|Home | About | Journals | Submit | Contact Us | Français|
The Osteoarthritis Initiative (OAI) is targeted at identifying sensitive biomarkers and risk factors of symptomatic knee OA onset and progression. Quantitative cartilage imaging in the OAI relies on validated FLASH sequences that suffer from relatively long acquisition times, and on a near-isotropic double echo steady state (DESS) sequence. We therefore directly compared the sensitivity to cartilage thickness changes and the correlation of these protocols longitudinally.
Baseline and 12 month followup data of 80 knees were acquired using 1.5mm coronal FLASH and 0.7mm sagittal DESS sequences. In these and in 1.5mm coronal multiplanar reconstructions (MPR) of the DESS the medial femorotibial cartilage was segmented with blinding to acquisition order. In the weight-bearing femoral condyle, a 60% (distance trochlear notch to posterior femur) and 75% region of interest (ROI) were studied.
The standardized response mean (SRM=mean change /standard deviation of change) in central medial femorotibial (cMFTC) cartilage thickness was −0.34 for coronal FLASH, −0.37 for coronal MPR DESS, −0.36 for sagittal DESS with the 60% ROI, and −0.38 for the 75% ROI. Using every 2nd 0.7mm sagittal slice (DESS) yielded similar SRMs in cMFTC for the 60% and 75% ROI from odd (−0.35/−0.36) and even slice numbers (−0.36/−0.39), respectively. Baseline cartilage thickness displayed high correlations (r≥0.94) between the three protocols; the correlations of longitudinal changes were ≥0.79 (Pearson) and ≥0.45 (Spearman).
Cartilage morphometry with FLASH and DESS display similar longitudinal sensitivity to change. Analysis of every second slice of the 0.7mm DESS provides adequate sensitivity to change.
The Osteoarthritis Initiative (OAI) is targeted at identifying sensitive biomarkers of symptomatic knee osteoarthritis (OA) and risk factors associated with the onset and progression of OA. Quantitative cartilage magnetic resonance imaging (qMRI) is a technology that is hoped to improve the ability to evaluate the response to treatment with disease modifying OA drugs with shorter observation periods than currently possible with radiography (<2 years) 1. qMRI in the OAI relies on validated fast low angle shot (FLASH) sequences and on a near-isotropic double echo in steady state (DESS) sequence 2.
It has been shown that FLASH with fat suppression or water excitation provides accurate and reproducible measures of cartilage volume and thickness at 1.5 3-6 and 3 Tesla 7-9. FLASH or similar spoiled gradient recalled (SPGR) sequences have the advantage that they are available from most MR vendors and scanners, but suffer from relatively long imaging times (~9 mins in the OAI) required for the acquisition of high resolution coronal imaging data with adequate contrast-to-noise ratios 10,11. In contrast to the FLASH, the DESS sequence with water excitation at 3 Tesla 12 offers a higher fluid-to-cartilage contrast 13 and allows for the acquisition of near-isotropic sagittal images with relatively low partial volume effects at comparable imaging times (~11 mins in the OAI). However, the longitudinal performance of the DESS has not been validated to date, by comparing rates of change and sensitivity to change with a previously validated standard protocol (FLASH).
The accuracy and precision of coronal FLASH and sagittal DESS protocols were found comparable for the analysis of femorotibial cartilages in the OAI pilot studies 8,14,15. Two recent publications on participants from the OAI progression subcohort reported similar magnitudes and spatial patterns of femorotibial cartilage loss over one year for coronal FLASH 16 and sagittal DESS (use of every 2nd slice) 17, but were not performed in identical knees and also utilized different analysis methods.
The purpose of the current study was to directly compare the longitudinal performance of FLASH and DESS in the same knees of OAI participants. The sensitivity to cartilage thickness changes over one year and the correlation of these changes between the different protocols was determined, to address whether longitudinal analyses in different subsets of the OAI participants (obtained with different OAI protocols) can be analyzed together (i.e. can be pooled), to gain power through larger statistical analyses, e.g. for the identification of OA risk factors. Because the greater number of slices acquired by the DESS increases the time and cost of image segmentation, we also evaluated whether the analysis of every 2nd slice or the analysis of 1.5mm coronal multiplanar reconstructions [MPR] of the 0.7mm sagittal DESS were associated with a deterioration of the longitudinal sensitivity to cartilage thickness changes or not. Additionally, we compared results for weight-bearing femoral regions of interest (ROI) covering 60% and 75% of the distance between the trochlear notch and the posterior ends of the femoral condyles.
