1. Top of page
  2. Abstract
  6. Acknowledgements


To estimate the incidence rates of serious and nonserious infections in patients with rheumatoid arthritis (RA) who start treatment with a biologic agent, and to compare these rates with those in patients with RA who receive conventional treatment.


Patients enrolled in the German biologics register between May 2001 and September 2003 were included. Treating rheumatologists assessed adverse events and serious adverse events. All adverse events and serious adverse events experienced within 12 months after study entry were analyzed. Propensity score methods were applied to estimate which part of a rate increase was likely to be attributable to differences in patient characteristics.


Data were available for 512 patients receiving etanercept, 346 patients receiving infliximab, 70 patients receiving anakinra, and 601 control patients treated with disease-modifying antirheumatic drugs. The total number of adverse events per 100 patient-years was 22.6 (95% confidence interval [95% CI] 18.7–27.2) among patients receiving etanercept, 28.3 (95% CI 23.1–34.7) among patients receiving infliximab, and 6.8 (95% CI 5.0–9.4) among controls (P < 0.0001). Significant differences in the rate of serious adverse events were also observed. For patients receiving etanercept, those receiving infliximab, and controls, the total numbers of serious adverse events per 100 patient-years were 6.4 (95% CI 4.5–9.1), 6.2 (95% CI 4.0–9.5), and 2.3 (95% CI 1.3–3.9), respectively (P = 0.0016). After adjusting for differences in the case patient mix, the relative risks of serious adverse events were 2.2 (95% CI 0.9–5.4) for patients receiving etanercept and 2.1 (95% CI 0.8–5.5) for patients receiving infliximab, compared with controls.


Patients treated with biologic agents have a higher a priori risk of infection. However, our data suggest that this risk is increased by treatment with tumor necrosis factor inhibitors.

Tumor necrosis factor α (TNFα) and interleukin- 1 (IL-1) have been shown to play an essential part in the pathogenetic mechanism of rheumatoid arthritis (RA). Drugs targeting TNFα and IL-1 have been developed in order to neutralize the effects of these proinflammatory cytokines. Randomized clinical trials have demonstrated the efficacy of cytokine inhibitors (infliximab, etanercept, adalimumab, anakinra) in reducing inflammatory activity as well as inhibiting joint destruction in patients with active RA (1–5). Today, these drugs play an important role in the treatment of RA, particularly in patients whose disease is not responding to treatment with conventional disease-modifying antirheumatic drugs (DMARDs).

Although trial-based evidence and clinical experience to date are encouraging, there is still insufficient information on the long-term safety and continuing efficacy of these biologic agents for daily care. In the long-term surveillance of patients treated with TNF inhibitors, serious adverse events, particularly intracellular organism infections such as tuberculosis and other granulomatous infections, were observed (6–9). Therefore, proposals to investigate the incidence of adverse events in patients with RA who are exposed to biologic agents, by means of long-term followup studies, were made by American and European rheumatologists (10–13). In 2001, the German Society of Rheumatology invited all rheumatologists to contribute to a national register in order to investigate the long-term safety, effectiveness, and costs of biologic therapies in RA. The Epidemiology Unit at the German Rheumatism Research Center was charged with maintaining the register, and an advisory board was established by the German Society of Rheumatology.

Here we present the first results from the prospective cohort study known as RABBIT, which is the German acronym for rheumatoid arthritis–observation of biologic therapy. We investigated the incidence of serious and nonserious infections within the first year of treatment with biologic agents.

It is known that patients with RA are at increased risk of development of infectious diseases (14, 15). Therefore, the observed number of events in patients receiving biologic agents must be compared with the number of events expected in patients receiving conventional treatment. For this purpose, we took advantage of a specific feature of the German biologics register. RABBIT has an inherent control group of patients treated with conventional DMARDs who are followed up according to the same protocol in the same rheumatology units. Therefore, we were able to contrast the adverse event rate in patients receiving biologic agents with the rate in patients treated with conventional DMARDs.


