Direct comparison of treatment responses, remission rates, and drug adherence in patients with rheumatoid arthritis treated with adalimumab, etanercept, or infliximab: Results from eight years of surveillance of clinical practice in the nationwide Danish DANBIO registry

Authors

  • Merete Lund Hetland,

    Corresponding author
    1. DANBIO Registry, and Copenhagen University Hospital, Hvidovre and Glostrup, Denmark
    • DANBIO, Copenhagen University Hospital at Hvidovre, Department of Rheumatology 232, Kettegård Alle 30, DK 2650 Hvidovre, Denmark
    Search for more papers by this author
    • Dr. Hetland has received consulting fees, speaking fees, and/or research grants from Abbott, Centocor, Roche, Schering-Plough, UCB-Nordic, and Wyeth (less than $10,000 each), and, on behalf of DANBIO, she has received grants from Abbott, Bristol-Myers Squibb, Roche, Schering-Plough, UCB-Nordic, and Wyeth (more than $10,000 each).

  • Ib Jarle Christensen,

    1. The Finsen Laboratory, Copenhagen University Hospital, Rigshospitalet, Denmark
    Search for more papers by this author
  • Ulrik Tarp,

    1. Aarhus University Hospital, Aarhus, Denmark
    Search for more papers by this author
    • Dr. Tarp has received consulting and/or speaking fees from Abbott, Genmab, Roche, Schering-Plough, and Wyeth (less than $10,000 each).

  • Lene Dreyer,

    1. Copenhagen University Hospital, Rigshospitalet, Denmark
    Search for more papers by this author
  • Annette Hansen,

    1. DANBIO Registry, and Copenhagen University Hospital, Gentofte, Denmark
    Search for more papers by this author
    • Dr. A. Hansen has received consulting fees, speaking fees, and/or research grants from Abbott and Schering-Plough (less than $10,000 each).

  • Ib Tønder Hansen,

    1. Viborg Hospital, Viborg, Denmark
    Search for more papers by this author
  • Gina Kollerup,

    1. Copenhagen University Hospital, Bispebjerg, Denmark
    Search for more papers by this author
    • Dr. Kollerup has received consulting fees from Abbott (less than $10,000).

  • Louise Linde,

    1. DANBIO Registry, and Copenhagen University Hospital, Hvidovre and Glostrup, Denmark
    Search for more papers by this author
    • Dr. Linde has received a study grant from Schering-Plough (more than $10,000).

  • Hanne M. Lindegaard,

    1. Odense University Hospital, Odense, Denmark
    Search for more papers by this author
    • Dr. Lindegaard has received consulting fees from Abbott, Roche, and Bristol-Myers Squibb (less than $10,000 each).

  • Uta Engling Poulsen,

    1. Rheumatism Hospital, University of Southern Denmark, Graasten, Denmark
    Search for more papers by this author
  • Annette Schlemmer,

    1. Aarhus University Hospital, Aalborg, Denmark
    Search for more papers by this author
    • Dr. Schlemmer has received speaking fees from Schering-Plough (less than $10,000).

  • Dorte Vendelbo Jensen,

    1. Hørsholm Hospital, Hørsholm, Denmark
    Search for more papers by this author
    • Dr. D. V. Jensen has received speaking fees from Wyeth (less than $10,000).

  • Signe Jensen,

    1. Copenhagen University Hospital, Gentofte, Denmark
    Search for more papers by this author
  • Gisela Hostenkamp,

    1. Institute for Public Health, Health Economic Unit, Odense, Denmark
    Search for more papers by this author
  • Mikkel ØStergaard

    1. Copenhagen University Hospital, Hvidovre and Glostrup, and Copenhagen University Hospital, Gentofte, Denmark
    Search for more papers by this author
    • Dr. Østergaard has received consulting fees, speaking fees, and/or research grants from Abbott, Amgen, Bristol-Myers Squibb, Centocor, Genmab, GlaxoSmithKline, Novo, Pfizer, Roche, Schering-Plough, UCB, and Wyeth (less than $10,000 each).


Abstract

Objective

To compare tumor necrosis factor α inhibitors directly regarding the rates of treatment response, remission, and the drug survival rate in patients with rheumatoid arthritis (RA), and to identify clinical prognostic factors for response.

