SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Objective

Adipose tissue can secrete soluble mediators (adipokines) with potent immune regulatory functions. Some adipokines have been previously associated with radiographic damage in patients with rheumatoid arthritis (RA). In the present study, we investigated the capacity of baseline adipokine levels to predict radiographic progression over a period of 4 years and studied their contribution relative to that of other known risk factors, such as anti–cyclic citrullinated peptide (anti-CCP) antibodies.

Methods

Serum concentrations of leptin, visfatin, resistin, adiponectin, adipsin, tumor necrosis factor α (TNFα), and interleukin-6 (IL-6) were determined in serum samples obtained at baseline from 253 patients with RA from the Early Arthritis Cohort. The association between levels of these adipokines and radiographic progression was determined using a multivariate normal regression model correcting for age, sex, treatment strategy, body mass index (BMI), and the presence of anti-CCP antibodies.

Results

Levels of IL-6, TNFα, visfatin, and adiponectin were positively associated with radiographic progression over 4 years. This association was independent of BMI. However, only adiponectin levels remained significantly associated with radiographic progression when the model was corrected for the presence of anti-CCP antibodies, whereas a trend was observed for IL-6. The association of both TNFα and visfatin with radiographic damage disappeared after correction for the presence of anti-CCP antibodies, which is consistent with the fact that the levels of both cytokines correlated significantly with anti-CCP antibody levels in these patients.

Conclusion

Our results indicate that adipokines are predictors of radiographic progression in RA, possibly through distinct underlying biologic mechanisms.

Rheumatoid arthritis (RA) is a systemic autoimmune disease primarily affecting the joints. Chronic inflammatory processes lead to cartilage damage and bone erosions, resulting in destruction of the total joint architecture. The rate of joint destruction, however, is variable within the population of patients with RA, and the factors contributing to disease progression are still incompletely understood. More rapid progression of the disease is associated with markers of inflammation, such as the erythrocyte sedimentation rate (ESR), as well as the presence of anti–cyclic citrullinated peptide (anti-CCP) antibodies (1, 2).

Body mass index (BMI) was recently shown to be inversely correlated with radiographic progression in RA (3, 4). Although the underlying mechanisms are unclear, it is possible that the soluble mediators secreted by adipose tissue (adipokines) play a role in this process. Indeed, it has become increasingly evident over the last several years that adipose tissue not only is responsible for the storage of lipids but also can function as an endocrine organ that can secrete several soluble mediators with potent regulatory effects, influencing whole-body metabolism (5). These mediators have profound effects not only on glucose homeostasis and the regulation of food intake but also on inflammatory responses (6).

Among the adipokines known to be secreted by adipose tissue, tumor necrosis factor α (TNFα), interleukin-6 (IL-6), leptin, resistin, and visfatin are considered to be proinflammatory, whereas adiponectin has been described to have antiinflammatory as well as proinflammatory properties, depending on its molecular form (6). The concentrations of some of these adipokines, such as IL-6, have recently been shown to be elevated in the serum of patients with RA and to be correlated with inflammation markers such as the ESR, the C-reactive protein (CRP) level, and disease activity scores (7–13). Other adipokines, such as leptin, have a less clear association with disease markers. Some investigators have described serum leptin levels to be up-regulated in patients with RA, whereas others could not confirm this observation (8, 14, 15).

Although adipokines originally were thought to be secreted by adipose tissue, they were later described as being secreted by several cell types, including cells present in the joint (6, 16–18). The precise role of adipokines in joint physiology is unknown, but it is believed that they can have proinflammatory actions by inducing the release of proinflammatory cytokines and metalloproteinases from synovial fibroblasts, chondrocytes, and lymphocytes (19, 20). Adiponectin, for instance, has been shown to enhance IL-6 production in synovial fibroblasts, suggesting a proinflammatory property of this adipokine in the joint (19, 20).