  1. Top of page
  2. Abstract
  6. Acknowledgements


Patients ages 18–75 years who meet the American College of Rheumatology (formerly, the American Rheumatism Association) criteria for a diagnosis of RA (16) are eligible as “cases” if new treatment with infliximab, etanercept, anakinra (since January 2003), or adalimumab (since September 2003) was started, and are classified as “controls” if conventional DMARD therapy was started after the failure of at least 1 other DMARD. Patients may also be enrolled in the control group if an additional DMARD is added to existing DMARD treatment. Patients are required to give written informed consent at the time of enrollment. The study protocol was approved by the ethics committee of the Charité Medical School, Berlin, and, when necessary, by the local ethics committee of the participating rheumatology unit. Patient recruitment is ongoing. In the following analyses, patients who were enrolled up to September 1, 2003 were included. All followup data available until September 2004 were used.


RABBIT is a long-term prospective cohort study of patients with RA who receive daily rheumatologic health care. The study protocol stipulates that treatment decisions are not to be influenced by the principal investigators, the scientific advisory board, or the pharmaceutical companies sponsoring the register. For reasons of full transparency, all participating rheumatologists received a copy of the contract between the German Rheumatism Research Center and the 4 pharmaceutical companies. The contract specifies that full responsibility for the conduct of the study, data ownership, and publication rights are in the hands of the principal investigators.


At baseline and at 3, 6, and 12 months of followup, the treating rheumatologist recorded results of a tender joint count and a swollen joint count (both of which were based on 28 joints), the erythrocyte sedimentation rate (ESR; Westergren's method), the C-reactive protein (CRP) level, morning stiffness, and DMARD and/or biologic therapy, including details of starting and ending therapy, reasons for treatment termination, concomitant therapies with glucocorticoids or nonsteroidal antiinflammatory drugs, and adverse events. Patients assessed their pain, general health, or fatigue on a numeric rating scale of 0 to 10. Disability was assessed by means of the Hannover Functional Status Questionnaire (17). The disease activity score in 28 joints (DAS28) was calculated (18).

Physicians assessed adverse events and serious adverse events according to accepted guidelines. For reasons of quality assurance, the rheumatologists were provided with the definitions of adverse events and serious adverse events according to International Conference on Harmonisation guideline E2A (19), as well as with the suggestion to grade adverse events as mild, moderate, or severe, according to the proposal of the Outcome Measures in Rheumatology Clinical Trials conference (20). These definitions were included in the case report form at each visit. To ensure complete registration of adverse events, training courses for physicians and study nurses were provided. The study physicians, AS and SK, coded the adverse event according to the Medical Dictionary for Regulatory Affairs (MedDRA; version 7.0) (21). Coding was carried out at the preferred term level. All codes were checked in a second coding stage. MedDRA provided a hierarchy for summarizing the symptoms or diagnoses to higher-level groups. Nevertheless, there is no simple or definite rule for summarizing the diagnoses. Different pathways were provided for a large number of preferred terms. All preferred terms resulting at any pathway in the systemic organ class “infections or infestations” (except chronic bronchitis, postherpetic neuralgia, rheumatic endocarditis, and skin papilloma) were included in the analysis. Influenza-like illness was also included.

Statistical analysis.

An intention-to-treat analysis was performed to compute adverse event/serious adverse event rates for patients in whom a new treatment was started. In every patient the “time under risk” began with the first application of the index drug (etanercept, infliximab, anakinra, or change in DMARD) and ended at the 12-month date or at the last available visit, depending on which came first. Whether or not the treatment was changed, all single episodes of infection observed within this time window were counted. The time windows were summed up to observed patient-years of followup.

To estimate the actual risk in the different groups, event rates per 100 patient-years and their 95% confidence intervals (95% CIs) (22) were calculated. Poisson regression analysis was performed to compare the event rates by means of a likelihood ratio test and to calculate the relative risk (RR) of infectious diseases in study patients compared with the risk in the control group. This computation was done by means of the GENMOD procedure (23). These RRs are useful for comparing the risk of specified infections in patients receiving etanercept or infliximab with the risk in patients receiving conventional DMARDs. However, these patients differed not only in the kind of treatment but also in factors known to be associated with the susceptibility to infection. Propensity score methods were applied to estimate which part of the RRs was likely to be related to differences in the patient characteristics and which part was related to mechanisms of action of the TNF inhibitors (24). The aim was to stratify patients according to their risk of being treated with TNF inhibitors instead of conventional DMARDs, and to use these groupings to adjust for susceptibility to infection.

