Downloaded from ard.bmj.com on August 21, 2014 - Published by group.bmj.com
ARD Online First, published on January 15, 2014 as 10.1136/annrheumdis-2013-204277 Clinical and epidemiological research
EXTENDED REPORT
Predicting the severity of joint damage in rheumatoid arthritis; the contribution of genetic factors Hanna W van Steenbergen,1 Roula Tsonaka,2 Tom WJ Huizinga,1 Saskia le Cessie,2,3 Annette HM van der Helm-van Mil1 Handling editor Tore K Kvien ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ annrheumdis-2013-204277). 1
Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands 2 Department of Medical Statistics, Leiden University Medical Center, Leiden, The Netherlands 3 Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, The Netherlands Correspondence to HW van Steenbergen, Department of Rheumatology, Leiden University Medical Center, P.O. Box 9600, Leiden 2300 RC, The Netherlands; [email protected] Received 12 July 2013 Revised 4 December 2013 Accepted 21 December 2013
To cite: van Steenbergen HW, Tsonaka R, Huizinga TWJ, et al. Ann Rheum Dis Published Online First: [please include Day Month Year] doi:10.1136/ annrheumdis-2013-204277
ABSTRACT Background The severity of radiologic progression is variable between rheumatoid arthritis (RA) patients. Recently, several genetic severity variants have been identified and were replicated, these belong to 12 loci. This study determined the contribution of the identified genetic factors to the explained variance in radiologic progression and whether genetic factors, in addition to traditional risk factors, improve the accuracy of predicting the severity of radiologic progression. Methods 426 early RA patients with yearly radiologic follow-up were studied. The main outcome measure was the progression in Sharp-van der Heijde score (SHS) over 6 years, assessed as continuous outcome or categorised in no/little, moderate or severe progression. Assessed were improved fit of a linear mixed model analysis on serial radiographs, R2 using linear regression analyses, C-statistic and the net proportion of patients that was additionally correctly classified when adding genetic risk factors to a model consisting of traditional risk factors. Results The genetic factors together explained 12–18%. When added to a model including traditional factors and treatment effects, the genetic factors additionally explained 3–7% of the variance ( p value R2change=0.056). The percentage of patients that was correctly classified increased from 56% to 62%; the net proportion of correct reclassifications 6% (95% CI 3 to 10%). The C-statistic increased from 0.78 to 0.82. Sensitivity analyses using imputation of missing radiographs yielded comparable results. Conclusions Genetic risk factors together explained 12–18% of the variance in radiologic progression. Adding genetic factors improved the predictive accuracy, but 38% of the patients were still incorrectly classified, limiting the value for use in clinical practice.
expression.3–6 Here, we aimed to explore the relevance of currently known genetic risk factors with regards to (1) explaining the interindividual variance in radiologic progression and (2) improving the accuracy of predicting radiologic progression for individual patients. Known traditional risk factors explain about one-third of the variance in joint damage after 5 years of disease; the majority of these risk factors were related to patient characteristics (age, gender), inflammation (acute phase reactants, swollen joint counts) and the presence of auto-antibodies.7 The contribution of the genetic risk factors to the explained variance has not been explored. Prediction of RA-severity on the level of individual patients is not yet accurate. Several matrices to predict rapid radiologic progression have been derived, consisting of three or four risk factors. Most of these matrices are not validated in the general RA population, and failed to correctly classify ∼50% of patients. In particular, the patients who developed progressive disease were not recognised.8–13 Consequently, the value of these matrices for clinical practice is still limited. Whether the addition of genetic factors improves prediction is unknown. This study examined the variance in joint damage progression explained by recently identified genetic risk factors and their value in improving the prediction of the severity of joint damage progression. We assessed traditional performance measures of prediction models and the net proportion of RA-patients that is additionally correctly classified when adding risk factors to a prediction model consisting of known risk factors.
