Objectives To compare radiographic progression during treatment with disease-modifying antirheumatic drugs (DMARD) and subsequent treatment with tumour necrosis factor α inhibitors (TNF-I) in rheumatoid arthritis (RA) patients in clinical practice.
Methods Conventional radiographs (x-rays) of hands and wrists were obtained ∼2 years before start (prebaseline), at baseline and ∼2 years after start (follow-up) of TNF-I. Clinical data were obtained from the DANBIO registry and the patient files. x-Rays were scored blinded to chronology according to the Sharp/van der Heijde method. Annual radiographic progression rates during the DMARD (prebaseline to baseline x-ray) and TNF-I (baseline to follow-up x-ray) periods were calculated.
Results 517 RA patients (76% women, 80% IgM rheumatoid factor positive, 65% anticyclic citrullinated peptide positive, 40% current smokers, age 54 years (range 21–86), median disease duration 5 years (range 0–57)) were included. Patients were treated with infliximab (61%), etanercept (15%) or adalimumab (24%). During the DMARD period 85% of patients received methotrexate, 51% sulphasalazine and 78% prednisolone. The median DMARD period was 733 days (IQR 484–1002) and the median TNF-I period was 562 days (IQR 405–766). The median radiographic progression rate decreased from 0.7 (IQR 0–2.9) total Sharp score units/year (dTSS) in the DMARD period to 0 (0–0.9) units/year in the TNF-I period (p<0.0001, Wilcoxon). Corresponding mean dTSS values were 2.1 (SD 3.7) versus 0.7 (SD 2.3) units/year (p<0.0001, paired t test). 305 patients progressed (dTSS >0) in the DMARD period compared with 158 patients in the TNF-I period (p<0.0001, χ2).
Conclusion This nationwide observational study of RA patients documented significantly reduced radiographic progression during TNF-I treatment compared with the previous period of DMARD treatment.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Rheumatoid arthritis (RA) is a chronic inflammatory disease, characterised by progressive destruction of joint bone and cartilage resulting in severe disability, increased morbidity and mortality.1 ,2 Inhibitors of tumour necrosis factor α (TNF-I), a key cytokine in RA pathogenesis, have markedly improved the treatment options in RA patients who fail treatment with synthetic disease-modifying antirheumatic drugs (DMARD).
TNF-I have been investigated in controlled trials, in which they have been shown to reduce the signs and symptoms of RA and halt the progression of joint destruction measured on conventional radiographs (x-rays) in a large fraction of patients.3,–,5
However, patients participating in clinical trials are selected to have high disease activity, no significant comorbidities and high compliance. These patients are therefore not representative of the RA patients who are treated with TNF-I in clinical settings.6 ,7 Data from registries, reflecting routine care outside trials, have demonstrated that TNF-I effectively relieve the signs and symptoms of RA, although not all patients respond to treatment.8,–,10
A key aspect of efficacy assessment in RA is the impact of treatment on progression of structural joint damage assessed by x-rays. Only one study has compared the radiographic progression before and during TNF-I treatment in clinical practice.11 Important limitations of that study was that progression rates before TNF-I treatment were estimated not measured, and that the study only included 99 infliximab-treated patients with a baseline disease activity score in 28 joints (DAS28) greater than 5.1.
The aim of the present study was to investigate the effect of TNF-I on radiographic progression in DMARD-resistant RA patients treated in routine clinical practice by comparing radiographic progression rates before and during TNF-I treatment.
Patients and methods
This is a retrospective longitudinal study of the course of structural joint damage assessed by x-rays in RA patients registered in the nationwide Danish registry, DANBIO, before and after the initiation of treatment with TNF-I. DANBIO covers over 90% of RA patients treated with TNF-I. In Denmark, TNF-I therapy is initiated after failure of one or more synthetic DMARD and is administered in routine regimens with DMARD and/or prednisolone decided by the treating rheumatologist. It is recommended that RA patients registered in DANBIO should have x-rays of hands, wrists and feet performed on treatment initiation and yearly thereafter. Details of the DANBIO registry and the Danish treatment strategy have been published elsewhere.10 ,12
We included all RA patients in DANBIO who were TNF-I naive, started treatment with adalimumab, etanercept or infliximab before 1 July 2007 and had three relevant hand x-rays (prebaseline, baseline and follow-up). A baseline x-ray had to precede the initiation of TNF-I treatment by less than 3 months (0–3 months after the start of TNF-I was preferred), while the follow-up x-ray had to be obtained over 6 months after the baseline x-ray (preferably 2 years after TNF-I start). The prebaseline x-ray preceded both TNF-I start and baseline x-ray by more than 6 months (preferably 2 years before TNF-I initiation).