The following patient characteristics, which were suspected to be predictors of infection in patients with RA (25) as well as predictors of TNF treatment, were included in a multivariate logistic regression analysis: age, number of DMARD failures, DAS28, CRP level, the presence of rheumatoid factor (RF), and disability. Using this analysis, the likelihood of being treated with etanercept or infliximab instead of conventional DMARDs was calculated. To assess how effectively the model described the outcome variable (TNF treatment: yes/no), the Hosmer-Lemeshow goodness-of-fit test was applied. Patients for whom the estimated likelihood of receiving etanercept or infliximab was <40% (9.2% of the etanercept/infliximab groups, 59.2% of the control group) were excluded, because they were not representative of those who receive TNF treatment. The other patients were grouped into those for whom the likelihood was 40–70% and those for whom the likelihood was >70% (etanercept/infliximab, 64.8%; control, 14.2%). These subgroups were used to compute the adjusted RRs for infectious diseases by means of Poisson regression. In order to detect an RR of 2.0 between 2 groups, with a power of 80%, at least 68 events (194 events for an RR of 1.5) in both groups must be observed (26). Thus, the study had a power of >80% to detect a 2-fold increase in the risk of any adverse event or any moderate/severe adverse event in 1 of the groups receiving biologic agents, compared with the control group. This applies to the entire sample as well as to the subsample of patients with a propensity score of >0.4. However, the power to detect an RR of 2.0 in the rate of serious adverse events was <60%.

To investigate the influence of risk factors for infection among completers, Cox regression analysis was applied. Using this method, the time to the first adverse event (and first serious adverse event) was compared between the completers in the anti-TNF groups and the control group. Adverse events experienced after treatment termination of the index biologic (or the index DMARD in the control group) were not considered. The following risk factors were investigated: propensity score, age, sex, number of DMARD failures, DAS28, CRP level, presence of RF, disability, prednisone dose at study entry, area under the curve of prednisone dose over 12 months, comorbidity (yes/no), and especially chronic lung disease (yes/no), diabetes (yes/no), and psoriasis (yes/no). Cox regression was applied to the entire sample and to the subsample of patients with a propensity score of >0.4. The power of Cox regression is comparable with that of Poisson regression. Depending on the sample size, 65–74 events are needed to detect a hazard ratio (HR) of at least 2 with a power of 80% (27). This condition was fulfilled for adverse events in total, whereas the number of patients who experienced serious adverse events was low. Therefore, the corresponding power of Cox regression with serious adverse event as the outcome was low (<40%).

The Mann-Whitney test and the chi-square test were used to compare baseline characteristics of the patients.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Between May 1, 2001 and September 1, 2003, a total of 1,529 patients from 109 centers were entered into the RABBIT database. In this population, 601 patients had a change in their conventional DMARD therapy (control group), and 512 patients started treatment with etanercept, 346 started with infliximab, and 70 started with anakinra. Seventy-four percent of these patients completed the 12 months of followup in September 2004. The dropout rate was low: 6.9% for the first 6 months and 11.1% in total. In 15% of the patients, the followup was not yet completed.

Patient characteristics.

Table 1 shows the baseline clinical status of patients treated with the individual biologic drugs compared with patients in the control group. Patients in the groups receiving biologic agents had highly active disease and low functional status. The DAS28, number of swollen joints, and CRP values were similar in the groups receiving biologic agents but were clearly different from those in the control group (P < 0.001). The vast majority of the patients receiving etanercept, infliximab, or anakinra had previously been treated with methotrexate (MTX), and approximately three-fourths had been treated with leflunomide. Other DMARDs such as sulfasalazine and antimalarials had been tried in >40%, and azathioprine or cyclosporin A had been tried in >18% of these patients, leading to an average of 3.6–4.2 previous DMARDs; this number was significantly lower in the control group. At baseline, the proportion of patients receiving glucocorticoids (dose >10 mg of prednisolone) was rather high (Table 1). However, this percentage was reduced to <9% in each of the groups within the following 3 months. There was no significant difference between groups in the frequency of comorbid conditions that likely predispose to infection (Table 1).

