Article Text

Download PDFPDF

The effectiveness and medication costs of three anti-tumour necrosis factor α agents in the treatment of rheumatoid arthritis from prospective clinical practice data
  1. W Kievit1,
  2. E M Adang2,
  3. J Fransen1,
  4. H H Kuper3,
  5. M A F J van de Laar3,
  6. T L Jansen4,
  7. C M A De Gendt5,
  8. D-J R A M De Rooij6,
  9. H L M Brus7,
  10. P C M Van Oijen8,
  11. P C L M Van Riel1
  1. 1
    Departments of Rheumatology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
  2. 2
    Department of Epidemiology, Biostatistics and Health Technology Assessment, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands
  3. 3
    Departments of Rheumatology, Medisch Spectrum Twente, Enschede, The Netherlands
  4. 4
    Departments of Rheumatology, Medical Centre Leeuwarden, Leeuwarden, The Netherlands
  5. 5
    Departments of Rheumatology, Rijnstate Hospital, Arnhem, The Netherlands
  6. 6
    Departments of Rheumatology, Sint Maartenskliniek, Nijmegen, The Netherlands
  7. 7
    Departments of Rheumatology, Twee Steden Hospital, Tilburg, The Netherlands
  8. 8
    Departments of Rheumatology, Jeroen Bosch Hospital, Den Bosch, The Netherlands
  1. W Kievit, Department of Rheumatology (470), Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands; w.kievit{at}


Aim: to evaluate the effects of adalimumab, etanercept and infliximab on disease activity, functional ability and quality of life and the medication costs in a naturalistic design.

Methods: All patients from the Dutch Rheumatoid Arthritis Monitoring (DREAM) register starting on tumour necrosis factor (TNF)α-blocking agents for the first time were monitored and assessed by trained research nurses every 3 months. The primary outcome was the Disease Activity Score (DAS28) course over the 12 months follow-up, analysed by linear mixed models. Secondary outcomes were the Health Assessment Questionnaire (HAQ), EuroQol five dimensions (EQ-5D) and the Short-Form 36 items (SF36) scores, and medication-related total costs.

Results: The DAS28 and SF-36 physical component scale decreased in all three medication groups over 12 months, but the decrease was larger for adalimumab and etanercept in comparison to infliximab (p<0.001). The analyses of the HAQ and the EQ-5D scores showed the same (non-significant) trend, namely that at 12 months, the functionality and quality of life was better for adalimumab and etanercept patients. With regard to the medication costs, infliximab treatment resulted in significantly higher costs over the follow-up period than treatments with either adalimumab or etanercept. The comparison between adalimumab and etanercept showed a significant difference in the 12-month DAS28 course (p = 0.031). There were no additional indications for differences in effectiveness or costs between adalimumab and etanercept.

Conclusion: The evaluation of the effectiveness and costs showed that adalimumab and etanercept are more or less equal and favourable compared to infliximab in the first year of treatment.

Statistics from

Request Permissions

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.

In 1997, new treatments were introduced for the management of rheumatoid arthritis (RA), which specifically target the cytokine tumour necrosis factor (TNF)α. These agents have revolutionised the treatment of RA. Adalimumab and infliximab are monoclonal antibodies, whereas etanercept is a soluble receptor blocking TNFα. Although the TNFα blocking agents differ in structure, they all block TNFα1 2 and their efficacy is considered to be comparable.3 Two meta-analyses compared the response percentages of the three TNFα blocking agents using data of published randomised controlled trials (RCT)4 5 and showed no difference in effectiveness. No head-to-head comparisons have been published to date.

However, the response observed in RCTs can be different from the response achieved in daily clinical practice.6 7 In daily clinical practice, patients generally are older, have more co-morbidities and lower disease activity. Furthermore, in daily practice there are more variations in dosing and co-medication, and less compliance to treatment. A clinical practice register is of value by providing insight in the true use of medication in daily practice, heterogeneity of the treated population and, ultimately, the related effectiveness and costs. Therefore, a comparison between anti-TNFα treatments in daily clinical practice provides additional information to meta-analyses of RCTs. Furthermore, the use of observational research data for the evaluation of the cost-effectiveness for reimbursement authorities is propagated.810 Consequently, the Dutch government funded a multi-centre, anti-TNFα register on which RA patients who started for the first time with one of the anti-TNFα agents have been included. These data were used to compare the effects of adalimumab, etanercept and infliximab on disease activity, functional ability, quality of life and medication costs in patients who failed at least two disease-modifying antirheumatic drugs (DMARDs).



