O

Ide

CMa

b

c

d

e

f

a

AAA

KKJTS

1

snooOto(

T

1d

Joint Bone Spine 80 (2013) 183–187

Available online at

www.sciencedirect.com

riginal article

n normal knees, joint space width (JSW) is correlated with the intermarginistance (IMD), a measure of medial tibial plateau alignment. Variations in IMDxplain variability in JSW in serial radiographs

atherine Merciera,b,c, Muriel Pipernoa,∗, Eric Vignona, Kenneth Brandtd,arc Hochberge, Marie-Pierre Hellio Le Graverandf

Universite Claude-Bernard Lyon-1, Lyon, FranceHospices Civils de Lyon, Service de Biostatistique, 69003 Lyon, FranceCNRS, UMR5558, Laboratoire de Biométrie et Biologie Evolutive, Equipe Biostatistique Santé, 69622 Villeurbanne, FranceUniversity of Kansas Medical Center, Kansas City, KS, USAUniversity of Maryland School of Medicine, Baltimore, MD, USAPfizer Global Research and Development, New London, CT, USA

r t i c l e i n f o

rticle history:ccepted 5 July 2012vailable online 19 September 2012

eywords:nee osteoarthritis

oint space measurementibial plateau alignmenttandardized radiography

a b s t r a c t

Objective: To ascertain the importance of alignment of the medial tibial plateau (MTP), as determined bythe distance between the anterior and posterior margins of the plateau (intermargin distance [IMD]), formeasurements of joint space width (JSW) in radiographs of normal knees.Methods: JSW and IMD were measured in paired baseline and 12-month knee films of 122 subjects fromthe osteoarthritis initiative (OAI). Relationships between JSW and IMD, and between the variation in JSWand variation in IMD, were evaluated.Results: In cross-sectional analysis, a non-linear relationship existed between JSW and the concurrentIMD. With poor MTP alignment (IMD > 1.7 mm), a 1.0-mm increase in IMD resulted in a 0.16-mm (95%CI:0.11–0.21) increase in JSW (P < 0.0001). In a longitudinal analysis, the effect of IMD variation on variationin JSW was also highly significant (P < 0.0001). A variation of 1 mm between IMDBaseline and IMD12month

was associated with a 0.10-mm (95% CI: 0.06–0.13) variation in JSW, with variations in JSW and IMD

occurring in the same direction. An IMD variation less than or equal to 1.0 mm was determined to beacceptable for accurate evaluation of JSW in serial radiographs.Conclusion: The error in measurement of JSW caused by variation in IMD in serial radiographs of normalknees can be as large, or larger, than the mean rate of 12-month joint space narrowing (JSN) in OA knees.MTP alignment and replication of alignment in serial knee films are required for accurate determinationof JSN in OA knee.

ed by

© 2012 Publish

. Introduction

In studies designed to assess the radiographic progression oftructural joint damage in subjects with knee osteoarthritis (OA),arrowing of the tibiofemoral joint space (JSN) in paired filmsbtained over time is generally considered to reflect both thinningf the articular cartilage and meniscal pathology associated withA progression [1,2] Protocols that are intended to standardize

he radioanatomic positioning of the knee have been developed toptimize the reproducibility of measurements of joint space widthJSW) [3,4]. Their development has been motivated, in part, by the

∗ Corresponding author. Centre Hospitalier Lyon Sud, 69495 Pierre Benite, France.él.: +33 47 88 61 688; Fax: +33 47 88 66 540.

E-mail address: muriel.piperno@chu-lyon.fr (M. Piperno).

297-319X/$ – see front matter © 2012 Published by Elsevier Masson SAS on behalf of thoi:10.1016/j.jbspin.2012.07.006

Elsevier Masson SAS on behalf of the Société Française de Rhumatologie.

fact that a reduction in measurement error in the placebo group inrandomized clinical trials (RCTs) of putative structure-modifyingOA drugs (SMOADs) [5] would permit a smaller sample size orshorten the treatment period.