The study was performed on the right knees from 80 participants from the first half (2678 cases) of the OAI cohort [http://www.oai.ucsf.edu/datarelease/): OAI public use data sets 0.1.1 (baseline clinical), 0.C.1 (baseline images) and 1.C.1 (12 month followup images)] selected for another study, which reported rates of progression in knees of participants with medial joint space narrowing (mJSN) in one, but no (or less) mJSN in the contralateral knee 18. Participants were chosen, if they had a body mass index (BMI) >25 kg/m2, a mJSN OARSI grade 1-3 19,20 in one knee, no or less mJSN in the contra-lateral knee, and no (or less than medial) lateral JSN in both knees 18. In addition, the participants displayed chronic pain in both knees (most days of a month within the last 12 months, 78 cases), or had occasional pain in one knee (pain in past 12 months but not most days of a month) in combination with a definite osteophyte (2 cases). The sample comprised 48 women with an age of 60.3±8.3 years (mean±standard deviation) and a BMI of 31.9±4.3 kg/m2, and 32 men with an age of 61.7±10.3 years and a BMI of 30.1±3.3 kg/m2 18. The right knees displayed no mJSN in 44 cases, grade 1 mJSN in 23, grade 2 in 11, and grade 3 mJSN in two cases.
Baseline and 12 month followup MR images had been acquired by the OAI using 3 Tesla MR systems (Siemens Magnetom Trio, Erlangen, Germany) and quadrature transmit-receive knee coils (USA Instruments, Aurora, OH). Sagittal DESS (sagDESS) images with water excitation and 0.7 mm slice thickness were acquired in both knees 2,8,17, whereas double oblique 1.5mm coronal 3D FLASH images (corFLASH) with water excitation were acquired only in the right knees 2,8,16. In addition, 1.5mm coronal multi-planar reconstructions were obtained from the sagittal DESS images (corMPR DESS) 8,14.
All analyses were funded by a consortium initiative of seven pharmaceutical industry partners (see acknowledgement), the OAI coordinating center at the University of California San Francisco, and an image analysis company (Chondrometrics GmbH). The analysis of the sagDESS was performed as part of a previously published study 18, and the analysis of the corMPR DESS was co-funded by the OAI coordinating center.
Images were shipped from the OAI coordinating center to the image analysis center (Chondrometrics GmbH, Ainring, Germany), where quality control was performed (M.H.). The medial femoral cartilage was excluded in one case, because the slice orientation of the corFLASH was not appropriately aligned with the posterior end of the femoral condyles. The analysis thus comprised 80 tibial and 79 femoral cartilages.
Baseline and 12 month follow-up images of each imaging protocol were analyzed as pairs by one of seven readers, each with more than 3 years experience in cartilage segmentation of both DESS and FLASH 8,14,16,18. The readers were blinded to the order of the image acquisition and to the clinical and radiographic data. The baseline and follow-up images for each of the three protocols were always analyzed by the same reader, but the sagDESS, corFLASH, and corMPR DESS baseline and follow-up pairs were not read on the same day or necessarily by the same reader.
Manual segmentation of the cartilage surface (AC) and of the total subchondral bone area (tAB) was performed in the medial tibia (MT) and the entire medial femoral condyle (MF) for the sagDESS 18, and segmentation of MT and the weight-bearing central part of the medial femoral condyle (cMF) was performed for corFLASH and corMPR DESS 8,16. The number of slices segmented per cartilage plate was on average 30 for MT and 14 for cMF with corFLASH and corMPR DESS, and was on average 41 for MT and 43 for cMF with sagDESS. The time required for segmentation varied between data sets (depending on disease state and individual image characteristics) and readers, but was generally similar for corFLASH and corMPR DESS (1.5mm) and twice as long for sagDESS when all 0.7 mm slices were segmented. All segmentations were quality controlled by one expert reader (S.M.) who was also blinded to the order of acquisition, and were revised by the original readers, when necessary.
To match the femoral regions of interest (ROI) of the sagDESS with the corFLASH / corMPR DESS, the mean cartilage thickness over the subchondral bone (ThCtAB) was determined in a femoral ROI extending 60% from a plane through the intercondylar notch to a plane tangential to the posterior ends of both the medial and lateral femoral condyles 8,16.