From February 2003 until September 2004, 11 rheumatology centres in the Netherlands started to include all RA patients who started for the first time on one of the anti-TNFα agents in a prospective register (the Dutch Rheumatoid Arthritis Monitoring (DREAM) register). The inclusion of patients still continues to the present day. In the Netherlands, patients can start with an anti-TNFα treatment when they have at least moderate disease activity (Disease Activity Score (DAS28)⩾3.2) and failed on at least two DMARDs including methotrexate (MTX) at an optimal dose of 25 mg/week. To get reimbursement of anti-TNFα therapy, it is obligatory for RA patients to fulfil these criteria. Patients in this study were not randomised to a specific treatment.


Although dosing was at the discretion of the attending rheumatologist, generally patients started the treatment with the Dutch labelled doses: adalimumab at 40 mg per 2 weeks, etanercept at 25 mg twice weekly, and infliximab at 3 mg/kg every 8 weeks after a loading dose (at weeks 0, 2 and 6). The TNFα blocking agents could be administered in combination with a DMARD, corticosteroids or other treatment. All start and stop dates, doses, changes in doses and reasons for change were registered.

Outcome measures

The baseline visit was at the moment of treatment initiation with the first anti-TNFα agents. Patients’ characteristics such as age, sex, disease duration, rheumatoid factor, previous DMARDs used and the presence of one or more erosions in hands or feet were registered. A trained research nurse assessed the patients at baseline and subsequently every 3 months. If necessary, patients were seen more frequently by the attending rheumatologist. Assessments included disease activity, functional ability, quality of life and costs.

Disease activity was measured by DAS2811 score and response was measured by means of the European League Against Rheumatism (EULAR) response criteria.12 Functional ability was assessed by means of the Health Assessment Questionnaire (HAQ).13 14 Quality of life was measured by the EuroQol five dimensions (EQ-5D)15 utility score and the physical and the mental component scales16 of the Short-Form 36 (SF-36).17

Because the medication costs were considered the cost-drivers, only medication related resources were included in the cost analysis. Cost of anti-TNFα, DMARDs, corticosteroid or NSAIDs were calculated by multiplying the true doses by the price per mg of the particular drugs. The prices per labelled dose of three drugs were based on the tariff lists (September 2006) published by the Dutch Health Care Insurance Board. A real cost price was calculated for the in-patient clinic administration of infliximab on basis of the involved personnel, materials and devices. Cost for overhead (35%) and housing (10%) were added to the total cost-price of the in-patient clinic administration according to the Dutch guidelines for cost-analyses in health care.18


Complete data of the first 12 months after treatment initiation with a TNF blocking agent were analysed on an intention-to-treat (ITT) basis. This means that all patients were analysed in the group of medication on which they initially started, regardless of whether they received or adhered that treatment for the full 12 months. Such an ITT was possible to perform because data collection was continued when patients had stopped using their initial anti-TNFα.

Because we were interested in the evaluation of the whole 12 months after starting with the first anti-TNFα agent, the primary outcome was the course of the DAS28 over the 12 months follow-up. This was analysed by means of a linear mixed model. After creating an unconditional growth model that corrected for the repeated observations within patients using the first order autoregressive correlation structure, the factor “medication” was included as a fixed effect. Next, a random component was added to the medication effect to adjust for possible sampling variation due to differences at baseline and in the course between the medication groups. Confounders could be added to the model if operational. This approach was repeated for the secondary outcomes.

For descriptive purposes, changes in DAS28 and HAQ from baseline and the EULAR response percentages were analysed at every timepoint. For this analysis, moderate and good EULAR responders were added and counted as responders. Possible differences in baseline characteristics between the three agents were statistically tested by one-way analysis of variance (ANOVA) procedures, or by non-parametric Kruskall–Wallis tests in case of continuous outcomes and by χ2-square tests for dichotomous outcomes. To give insight in the duration of drug use, a Kaplan–Meijer survival analysis was performed and the medication use after withdrawl of the initially-started anti-TNFα agents was described. The difference in drug survival curves was tested by means of a log-rank test.

Differences in prognostic factors for the treatment effect were considered as possible confounders. A propensity score was estimated on the basis of the baseline values: DAS28, HAQ, age, disease duration, rheumatoid factor, sex, number of previous DMARDs and the presence of at least one erosion. However, there were no factors found that could predict treatment.