In standardized radiographic protocols, however, the excellentreliability of measurement of minimum JSW of the medial com-partment of the knee, as evaluated by repeated measurements onthe same radiograph, contrasts strikingly with the comparativelylarge error of measurement of the method in the determinationof JSN in serial radiographs of the same joint. In our hands, thestandard deviation (SD) of the difference between two measure-ments of JSW performed on the same radiograph does not exceed

0.04 mm, while the SD of the difference in JSW between pairedradiographs of non-arthritic control knees obtained at a 1-yearinterval was 0.32 mm [6]. This indicates that the measurementerror of JSN is due mainly to a lack of reproducible positioning of

e Société Française de Rhumatologie.

184 C. Mercier et al. / Joint Bone Spine 80 (2013) 183–187

Fig. 1. Illustration of intermargin distance (IMD), according to alignment of medialtibial plateau (MTP) with the orientation of the X-ray beam. Perfect MTP alignmentleads to a small IMD (A), i.e. superimposition of the anterior and posterior marginso

tabdmimsrbtr

htmstmflwttcg(

pers1a0stbttt

Fig. 2. Illustration of the measurement of JSW. The program detects both the artic-ular bone edges and the minimum distance along a vertical line between the bone

f the MTP. Poor MTP alignment results in a large IMD (B) in the same knee.

he joint in serial radiographs. Angle of knee flexion, knee rotationnd alignment of the MTP with the X-ray beam have been shown toe important for good joint repositioning [7]. Alignment is usuallyetermined from the distance between the anterior and posteriorargins of the MTP in the radiograph (Fig. 1). “Good” alignment

s defined as superimposition of the two margins or by an inter-argin distance less than 1.0 to 1.7 mm [2,8]. Several protocols for

tandardizing the radioanatomic positioning of the knee non fluo-oscopically can successfully reproduce knee flexion and rotation,ut not good alignment of the medial tibial plateau (MTP) withhe central X-ray beam, the consistent achievement of which hasequired fluoroscopy [3,4,7].

Reproduction of alignment of the MTP with the X-ray beamas been repeatedly found to be essential for adequate evalua-ion of JSN [6–9]. For example, when the anterior and posterior

argins of the MTP were superimposed (IMD ± 1 mm) in pairedtanding anteroposterior films of subjects with knee OA, JSN waswice as rapid, and only one-fourth as variable, as when MTP align-

ent did not meet that criterion [9]. However, among paired fixedexion (FF) views of OA knees obtained at a 12-month interval,hen the IMD in the follow-up film (IMD12mos) was smaller than

hat in the baseline film (IMDbl), JSN was significantly more rapidhan when IMD12mos was larger than IMDbl. [6]. We have recentlyonfirmed this artefactual effect of MTP alignment on JSN in radio-raphs of OA knees of subjects in the Osteoarthritis InitiativeOAI) [8].

In OA knees, however, JSN may vary with the progression ofathology within the joint. It is difficult, therefore, to ascertain theffect on the error of measurement of JSN that is attributable to pooreproduction of MTP alignment in serial radiographs. In the presenttudy, therefore, we have measured JSW and IMD in baseline and2-month follow-up images of both knees of 122 asymptomaticnd radiographically normal (i.e., Kellgren-Lawrence Grade [KLG]) [10] subjects enrolled in the control subcohort of the OAI, i.e.,ubjects in whom baseline JSW was unlikely to diminish overhe next 12 months. Our aims were to evaluate the relationship

etween IMD and JSW at the same time point; the magnitude ofhe effect that the difference between IMD12mos and IMDbl had onhe difference between JSW12mos and JSWbl, and the magnitude ofhe variation in IMD below which JSN is “not” significantly affected.

edges. The width of the joint space is calculated by the computer and the loca-tion of the measurement is shown on the screen. Measurement of IMD requiresintervention of the observer for selection of tibial margins location.

2. Methods

Subjects in this analysis are a subset of the 4796 participantsin the OAI, an ongoing 8-year, multicenter, longitudinal, prospec-tive observational cohort study focusing primarily on identifyingbiomarkers for the development and progression of knee OA. Thestudy protocol, amendments, and informed consent documen-tation were reviewed and approved by the institutional reviewboards at the four participating clinical centers and the coordinat-ing center. Data used in this analysis were obtained from the publicuse database for the OAI control subcohort [11]. The 122 subjects(75 women) enrolled in the control subcohort had a mean age of 55years (range, 45–78 years) and mean baseline BMI of 24.3 (range,17.7–31.4). They did not report frequent symptoms of OA in eitherknee (pain, aching or stiffness on most days of a month in the pastyear) at their baseline or 12-month follow-up visit. The data for thecontrol subsample used in this analysis are from clinical datasets0.1.1 and 1.2.1 and image release 0.C.1 and 1.C.1. All radiographswere obtained by trained and certified radiology technicians, whofollowed a standard protocol [11].