It was recently reported that serum IL-6, visfatin, and adiponectin levels in patients with RA were positively associated with radiographic joint damage in cross-sectional studies (11–13, 21, 22). However, their relationship with radiographic progression is still unknown. In the present study, we investigated whether serum adipokine levels at baseline are able to predict radiographic damage over a period of 4 years independently of other potent predictors for progression, such as BMI and the presence of anti-CCP antibodies. Moreover, we investigated the predictive capacity of adipokines in anti-CCP antibody–positive and anti-CCP antibody–negative patients with RA.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Patients participating in the Leiden Early Arthritis Cohort (EAC) were included in this study (23). All of the patients fulfilled the American College of Rheumatology (ACR) 1987 revised criteria for the classification of RA (24) within the first year of followup (n = 253) and presented to the Leiden EAC between 1993 and 2002. Radiographs of the hands and feet were obtained at baseline and yearly thereafter and scored in chronologic order for erosions and joint space narrowing according to the Sharp/van der Heijde method (25). All radiographs were scored by one experienced scorer who was blinded with respect to the patient's autoantibody status, treatment, and clinical outcome. The intrareader variability described by the intraclass correlation coefficient was 0.97 for the radiographic progression rate.

Analysis of radiographic progression stratified for the year of inclusion led to the identification of 3 distinct groups of patients with different progression rates: patients included before 1996, patients included between 1996 and 1999, and patients included after 1999. These periods coincided with 3 different treatment strategies. Patients included before 1996 were treated initially with analgesics and subsequently with chloroquine or sulfasalazine if they had persistent active disease (delayed treatment). Between 1996 and 1998, patients with RA were treated promptly with either chloroquine or sulfasalazine, and from 1999 onward patients were promptly treated with either sulfasalazine or methotrexate (26). Based on these analyses, treatment strategy was considered a possible confounder and was corrected for in subsequent analyses, by modeling it into a categorical variable that can have 3 values, depending on the period in which inclusion of the patient took place.

Laboratory assessments.

Serum samples were obtained at baseline and stored at −80°C. None of the patients included in this study were being treated with antirheumatic drugs at the moment when serum samples were obtained. The ESR (mm/hour) and the serum CRP level (mg/liter) were measured at the Clinical and Chemical laboratory at the Leiden University Medical Center. Anti-CCP antibody levels were measured using enzyme-linked immunosorbent assays (Euro-Diagnostica), with seropositivity defined above a cut-off level of 25 units. The concentrations of serum adipokines (leptin [ng/ml], adiponectin [μg/ml], adipsin [μg/ml], visfatin [ng/ml], resistin [ng/ml], IL-6 [pg/ml], and TNFα [pg/ml]) were measured using the Bio-Plex Pro Human Diabetes kit, the Bio-Plex array reader, and Bio-Plex software (Bio-Rad), according to the manufacturers' instructions.

Statistical analysis.

The BMI (kg/m2) was normally distributed in the study population. Correlations between BMI, inflammation markers, and adipokine levels were calculated using Spearman's rank correlation test.

The association between adipokine levels at baseline and the rate of radiographic progression (ΔSvdH × years = SvdH year x – SvdH baseline, where SvdH = Sharp/van der Heijde score) over 4 years was tested using a multivariate normal regression model. This repeated measurements analysis takes advantage of within-patient correlations in serial radiography data and does not exclude patients with missing radiographs (27).

All analyses were adjusted for age, sex, and treatment strategy, as previously described (28). In addition to this basic prediction model (model 1), adjustments were also made for BMI and the presence of anti-CCP antibodies, as indicated (models 2–5). Adipokine levels, except for adiponectin and adipsin, as well as Sharp/van der Heijde scores, were logarithmically transformed to meet the assumptions of linear regression.

The estimates (i.e., effect size) resulting from the analyses were recalculated to reflect the relationship between radiographic progression and adipokine levels rather than the relationship between their log-transformed values. For all adipokines except adipsin and adiponectin, the depicted estimates represent the relative change in the rate of joint destruction (ΔSvdH) over 4 years when the level of the adipokine doubles. As an example, an estimate of 1.5 for the association of IL-6 with joint destruction means that the rate of joint destruction (ΔSvdH over 4 years) increases 1.5-fold when IL-6 levels double. For adipsin and adiponectin, the estimates represent the relative change in the rate of joint destruction (ΔSvdH) over 4 years when the adipokine level increases 50 μg/ml. For BMI, the estimate represents the relative change in the rate of joint destruction (ΔSvdH) over 4 years when the BMI increases 1 unit. Finally, for anti-CCP antibodies, the estimate represents the relative change in the rate of joint destruction (ΔSvdH) over 4 years when the patient is anti-CCP antibody positive.