Methods

The nationwide DANBIO registry collects data on rheumatology patients receiving routine care. For the present study, we included patients from DANBIO who had RA (n = 2,326) in whom the first biologic treatment was initiated (29% received adalimumab, 22% received etanercept, and 49% received infliximab). Baseline predictors of treatment response were identified. The odds ratios (ORs) for clinical responses and remission and hazard ratios (HRs) for drug withdrawal were calculated, corrected for age, disease duration, the Disease Activity Score in 28 joints (DAS28), seropositivity, concomitant methotrexate and prednisolone, number of previous disease-modifying drugs, center, and functional status (Health Assessment Questionnaire score).

Results

Seventy percent improvement according to the American College of Rheumatology criteria (an ACR70 response) was achieved in 19% of patients after 6 months. Older age, concomitant prednisolone treatment, and low functional status at baseline were negative predictors. The ORs (95% confidence intervals [95% CIs]) for an ACR70 response were 2.05 (95% CI 1.52–2.76) for adalimumab versus infliximab, 1.78 (95% CI 1.28–2.50) for etanercept versus infliximab, and 1.15 (95% CI 0.82–1.60) for adalimumab versus etanercept. Similar predictors and ORs were observed for a good response according to the European League Against Rheumatism criteria, DAS28 remission, and Clinical Disease Activity Index remission. At 48 months, the HRs for drug withdrawal were 1.98 for infliximab versus etanercept (95% 1.63–2.40), 1.35 for infliximab versus adalimumab (95% CI 1.15–1.58), and 1.47 for adalimumab versus etanercept (95% CI 1.20–1.80).

Conclusion

Older age, low functional status, and concomitant prednisolone treatment were negative predictors of a clinical response and remission. Infliximab had the lowest rates of treatment response, disease remission, and drug adherence, adalimumab had the highest rates of treatment response and disease remission, and etanercept had the longest drug survival rates. These findings were consistent after correction for confounders and sensitivity analyses and across outcome measures and followup times.

The tumor necrosis factor α (TNFα) inhibitors adalimumab, etanercept, and infliximab have dramatically improved the treatment of rheumatoid arthritis (RA). The efficacy of these agents has been demonstrated in large randomized clinical trials (RCTs) (1–3). Today, many clinicians aim at achieving a good response according to the European League Against Rheumatism (EULAR) criteria (EULAR good response) (4), or even remission. Because the treatment responses in 60–80% of TNFα inhibitor–treated patients do not meet these goals, identification of predictors of a beneficial outcome are needed (5). Previous studies have suggested that concomitant treatment with methotrexate (MTX) adds to the TNFα inhibitor treatment response, and low functional status decreases this response (6, 7).

It is debated whether the different TNFα inhibitors have similar efficacy. Meta-analyses of RCTs have demonstrated equal clinical efficacy, although no RCTs have compared the effectiveness of the individual TNFα inhibitors (8–11). Indirect comparisons of RCTs have been inconclusive, partly because of a lack of statistical power (8–11). However, emerging evidence from observational registries suggests that more patients achieve a moderate response to adalimumab or etanercept compared with infliximab (12), and that adherence rates were higher for etanercept compared with infliximab in one cohort (13, 14) but not in another cohort (15).

Most likely, randomized, controlled, head-to-head comparisons of the outcomes of TNFα inhibitor therapy will never be carried out. Consequently, observational studies, e.g., from clinical registries, are needed to gain evidence of potential differences between the 3 TNFα inhibitors regarding their ability to induce satisfactory treatment responses (70% improvement according to the American College of Rheumatology criteria [an ACR70 response] [16], EULAR good response [4]) or remission (Clinical Disease Activity Index [CDAI] [17], Disease Activity Score in 28 joints [DAS28] [18]) in real-life settings. A further advantage of using clinical practice data to assess real-life efficacy is that strict inclusion and exclusion criteria in RCTs make results less generalizable to routine care (5, 19).

The aims of the present study of TNFα inhibitor–naive patients with RA receiving routine care were to identify clinical prognostic factors for response and to compare directly the 3 TNFα inhibitors regarding treatment responses, induction of disease remission, and long-term drug survival rates. The study is based on data from the nationwide Danish DANBIO registry. Coverage is >90% (5), and results can be considered representative of patients with RA who are treated with routine care across a country.

PATIENTS AND METHODS

The DANBIO registry.