Baseline and 12-month follow-up FF radiographs [12] of bothknees of each subject were evaluated. Ten of the 122 subjectswere eliminated from the analysis of JSN because the 12-monthX-ray was missing. Therefore, in the cross sectional study a totalof 468 knee radiographs were analyzed (244 baseline films and224 follow-up films); in the longitudinal study, 224 knees wereanalyzed.

All 468 knee radiographs were read for features of OA by thesame experienced observer (EV), who was blinded to the resultsof the measurement of JSW; all knees were judged to be KLG 0at both time points. Measurements of IMD and JSW were per-formed in both knees of each subject with Holy’s software, exactlyas described previously (Fig. 2) [6]. The intraobserver reproducibil-ity of JSW measurement on the same radiograph is illustrated by a0.00 ± 0.04 mm mean (SD) difference between two blinded mea-surements. The interobserver reproducibility is identical to thatof the intraobserver reproducibility. Radiographs were read withthe observer blinded to the sequence, but were read by pair to

ensure that measurement of JSW was made at the same site inthe medial compartment in the baseline and follow-up film. JSWmeasurement was determined by computer; intervention of theobserver was not required in the present series. Within each pair

one Spine 80 (2013) 183–187 185

oml“aiId

3

oe(trabtt((uchomamkeftvt

4

4

a(aa(seid

esrtd(awigIII

Fig. 3. Relationship between JSW (mm) and IMD (mm) in the same FF radiograph.The scatter plot represents the observed values of JSW and IMD, with each point

C. Mercier et al. / Joint B

f films we analyzed the difference in the magnitude of the IMD, inm, in the follow-up radiograph (IMD12mos), relative to the base-

ine image (IMDBL). A positive value signifies an IMD12mos that waslarger” than the corresponding IMDBL; a negative value signifiesn IMD12mos “smaller” than the corresponding IMDBL. The variationn JSW was calculated as JSWBL–JSW12 mos. Differences in JSW andMD between the follow-up and baseline radiograph are hereafteresignated “variations”.

. Statistical analysis

All statistical analyses were performed using a mixed modelf regression in which the subject was considered as a randomffect, so as to take into account the within-subject correlationtwo knees of the same subject). Covariates (patient characteris-ics) were considered as fixed effects. The empirical curve of theelationship between JSW and concurrent IMD was obtained with

Loess smoothing function. The effect of IMD on JSW was testedy a likelihood ratio between the base model (without IMD) andhe model with IMD using linear and quadratic terms to reproducehe curvilinear relation. In the base model, patient characteristicssex, age, height, weight, BMI) and the sequence of the radiographbaseline or follow-up) were used as covariates and the statisticalnit of analysis was the radiograph. Quantitative covariates wereentered. Two base models were analyzed, one with weight andeight and the other with BMI. Because they were similar, and inrder to maintain consistency with our previous report [8], theodel with BMI was used as the base model. The effect of vari-

tion in IMD on variation in JSW was tested as a fixed effect in aixed model of regression in which the unit of analysis was the

nee. The same patient characteristics were initially tested as fixedffects and only the significant effects were retained in the modelor adjustment. Variance heterogeneity was taken into account inhe model to insure independence between predicted and residualalues. Statistical analysis was performed in R version 2.10.0, usinghe nonlinear mixed-effects model [13].

. Results

.1. Relationship between JSW and IMD in the same radiograph

All 468 normal knees were used for this analysis. Mean JSWnd mean IMD were 4.53 mm (range, 3.08–7.75 mm) and 1.93 mmrange, 0.06–7.43 mm), respectively. Mean JSW (with the meange and the mean BMI held constant) was 4.14 mm (3.99–4.27)mong women and was 0.78 mm (0.57–0.99) greater for menP < 0.0001). For every ten-year increase in age of the populationample, JSW decreased by −0.14 mm (−0.2–−0.02) (P = 0.02). Theffect of BMI (0.02 [−0.05–0.01] mm) was not statistically signif-cant. The sequence of the radiograph (i.e., follow-up or baseline)id not affect the results.