Since segmentations of the entire femoral condyle were available for the sagDESS, the 60% ROI was compared with a 75% ROI used in the previous study18, to explore whether the sensitivity to change was different for femoral ROI extending further posterior on the condyle. To evaluate whether the analysis of only every 2nd slice of the near-isotropic, 0.7mm sagDESS was sufficient for characterizing longitudinal cartilage loss, computations (of ThCtAB) from the sagDESS were repeated by including only “odd” or “even” segmented slice numbers in the analysis, respectively.
Since recent studies on OA progression in the femorotibial joint have focused on anatomical subregions 21-24, and because sagittal and coronal imaging protocols have larger partial volume effects in different areas of the femorotibial cartilages, we also determined longitudinal changes of ThCtAB in central, external, internal, anterior and posterior subregions of MT, and in central, external, internal subregions of cMF 23,24. The central subregion occupied 20% of the tAB in MT, and 33% in cMF, respectively. Because a recent study demonstrated an increased capability of detecting differences in rates of progression between different subcohorts when using an ordered values (ranking) system derived from the above subregional measurements, we also computed the mean subregional changes across the 79 knees with complete data for MT and cMF for rank 1 (subregions with the greatest reduction in ThCtAB [in μm], rank 2 (subregions with the second greatest reduction), rank 3, etc., up to rank 8 (subregions with the smallest reduction (or greatest increase) in ThCtAB 25.
To obtain a single integral measure for the weight-bearing part of the femorotibial joint (MFTC), ThCtAB of MT and cMF was added at baseline and follow up. Similarly, the central subregions of MT and cMF (cMT and ccMF) were combined to derive thickness values for the central femorotibial compartment (cMFTC).
As a measure of progression, the mean change (MC in μm) in ThCtAB was determined between baseline and followup. The percent mean change (MC%) was calculated by relating the MC across all participants to the mean ThCtAB at baseline. To estimate the sensitivity to change, the standardized response mean (SRM=MC/SD of change in μm) was computed.
All statistical analyses were performed using SPSS 15 software(SPSS Inc., Chicago, IL, USA). A paired t-test was applied to evaluate whether changes in cartilage thickness between baseline and 12 months were statistically significant. To determine the correlation of the baseline measurements (ThCtAB) between the three protocols, the Pearson correlation coefficient was used and differences between the protocols were assessed using Bland and Altman plots. The correlation of the longitudinal changes between the protocols (delta ThCtAB) was calculated using both the Pearson correlation coefficient and the Spearman rank correlation coefficient (applicable for non-normally distributed data). The number of “progressors”, i.e. the number of participants in which the rate of change in MFTC over one year was higher than the smallest detectable difference (SDD), was determined for each imaging protocol. The SDD was taken to be 2.8 × the intra-reader precision error determined for paired readings of test-retest data published in a previous study 15.
The cartilage thickness values were similar for the three protocols for the compartment, plates and subregions (Tables (Tables11 to to3,3, Fig. 1). The differences (errors) between the protocols did not show a relationship with the magnitude of the baseline cartilage thickness (Fig. 1). Central subregions displayed a greater ThCtAB than the peripheral ones. The correlation (Pearson) for ThCtAB at baseline between the protocols was ≥0.94 for MFTC and cMFTC, ≥0.93 in MT and cMT, and ≥0.90 in cMF and ccMF. The lowest correlation was observed for the internal MT (corMPR DESS vs. sagDESS; r=0.79) and in the internal cMF (corMPR DESS vs. sagDESS; r=0.82). The lowest correlation between corFLASH and sagDESS was observed in the internal MT (r=0.83).
The size of the subchondral bone area for the femoral ROI was the same for corFLASH (5.9±1.1 cm2) and corMPR DESS (5.9±1.2 cm2), and was similar for the 60% ROI of the sagDESS (6.3±1.3 cm2). The 75% ROI of the sagDESS was about 30% larger (8.2±1.6 cm2). Whereas the 60% ROI covered 36±1.8% of the entire femoral condyle (32%-41%), the 75% ROI covered 47±2.0% (43%-52%).
The rate of change in the entire MFTC was −63μm (−1.9%) for corFLASH, −86μm (−2.5%) for corMPR DESS, and −97μm (−2.9%) for sagDESS (Table 1). The sensitivity to change (SRM) varied between −0.28 (corFLASH) and −0.35 (sagDESS). In cMFTC, the SRM was somewhat greater and more consistent between the protocols (−0.34 to −0.37; Table 1).