The Z statistic was used to compare estimates between groups, and a Z score greater than or equal to 1.96 was considered significant.

SPSS version 17.0 was used to analyze the data. Bonferroni correction was performed when assessing the significance of the studied associations. Adjusted P values of ≤0.007, ≤0.013, or ≤0.05 were considered significant, as indicated in each case.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

Serum levels of soluble factors secreted by adipose tissue, collectively called adipokines, were measured in a cohort of 253 patients with RA. The characteristics of the patients, Sharp/van der Heijde scores, cytokine levels, and adipokine levels are shown in Table 1. All patients fulfilled the ACR 1987 revised criteria for the classification of RA within the first year of followup (24). BMI, the ESR, and the presence of anti-CCP antibodies were previously associated with radiographic progression (1–4). Therefore, we first studied their correlations with radiographic progression and baseline adipokine levels in our study population (Figure 1).

Table 1. Characteristics of the patients with rheumatoid arthritis*
  • *

    Except where indicated otherwise, values are the median (interquartile range). BMI = body mass index; anti-CCP = anti–cyclic citrullinated peptide; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; IL-6 = interleukin-6; TNFα = tumor necrosis factor α; SHS = Sharp/van der Heijde score.

Age, mean ± SD years56.1 ± 15
Female sex, %68.8
BMI, mean ± SD kg/m225.6 ± 3.5
Nonsmoker, %56.4
Anti-CCP antibody positive, %59
ESR, mm/hour37 (22−59)
CRP, mg/liter20 (8−45)
Cytokines 
 IL-6, pg/ml28.9 (14.7–56.4)
 TNFα, pg/ml5.4 (2.9–21.0)
Adipokines 
 Resistin, ng/ml1.7 (1.1–2.4)
 Visfatin, ng/ml10.0 (2.7−91.3)
 Leptin, ng/ml12.9 (4.6–47.4)
 Adipsin, μg/ml1.0 (0.8−1.7)
 Adiponectin, μg/ml28.2 (15.6−47.5)
Total SHS (range 0–448) 
 After 1 year3 (0–8.8)
 After 2 years5 (1–15)
 After 3 years7 (2–19.8)
 After 4 years9.5 (3−27.3)
thumbnail image

Figure 1. A, Correlations between adipokine levels, body mass index (BMI), and markers of inflammation. Values are Spearman's rank correlation coefficients. ∗ = P < 0.05; ∗∗ = P ≤ 0.001, by Spearman's rank correlation test. B, Differences in baseline adipokine levels, inflammation marker levels, and BMI between anti–cyclic citrullinated peptide (anti-CCP) antibody–positive and anti-CCP antibody–negative patients. Bars show the medians. IL-6 = interleukin-6; TNFα = tumor necrosis factor α; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; Conc = concentration.

Download figure to PowerPoint

Acute-phase reactants, BMI, anti-CCP antibodies, and radiographic progression.

First, we investigated whether the ESR, the CRP level, BMI, and the presence of anti-CCP antibodies were associated with radiographic progression, using a multivariate normal regression model. The presence of anti-CCP antibodies was significantly associated with an increased rate of progression over 4 years in our cohort (association estimate 2.21; P < 0.0001), whereas the ESR and the CRP level were not. We also observed a tendency toward an inverse association with BMI, although this did not reach significance (association estimate 0.97; P = 0.057).

Association of adipokine levels with BMI and inflammation at baseline.

Resistin, leptin, and adipsin correlated positively with BMI, whereas serum adiponectin levels correlated negatively with BMI (Figure 1A). As expected, IL-6 levels correlated significantly with the clinical markers of inflammation, the ESR and the CRP level. The other adipokines did not correlate with these acute-phase markers. Because IL-6 and TNFα are secreted by adipose tissue but also are often used as markers of inflammation, we studied their correlation with the other adipokines. Except for leptin, all other adipokines associated with at least 1 of the inflammatory cytokines (IL-6 or TNFα), indicating inflammation as a possible mechanism involved in the association between adipokines and radiographic progression.