Table 1. Characteristics of patients and controls at baseline*
CharacteristicEtanercept (n = 512)Infliximab (n = 346)Anakinra (n = 70)Control (n = 601)
  • *

    Except where indicated otherwise, values are the mean ± SD. IQR = interquartile range; RF = rheumatoid factor; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; DAS28 = Disease Activity Score in 28 joints; DMARDs = disease-modifying antirheumatic drugs; MTX = methotrexate.

  • Full function was assessed by means of the Hannover Functional Status Questionnaire, which measures the ability to perform activities of daily living.

Age, years53.7 ± 12.653.6 ± 12.654.3 ± 11.656.5 ± 11.4
Female, %78.170.877.182.7
Body mass index, kg/m225.6 ± 5.025.2 ± 4.224.9 ± 3.826.4 ± 4.7
Disease duration, median (IQR) years9.0 (5.0–16.5)8.0 (4.0–14.0)13.0 (7.0–22.0)6.0 (3.0–13.0)
RF positive, %80.878.977.175.0
Swollen joint count10.5 ± 6.310.8 ± 6.510.2 ± 6.37.7 ± 5.3
Tender joint count13.3 ± 7.412.7 ± 7.612.6 ± 7.110.0 ± 6.6
ESR, median (IQR) mm/hour34.0 (20.0–54.0)35.0 (17.5–58.0)37.0 (18.5–56.0)24.0 (14.0–40.0)
CRP, median (IQR) mg/liter20.0 (7.0–44.0)20.0 (8.0–50.3)21.0 (8.1–35.5)11.0 (4.0–25.0)
DAS286.1 ± 1.26.0 ± 1.26.1 ± 1.25.4 ± 1.2
% full function52.7 ± 23.453.9 ± 21.752.2 ± 24.263.4 ± 22.2
No. of previous DMARDs3.7 ± 1.53.6 ± 1.64.2 ± 1.92.0 ± 1.1
Current DMARDs, %    
 MTX alone33.064.561.420.1
 Other single DMARD11.315.07.124.6
 Combination of 2 DMARDs5.18.12.947.8
 Combination of 3 DMARDs1.
 No DMARDs49.410.428.60.0
Glucocorticoids, any dose, %87.485.287.077.2
 >10 mg prednisolone/day17.720.617.48.7
Comorbid conditions, %    
 Chronic lung disease9.


Infections were observed in 204 of 1,529 patients (13%) (15% of those receiving etanercept, 21% of those receiving infliximab, 13% of those receiving anakinra, and 6% of control patients). Seventy-two of 251 adverse events (28.7%) were graded as mild, and 66 adverse events (26.3%) fulfilled the criteria for seriousness, because they resulted in death (n = 4), were life-threatening (n = 9), inpatient treatment was required (n = 41), or the event led to significant disability/incapacity or a comparable significant risk (n = 12). A detailed list of adverse events (and serious adverse events) is provided in Table 2. Table 3 shows the number of events per 100 patient-years among patients receiving the various drugs (etanercept 483.2 patient-years, infliximab 325.4 patient-years, controls 570.9 patient-years). The relative risks of higher or lower event rates in the etanercept or infliximab group compared with the control group and their 95% CIs are shown in Tables 4 and 5. Below, we report the results for different subgroups of adverse events.

Table 2. Serious and nonserious infections in study patients and controls*
  • *

    Values are the total number of events (number of serious events).