In MT, corFLASH showed lower rates of progression and sensitivity to change than corMPR DESS and sagDESS, except for the external subregion, where the rate of change and SRMs were greatest in comparison with the other subregions (Table 2). Here, rates of change and SRMs were similar between corFLASH (−4.1%, −0.33), corMPR DESS (−3.4%, −0.28) and sagDESS (−4.7%, −0.39). In cMF, the rates of change (but not the SRMs) were >2-fold greater than in MT (Tables (Tables22 and and3).3). The rates of change and SRMs in cMF (Table 3) were similar between corFLASH, corMPR DESS and sagDESS (−0.29 to −0.30) (Table 3). The number of participants in which the observed rate of change in ThCtAB of MFTC exceeded the smallest detectable difference (SDD, 2.8 × the precision error) was 12/79 (15%) for the corFLASH, 10/79 (13%) for the corMPR DESS, 21/79 (27%) for the sagDESS.
When examining ordered values of the subregional medial femorotibial cartilage changes, 5 of the 8 ranks showed negative values (i.e. cartilage thinning) in all MR protocols. The rate of change tended to be greater for the sagDESS and the corDESS than for the corFLASH (Table 4). In the 1st rank (regions with the greatest thinning), the rate of change varied between −155 μm (−8.5%) for the corFLASH and −224 μm (−12.6%) for the sagDESS (60% ROI, Table 4). Because not only the rate of change but also the variability of the change was higher for the sagDESS and corMPR DESS, the SRM, however, differed only slightly between the three protocols (Table 4).
Whereas many participants displayed little change between baseline and 12 months, some participants displayed relatively large reductions in ThCtAB (Figure 2). In seven knees, the change in ThCtAB of MFTC exceeded the SDD in all 3 imaging protocols. In three knees this was the case for two of the three protocols (one for each of 3 possible combinations), and in 16 knees this was the case for only one protocol (3x corFLASH, 1x corMPR DESS, and 12x sagDESS). The correlations between the longitudinal changes were moderate between the protocols, with coefficients ranging from 0.55 to 0.90 (Pearson) and from 0.28 to 0.60 (Spearman rho) in the compartment and plates (Table 5). In the subregions, the Pearson correlation coefficients ranged from −0.02 (corFLASH vs. corMPR DESS, posterior subregion of MT) to 0.89 (corFLASH vs. corMPR DESS; external subregion of MT). The Spearman rho correlation coefficients ranged from 0.02 (corFLASH vs. corMPR DESS; posterior subregion of MT) to 0.55 (corMPR DESS vs. sagDESS; central subregion of cMF). The ordered values of subregional thickness changes showed high Pearson correlation coefficients (up to r=0.90) and moderate Spearman correlation coefficients (up to 0.56) between the protocols for the lower ordered values (ranks), i.e. the regions with the greatest changes (Table 5).
Extension of the femoral ROI from 60% to 75% only marginally affected the rate of change observed in cMF (Table 3), and there only was a slight increase in the SRM for 75% (−0.32) versus 60% (−0.30). Analysis of every 2nd slice of the sagDESS had only a marginal impact on the rates of change and sensitivity to change (SRM −0.35 for all, −0.34 for odd and −0.34 for even slices for MFTC, one participant less exceeding SDD for odd slices). This also applied to the tibia and femur, to the 60% and 75% cMF ROI, and to the subregions (Tables (Tables11 to to33).
In this study we have compared the longitudinal performance of MR sequences of the OAI imaging protocol2 in identical knees, specifically a near-isotropic 0.7mm sagDESS 8,12-14 with a previously validated 1.5mm corFLASH 3-9. Also, this is the first study to explore the impact of analyzing every 2nd slice of the 0.7mm sagDESS compared with the analysis of every slice, and the sensitivity to change for a 60% and a 75% region of interest in the weight-bearing medial femoral condyle. A high agreement was found for (cross sectional) baseline cartilage thickness measurements between corFLASH, sagDESS and the corMPR DESS. Longitudinally, the rate of and the sensitivity to change were similar for the three protocols for the (central) medial femorotibial compartment and for the weight-bearing femur. In the tibia, the rate of and the sensitivity to change in cartilage thickness were less for the corFLASH than for the sagDESS or corMPR DESS, except for the external subregion, where the greatest changes were observed throughout MT. Correlations of the baseline values were high and those of the longitudinal changes were moderate between the protocols. The rate of and the sensitivity to change were not affected by the dimension of the femoral ROI (60% or 75%), or by using only every 2nd slice of the 0.7mm sagittal DESS.