Differences in baseline adipokine levels between anti-CCP antibody–positive and anti-CCP antibody–negative patients.

The presence of anti-CCP antibodies accounts for a phenotype of more progressive RA. Therefore, we investigated whether the presence of anti-CCP antibodies is associated with differences in adipokine levels at baseline. By comparing anti-CCP antibody–positive and anti-CCP antibody–negative patients, we showed that serum levels of TNFα (P < 0.0001) and visfatin (P < 0.0001) were significantly higher in anti-CCP antibody–positive patients (Figure 1B). No significant differences were observed in baseline levels of IL-6, resistin, leptin, adiponectin, adipsin, the CRP level, the ESR, or BMI between these groups (Figure 1B).

The fact that BMI and the presence of anti-CCP antibodies were correlated both with adipokine levels at baseline and radiographic progression over 4 years suggested that these factors could be involved in the association between adipokines and radiographic progression. Therefore, we corrected for these factors in further analyses, as indicated.

Adipokine levels and radiographic progression.

Next, we investigated whether baseline adipokine levels are able to predict radiographic progression over 4 years, using repeated measurements analyses in which corrections were made for age, sex, and treatment strategy (model 1; basic) in addition to BMI (model 2; basic plus BMI), cytokines/adipokines (model 3; basic plus BMI plus TNFα plus IL-6 plus visfatin plus adiponectin), and the presence of anti-CCP antibodies (models 4 and 5; basic plus anti-CCP antibodies and basic plus BMI and anti-CCP antibodies, respectively) (Table 2). Sharp/van der Heijde scores were logarithmically transformed to normalize their distribution. Likewise, adipokine levels, except for adiponectin and adipsin, were logarithmically transformed to meet the assumptions of linear regression. As shown in Table 2, levels of TNFα, IL-6, visfatin, and adiponectin were associated with radiographic progression over 4 years (model 1). Except for IL-6, this association remained when all 4 adipokines were modeled together (model 3). The individual association of these adipokines with progression was independent of BMI (model 2). However, only adiponectin remained positively associated with radiographic progression, whereas a trend was observed for IL-6 levels, when corrections were made for the presence of anti-CCP antibodies alone (model 4) or in addition to BMI (model 5).

Table 2. Association between baseline adipokine levels and the rate of radiographic progression over 4 years*
 EstimateInterval
Lower boundaryUpper boundaryP
  • *

    Repeated measurements analysis was performed, with Sharp/van der Heijde scores over 4 years as the outcome. The indicated adipokines were modeled individually (except for model 3, in which all adipokines were modeled together), and the analyses were corrected for age, sex, treatment strategy (model 1; basic model) and for body mass index (BMI) (model 2; basic plus BMI), anti–cyclic citrullinated peptide (anti-CCP) antibodies (model 4; basic plus anti-CCP antibodies), or both (model 5; basic plus BMI and anti-CCP antibodies). Estimates were calculated as described in Patients and Methods. TNFα = tumor necrosis factor α; IL-6 = interleukin-6.

  • After Bonferroni adjustment, P values ≤0.007 (model 1), ≤0.013 (models 2, 4, and 5), and ≤0.05 (model 3) were significant.

Model 1    
 TNFα1.091.041.14<0.001
 IL-61.101.031.160.003
 Visfatin1.031.011.05<0.001
 Resistin0.950.881.020.126
 Leptin1.041.001.080.046
 Adipsin3.980.011,0000.669
 Adiponectin1.281.101.490.002
Model 2    
 TNFα1.091.041.14<0.001
 IL-61.101.031.160.003
 Visfatin1.031.021.05<0.001
 Adiponectin1.261.091.470.003
Model 3    
 TNFα1.081.021.150.014
 IL-61.020.951.100.52
 Visfatin1.021.001.040.032
 Adiponectin1.301.111.530.001
Model 4    
 TNFα1.040.991.100.123
 IL-61.081.011.150.019
 Visfatin1.010.991.030.129
 Adiponectin1.241.061.450.008
Model 5    
 TNFα1.050.991.100.100
 IL-61.081.011.150.021
 Visfatin1.021.001.040.098
 Adiponectin1.221.051.440.012