Fig. 3 shows the relationship between JSW and IMD in whichach point represents a single measurement. The empirical curvehows a curvilinear relationship between JSW and IMD, with a weakelationship for small values of IMD, but a strong and direct rela-ionship for larger IMD values. The overall effect of IMD on JSW, asetermined by the likelihood ratio test (61.5), was highly significantP < 0.0001). To determine whether the relationship between JSWnd IMD exists only in knees with poor MTP alignment, the analysisas performed by cutting the general model of quadratic relation

nto two sub- models of linear relation, with one slope for radio-

raphs with an IMD ≤ 1.7 mm and a second slope for those with anMD > 1.7 mm, following which the two were estimated separately.n radiographs with an IMD ≤ 1.7 mm (49%), a 1-mm increase inMD led to a non-significant (P = 0.051) increase in JSW (0.07 mm

representing a single radiograph. Shaded circles represent baseline radiographs andopen circles represent 1-year radiographs. The smoothing curve (loess method)shows a nonlinear relationship.

[95%CI: 0.000–0.152]). However, in the 51% of radiographs in whichthe IMD was greater than 1.7 mm, a significant (P < 0.0001) effectof IMD on JSW existed, with a 1 mm increase in IMD leading to a0.16-mm (95%CI: 0.11–0.21) increase in JSW.

4.2. Effect of variation in IMD on variation in JSW in pairedradiographs of normal knees

Variation in JSW and IMD between the follow-up and baselineradiographs was analyzed in both knees in the 224-paired radio-graphs (112 patients). The observed mean variations in JSW andin IMD were −0.05 mm (range: −1.75 to 1.34 mm) and −0.26 mm(range: −4.07 to 3.51 mm), respectively. The correlation betweenthe variation in JSW and the variation in IMD is shown in Fig. 4.Although variation in JSW was not significantly related to age, sexor BMI, sex was retained in the basis model because the variationin JSW was 0.10 ± 0.07 mm larger in men than in women (P = 0.09).The effect of IMD variation on the variation in JSW remained afteradjustment for sex, was highly significant (P < 0.0001), and indi-cated that a 1-mm variation in IMD was associated with a 0.10 mm(95% CI: 0.06–0.13) variation in JSW, with the variations in JSW andIMD both occurring in the same direction (i.e., increase or decrease).

4.3. Determination of the magnitude of variation in IMD“required” for accurate evaluation of JSN in paired radiographs

The influence of the magnitude of variation between IMDBL andIMD12mos on the variation in JSW was analyzed in various sub-groups. Because estimates that were made by taking into accountthe direction of the variation led to similar results, analyses wereperformed using the absolute value of the IMD variation. Table 1shows that with IMD variations less than 1.0 mm, JSW did not varysignificantly.

5. Discussion

Computerized measurement of minimum JSW in the medialtibiofemoral compartment of the knee is highly reliable [6] but,due to variations in knee repositioning, the measurement error in

186 C. Mercier et al. / Joint Bone Spine 80 (2013) 183–187

Table 1Mean difference between JSWBL and JSW12 mos (95%CI) (mm) according to variouscut-off values for the difference between IMDBL and IMD12 mos (mm) in non-arthriticKLG 0 knees.

IMD variation,mm

Number ofknees

Mean variation in JSW(95%CI), mm

P

< 0.5 68 −0.04 (−0.24–0.16) 0.64< 1 129 0.06 (−0.03–0.15) 0.16< 1.5 161 0.06 (0.00–0.12) 0.049< 2 190 0.07 (0.03–0.12) 0.0012

Ca

“flbipspitiToaa[

ttwJatvTtIit

dvBOtvsvatiwJdttcfe

Jbic

Fig. 4. The effect of variation in IMD on variation in JSW in paired FF radiographs. Thescatter plot represents the observed variations in JSW and IMD between baseline and

omparison of the measurement error of JSW due to IMD variation in normal, non-rthritic knees, with that of JSN determination in OA knees.