The intra-reader test retest precision errors for the three protocols used here have been thoroughly examined in the OAI pilot studies 8,14,15 and were reported to range from 1.3% to 1.9% of the baseline cartilage thickness in MFTC for paired analysis. A recent study by Bae et al. (using a semi-automated approach) reported inter-reader precision errors in the same range as intra-reader errors for cartilage volume with the sagittal DESS26.
Limitations of the study include that no formal assessment of the inter-reader precision errors was performed, as these may differ between the three protocols. Although the data set pairs (baseline and follow-up) for each of the three protocols were not processed at the same point in time and not necessarily by the same person, the baseline and follow-up images of each protocol were always analyzed by the same reader, and quality control readings were performed for all segmented slices of all data sets by a single expert, to ensure consistency between the readers. Further limitations of the study include the limited number of analyzed knees, and the small number of participants showing change in cartilage thickness over one year (i.e. larger than the SDD). This is the first study, however, to provide a thorough validation of the DESS for the longitudinal measurement of cartilage loss in comparison with the previously validated coronal FLASH protocol 3-9. Because few subjects displayed cartilage loss (i.e. larger than the SDD) over one year, the Pearson correlation coefficients of longitudinal changes between the protocols were greater than the Spearman coefficients, the reason being that the Spearman coefficients are determined to a lesser degree than the Pearson coefficient by those cases with the greatest reductions in cartilage thickness. Also, another limitation is that the current analysis was confined to the medial femorotibial compartment, because the participants had medial disease and therefore this compartment was of primary interest 18. The results cannot therefore be extrapolated to the lateral compartment.
The high agreement seen for the baseline values confirms earlier cross sectional findings8 and suggests that quantitative data from FLASH and DESS from different subcohorts of the OAI can be analyzed together (pooled) for the purpose of cross-sectional analyses. The longitudinal performance of the protocols is in agreement with the observation of nine participants over two years in the OAI pilot study 15, in which the rate of change in cartilage thickness was also somewhat higher for the sagDESS than for the corFLASH or for the corMPR DESS. The results are also in principle agreement with two recent studies in the first release of the OAI 16,17 that showed similar spatial distribution patterns of thickness changes between femorotibial cartilage plates for corFLASH 16 and sagDESS 17.
The current study shows a similar performance (sensitivity to change) of FLASH and DESS in joint regions where high rates of change occur (cMF and subregions, cMFTC, external MT, and lowest ranks (=ordered values with greatest changes), whereas a higher sensitivity to change for the sagittal DESS was observed in regions where the changes were small. However, these findings should be confirmed in other (and possibly larger) cohorts, before they are generalized. The comparison of ordered values (ranks) is particularly useful in revealing these relationships, because the subregions with the greatest (or lowest) longitudinal changes are averaged across individuals independent of their anatomical locations. This is done in order to account for the fact that cartilage loss is spatially heterogeneous, depending on individual risk factors and patho-physiology, and that therefore the subregions with the greatest cartilage loss vary substantially between participants.
The correlations of the longitudinal changes between the protocols were only moderate, but were generally higher in cartilage plates, subregions or ranks showing relatively high changes. This is because the relationship between the magnitude of the actual changes and the precision errors is likely more favorable for regions or ranks with relatively large changes, whereas in subregions and ranks with only small change the correlations are more strongly affected by the precision errors. Additionally, the lower correlations in some of the peripheral subregions may result from partial volume effects that are higher in the anterior and posterior tibia with coronal, and higher in the internal tibia with sagittal protocols. Still, the current data indicate that for total and central regions of the medial femorotibial compartment and for ordered values with the relatively greatest reduction in cartilage thickness, longitudinal analyses from different OAI subcohorts (analyzed with different OAI protocols) may be combined (pooled), in order to gain increased statistical power for the identification of risk factors of OA progression.
Other isotropic or near isotropic options for cartilage imaging where fluid is delineated as hyperintense have been proposed, i.e. balanced steady state free precession (bSSFP)28, vastly undersampled isotropic projection steady-state free precession (VIPR)29, and 3D isotropic resolution fast spin-echo MR imaging (3D-FSE)30. Given the encouraging results with DESS, these may provide possible future choices for quantitative cartilage analysis, once they are validated for this purpose.