Adiponectin remained significantly associated with radiographic progression when additional corrections were made for TNFα (association estimate 1.29; P = 0.002) or for IL-6 (estimate 1.24; P = 0.008) in model 5. Likewise, visfatin remained significantly associated with radiographic progression when additional corrections were made for IL-6 and TNFα in model 2 (estimate 1.03; P = 0.007). These data indicated that the association of adiponectin and visfatin with radiographic progression was, at least partially, independent of inflammatory cytokines.

Adipokine levels and radiographic progression in anti-CCP antibody–positive and anti-CCP antibody–negative patients.

To gain more insight into the association between IL-6 and adiponectin and anti-CCP antibodies, we studied the predictive value of these adipokines in anti-CCP antibody–positive and anti-CCP antibody–negative patients. Although the effect size of the association between IL-6 and radiographic progression changed only marginally upon stratification for the presence of anti-CCP antibodies, this association lost significance in both groups (for anti-CCP antibody–positive patients, association estimate 1.07 [P = 0.14]; for anti-CCP antibody–negative patients, association estimate 1.07 [P = 0.12]).

In contrast, adiponectin levels remained positively associated with radiographic progression only in the anti-CCP antibody–positive patient population (association estimate 1.34; P = 0.005). However, the difference between the effect sizes of this association in the anti-CCP antibody–positive and anti-CCP antibody–negative patient groups did not reach significance (z = 1.71, P = 0.09), indicating a possible lack of statistical power to observe a significant effect.

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

In this study, we investigated whether baseline serum levels of adipokines are able to predict radiographic progression in patients with early RA. Our results indicate that baseline levels of adiponectin can predict radiographic progression over a period of up to 4 years independently of the presence of anti-CCP antibodies and BMI. A similar trend was observed for IL-6. TNFα and visfatin can also predict radiographic progression independently of BMI, but their association was overruled by the presence of anti-CCP antibodies, indicating that anti-CCP antibodies might be in the causal path underlying these associations. Therefore, our data indicate that adipokines are associated with radiographic progression in RA, possibly through distinct biologic mechanisms.

Adiponectin is an adipokine with strong structural homologies to C1q and is mainly produced by adipocytes (29, 30). It has been described to have primarily antiinflammatory and antiatherogenic properties, and serum levels are generally inversely correlated with BMI (ρ = −0.153, P = 0.017 in our study) (31). Previous cross-sectional studies have shown that serum adiponectin levels are elevated in RA and correlate with disease activity, erosions, and joint space narrowing (12, 21). Our study showed that adiponectin is a predictive factor for radiographic progression, independently of BMI. More importantly, this association was also independent of anti-CCP antibodies, the strongest known predictors of RA progression, suggesting the involvement of antibody-independent pathways in radiographic damage (2). Although stratification for the presence of anti-CCP antibodies indicated that adiponectin remained significantly associated with radiographic progression only in anti-CCP antibody–positive patients, the effect size of this association was not significantly different from that in the anti-CCP antibody–negative group (P = 0.09), possibly because of a lack of statistical power. Therefore, we cannot conclude that adiponectin showed differential effects in anti-CCP antibody–positive and anti-CCP antibody–negative disease, but our data did show a clear association of adiponectin levels and radiographic progression in anti-CCP antibody–positive patients with RA.

The precise nature of the mechanisms involved in this association remains unclear. Despite several reports indicating antiinflammatory properties of adiponectin, some proinflammatory actions of adiponectin in the joint were also previously shown. For example, adiponectin is able to up-regulate IL-6 and matrix metalloproteinases in RA synovial fibroblasts (19, 20, 32) and can promote osteoclast formation in vitro (33). Although serum levels do not always reflect synovial fluid levels of adiponectin, these observations indicate antibody-independent mechanisms through which adiponectin could contribute to disease progression.

