOKS-Effusion: effusion according to the Boston-Leeds osteoarthritis knee score; CE-Synovitis: whole-knee synovitis score according to Guermazi et al.; Nvoxel: sum of voxels with plateau and washout enhancement patterns; ME: maximal enhancement; IRE: initial rate of enhancement; Ktrans: volume exchange coefficient; iAUGC: initial area under the Gd curve
Article info
Publication history
82
Identification
Copyright
© 2016 Published by Elsevier Inc.
User license
Elsevier user license | How you can reuse 
Elsevier's open access license policy
Elsevier user license
Permitted
For non-commercial purposes:
- Read, print & download
- Text & data mine
- Translate the article
Not Permitted
- Reuse portions or extracts from the article in other works
- Redistribute or republish the final article
- Sell or re-use for commercial purposes
Elsevier's open access license policy