Baseline and one year follow-up JSN readings (OARSI atlas) for 59 of the 80 knees studied here have recently been made publicly available. Two knees (3.4%) showed an increase by two medial JSN grades, and both were identified as “progressors” (change in MFTC larger than SDD) by all three MRI protocols used. Seven knees (11.9%) displayed an increase by one medial JSN grade: three of those were identified as progressors by all three protocols, one by corFLASH only, and three by none of the protocols. Fifty (84.7%) knees maintained the same medial JSN grade: Of those six were identified as “progressors” by corFLASH, five by corMPR DESS, and 16 by sagDESS. These findings suggest a somewhat higher agreement between corFLASH “progression” with increases in radiographic JSN than for sagDESS.
Although the larger femoral ROI (75%) of the sagDESS covered a greater (47%) part of the femoral condyle, the rate and sensitivity to change was similar to the 60% ROI, corresponding with the region measured in coronal views 8. The current study also showed that analysis of every 2nd slice in the sagDESS was sufficient to adequately characterize longitudinal changes, and the rates and sensitivity to change were not lower than for analyses covering every slice. This permits to cut segmentation time and cost substantially, as long as fully automated segmentation algorithms are not available and validated.
In conclusion, the results of this study suggest that cartilage morphometry with FLASH and DESS display similar longitudinal sensitivity to change in cartilage thickness in anatomical subregions of the femorotibial joint that display the greatest change over time, or in ordered values (ranks) which average the greatest magnitude of subregional change across subjects, independent of the anatomical location. The high correlations for the baseline measurements show that data from the different OAI subcohort analyzed with different protocols can be combined (pooled) for cross-sectional analyses. The correlations for longitudinal thickness changes were moderate to high, indicating that pooling data from different subcohorts analyzed with different OAI protocols may also be feasible for longitudinal studies. Segmentation of every 2nd slice of the 0.7mm sagittal DESS is adequately to characterize cartilage loss longitudinally, allowing for considerable savings in segmentation time and cost.
We would like to thank John Lynch for his help in working with the OAI images and the readers at Chondrometrics GmbH, Gudrun Goldmann, Linda Jakobi, Manuela Kunz, Dr. Susanne Maschek, Sabine M hlsimer, Annette Thebis, and Dr. Barbara Wehr for dedicated data segmentation. We would also like to thank the OAI investigators and technicians for providing high quality images and the funding sources for their support.
Funding sources: Data acquisition: The OAI is a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health.
Image analysis: Funding was provided by a consortium of the OAI coordinating center at the University of California, San Francisco (UCSF Vendor ContractNo. 9000011571) and seven industry partners: Pfizer Inc., Eli Lilly & Co, Merck Serono SA, Glaxo Smith Kline, Wyeth Research, Centocor, and Novartis Pharma AG. This manuscript has received the approval of the OAI Publications Committee based on a review of its scientific content and data interpretation.
Funding Source: The study and image acquisition was funded by the Osteoarthritis initiative, a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262).
The image analysis was funded by a consortium of the OAI coordinating center at the University of California, San Franciso (UCSF Vendor ContractNo. 9000011571) and seven industry partners: Pfizer Inc., Eli Lilly & Co, Merck Serono SA, Glaxo Smith Kline, Wyeth Research, Centocor, and Novartis Pharma AG
Conflicts of Interest: Wolfgang Wirth, Susanne Maschek and Martin Hudelmaier have part time appointments with Chondrometrics GmbH. Felix Eckstein is CEO of Chondrometrics GmbH, a company providing MR image analysis services. He provides consulting services to Pfizer, MerckSerono, Novo Nordisk, Wyeth, and Novartis. Marie-Pierre Hellio Le Graverand has a full time employment with Pfizer Inc., Olivier Benichou with Eli Lilly, Donatus Dreher with MerckSerono, Richard Y. Davies with Glaxo Smith Kline, Jennifer Lee with Wyeth, Kristen Picha with Centocor, and Alberto Gimona with Novartis. Michael Nevitt has no competing interests.
Ethics approval: The study was conducted in compliance with the ethical principles derived from the Declaration of Helsinki and in compliance with local Institutional Review Board, informed consent regulations, and International Conference on Harmonization Good Clinical Practices Guidelines.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.