INTRODUCTION
PATIENTS AND METHODS Patients
The severity of rheumatoid arthritis (RA) is commonly expressed by the extent of damage of hand and feet joints. Joint damage can be measured objectively with validated scoring methods and is associated with long-term functional disability.1 The severity is highly variable between patients; many patients show mild progression and few severe progression. The processes underlying these differences are partly understood. The observation that the heritability of radiologic progression is 45– 58%2 underlined the notion that genetic factors play a role. Presently, several genetic risk factors for radiologic progression have been identified and replicated. Some of these variants were also associated with differences in mRNA or protein
Between 1993 and 2006, 600 RA patients (1987-ACR-criteria) were included in the Leiden Early Arthritis Clinic (EAC).7 Inclusion in the EAC took place when arthritis was confirmed at physical examination and symptom duration was less than 2 years. At first visit, patients and rheumatologists filled questionnaires, 66-swollen and 68-tender joint counts were performed (66-SJC and 68-TJC14), and blood samples were taken. Patients were followed yearly. The initial treatment strategy differed for different inclusion periods: patients included in 1993–1995 were initially treated with NSAIDs, patients included in 1996–1998 were initially treated with hydroxychloroquine or sulphasalazine, and patients included since 1999 were
van Steenbergen HW, author et al. Ann Rheum Dis 2014;0:1–7. doi:10.1136/annrheumdis-2013-204277 1 Copyright Article (or their employer) 2014. Produced by BMJ Publishing Group Ltd (& EULAR) under licence.
Downloaded from ard.bmj.com on August 21, 2014 - Published by group.bmj.com
Clinical and epidemiological research promptly treated with methotrexate. The severity of radiologic progression differed for these three treatment groups; therefore, treatment effects were incorporated in the analyses. The traditional risk factors studied were age, gender, symptom duration at first visit, localisation initial joint symptoms, 66-SJC, presence of anti-citrullinated-peptide-antibodies (ACPA), presence of rheumatoid factor (RF) and erythrocyte sedimentation rate (ESR).
Selection of genetic risk factors and genotyping We selected single nucleotide polymorphisms (SNPs) using the following criteria: the SNP was studied in relation to the severity of radiologic progression in several cohorts and the association was independently replicated or found significant in a meta-analysis including all published data. Based on these criteria, we came to a selection of genetic variants that is presented in table 1. Notably, rs4810485 in CD40 and rs7607479 in SPAG16 were identified as risk factors for radiologic progression only in ACPA-positive RA. Genotypings in the EAC were done with allele-specific kinetic PCR analysis,15 Illumina Golden Gate platform,3 4 16 17 Illumina Immunochip,5 18 Sequenom iPLEX6 and LightSnp (Roche).19 Quality control of genotyping was performed as described previously.3–6 15–19 426 patients had complete data on all evaluated traditional and genetic risk factors (figure 1).
Radiologic outcome
Figure 1 Flow chart of patient selection. RA, rheumatoid arthritis; EAC, early arthritis clinic; RMA, repeated measurement analysis; R2, proportion of explained variance. Baseline characteristics of the included (n=426) and excluded patients (n=174) were not different (data not shown). The patients with follow-up until six years (n=239) were younger compared to the patients without complete follow-up until six years (n=187) (mean (SD) 53.9 (14.5) versus 60.0 (15.7) years, p30 units, indicating no/little, moderate and severe radiologic progression (figure 3A). The first cut-off was chosen as progression of ≤1
Table 1 Genetic variants studied, and the R2 of each variant for radiologic progression over six years Genetic variant (risk allele)
Located in/nearby gene(s)
Chr.
MAF*
Tested model*
R2 ΔSHS0–6 years (%) in RA (n=239)
R2 ΔSHS0–6 years (%) in ACPA-pos RA (n=144)
SE (28) rs4810485 (T) (15) rs7667746 (G) (16) rs7665842 (G) (16) rs4371699 (A) (16) rs6821171 (C) (16) rs1896368 (G) (4) rs1896367 (A) (4) rs1528873 (A) (4) rs2104286 (C) (18) rs8192916 (A) (3) rs1119132 (A) (17) rs7607479 (C) (6) rs26232 (T) (19) rs11908352 (A) (5) rs451066 (A) (5) rs1485305 (T) (Knevel, 2013, unpublished)
HLA-DRB1 CD40 IL-15 IL-15 IL-15 IL-15 DKK-1 DKK-1 DKK-1 IL2RA GRZB IL-4R SPAG16 C5orf30 MMP-9 rs1465788 OPG
6 20 4 4 4 4 10 10 10 10 14 16 2 5 20 14 8
0.39 0.24 0.33 0.40 0.19 0.29 0.47 0.41 0.47 0.24 0.42 0.13 0.33 0.29 0.21 0.20 0.44
Add Rec Rec Rec Rec Rec Add Add Add Add Rec Rec Add Add Add Add Add