serial” radiographs of the same joint is large. The angle of kneeexion, knee rotation and alignment of the MTP with the X-rayeam all have been shown to be important in good reposition-

ng of the knee [14]. Protocols to standardize the radioanatomicositioning of the knee have been developed. They were shown touccessfully reproduce knee flexion and rotation but offer both aoor alignment and a poor reproduction of alignment. By analyz-

ng the present series of FF radiographs from the OAI we were ableo confirm that variations in knee flexion and rotation were min-mal and were unrelated to variation in JSW (data not reported).he present analysis of baseline and 12-month follow-up X-raysf radiographically normal knees of non-arthritic control subjectslso confirms that the FF protocol commonly results in poor MTPlignment and poor reproduction of alignment in serial radiographs6,8].

In serial radiographs of OA knees, variations in IMD were foundo interfere importantly with JSN determination. In a recent head-o-head comparison of the FF and Lyon schuss views of OA knees [6],e found an inverse correlation between JSN and IMD variation, i.e.,

SN decreased when IMD12mos was larger at follow-up than IMDBLnd JSN increased when IMD12mos was smaller than IMDbl. Notably,he correlation was seen only in paired FF radiographs in which theariation in IMD was much larger than that in the Lyon schuss view.his was confirmed in an analysis of a series of FF radiographs fromhe OAI Progression cohort, in which a 1-mm variation betweenMDbl and IMD12mos was associated with a 0.07-mm decrease orncrease in JSN, depending on the direction of change in IMD in thewo exams [8].

The study of asymptomatic knees that had no radiographic evi-ence of OA and in which, therefore, JSW was not expected toary over the 12-month follow-up period was of major importance.ecause baseline JSW and JSN both vary largely among individualA knees, the study permitted an accurate evaluation of the rela-

ionship between JSW and IMD and between variation in IMD andariation in JSW. The results confirmed that variations in IMD inerial radiographs of the same knee are strongly correlated withariations in JSW. They also indicated that the impact of IMD vari-tion on measurement of JSW in normal knees was greater thanhat we had found to exist in OA knees. In the normal knees exam-ned in the present study, a 1-mm variation in IMD was associated

ith a 0.10-mm (95% CI: 0.06–0.13) variation in JSW. Variations inSW and IMD both occurred in the same direction (i.e., increase orecrease). This explains why we previously observed in OA kneeshat the rate of JSN in OA knees was more rapid when the IMD inhe follow-up exam was smaller than that in the baseline exam and,ontrariwise, that the rate of JSN was slower when the IMD in theollow-up exam was larger than that in the corresponding baselinexam [6,9].

The cross-sectional study of the relationship between IMD and

SW in normal knees offers an explanation for the relationshipetween variation in IMD and variation in JSN in serial radiographs

nsofar as a strong positive relation was found between IMD andoncurrent JSW. A curvilinear relationship existed between JSW

follow-up, with each point representing a single knee. The smoothing curve (loessmethod) shows a linear relationship with more variability for extreme values.

and IMD, with a weak relationship for small values of IMD, but astrong and direct relationship for larger IMD values (Fig. 3).

Fig. 3 indicates that the relationship between JSW and concur-rent IMD became significant with an IMD approximating 2 mm. Aprecise threshold could not be determined but a highly significant(P < 0.0001) effect of IMD on JSW existed when IMD was less than1.7 mm, with a 1-mm increase in IMD leading to a 0.16 (95% CI:0.11–0.21) mm increase in JSW. The selected threshold (1.7 mm)for this analysis was made on the basis of a previous study in whichwe used a 1.7 mm IMD cut-off to differentiate OA knees with “good”alignment from those with “poor” alignment [8].