Since October 2000, Danish rheumatologists have monitored and reported details of TNFα inhibitor therapy for patients with RA to the DANBIO registry. The registry is Web-based (www.danbio-online.dk), which ensures immediate feedback to the treating physician regarding disease activity, present and past medication, and adverse events. Details of the variables that are registered have been provided elsewhere (20). The registry is based on open-source software, which means that it has no licensing costs and undergoes rapid development cycles (21).

DANBIO has been approved by the National Board of Health as a national quality registry. It is sponsored by the hospital owners (Danish Regions) and by the pharmaceutical companies that offer biologic treatments for rheumatic diseases. The sponsors have no influence on the registry set-up, data collection, data analysis, or publication of results. These issues are all administered by an independent steering committee with representatives from the Danish Society of Rheumatology, the Institute for Rational Pharmacotherapy, and the hospital owners.

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.

Treatment strategy and followup.

The patients included in the study had RA, as diagnosed by the treating rheumatologist. All patients had been treated with ≥1 conventional disease-modifying antirheumatic drug (DMARD), but treatment failed to such an extent that therapy with either adalimumab, etanercept, or infliximab was initiated as the first biologic agent. The treatment regimens reflected routine care: concomitant MTX (or other DMARD) and prednisolone were administered according to the decision of the treating rheumatologist. The TNFα inhibitors were prescribed in standard doses unless the rheumatologist decided otherwise. The rheumatologist recorded information on the type of drug, start and stop dates (date of first missed dose), reasons for withdrawal, the swollen joint count in 28 joints and the tender joint count in 28 joints, the physician's global score, and the serum C-reactive protein (CRP) level. The patient completed the Danish version of the Health Assessment Questionnaire (HAQ) (22) together with the patient's global and pain assessment, either by touch screen or using paper forms (21). Patients were registered at least twice yearly.

Treatment outcomes.

The following treatment responses were assessed in each patient after 6 months and 12 months: ACR70 response (16), a EULAR good response (4), CDAI remission (17), and DAS28 remission (DAS28 <2.6) (18). In addition to the crude response rates, the LUNDEX-corrected responses ([fraction of starters still in the study after x months] × [fraction responding at x months]) (14) were calculated. The number of patients varies with the efficacy measure, because some patients lacked ≥1 variable at followup. To allow a comparison with previous studies, ACR50 and EULAR good/moderate responses are also presented in Table 3 and Figure 2 but are not discussed any further, because the focus was on the stricter response criteria.

Table 3. Treatment responses after 6 months and 12 months of treatment*
 AdalimumabEtanerceptInfliximabP
 6 months12 months6 months12 months6 months12 months6 months12 months
  • *

    Values are the percent. LUNDEX-corrected responses were calculated as follows: (fraction of starters still in the study after x months) × (fraction responding at x months). EULAR = European League Against Rheumatism; DAS28 = Disease Activity Score in 28 joints; ACR50 = American College of Rheumatology 50% response; CDAI = Clinical Disease Activity Index.

EULAR response        
 No. of patients536444414377889690  
 Good525742493440<0.0001<0.0001
 Moderate333039323839
 No response151219202921
LUNDEX-corrected EULAR response        
 No. of patients6676605065091,1151,115  
 Good413934362725<0.0001<0.0001
 Moderate272132243024
 No response324033404351
DAS28 remission        
 No. of patients536444414377889690  
 Remission323926332127<0.0001<0.0001
 LUNDEX corrected262721241716<0.0001<0.0001
ACR response        
 No. of patients519426383346852660  
 ACR50455340453138<0.0001<0.0001
 ACR70243021271417<0.0001<0.0001
LUNDEX-corrected ACR response        
 No. of patients6506424754781,0781,085  
 ACR50363532322523<0.0001<0.0001
 ACR70192017191111<0.0001<0.0001
CDAI remission        
 No. of patients539448412373888687  
 Remission1825131810160.00010.0003
 LUNDEX corrected15171013810<0.00010.0001
Figure 2.

Odds of achieving a treatment response after 26 weeks, adjusted for age, disease duration, disease activity, concomitant methotrexate and prednisolone treatment, number of previous DMARDs, center, and HAQ score at baseline. See Figure 1 for definitions.

Statistical analysis.

Descriptive statistics for continuous variables are presented as medians and ranges; categorical variables are presented as frequencies and percentages. Differences between groups were analyzed using rank statistics (Kruskal-Wallis, Wilcoxon's rank sum test, Wilcoxon's signed rank test for paired data, and the chi-square test for independence).