IL-6 is a pleiotropic cytokine with a wide range of biologic activities. Although IL-6 is known as an acute-phase protein intimately associated with inflammation, it is also abundantly secreted by adipose tissue. In fact, it is estimated that in healthy individuals, one-third of the circulating IL-6 levels are derived from adipose tissue (34). Although the precise origin of serum IL-6 in our cohort was unclear, IL-6 levels were correlated with acute-phase reactants (ESR and CRP level) and not with BMI, indicating IL-6 as a marker of the inflammatory processes in patients with RA, as suggested in previous studies in which a correlation between the levels of soluble IL-6 receptor and IL-6 with clinical disease activity was observed (9, 13, 35–37). In our study, the association between IL-6 and radiographic progression was independent of the ESR, the CRP level (data not shown), and BMI. Although the association of IL-6 with radiographic progression appeared to be at least partially dependent on the presence of anti-CCP antibodies, the involvement of antibody-independent pathways in this association cannot be excluded, because the effect size was largely unaffected by the correction for anti-CCP antibodies (an association estimate of 1.08 when a correction was made for anti-CCP antibodies versus an estimate of 1.10 without such a correction) (Table 2).

The biologic mechanisms that could be involved in the association of IL-6 with radiographic progression are unclear, because both protective and destructive effects of IL-6 on joint structures in vitro have been described (38–40). Furthermore, a possible underlying mechanism could be the effect of IL-6 on anti-CCP antibody–producing B cells, because IL-6 is a well-known growth factor for B cells and has been shown to play a role in mouse models of antibody-mediated arthritis (41, 42). The relative importance of IL-6–mediated antibody-dependent and antibody-independent mechanisms in radiographic damage in patients with RA remains to be further explored.

Visfatin, or pre-B cell colony–enhancing factor, is secreted by adipocytes but mostly by macrophages (6, 43). TNFα is a cytokine with a broad range of proinflammatory and immunostimulatory actions and has been shown to play a pivotal role in the pathogenesis of RA, because TNFα inhibitors have been shown to be very effective treatment strategies, in combination or as monotherapy, reducing signs and symptoms of disease (44–46). In our cohort, neither visfatin nor TNFα correlated with BMI (Figure 1A). However, both correlated with IL-6, indicating that these adipokines are markers of inflammation, as previously reported (11, 14, 15). Furthermore, their association with radiographic progression seemed to be entirely dependent on anti-CCP antibodies, indicating either a strong correlation between these adipokines and anti-CCP antibodies or that their predictive capacity is weaker than that of anti-CCP antibodies.

Although it is difficult to distinguish between these possibilities, the fact that both visfatin and TNFα levels were correlated with anti-CCP antibody levels in our cohort (for visfatin, ρ = 0.340, P < 0.001; for TNFα, ρ = 0.294, P < 0.001) would indicate that their association with radiographic progression could be in fact mediated by anti-CCP antibodies. Indeed, visfatin was previously reported to be involved in B cell development, which might explain the association with anti-CCP antibodies in our study (47). Likewise, TNFα release could be a direct result of the presence of anti-CCP antibodies. This is a tempting speculation, in view of previously published data showing that activation of macrophages by anti-CCP antibody–containing immune complexes can lead to TNFα release (48).

In conclusion, we observed that baseline levels of several adipokines were predictive of radiographic progression in RA. Although the precise mechanisms involved remain to be elucidated, our study indicated that distinct, antibody-dependent or antibody-independent mechanisms could be employed by different adipokines. Importantly, an antibody-independent association was shown for adiponectin, which suggested that this adipokine may potentially be a valuable prediction marker for radiographic joint damage and disease progression in patients with early RA.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Ioan-Facsinay had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ioan-Facsinay.

Acquisition of data. Klein-Wieringa, van der Linden, Kwekkeboom, van Beelen, van der Helm-van Mil.

Analysis and interpretation of data. Klein-Wieringa, van der Linden, Knevel, Huizinga, van der Helm-van Mil, Kloppenburg, Toes, Ioan-Facsinay.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES

We would like to thank Saskia le Cessie and Fina Kurreeman for their help with the statistical analyses.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. AUTHOR CONTRIBUTIONS
  7. Acknowledgements
  8. REFERENCES