x-Rays were collected, digitised and anonymised. An experienced reader, blinded to patient identity and image sequence, read the x-rays according to the Sharp/van der Heijde method. Joint erosions were assessed at 15 sites in each hand (0–5 scale), while joint space narrowing was assessed at 16 sites in each hand (0–4 scale), and a total Sharp score (TSS) was calculated.13 Annual radiographic progression rates before baseline (DMARD period) were calculated by subtracting TSS at prebaseline x-ray from TSS at baseline x-ray and dividing the change in TSS with the number of days between the two x-rays and multiplying by 365 days. Similarly, we calculated annual radiographic progression rates between baseline x-ray and follow-up x-ray (TNF-I period).
Intraobserver reliability was estimated from re-evaluation of x-rays of 61 patients with a minimum of 6 months between evaluations, representing high and low baseline scores as well as high and low progression rates. The intraobserver intraclass correlation coefficient (one-way random effects model)14 for status scores at baseline was 0.95, while the intraclass correlation coefficient for TSS change in the TNF-I period was 0.34. The smallest detectable change (SDC) for TSS change in the TNF-I period was 4.9 TSS units/year.15
Data from the clinical visits (28 swollen joint count, 28 tender joint count, C-reactive protein; CRP) closest in time to the prebaseline x-ray (prebaseline visit) and follow-up x-ray (follow-up visit) were obtained from the DANBIO registry or the patient files, and DAS28 based on three variables (DAS28–CRP(3)) was calculated. For baseline x-ray the clinical visit (baseline visit) closest to the date of TNF-I initiation was used.
Patient files were reviewed and data on concomitant DMARD and glucocorticoid treatment (oral, intramuscular, intraarticular or intravenous) were registered. Administered glucocorticoids were converted into corresponding prednisolone dosages, based on the assumption that 3 mg betamethasone, 20 mg methylprednisolone and 20 mg triamcinolone are all equivalent to 25 mg prednisolone.
For each patient all CRP measurements were collected from the patient files and the time-averaged CRP level was calculated for the DMARD period (available in 363 patients, median number of measurements 19 (IQR 12–29)) and the TNF-I period (available in 370 patients, median number of measurements 16 (IQR 11–23)) to provide an estimate of the inflammatory burden during the two periods.16
Descriptive statistics for continuous variables are presented as medians with ranges or IQR in parentheses, categorical variables are presented as frequencies with percentages in parentheses. Annual radiographic progression rates were compared using parametric and non-parametric tests.
We performed sensitivity analyses on the 84 patients who started TNF-I treatment before 1 January 2003 (at which time adalimumab was marketed) in order to address the problem of confounding by chronology, and analyses of patients who had a disease duration of more than 10 years (n=160) versus less than 10 years to address the problem of channelling bias.17 These analyses gave similar results (data not shown).
Analyses were two-sided and p values less than 0.05 were considered significant. Analyses were performed using R version 2.9.0 (R Foundation for Statistical Computing, Vienna, Austria).18
Sample size considerations
We assumed a within-subject SD of TSS change of five units, based on analysis of our first 157 patients. A sample size of 199 patients would have 80% power to detect a difference of 1 TSS unit/year between the DMARD period and the TNF-I period with a paired t test and α=0.05.19
Two thousand five hundred and ninety-nine TNF-I-naive RA patients were identified in DANBIO, and 622 patients had three x-rays that fulfilled the selection criteria. It was technically possible to read all three x-rays in 573 patients. After review of the 573 patient files, 56 patients were excluded due to errors in the original database entry (erroneous diagnosis, erroneous registration of biological treatment, etc.). Demographic characteristics of the 517 included patients are presented in table 1.
The 2026 RA patients registered in DANBIO, from whom we were unable to obtain three relevant x-rays, were similar to the study population regarding sex, age, IgM rheumatoid factor and anticyclic citrullinated peptide positivity, number of previous DMARD, DAS28–CRP(3) and CRP levels at initiation of TNF-I treatment (all p>0.05, Mann–Whitney). The study population had a shorter disease duration (median 5 (IQR 1–13) versus 9 (IQR 4–17) years, p<0.001, Mann–Whitney).