Logistic regression was used for the prediction model, stratifying by drug and the DAS28 at baseline. The probability of a response was modeled. The results of these analyses are presented as odds ratios (ORs) with 95% confidence intervals (95% CIs).

The ORs for achieving the different treatment responses after 26 weeks and 52 weeks were adjusted for age, disease duration, disease activity, concomitant MTX and prednisolone, number of previous DMARDs, center, and HAQ score at baseline. The comparison of the drugs was done using logistic regression analysis, adjusting for baseline variables as described above. Propensity scores were also used for this analysis (23).

Additional sensitivity analyses included the following: analyses on unadjusted data, analysis in which all withdrawers were classified as nonresponders, and analysis including only patients who started treatment after January 1, 2003 (at which time adalimumab was marketed); all gave similar results (data not shown). No evidence of interaction between the drugs and the covariates (e.g., concomitant MTX) was found.

Kaplan-Meier estimates of the probability for drug survival were used, and drugs were compared using the log rank statistic. Hazard ratios (HRs) were calculated using the proportional hazards model for drug withdrawal. These were corrected for baseline disease activity, age, disease duration, concomitant MTX and prednisolone, number of previous DMARDs, HAQ score, and center. Sensitivity analyses using propensity scores gave similar results (data not shown).

All data were analyzed using SAS version 9.1 (SAS Institute, Cary, NC) or R version 2.9.0 (R Foundation for Statistical Computing, Vienna, Austria). P values less than 0.05 were considered significant.

RESULTS

Study population.

Between October 2000 and December 31, 2007, a total of 8,074 patients had been registered in DANBIO. Of these, 5,748 patients were not included in the present study: 1,997 patients had diagnoses other than RA, 3,225 patients with RA were treated with DMARDs, 72 patients with RA received a biologic agent other than a TNFα inhibitor as the first biologic treatment, and 454 patients had a baseline DAS28 <3.2 (CRP-based, 4 variables) or lacked a baseline DAS28. Thus, 2,326 patients with RA who started their first TNFα inhibitor treatment with adalimumab (n = 675 [29%]), etanercept (n = 517 [22%]), or infliximab (n = 1,134 [49%]) were eligible. The patients were followed up until April 3, 2009 or until they withdrew from treatment, whichever came first. The median (interquartile range [IQR]) followup times were as follows: for adalimumab, 20 months (IQR 7–39); for etanercept, 21 months (IQR 9–42); for infliximab, 16 months (IQR 5–36). The cumulative numbers of patient-years of treatment with each drug were 1,349, 1,161, and 2,286, respectively.

Patients who withdrew from treatment before 6 months (n = 449) were excluded from the crude treatment responses but were included in the LUNDEX-corrected response and in the drug survival analyses. The main reasons for withdrawal during the first 6 months were lack of efficacy (52%) and adverse events (38%). The patients who withdrew were similar to the study population regarding sex, age, IgM rheumatoid factor seropositivity, number of previous DMARDs, and concomitant prednisolone (Table 1). Fewer patients who withdrew received concomitant MTX, their disease duration was shorter, and their DAS28 and HAQ score were slightly higher.

Table 1. Baseline characteristics of patients in the study population and patients who withdrew*
CharacteristicStudy population (n = 1,877)Withdrawers (n = 449)P
  • *

    Except where indicated otherwise, values are the median (range). The Disease Activity Score in 28 joints (DAS28) was based on the C-reactive protein (CRP) level with 4 variables. RF = rheumatoid factor; DMARDs = disease-modifying antirheumatic drugs; MTX = methotrexate; HAQ = Health Assessment Questionnaire.

  • By chi-square test for dichotomous variables and Wilcoxon's 2-sample test for continuous variables.

  • Median (interquartile range) among patients receiving the drug.

Male sex, %26.724.50.33
Age, years57 (15–89)57 (19–88)0.89
IgM RF seropositive, %80.278.30.42
Disease duration, years9 (0–68)7 (0–63)0.0009
No. of previous DMARDs3 (0–9)3 (0–9)0.60
DAS285.4 (3.3–8.4)5.6 (3.3–8.2)0.0005
Concomitant MTX, %76.465.0<0.0001
 MTX dosage, mg/week16 (10–20)15 (10–20)0.20
Concomitant prednisolone, %45.446.20.77
 Prednisolone dosage, mg/day7.5 (5–10)7.5 (5–10)0.24
HAQ score1.375 (0–3)1.500 (0–3)<0.0001

Predictors of treatment response.