The relationship that we observed between variations in IMDand variations in JSW in serial radiographs of normal knees per-mitted the evaluation of the magnitude of the bias that was dueto variability in IMD in evaluation of JSN in OA joints. Consider-ing that a 1-mm variation in IMD in normal knees could result in0.10–0.13 mm variation in JSW and that IMD variation betweenpaired FF views could be as large as 4 mm (Figs. 3 and 4), the mea-surement error of JSW in serial radiographs of normal knees canbe as large as 0.40–0.52 mm, i.e., much larger than the 0.15 mmmean rate of JSN over 1 year we reported for KLG 2/3 OA kneesthat exhibited good MTP alignment at both baseline and follow-up(i.e., IMD at both exams ≤ 1.70 mm). We, therefore, more accuratelyrecalculated JSN in these OA knees, relative to that in nonarthriticcontrol KLG 0 knees and after adjustment for sex and for the varia-tion in IMD. The recalculated mean 12-month JSN in the OA kneeswas 0.16 mm (95% CI: 0.06–0.27). Thus, the error in measurementof JSW that can be caused by variation in IMD can approximateor exceed the magnitude of 12-month JSN in OA knees. Clearly,the magnitude of the artefactual modification of JSW that resultsfrom variations in IMD in serial radiographs may jeopardize theinterpretation of data on the relation between JSN and a variety ofindependent variables (risk factors for OA progression), e.g., age,BMI, periarticular muscle weakness, sex or treatment in an RCT ofan SMOAD.

Determination of the acceptable magnitude of the variationin IMD that does not induce statistically significant variation in

JSW was also of interest. Subgroup analysis of the normal kneesindicated that the largest IMD variation that did not significantlyaffect JSW variation was less than 1.0 mm (Table 1). Difficulty in

one Sp

ddiwe

rtaiibsiwcalgttyraradaydro

mtbMmd

A

m

m

r

t

t

D

c

[

[[

[[

[

[

C. Mercier et al. / Joint B

etermining the threshold in IMD variation more precisely wasue both to the sample size, which decreased as the threshold

ncreased (enlarging the CI), and to the variance, which increasedith large variations in IMD (i.e., with more points representing

xtreme IMD values far from the curve in Fig. 3).Although consistent achievement of good MTP alignment

equires radiographic protocols using fluoroscopy [3,4], the limita-ion imposed by fluoroscopy in a multicenter RCT is significant. Anlternative, however, might be the use of a non-fluoroscopic mod-fication of the conventional fluoroscopically-assisted LS protocol,n which an X-ray of the knee is obtained after modifications of theeam angle have resulted in acquisition of a radiograph showing amall IMD [15], which generally requires no more than two or threeterations. In a preliminary study, a mean IMD of 0.89 ± 0.44 mm

as achievable by this method, which affords sensitivity to JSNomparable to that achieved with the original fluoroscopically-ssisted LS protocol [15]. The radiation dose of a knee radiograph isess than 0.1 mSv. Therefore, the total dose of the two or three radio-raphs possible with the modified LS protocol largely falls withinhe restrictions of the International Commission on Radiation Pro-ection on annual radiation dosage of public exposure of 1.0 mSv perear. Also, a technique for determining the optimal angle of the X-ay beam from the initial non-fluoroscopic LS iteration, permitting

reduction in time and radiation exposure of the subject, has beeneported recently and requires further evaluation [16]. Anotherlternative could be the use of magnetic resonance imaging (MRI) toirectly measure cartilage volume and/or thickness. Use of MRI tossess tissues within the knee joint has greatly expanded in recentears for research purposes [17] but further research is needed toetermine which MRI parameter(s) might be more sensitive thanadiographic JSN is for evaluation of compounds for the indicationf structure modification in OA.

In conclusion, the present work demonstrated that measure-ent of knee JSW is largely dependent upon MTP alignment and

hat change in alignment in serial radiographs of OA knees coulde responsible of a large error in evaluation of JSN. Thus, goodTP alignment and replication of alignment in serial knee filmsust absolutely be achieved in any OA knee study for an accurate

etermination of JSN.

uthor contributions statement

C. Mercier performed statistical analysis and redaction ofanuscript.E. Vignon performed measurements and redaction of

anuscript.K. Brandt performed the study plan and redaction of manuscript.M. Pipermo performed measurements of IMD and JSW and

edaction of manuscript.M. Hochberg performed manuscript review for the OAI Publica-

ions Committee.M.P. Hellio Le Graverand performed the study plan and revised

he manuscript.

isclosure of interest

The authors declare that they have no conflicts of interest con-erning this article.