The median interval between prebaseline and baseline x-rays (DMARD period) was 733 days (IQR 484–1002 days), and the most prevalent DMARD was methotrexate. A total of 456 (88%) patients received methotrexate at some time point during the period (time-averaged dose 14 mg/week) and 211 (41%) patients received parenteral methotrexate at least once. During the period, patients were treated with two (range one to six) different DMARD and 424 (82%) patients received glucocorticoids (time-averaged prednisolone dose 2.1 mg/day). From prebaseline visit to baseline visit (median interval 596 days (IQR 404–803)), the median DAS28–CRP(3) increased from 4.3 (IQR 3.0–5.3) to 5.0 (IQR 4.2–5.7) (p<0.0001, Wilcoxon). The mean time-averaged CRP for the period was 25 mg/l (SD 21).
The median interval between baseline and follow-up x-rays (TNF-I period) was 562 days (IQR 405–766 days). At baseline, treatment with infliximab (61%), etanercept (15%) or adalimumab (24%) was started. At follow-up x-ray, 310 (60%) patients were treated with their initial drug, while 150 patients (29%) had switched to another biological drug (137 of those switched to a different TNF-I) and 57 patients (11%) had withdrawn from biological treatment.
Methotrexate was the most prevalent DMARD (435 patients, 84%). Glucocorticoids were administered to 370 (72%) patients in a time-averaged dose of 1.5 mg/day prednisolone, which was lower than during the DMARD period (p<0.001, Wilcoxon). See table 2 for details on DMARD treatment in both periods, and table 3 for details on TNF-I treatment.
From baseline visit to follow-up visit (median interval 727 days (IQR 538–938)), the median DAS28–CRP(3) decreased from 5.0 (IQR 4.2–5.7) to 3.0 (IQR 2.0–3.9) (p<0.0001, Wilcoxon). The mean CRP level decreased from 31 to 15 mg/l (p<0.001, paired t test). Time-averaged CRP for the TNF-I period was significantly lower than for the DMARD period (15 mg/l vs 25 mg/l, p<0.001, paired t test).
Course of radiographic progression during DMARD treatment and subsequent TNF-I treatment
In the DMARD period, the median TSS increased from 7 (prebaseline x-ray) to 13 (baseline x-ray) (p=0.0005, Wilcoxon). At the end of the TNF-I period (follow-up x-ray) the median TSS was 14 (baseline TSS vs follow-up TSS, p=0.53, Wilcoxon) (figure 1). The annual rate of radiographic progression decreased from a median 0.7 (mean 2.1) TSS units/year in the DMARD period to 0 (0.7) TSS units/year in the TNF-I period (p<0.0001, Wilcoxon, paired t test). Significant decreases in progression rates in erosion and joint space narrowing scores were also found (table 4).
As shown in figure 2, a majority of patients (305, 59%) progressed radiographically (change in TSS >0) in the DMARD period, while 158 patients (31%) progressed in the TNF-I period (p<0.0001, χ2).
In the 158 patients who continued to progress during the TNF-I period the annual rate was a median 1.7 (mean 2.8) TSS units/year. In the DMARD period, the same patients progressed with a median 2.5 (mean 3.9) TSS units/year (p=0.007, Wilcoxon, p=0.006, paired t test), ie, the progression rate was significantly reduced.
Seventy-one patients had a high progression rate, (ie, >SDC 4.9 TSS units/year) in the DMARD period compared with 22 patients in the TNF-I period (p<0.0001, χ2).
Radiographic progression in the DMARD period was weakly correlated with progression in the TNF-I period (ρ=0.26, p<0.001, Spearman).
Clinical non-response to first TNF-I and radiographic progression
Radiographic progression in two subgroups of patients, who withdrew or switched from their first TNF-I due to lack/loss of effect or adverse events, was analysed separately and is presented in table 4. Demographic characteristics were similar in the two groups (p>0.05, Mann–Whitney) and a significantly lower rate of radiographic progression was found in the TNF-I period compared with the DMARD period in both groups.
Relation between radiographic progression and inflammatory activity in the two periods
In both periods, radiographic progression had a weak correlation with time-averaged CRP (DMARD period: ρ=0.18, p<0.001; TNF-I period: ρ=0.20, p<0.001, Spearman). In the DMARD period, the 184 patients with a time-averaged CRP below the median value (18 mg/l) progressed less (median 0.2 (IQR 0–1.8), mean 1.7 (SD 2.2) TSS units/year) than the 179 patients with a time-averaged CRP above the median value (median 1.2 (0–4.7), mean 2.7 (3.8) units/year) (p=0.002, Mann–Whitney). A similar result was found in the TNF-I period, in which the 182 patients with a time-averaged CRP below the median (10 mg/l) progressed less (median 0 (0–0), mean 0.3 (1.6)) than the 188 patients with a time-averaged CRP above the median (median 0 (0–1.3), mean 1.0 (2.6) TSS units/year) (p=0.001, Mann–Whitney).