After 6 months, among the patients who had not withdrawn, an ACR70 response had been achieved in 19%, a EULAR good response in 41%, DAS remission in 25%, and CDAI remission in 13%. Older age, low functional status (high HAQ score), and concomitant prednisolone treatment were negative predictors of an ACR70 response (Figure 1A), whereas concomitant MTX, male sex, number of previous DMARD treatments, and disease duration were not predictors. Based on the ORs presented in Figure 1, it can be deduced that the OR for an ACR70 response in an 80-year-old patient with an HAQ score of 2 who received concomitant prednisolone (compared with a patient who was 10 years younger, had an HAQ score of 1, and did not receive prednisolone) would be reduced to 0.47. For all outcome measures, the patterns were largely similar, with statistically significant ORs <1 for older age, low functional status, concomitant prednisolone treatment, and number of previous DMARDs, an OR of ∼1 for male sex and disease duration, and ORs >1 (statistically significant only for a EULAR good response) for concomitant MTX (Figures 1B–D). Expanding the example described above, the odds for achieving a EULAR good response, DAS28 remission, and CDAI remission would be reduced to 62%, 57%, and 40%, respectively. Results at the 12-month followup were similar (data not shown).

Figure 1.

Predictors of treatment response after 26 weeks. ACR70 = American College of Rheumatology 70% response; OR = odds ratio; 95 pct CI = 95% confidence interval; MTX = methotrexate; HAQ = Health Assessment Questionnaire; DMARDs = disease-modifying antirheumatic drugs; EULAR = European League Against Rheumatism; DAS28 = Disease Activity Score in 28 joints; CDAI = Clinical Disease Activity Index.

Treatment responses to adalimumab, etanercept, and infliximab.

There were no clinically relevant differences in the distribution of age, sex, disease activity, and disease duration between the patients who received the 3 TNFα inhibitor treatments (Table 2). Fewer patients treated with etanercept received concomitant MTX, and more patients receiving infliximab were receiving concomitant MTX and prednisolone.

Table 2. Baseline characteristics and disease activity of the patients according to treatment*
 Adalimumab (n = 544)Etanercept (n = 425)Infliximab (n = 908)P
  • *

    Except where indicated otherwise, values are the median (range). The Disease Activity Score in 28 joints (DAS28) was based on the C-reactive protein (CRP) level with 4 variables. IgM-RF = IgM rheumatoid factor; DMARDs = disease-modifying antirheumatic drugs; MTX = methotrexate; TNF = tumor necrosis factor.

  • By chi-square test for dichotomous variables and Kruskal-Wallis test for continuous variables.

  • Median (interquartile range) among patients receiving the drug.

  • §

    Average body weight 72.6 kg.

Male sex, %2528270.58
Age, years56 (15–85)58 (19–89)57 (17–85)0.30
IgM-RF positive, %8084780.10
Disease duration, years9 (0–51)8 (0–47)9 (0–68)0.24
No. of previous DMARDs3 (0–8)3 (0–8)3 (0–9)0.0044
DAS28 at baseline5.3 (3.3–8.3)5.4 (3.3–8.4)5.4 (3.3–8.3)0.035
Concomitant MTX, %706187<0.0001
 MTX dosage, mg/week20 (12.5–25)15 (12.5–20)15 (10–20)<0.0001
Concomitant prednisolone, %4043500.0003
 Prednisolone dosage, mg/day7.5 (5–10)7.5 (5–10)7.5 (5–10)<0.0001
TNF inhibitor dose, mean ± SD§    
 First dose, mg/treatment40 ± 245 ± 10229 ± 55
 Maintenance dose, mg/treatment40 ± 245 ± 11257 ± 84
 Maintenance frequency, weekly interval1.9 ± 0.50.9 ± 0.36.9 ± 1.6
 Dose at last registered visit, mg/treatment40 ± 344 ± 11259 ± 87

Overall, the crude treatment response rates after 6 months and 12 months were highest for adalimumab, intermediate for etanercept, and poorest for infliximab (Table 3). After correction using the LUNDEX, 19% of patients receiving adalimumab, 17% of patients receiving etanercept, and 11% of those receiving infliximab had achieved an ACR70 response, and 41%, 34%, and 27%, respectively, had a EULAR good response after 6 months. Similarly, 26%, 21%, and 17% of these patients, respectively, had achieved DAS28 remission, and 15%, 10%, and 8%, respectively, had achieved CDAI remission.