[

ine 80 (2013) 183–187 187

Acknowledgments

The OAI is a public-private partnership comprised of five con-tracts (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 ofHealth, a branch of the Department of Health and Human Ser-vices, and conducted by the OAI Study Investigators. Private fundingpartners include Pfizer, Inc.; Novartis Pharmaceuticals Corporation;Merck Research Laboratories; and GlaxoSmithKline. Private sectorfunding for the OAI is managed by the Foundation for the NationalInstitutes of Health. This manuscript has received the approval ofthe OAI Publications Committee based on a review of its scientificcontent and data interpretation.

References

[1] Hunter DJ, Zhang YQ, Tu X, et al. Change in joint space width: hyalinearticular cartilage loss or alteration in meniscus? Arthritis Rheum 2006;54:2488–95.

[2] Buckland-Wright JC, Macfarlane DG, Lynch JA, et al. Joint space width meas-ures cartilage thickness in osteoarthritis of the knee: high-resolution plainfilm and double contrast macroradiographic investigation. Ann Rheum Dis1995;54:63–8.

[3] Vignon E. Radiographic issues in imaging the progression of hip and kneeosteoarthritis. J Rheumatol 2004;31:36–44.

[4] Vignon E, Conrozier T, Hellio Le Graverand MP. Advances in radiographicimaging of progression of hip and knee osteoarthritis. J Rheumatol 2005;32:1143–5.

[5] Brandt KD, Mazzuca SA, Conrozier T, et al. Which is the best radiographic pro-tocol for a clinical trial of a structure modifying drug in patients with kneeosteoarthritis? J Rheum 2002;29:1308–20.

[6] Hellio Le Graverand MP, Vignon E, Brandt KD, et al. Head- to-head comparisonof the Lyon schuss and fixed flexion radiographic techniques. Long-term repro-ducibility in normal knees and sensitivity to change in osteoarthritic knees. AnnRheum Dis 2008;67:1562–7.

[7] Hellio Le Graverand MP, Mazzuca S, Lassere M, et al. Radiography work-ing group of the OARSI-OMERACT imaging workshop. Assessment ofthe radioanatomic positioning of the osteoarthritic knee in serial radio-graphs: Comparison of three acquisition techniques. Osteoarthritis Cartilage2006;14:37–43.

[8] Vignon E, Brandt K, Mercier C, et al. Alignment of the medial tibial plateauaffects the rate of joint space narrowing in the osteoarthritic knee. Osteoarthri-tis Cartilage 2010;18:1436–40.

[9] Mazzuca SA, Brandt KD, Dieppe PA, et al. Effect of alignment of the medialtibial plateau and X-ray beam on apparent progression of osteoarthritisin the standing anteroposterior knee radiograph. Arthritis Rheum 2001;44:1786–94.

10] Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. AnnRheum Dis 1957;16:494–502.

11] http://www.oai.ucsf.edu/.12] Peterfy C, Li J, Zaim S, et al. Comparison of fixed-flexion positioning with

fluoroscopic semi-flexed positioning for quantifying radiographic joint spacewidth in the knee: test-retest reproducibility. Skeletal Radio1 2003;32:128–32.

13] Pinheiro JC, Bates DM. Mixed-Effects Models in S and S-PLUS. Springer; 2000.14] Vignon E, Hellio Le Graverand MP. Update on radiographic imaging in

osteoarthritis of the hip and knee. In: Sharma L, Berenbaum F, editors.Osteoarthritis: a companion to rheumatology. Philadelphia: Mosby Elsevier;2007. p. 316–28.

15] Mazzuca SA, Hellio Le Graverand M-P, Vignon E, et al. Performance ofa non-fluoroscopically assisted substitute for the Lyon schuss knee radio-graph: quality and reproducibility of positioning and sensitivity to jointspace narrowing in osteoarthritic knees. Osteoarthritis Cartilage 2008;16:1555–9.

16] Hellio Le Graverand MP, Powell K, Vignon E, et al. Technique for determin-ing optimal Lyon schuss X-ray beam angle. Osteoarthritis Cartilage 2010;18:

S188.

17] Duryea J, Neumann G, Niu J, et al. Comparison of radiographic joint spacewidth with magnetic resonance imaging cartilage morphometry: analysis oflongitudinal data from the Osteoarthritis Initiative. Arthritis Care Res 2010;62:932–7.