In case of missing values for the DAS28 caused by a missing value for the ESR, the ESR was imputed by means of linear multivariate regression using the values of the SW28 (square root of number of swollen joints from 28 joints), 28-joint-count for tenderness (TEN28) and the visual analogue scale assessment of general health (VAS-GH) of the patient.

All analyses were performed using SPSS V14.0 (SPSS Inc., Chicago, Illinois, USA).


Baseline values

Between February 2003 and August 2007, 916 patients were included. In total, 707 patients had at least 1-year follow-up and fully accessible data at time of analysis (August 2007). In the first year of follow-up, 33 (4.7%) patients were no longer able or willing to participate in the data collection. Data of these patients were used for analysis until the moment of exclusion. Concerning the included patients, 267 (38%) patients started with adalimumab, 289 (41%) with etanercept and 151 (21%) patients with infliximab.

Table 1 shows the baseline characteristics of the patients in the three drug groups. Patients starting with anti-TNFα had a severe disease activity and relative long disease duration. There were no significant differences between the three groups of patients on all measured baseline characteristics, except for the percentage of patient with positive rheumatoid factor. However, rheumatoid factor was not considered as a confounder, since it was not associated with response on treatment (p = 0.626).

Table 1 Baseline characteristics in the three drug groups


Figure 1 shows drug survival curves for the three medication groups. Within the follow-up timeframe of 12 months, 22%, 21% and 31% of the patients discontinued the treatment with adalimumab, etanercept or infliximab, respectively. This percentage was significantly higher for infliximab patients compared to adalimumab patients (p = 0.049) and etanercept patients (p = 0.024). The occurrence of adverse events was the reason for discontinuation in 48% of the patients, in 33% the reason was a lack of efficacy, and in 18% there was another unspecified reason; there were no differences between the three groups of patients. After discontinuation of the initial anti-TNFα treatment, 42% of adalimumab and etanercept users switched to another anti-TNFα treatment, and 67% of infliximab switched. The same anti-TNFα treatment was re-challenged in 8%, 15% and in 2%, respectively. Another option was to start with one of the conventional DMARD therapies, which occurred in 27% of the adalimumab patients, 25% of the etanercept patients and 20% of the infliximab patients. The remaining patients did not start with another consecutive treatment after stopping with the initial anti-TNFα treatment, but they remained on their co-medication.

Figure 1 drug survival in the three treatment groups.

Patients starting with adalimumab and etanercept did so at the registered dose in 97% and 98% of the cases, respectively. Infliximab patients, however, started in 80% of the cases with an average dose of 3 mg/kg per 8 weeks, in 18% of the cases with an average dose of 4 mg/kg per 8 weeks and 2% started with higher dosages. All patients who started with the average dose of 4 mg/kg dose did so because, according to local standards, the number of vials (of 100 mg each) were rounded up when the patient needed more than 220 ml (that is, a weight higher than 73 kg) in order to prevent wastage.

Within 12 months, 7.1% and 4.8% of adalimumab and etanercept patients had a dose increase, whereas 31.1% of infliximab patients had a dose increase. The median of the 12-month averaged dose was 40 mg per 2 weeks for adalimumab, 50 mg per week for etanercept and 3.3 mg/kg per 8 weeks for infliximab.

With regard to co-medication (table 2), 78% to 87% (p = 0.02) of the patients used at least one DMARD as concomitant medication. This was in 79% to 87% MTX, in favour of adalimumab. A third of the patients used oral prednisone at baseline, with the higher percentages in the adalimumab group. Patients treated with adalimumab could decrease the DMARD dose (p = 0.039) significantly more often than infliximab patients. In total, 12–19% of the patients received an intramuscular or intra-articular prednisone injection during the first 12 months, which was not significantly different between the three groups (p = 0.076).

Table 2 Co-medication used and changes within 12 months

Primary and secondary outcomes

Figure 2 shows mean values for the DAS28, the HAQ, the EQ-5D and the SF-36 at different timepoints. The repeated measures analyses (linear mixed model) showed significant differences in the DAS28 course over 12 months between infliximab and both adalimumab and etanercept patients (p<0.001) and between adalimumab and etanercept patients (p = 0.031). The figures for the HAQ and the EQ-5D show the same trend: the functionality and quality of life was best for adalimumab patients. However, the repeated measures analyses did not show any significant differences.