After correction for differences in sex, age, disease duration, seropositivity, the DAS28, concomitant MTX and prednisolone treatment, the number of previous DMARDs, the HAQ score at baseline, and center, the ORs for achieving an ACR70 response after 6 months of treatment were 2.05 for adalimumab and 1.78 for etanercept, with infliximab as the reference drug (Figure 2). There was no significant difference between adalimumab and etanercept (OR 1.15, with etanercept as the reference). The ORs for adalimumab versus infliximab ranged from 1.78 to 2.76 and were statistically significant for all outcome measures (Figure 2). For etanercept versus infliximab, the ORs ranged from 1.16 to 1.99 and were statistically significant for all outcomes except DAS28 remission and CDAI remission. For adalimumab compared with etanercept, the ORs ranged from 1.15 to 1.58 and were significant for a EULAR good response and CDAI remission. The ORs at 12 months were similar (data not shown).

The serum CRP level decreased less in infliximab-treated patients (mean decrease 14 mg/ml [median 7 mg/ml]) than in adalimumab-treated patients (mean decrease 18 mg/ml [median 9 mg/ml]) and etanercept-treated patients (mean decrease 15 mg/ml [median 6 mg/ml]) (P = 0.035) (for infliximab versus etanercept, P = 0.58; for infliximab versus adalimumab, P = 0.01; for etanercept versus adalimumab, P = 0.07).

The dosages at the initiation of treatment, the maintenance dosages, and the dosages at the last registered visit were calculated for patients who received >1 dose of drug (or >6 weeks of treatment in those receiving infliximab) and for those in whom complete dosage registrations were available (Tables 2 and 3). The annualized mean ± SEM maintenance doses were as follows: for adalimumab, 1,099 ± 9.5 mg; for etanercept, 2,533 ± 6.2 mg; for infliximab, 1,949 ± 40 mg. The corresponding standard maintenance doses were 1,040 mg, 2,600 mg, and 1,417 mg (assuming 6.5 infliximab treatments per year), respectively.

The dose of infliximab at the time of withdrawal was investigated in the patients who had a recorded dose for the first and the last infusions (n = 591). In total, 51% of patients who withdrew due to lack of efficacy after week 26 were receiving an increased dose corresponding to, on average, 169% of the standard dose. This corresponds roughly to increasing the dose from 3 mg/kg to 5 mg/kg or to decreasing the dosing intervals from every 8 weeks to every 5 weeks.

Drug adherence.

The drug survival rate was highest for etanercept and lowest for infliximab (Figure 3). At 48 months, the unadjusted drug adherence rates were as follows: for adalimumab, 52% (95% CI 46–57%); for etanercept, 56% (95% CI 51–62%); for infliximab, 41% (95% CI 37–44%) (P < 0.0001, by log rank test). Patients still receiving the medications on April 3, 2009 were censored.

Figure 3.

Drug survival rate for each tumor necrosis factor α inhibitor. The number of withdrawals and the number of patients still receiving each drug at different time points are shown.

The HR for drug withdrawal, adjusting for baseline DAS28, age, disease duration, seropositivity, concomitant MTX and prednisolone treatment, number of previous DMARDs, HAQ score, and center, was overall highest for infliximab and lowest for etanercept, regardless of the reason for withdrawal (Table 4).

Table 4. Hazard ratios (95% confidence intervals) for drug withdrawal*
 Infliximab versus adalimumabInfliximab versus etanerceptAdalimumab versus etanercept
  • *

    Hazard ratios were adjusted for baseline disease activity, age, disease duration, concomitant methotrexate and prednisolone treatment, number of previous disease-modifying antirheumatic drugs, Health Assessment Questionnaire score, and center.