4.0 0.1 2.6 2.7 0.3 0.1 0.3 0.4 2.1 0.3 0.8 0.5 0.6 0.3 4.7 1.1 1.4
ARD Online First, published on January 15, 2014 as 10.1136/annrheumdis-2013-204277 Clinical and epidemiological research
EXTENDED REPORT
Predicting the severity of joint damage in rheumatoid arthritis; the contribution of genetic factors Hanna W van Steenbergen,1 Roula Tsonaka,2 Tom WJ Huizinga,1 Saskia le Cessie,2,3 Annette HM van der Helm-van Mil1 Handling editor Tore K Kvien ▸ Additional material is published online only. To view please visit the journal online (http://dx.doi.org/10.1136/ annrheumdis-2013-204277). 1
Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands 2 Department of Medical Statistics, Leiden University Medical Center, Leiden, The Netherlands 3 Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, The Netherlands Correspondence to HW van Steenbergen, Department of Rheumatology, Leiden University Medical Center, P.O. Box 9600, Leiden 2300 RC, The Netherlands; [email protected] Received 12 July 2013 Revised 4 December 2013 Accepted 21 December 2013
To cite: van Steenbergen HW, Tsonaka R, Huizinga TWJ, et al. Ann Rheum Dis Published Online First: [please include Day Month Year] doi:10.1136/ annrheumdis-2013-204277
ABSTRACT Background The severity of radiologic progression is variable between rheumatoid arthritis (RA) patients. Recently, several genetic severity variants have been identified and were replicated, these belong to 12 loci. This study determined the contribution of the identified genetic factors to the explained variance in radiologic progression and whether genetic factors, in addition to traditional risk factors, improve the accuracy of predicting the severity of radiologic progression. Methods 426 early RA patients with yearly radiologic follow-up were studied. The main outcome measure was the progression in Sharp-van der Heijde score (SHS) over 6 years, assessed as continuous outcome or categorised in no/little, moderate or severe progression. Assessed were improved fit of a linear mixed model analysis on serial radiographs, R2 using linear regression analyses, C-statistic and the net proportion of patients that was additionally correctly classified when adding genetic risk factors to a model consisting of traditional risk factors. Results The genetic factors together explained 12–18%. When added to a model including traditional factors and treatment effects, the genetic factors additionally explained 3–7% of the variance ( p value R2change=0.056). The percentage of patients that was correctly classified increased from 56% to 62%; the net proportion of correct reclassifications 6% (95% CI 3 to 10%). The C-statistic increased from 0.78 to 0.82. Sensitivity analyses using imputation of missing radiographs yielded comparable results. Conclusions Genetic risk factors together explained 12–18% of the variance in radiologic progression. Adding genetic factors improved the predictive accuracy, but 38% of the patients were still incorrectly classified, limiting the value for use in clinical practice.
expression.3–6 Here, we aimed to explore the relevance of currently known genetic risk factors with regards to (1) explaining the interindividual variance in radiologic progression and (2) improving the accuracy of predicting radiologic progression for individual patients. Known traditional risk factors explain about one-third of the variance in joint damage after 5 years of disease; the majority of these risk factors were related to patient characteristics (age, gender), inflammation (acute phase reactants, swollen joint counts) and the presence of auto-antibodies.7 The contribution of the genetic risk factors to the explained variance has not been explored. Prediction of RA-severity on the level of individual patients is not yet accurate. Several matrices to predict rapid radiologic progression have been derived, consisting of three or four risk factors. Most of these matrices are not validated in the general RA population, and failed to correctly classify ∼50% of patients. In particular, the patients who developed progressive disease were not recognised.8–13 Consequently, the value of these matrices for clinical practice is still limited. Whether the addition of genetic factors improves prediction is unknown. This study examined the variance in joint damage progression explained by recently identified genetic risk factors and their value in improving the prediction of the severity of joint damage progression. We assessed traditional performance measures of prediction models and the net proportion of RA-patients that is additionally correctly classified when adding risk factors to a prediction model consisting of known risk factors.