Figure 2 The mean score and 95% confidence intervals at every follow-up moment after start with anti-tumour necrosis factor (TNF)α for the Disease Activity Score (DAS) (A), Health Assessment Questionnaire (HAQ) (B), EuroQol five dimensions (EQ-5D) utility score (C) and Short-Form 36 (SF-36) physical component scale (PCS). *Significant difference (<0.05) compared with infliximab patients. †Significant difference (<0.05) compared to infliximab patients. Missing values at baseline, 3, 6, 9, 12 months, respectively: DAS28, 0.5%, 3.4%, 6.6%, 7.9%, 5.9%; HAQ, 7.7%, 6.3%, 9.1%, 10.9%, 6.8%; EQ5D, 15.0%, 18.4%, 23.6%, 24.7%, 28.6%; SF-36 PCS, 20.2%, 21.8%, 23.6%, 24.7%, 31.3%.

Concerning the SF-36, patients did have mental component scores comparable to the US population norm without changes, and therefore only the physical component scale (PCS) score is presented. The mean baseline score on the PCS was 30, which is 2 standard deviations worse than the US population norm.16 Adalimumab and etanercept patients improved after baseline and the course over 12 months was significantly better (p = 0.001) than the course of infliximab patients, as tested by the repeated measures analyses.

Table 3 shows the outcomes with regard to the improvements on the DAS28 and the HAQ and the percentage of patients with a EULAR response. Adalimumab and etanercept patients had significant larger improvements on the DAS28 than infliximab patients on all measurement points. For the changes on HAQ, a significant difference was noticed when adalimumab was compared with infliximab. The percentage of patients with a response on infliximab was significantly smaller on all timepoints than the percentage of patients with a response on adalimumab and etanercept. Occurrence of missing values was not different for the three medication groups.

Table 3 absolute differences in Disease Activity Score (DAS)28 and Health Assessment Questionnaire (HAQ) results from baseline and the European League Against Rheumatism (EULAR) response percentages (good and moderate added) at 3, 6, 9 and 12 months

As mentioned before, a group of patients (n = 26) was treated with a higher infliximab dose. Those who started with a higher dose at baseline tended to have a higher (p = 0.065) response rate at 3 months (73.1% vs 53.2%) than patients who started with the labelled dose of 3 mg/kg. It was not possible to analyse this on later timepoints, because after 3 months the dose of infliximab was often increased as shown before.


The median anti-TNFα, DMARDs, corticosteroids, NSAIDs and total medication costs with their interquartile ranges are presented in table 4 for five categories; 98% of the total costs could be ascribed to cost related to the anti-TNFα treatment itself. The total and anti-TNFα medication costs for infliximab were significant higher than these costs for adalimumab and etanercept. Note that the calculation of costs for infliximab comprised the costs for administration in the in-patient clinic (€80.73 per administration) and the increase of dose over time to a median dose of 4 mg/kg per 8 weeks. Furthermore, the higher costs for infliximab were entirely driven by the loading infusions at week 2 and week 6. Costs made in later months were not significant different from the costs of adalimumab or etanercept.

Table 4 median values (interquartile ranges) of medication costs (in Euros, €) in the first year after baseline


This evaluation of the effectiveness and costs showed that adalimumab and etanercept revealed better outcomes compared to infliximab. Significant different effects between the three TNFα blockers on disease activity, the physical component scale of the SF36 and the use of co-medication were found in favour of adalimumab and etanercept. Furthermore, data with regard to HAQ and the EQ-5D showed the same tendency but differences were not statistically significant. With regard to the medication costs, infliximab treatment resulted in higher costs over the follow-up period than treatments with either adalimumab or etanercept, mainly due to the loading gifts at 2 and 6 weeks. There was also a difference in disease activity between adalimumab and etanercept in favour of adalimumab but not on any of other secondary outcomes measures.