All patients (n = 1,089 withdrawals)1.35 (1.15–1.58)1.98 (1.63–2.40)1.47 (1.20–1.80)
Lack of efficacy (n = 727 withdrawals)1.16 (0.95–1.41)1.70 (1.35–2.15)1.47 (1.15–1.87)
Adverse effects (n = 327 withdrawals)1.77 (1.34–2.34)2.65 (1.88–3.73)1.50 (1.04–2.16)

DISCUSSION

To our knowledge, this is the first study directly comparing adalimumab, etanercept, and infliximab and the ability of these agents to elicit treatment responses that reflect modern treatment goals, including remission. Our main finding was that in DMARD-treated but hitherto TNFα inhibitor–naive patients with RA, ACR70 treatment responses and remission rates were lowest for infliximab, intermediate for etanercept, and highest for adalimumab. The findings persisted after correction for a large number of confounders and various sensitivity analyses and were consistent across different outcome measures and different followup times.

In the present study, which was based on longitudinal data from a nationwide cohort, two-thirds of the patients with RA who started TNFα inhibitor treatment had not achieved a EULAR good response after 6 months. This finding, which is in agreement with other reports (14, 24), highlights the opportunity for therapeutic improvement despite significant advances during the last decade and also the importance of identifying predictors of a beneficial outcome. Older age, low functional status, and concomitant prednisolone treatment were negative predictors, more than halving the probability of achieving an ACR70 response. Other investigators have reported concomitant MTX treatment to be a positive predictor (6, 7), but we observed this to be the case only for a EULAR good response.

The effectiveness of treatment was assessed by treatment responses (the ACR70 and a EULAR good response) and by the proportions of patients achieving remission (DAS28 or CDAI remission). The former reflects the magnitude of change in disease activity (either relative or absolute), and the latter reflects whether disease activity is suppressed below a certain threshold. After adjusting for differences in baseline characteristics, the ORs for any of the selected treatment responses or remission criteria were 1.8–2.1 for adalimumab and 1.2–1.8 for etanercept, with infliximab as the reference drug, and 1.2–1.6 for adalimumab versus etanercept. Retention rates, which may be considered a surrogate marker for drug efficacy, were lowest for infliximab, intermediate for adalimumab, and highest for etanercept. When the different aspects of drug efficacy were LUNDEX-adjusted for withdrawal from TNF inhibitor treatment, the differences in the clinical efficacy of the drugs persisted.

Considering the widespread clinical use of TNFα inhibitors and the huge economic implications for health care, surprisingly few studies have attempted to compare the efficacy of the individual TNFα inhibitors. All RCTs of efficacy have been funded by the pharmaceutical industry, and no head-to-head RCTs have been published. A limited number of adjusted indirect comparisons of RCTs have been performed, but with inconclusive results (8–11). One meta-analysis, including 26 published placebo-controlled RCTs of patients with RA in MTX-resistant populations, was not able to show any difference in efficacy among the 3 TNFα inhibitors (8). However, the CIs of the risk ratio (RR) estimates were so wide that clinically significant differences might have been missed. In another study, the RR estimates for achieving an ACR50 response suggested that etanercept might be more efficacious than both adalimumab and infliximab, but that study also lacked statistical power (9). A third study demonstrated a tendency toward lower efficacy for etanercept than for infliximab and adalimumab, but the selection of patients to receive etanercept (MTX naive) differed from the selection of patients to receive the other drugs (MTX resistant), which hampers the results and illustrates that comparisons between RCTs have potentially significant flaws (11). The latest meta-analysis concluded that all TNFα inhibitors were not different from each other (10). That study also had a lack of statistical power, with wide CIs.

The results from RCTs cannot easily be extrapolated to routine care. Several studies have documented that patients in RCTs have higher disease activity and fewer comorbidities than those treated in clinical practice, and that the prescribing practice in routine care has become less stringent over time (5, 19, 25). Some investigators claim that this leads to a poorer treatment response outside RCTs (26), while others, including our group, have reported similar response rates in RCTs and clinical practice (5).

Observational studies of cohorts of unselected patients receiving routine care allow direct comparisons of the drugs, although the lack of randomization should be kept in mind when interpreting the results. A smaller Dutch observational study of 770 patients with RA (12) showed that the performance of infliximab was poorer than that of adalimumab and etanercept (which performed equally well) with regard to achieving a moderate-to-good EULAR response as well as the drug survival rate. This contrasts with our finding of a significantly better outcome with adalimumab compared with etanercept regarding a EULAR good response and better retention rates for etanercept than for adalimumab. However, it was not specified which potential baseline biases were corrected for in the Dutch study, and it did not include strict response criteria such as ACR70 or EULAR good responses or remission rates.