Methotrexate dosage and radiographic progression
To investigate the effect of methotrexate treatment on radiographic progression, we subgrouped the patients into groups that had received less than 7.5 mg, 7.5–15 mg and more than 15 mg methotrexate per week (time-averaged dose). In the DMARD period there was a trend towards lower progression rates with increasing methotrexate dosage (medians 1 (IQR 0–3.3), 0.5 (IQR 0–2) and 0.4 (IQR 0–1.8)) (p=0.07, Kruskal–Wallis). Corresponding mean values were: 2.3 (SD 3.5), 2.1 (SD 3.7) and 1.9 (SD 4.2). In the TNF-I period no such trend was found (medians 0 (IQR 0–1.3), 0 (IQR 0–0.6) and 0 (IQR 0–0.7)). Mean values were 0.9 (SD 2.6), 0.6 (SD 2.1) and 0.8 (SD 2.5).
This is the first study to compare measured radiographic progression rates in individual patients before and during treatment with TNF-I in an observational nationwide cohort. Our main finding was that in patients with an insufficient response to synthetic DMARD, the rate of radiographic progression decreased markedly after the initiation of TNF-I treatment. Our findings were consistent in subanalyses of patients with varying disease durations and calendar years of treatment start. The cohort was representative of the entire DANBIO biological cohort regarding age, gender, IgM rheumatoid factor positivity and disease activity. A shorter disease duration in the study cohort can be ascribed to changing prescription practice.20 Consequently, the results can be considered representative of all TNF-I-treated patients with RA registered in DANBIO and thus of all Danish patients treated with TNF-I.
In randomised controlled trials (RCT) of patients failing synthetic DMARD, TNF-I have been shown to halt radiographic progression compared with methotrexate monotherapy, with mean progression rates during TNF-I treatment between −0.7 and 1.6 TSS units/year.3,–,5 In our observational cohort the mean radiographic progression rate after TNF-I initiation was 0.7 TSS units/year, suggesting that the benefits of TNF-I in patients treated in clinical practice corresponds to that reported in RCT. In a study from the Czech National Registry, infliximab slowed the annual radiographic progression rate from 8.56 (estimate based on disease duration) to 2.0. Compared with patients who receive TNF-I treatment in routine care, the 99 patients in the Czech study were highly selected with high disease activity (DAS28 >5.1) and extensive radiographic damage (mean TSS 90.1), limiting the generalisability of the results.11 A Swiss registry study21 included all patients who had received infliximab or etanercept for at least 10 months without major interruption (<4 months). The two TNF-I were equally effective in suppressing radiographic progression when given in combination with methotrexate. Due to the lack of knowledge on progression rates before TNF-I initiation and differences in scoring methods (Ratingen score for erosions and a computer-based method for joint space width assessment), the results cannot be compared with our findings.
Progression rates before study entrance have traditionally been estimated by dividing baseline TSS with disease duration.11 ,22 ,23 Large long-term cohort studies have documented that, on a group level, radiographic joint destruction in RA patients occurs at a linear rate from disease onset.24,–,27 Estimated prebaseline progression rate in our cohort was 1.8 TSS units/year (median), while the measured median progression rate was 0.7 TSS units/year. The use of scores based on direct evaluation of x-rays in this study strengthens the validity of our results.
The low rate of radiographic progression during DMARD treatment that we found reflects modern treatment strategies with aggressive use of DMARD aiming at remission. Most patients in our cohort received adequate doses of several DMARD and glucocorticoids in the period before the initiation of TNF-I therapy, a strategy known to limit radiographic progression.28 ,29 Nevertheless, the treating physician concluded that treatment with DMARD was not sufficient and initiated TNF-I therapy. The reduction of radiographic progression during the subsequent TNF-I therapy confirms that disease control before TNF-I treatment was not optimal.
The retrospective observational design of our study has potential weaknesses. The lack of a control group introduces the possibility of selection bias, but by letting every patient serve as his/her own control, we have limited this problem as much as possible in a non-randomised study. The inclusion criteria that required three x-rays from specific date intervals may have resulted in the selection of a cohort of patients who had been monitored closer than the average RA patient. However, no differences in baseline characteristics known to influence radiographic progression were found between our cohort and the entire DANBIO biological cohort. The clinical data from the DANBIO registry were collected prospectively, which strengthens the data quality. Another quality of our study design was the‘intention-to-treat’ analysis. All patients were thus included in the analysis irrespective of the duration of their TNF-I treatment. This design reflects real life in which some patients pause, switch or withdraw from treatment.