Better effectiveness of etanercept in comparison to infliximab was demonstrated by one other observational study.19 Adalimumab was not studied. Better ACR20 response rates at the 3-month and 6-month follow-ups, and better drug survival, was found for etanercept. Increases of the dose of infliximab were needed in 57% of the patients in that study. Lower costs were also reported for etanercept monotherapy compared with infliximab combined with MTX by Nuyten et al.20

The results of our study are in contrast with the results of two known meta-analyses,4 5 from which was concluded that there are no differences in effectiveness between the three anti-TNFα agents. However, in both meta-analyses the effectiveness in all infliximab dose groups of the ATTRACT (anti-TNF trial in rheumatoid arthritis with concomitant therapy) trial21 were combined. The Dutch labelled dose of infliximab, analysed in our study, is the lowest dose in the ATTRACT trial21 and thus not comparable with the mean infliximab dose analysed in both meta-analyses. The ATTRACT trial did not show clear (significant) differences in effectiveness between the four dose groups, but they did show a trend for increased efficacy in higher dose groups.21 Furthermore, in the present study a trend was shown for a higher response percentage in patients starting on a higher dose of infliximab. Therefore, it is very plausible that the relatively low dose of infliximab as analysed in this study might be the explanation for the difference in effectiveness compared to adalimumab and etanercept.

According to guidelines for cost analyses in several countries,18 22 costs should be calculated from a societal perspective. This would mean that besides medication, also other medical consumption, travelling and absence from work should be considered. Travel costs could be different between the three treatment groups due to the in-patients administration of infliximab. However, travel costs were not considered as the cost-drivers; neither were costs due to absence of paid work. Because of the severity of the RA and the age, most patients did not perform paid work (80% of the patients had either an elderly pension or an incapacity payment), which was not considered to change during the 12 months of follow-up. Medical consumption as a consequence of adverse events could have played a role. However, this factor could not be considered in this study because of a lack of statistical power to evaluate a possible difference of the occurrence of adverse events between the agents. Long-term cost savings or benefits were not within the scope of the study, although it is quite possible that the presented differences result in differences in quality of life, functionality, and costs-savings in the future.

Furthermore, it should be pointed out that patients were not randomised between the three drug groups in this study, which could possibly have led to confounding by indication. Confounding by indication can play a role when there are differences at baseline in factors related to the outcome measure of the study. We tested possible differences between treatment groups at baseline in known prognostic factors for effectiveness, such as all baseline values and patient’s characteristics, where we did not find clinically relevant differences. Other possible factors that might have played a role in the decision between the three anti-TNFα agents are a higher baseline risk for side effects, or the patient’s preference for one of the drugs. However, it is not likely that these factors could have influenced the effectiveness of the drugs as measured in this study. The percentage of patients who stopped due to side effects was not different between the three agents and will not have affected the difference in effectiveness between the agents. Patient preferences for one agent could influence the effectiveness when a patient is not treated with a treatment of choice. However, in this study patients were not randomised and were therefore treated with the treatment of choice, optimising the adherence to treatment and therefore the effectiveness of treatment. Finally, a post-study survey revealed that the choice between the three treatments was mostly time- and place-dependent in the way that the three drugs were differently available or reimbursed during the inclusion period.

We would like to emphasise that a randomised head-to-head study would be the best way to address the hypothesis of equal effectiveness of the three anti-TNFα agents from a methodological point of view (minimising confounding). However, such a study is not likely to be ever performed because there are no parties to fund such a study. In the absence of a randomised head-to-head comparison, in our opinion this observational study provides a valid answer to the rheumatologist on the question of which one of the three TNFα blocking agents should be prescribed. Observational data can even be of extra value, providing data on diversity in doses, effectiveness and costs in clinical practice compared to RCTs. There are other biological registries in Europe,2325 which could also provide real life data on anti-TNFα doses, drug survival and effectiveness. However, they did not report on the comparison of effectiveness between all three anti-TNFα agents.

Overall, this evaluation of the effectiveness and costs showed that adalimumab and etanercept revealed better outcomes than infliximab in the first year of treatment. Future research should consider whether a different dosing scheme for infliximab (higher doses or shorter intervals) is equally effective in daily clinical practice as adalimumab or etanercept.


We are indebted to all research nurses and rheumatologists of the 11 departments of rheumatology involved for their participation and contribution in the data collection. Furthermore, we would like to acknowledge Thea van Gaalen, Lia Schalkwijk and Carien Versteegden for the data processing and Hans Groenewoud for assistance in managing the database.



  • Funding: Funding from the Dutch National Health Insurance Board and the Dutch affiliations of Wyeth Pharmaceuticals, Abbott Pharmaceuticals and Roche Pharmaceuticals enabled the data collection for the DREAM cohort. The sponsors had no influence on the content of this manuscript

  • Competing interests: None declared.