INTRODUCTION
PATIENTS AND METHODS Patients
The severity of rheumatoid arthritis (RA) is commonly expressed by the extent of damage of hand and feet joints. Joint damage can be measured objectively with validated scoring methods and is associated with long-term functional disability.1 The severity is highly variable between patients; many patients show mild progression and few severe progression. The processes underlying these differences are partly understood. The observation that the heritability of radiologic progression is 45– 58%2 underlined the notion that genetic factors play a role. Presently, several genetic risk factors for radiologic progression have been identified and replicated. Some of these variants were also associated with differences in mRNA or protein
Between 1993 and 2006, 600 RA patients (1987-ACR-criteria) were included in the Leiden Early Arthritis Clinic (EAC).7 Inclusion in the EAC took place when arthritis was confirmed at physical examination and symptom duration was less than 2 years. At first visit, patients and rheumatologists filled questionnaires, 66-swollen and 68-tender joint counts were performed (66-SJC and 68-TJC14), and blood samples were taken. Patients were followed yearly. The initial treatment strategy differed for different inclusion periods: patients included in 1993–1995 were initially treated with NSAIDs, patients included in 1996–1998 were initially treated with hydroxychloroquine or sulphasalazine, and patients included since 1999 were
van Steenbergen HW, author et al. Ann Rheum Dis 2014;0:1–7. doi:10.1136/annrheumdis-2013-204277 1 Copyright Article (or their employer) 2014. Produced by BMJ Publishing Group Ltd (& EULAR) under licence.
Downloaded from ard.bmj.com on August 21, 2014 - Published by group.bmj.com
Clinical and epidemiological research promptly treated with methotrexate. The severity of radiologic progression differed for these three treatment groups; therefore, treatment effects were incorporated in the analyses. The traditional risk factors studied were age, gender, symptom duration at first visit, localisation initial joint symptoms, 66-SJC, presence of anti-citrullinated-peptide-antibodies (ACPA), presence of rheumatoid factor (RF) and erythrocyte sedimentation rate (ESR).
Selection of genetic risk factors and genotyping We selected single nucleotide polymorphisms (SNPs) using the following criteria: the SNP was studied in relation to the severity of radiologic progression in several cohorts and the association was independently replicated or found significant in a meta-analysis including all published data. Based on these criteria, we came to a selection of genetic variants that is presented in table 1. Notably, rs4810485 in CD40 and rs7607479 in SPAG16 were identified as risk factors for radiologic progression only in ACPA-positive RA. Genotypings in the EAC were done with allele-specific kinetic PCR analysis,15 Illumina Golden Gate platform,3 4 16 17 Illumina Immunochip,5 18 Sequenom iPLEX6 and LightSnp (Roche).19 Quality control of genotyping was performed as described previously.3–6 15–19 426 patients had complete data on all evaluated traditional and genetic risk factors (figure 1).
Radiologic outcome
Figure 1 Flow chart of patient selection. RA, rheumatoid arthritis; EAC, early arthritis clinic; RMA, repeated measurement analysis; R2, proportion of explained variance. Baseline characteristics of the included (n=426) and excluded patients (n=174) were not different (data not shown). The patients with follow-up until six years (n=239) were younger compared to the patients without complete follow-up until six years (n=187) (mean (SD) 53.9 (14.5) versus 60.0 (15.7) years, p30 units, indicating no/little, moderate and severe radiologic progression (figure 3A). The first cut-off was chosen as progression of ≤1
Table 1 Genetic variants studied, and the R2 of each variant for radiologic progression over six years Genetic variant (risk allele)
Located in/nearby gene(s)
Chr.
MAF*
Tested model*
R2 ΔSHS0–6 years (%) in RA (n=239)
R2 ΔSHS0–6 years (%) in ACPA-pos RA (n=144)
SE (28) rs4810485 (T) (15) rs7667746 (G) (16) rs7665842 (G) (16) rs4371699 (A) (16) rs6821171 (C) (16) rs1896368 (G) (4) rs1896367 (A) (4) rs1528873 (A) (4) rs2104286 (C) (18) rs8192916 (A) (3) rs1119132 (A) (17) rs7607479 (C) (6) rs26232 (T) (19) rs11908352 (A) (5) rs451066 (A) (5) rs1485305 (T) (Knevel, 2013, unpublished)
HLA-DRB1 CD40 IL-15 IL-15 IL-15 IL-15 DKK-1 DKK-1 DKK-1 IL2RA GRZB IL-4R SPAG16 C5orf30 MMP-9 rs1465788 OPG
6 20 4 4 4 4 10 10 10 10 14 16 2 5 20 14 8
0.39 0.24 0.33 0.40 0.19 0.29 0.47 0.41 0.47 0.24 0.42 0.13 0.33 0.29 0.21 0.20 0.44
Add Rec Rec Rec Rec Rec Add Add Add Add Rec Rec Add Add Add Add Add
4.0 0.1 2.6 2.7 0.3 0.1 0.3 0.4 2.1 0.3 0.8 0.5 0.6 0.3 4.7 1.1 1.4