It should be emphasised that the present study deals with patients with insufficient response to synthetic DMARD and that it therefore provides no evidence of an improved effect of TNF-I compared with synthetic DMARD in an unselected DMARD-naive population.
Due to the observational design of the study, no conclusions as to the cause of the decrease in radiographic progression rates can be drawn. The inflammatory load during TNF-I treatment was significantly lower than during DMARD treatment (as assessed by the time-averaged CRP), and DAS28–CRP(3) and CRP levels decreased significantly between the initiation of TNF-I treatment and the follow-up visit, supporting a hypothesis of reduced inflammation resulting in reduced joint damage. Conversely, time-averaged CRP only explained a fraction of radiographic progression both before and after the initiation of TNF-I therapy, thus indicating that TNF-I may also reduce joint damage through other mechanisms. Our data do not allow us to explore this further.
We have analysed the impact of TNF-I treatment as a class effect of the three TNF-I avaliable in 2007. The cohort was distributed unequally between the three drugs, leaving the study underpowered to compare the drugs. Future studies should aim at comparing the effectiveness of different TNF-I in suppressing radiographic joint destruction.
In conclusion, this nationwide observational study of RA patients with insufficient response to synthetic DMARD documented reduced radiographic progression during TNF-I treatment compared with the previous period of DMARD treatment. Progression rates during TNF-I treatment were similar to rates reported in RCT. Results were representative of the Danish RA population and robust in sensitivity analyses.
The DANBIO x-ray study group investigators: Hvidovre: Lykke Midtbøll Ørnbjerg, Merete Lund Hetland, Mikkel Østergaard, Anja Thormann; Århus: Ulrik Tarp; Gråsten: Uta Engling Poulsen; Vejle: Jakob Espesen; Aalborg: Vibeke Stevenius Ringsdal, Anette Schlemmer; Rigshospitalet: Niels Graudal; Frederiksberg: Gina Kollerup; Helsingør: Dorte Vendelbo Jensen; Gentofte: Ole Rintek Madsen, Annette Hansen; Holbæk: Bente Glintborg, Randi Pelck; Slagelse: Torben Grube Christensen; Odense: Hanne Lindegaard; Esbjerg: Ditte Dencker, Wolfgang Bøhme; Glostrup: Anne Rødgaard Andersen. The technical and statistical support from Niels Steen Krogh is highly appreciated. The DANBIO secretariat (Hanne Bagger Christiansen, Sandra Zbinden Pedersen, Cecilie Lindstrøm Egholm) are acknowledged for their assistance. The Departments of Radiology at the study hospitals are acknowledged for their help in collecting analogue and digitised radiographs.
Funding Danish Regions (ie, the hospital owners) gave financial support to DANBIO. Janssen Biologics (formerly Centocor) supported the present study with an unrestricted grant, while Abbott, Pfizer (formerly Wyeth) and MSD (formerly Schering-Plough) (since 2004), Bristol-Myers Squibb and Roche (since 2006), and UCB-Nordic (since 2007) have supported DANBIO with unrestricted grants. Janssen Biologics were allowed to comment on the work, but the authors had the full right to accept or refuse these comments. Except for this, the sponsors have had no influence on data collection, analysis or publication.
Competing interests MØ has received consulting fees, speaking fees or research grants from Abbott, Amgen, Bristol-Myers Squibb, Centocor/Janssen, Genmab, Glaxo-Smith-Kline, Mundipharma, Novo, Pfizer, Roche, Schering-Plough, UCB and Wyeth. UT has received consulting fees, speaking fees or research grants from Roche, Schering-Plough, Abbott and MSD. GK has received consulting fees, speaking fees or research grants from MSD. ORM has received consulting fees, speaking fees and research grants from Abbott, MSD, Pfizer, UCB, BMS, Amgen and Roche. HL has received consulting fees, speaking fees or research grants from Roche. AH has received consulting fees, speaking fees or research grants from Abbott, MSD and BMS. MLH has received consulting fees, speaking fees or research grants from Abbott, Bristol-Myers Squibb, Centocor/Janssen, Glaxo-Smith-Kline, MSD/Schering-Plough, Pfizer/Wyeth, Roche and UCB The remaining authors had no completing interests.
Ethics approval The study was based on data from the nationwide Danish DANBIO registry. DANBIO has been approved by The Danish Data Registry since the year 2000 (j.nr. 2007-58-0014 and j.nr. 2007-58-0006), and since October 2006 as a national quality registry by the National Board of Health (j.nr. 7-201-03-12/1). According to Danish law, informed consent and ethics approval were not required for the present study.
Provenance and peer review Not commissioned; externally peer reviewed.