Table 1

Synthesis of results for studies that estimate the correlation between radiographic damage at baseline and HAQ at the end of follow-up (type A comparison)

CitationVariables correlatedType of statistical analysisNo of patients for correlation (% baseline)Correlation score (year)p ValueAdjustments for confounders
Univariate analysis
Capell 199141§Modified Sharp – Lee functional indexSpearman's correlation58 (47.1%)0.352 (10 years)0.008*No
Van Zeben 199350Erosion score JSNSpearman's correlation132 (100%)Not reported>0.1No
Mottonen 199855LarsenSpearman's correlation130 (91.5%)0.01 (6 years)>0.05No
Jansen 200058Presence of erosionPearson's correlation111 (83.4%)−0.17 (1 year)>0.05No
Combe 200363Sharp/vdHSpearman's correlation156 (81.6%)0.167 (5 years)0.04*No
Erosion score0.186 (5 years)0.03*
Jansen 200365Sharp/vdHStudent's t test or chi-squared test282 (77.9%)Not relevant<0.05*No
Tengstrand 200468LarsenSpearman's correlation329 (39.0%)Not reported>0.05No
Zheng 200673MRI erosion scoreSpearman's correlation28 (66.6%)Not reported0.01 (MRI tendon)*No
Sharp – Hand and tendon function>0.2 (Sharp)
Multivariate analysis
van der Heijde 199246Sharp/vdHDiscriminant analysis142 (100%)Not relevantNSNo
Miyagi 199854Steinbrocker's stage – QoLFactor analysis45 (60.8%)Not relevantSig*No
Uhlig 200059Modified SharpBivariate regression149 (62.6%)Not relevant0.01*Yes (linear)
Linear regression>0.05
Landewe 200261Sharp/vdHGeneralized estimating equations148 (94.9%)Not relevant0.684No
Berglin 200362Modified LarsenSpearman's correlation43 (100%)0.385 (2 years)<0.05*Yes
Multivariate analysisNot relevant<0.05*
Benton 200466MRI modified SharpSpearman's correlation34 (81.0%)Not reported0.01*No
Logistic regression model0.5
Breedveld 200538Sharp/vdHRegression model using stepwise model selection428 (100%)Not relevant<0.001*No
Aletaha 200637§Sharp and Sharp/vdHANOVA testing for linear trend270 (91.5%)Not relevant<0.001*Yes
Linear regression model
Bansback 200669LarsenMultivariate logistic regression705 (71.6%)Not relevant0.2No
Welsing 200660Modified SharpMixed linear regression model185 (100%)Not relevant<0.001*Yes
van der Heijde 200875Sharp/vdHGeneralised mixed linear model622 (90.7%)Not relevant<0.001*Yes
  • In the column named ‘variables correlated’ the measure of damage is always reported while the measure of disability is reported only when different from HAQ (in italics).

  • In the column named ‘correlation score’ the year in brackets represents the end of follow-up (when disability was measured and correlated to baseline damage).

  • * Statistically significant correlation at p<0.05 or when the authors stated that the association was significant (sig) but did not provide a p value;

  • <1 year disease duration;

  • ‡1 –3 years disease duration;

  • § 3–6 years disease duration;

  • >6 years disease duration;

  • see table 2 for details on disease duration.

  • ANOVA, analysis of variance; HAQ, health assessment questionnaire; NS, not significant; sig, significant; vdH, van der Heide.