Infliximab had overall lower drug continuation rates compared with adalimumab and etanercept. The difference was most pronounced for withdrawal due to adverse events (HR 1.8–2.7) but was also significant for infliximab compared with etanercept in patients who withdrew due to lack of efficacy. A recent study from the Swiss registry showed poorer drug survival rates for infliximab, mainly due to an increased risk of adverse events (27). The drug survival rate of infliximab is in accordance with that in our study, with a half-life of ∼2 years. In contrast to the Swiss study, we report better overall retention rates for etanercept compared with adalimumab and infliximab, regardless of the reason for withdrawal. A smaller French study of 304 patients with RA supported our findings (28). A previous study from the southern Swedish registry compared drug survival and efficacy in TNFα inhibitor–naive RA patients treated with infliximab (640 patients) and those treated with etanercept (309 patients) (14). The etanercept-treated patients had higher adherence rates than did the infliximab-treated patients and tended to have a higher treatment response, although this was not a consistent finding. In an early study from the German registry, short-term drug survival rates were similar for etanercept and infliximab (29).

The observational design of the present study has some weaknesses. For example, the lack of randomization and blinding may have resulted in bias by indication, channeling bias, and performance bias. However, no clinically significant differences in disease duration or disease activity were seen at baseline, and the results were very robust in a variety of sensitivity analyses. In addition, the CRP level, which is an objective marker of disease, decreased less in infliximab-treated patients than in adalimumab-treated patients. However, the possibility cannot be excluded that differences in the timing of clinical assessments may potentially have biased the treatment outcomes, because infliximab-treated patients were often scored on the day of infusion (at trough drug levels), whereas the subcutaneously treated patients were scored independently of the day of injection. It cannot be ruled out that the infliximab dosage was insufficient in some patients, and that higher dosages would have improved the outcome. However, the patients were treated according to standard recommendations, and the dose of infliximab had been increased by 69% compared with the standard dose in the majority of patients who withdrew due to lack of efficacy. Furthermore, the adjusted HR for drug withdrawal was highest for infliximab compared with adalimumab and etanercept, and this ratio was independent of differences in the timing of the clinical assessment.

Future studies should address the question as to what degree the 3 TNFα inhibitors suppress radiographic destruction in patients treated in routine care, and whether this is a class effect, or whether the different drugs perform differently.

We conclude that in this large, nationwide cohort of TNFα inhibitor–naive patients with RA treated in routine clinical practice, significant differences in the efficacy of and adherence to therapy with adalimumab, etanercept, and infliximab were observed. Infliximab had the lowest treatment responses, disease remission rates, and drug adherence rates. Adalimumab had the highest treatment responses and remission rates, whereas etanercept had the longest drug survival rates. These findings were consistent across outcome measures and various sensitivity analyses. Furthermore, we confirmed that older age, low functional status, and concomitant prednisolone treatment were negative predictors of a treatment response and disease remission.

AUTHOR CONTRIBUTIONS

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. Hetland 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. Hetland, Tarp, Dreyer, Ib Tønder Hansen, Østergaard.

Acquisition of data. Hetland, Tarp, Dreyer, Annette Hansen, Ib Tønder Hansen, Kollerup, Lindegaard, Poulsen, Schlemmer, Dorte Vendelbo Jensen, Signe Jensen.

Analysis and interpretation of data. Hetland, Christensen, Tarp, Dreyer, Ib Tønder Hansen, Kollerup, Linde, Hostenkamp, Østergaard.

ROLE OF THE STUDY SPONSORS

Abbott, Wyeth, Schering-Plough, Bristol-Myers Squibb, Roche, and UCB-Nordic had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by the study sponsors.

Acknowledgements

The DANBIO study group comprised the departments of rheumatology at the following hospitals in Denmark: Ålborg, Århus, Bispebjerg, Esbjerg, Frederiksberg, Gentofte, Glostrup, Gråsten, Hjørring, Holbæk, Holstebro, Horsens, Hvidovre, Hørsholm, Kolding, Næstved/Nykøbing Falster, Odense, Randers, Rigshospitalet, Roskilde/Køge, Silkeborg, Slagelse, Svendborg/Fåborg, Vejle, and Viborg. DANBIO is indebted to the head of the Institute for Rational Pharmacotherapy, Jens Peter Kampmann, MD, DMSc, who hosted and financed the database from 2000 to 2003. Statistical analyses were performed by Ib Jarle Christensen and Zitelab Aps.

Ancillary