Upper respiratory tract infection15171 (1)
Bronchitis11 (2)7 (2)4 (1)
Pneumonia6 (5)8 (8)3 (3)
Lung abscess1 (1)  
Pleural infection1 (1) 1
Pulmonary tuberculosis 1 (1) 
Other respiratory tract infection 41
Influenza-like illness13134
Herpes simplex461
Herpes zoster5 (3)5 (2)7 (2)
Erysipelas4 (3)  
Cellulitis3 (2)12 (1)
Furuncle/folliculitis4 (1)2 (1)2
Abscess35 (1)1
Paronychia33 (1) 
Wound infection 1 
Other bacterial skin infection1 (1)1 (1)2 (1)
Fungal skin infection2 (2)1 
Sialadenitis 1 
Dental/periodontal infection 2 
Oral mycosis32 
Esophageal candidiasis  1 (1)
Gastrointestinal infection3 (1)21 (1)
Bacterial peritonitis1 (1)  
Infective arthritis5 (5)1 (1)1 (1)
Infective bursitis 1 
Acute osteomyelitis   
Endometritis 1 
Vaginal mycosis/candidiasis4 2
Urinary tract infection8 (2)32
Otitis media1  
Parapharyngeal abscess 1 
Thyroid gland abscess 1 
Sepsis/urosepsis3 (3) 2 (2)
Conjunctivitis  1
Borrelia infection 1 
Total109 (31)92 (20)39 (13)
Table 3. Frequency of adverse events in the study patients and controls*
  • *

    Values are the number of events per 100 patient-years (95% confidence interval). Lower respiratory tract infections comprised bronchitis, pneumonia, lung abscess, pleural infection, and pulmonary tuberculosis (TB); total respiratory tract infections comprised upper, lower, and other respiratory tract infections; herpesvirus infections comprised herpes simplex and herpes zoster infection; bacterial skin and subcutaneous tissue infections comprised erysipelas, cellulitis, furuncle/folliculitis, abscess, paronychia, wound infection, and other bacterial skin infection; gastrointestinal and oral soft tissue infections comprised sialadenitis, dental/periodontal infection, oral mycosis, esophageal candidiasis, and bacterial peritonitis; bone and joint infections comprised infective arthritis, infective bursitis, and acute osteomyelitis; urogenital tract infections comprised endometritis, vaginal mycosis/candidiasis, urinary tract infection, and pyelonephritis.

All adverse events    
 Upper respiratory tract infections3.10 (1.9–5.2)5.22 (3.3–8.4)0.18 (0.004–1.0)<0.0001
 Pneumonia1.24 (0.6–2.8)2.46 (1.2–4.9)0.53 (0.1–1.5)0.0490
 Lower respiratory tract infections, excluding TB3.93 (2.5–6.2)4.61 (2.8–7.7)1.40 (0.7–2.8)0.0043
 Pulmonary TB0 (0–0.8)0.31 (0.01–1.7)0 (0–0.5)0.24
 Respiratory tract infections, total7.04 (5.0–9.9)11.37 (8.2–15.7)1.75 (0.9–3.3)<0.0001
 Influenza-like illness2.69 (1.6–4.6)3.99 (2.3–6.9)0.70 (0.2–1.8)0.0038
 Herpesvirus infections, total1.86 (1.0–3.6)3.38 (1.9–6.1)1.40 (0.7–2.8)0.15
 Bacterial skin and subcutaneous tissue infections3.73 (2.4–5.9)3.99 (2.3–6.9)1.23 (0.6–2.6)0.01
 Skin and subcutaneous tissue infections, total6.00 (4.2–8.6)7.68 (5.2–11.4)2.63 (1.6–4.4)0.0017
 Gastrointestinal and oral soft tissue infections1.45 (0.7–3.0)2.15 (1.0–4.5)0.35 (0.04–1.3)0.031
 Bone and joint infections1.03 (0.4–2.5)0.61 (0.1–2.2)0.18 (0.004–1.0)0.16
 Urogenital tract infections2.69 (1.6–3.6)1.54 (0.6–3.7)0.70 (0.2–1.8)0.036
 Sepsis/urosepsis0.62 (0.1–1.8)0 (0–0.9)0.35 (0.04–1.3)0.53
 Total22.56 (18.7–27.2)28.27 (23.1–34.7)6.83 (5.0–9.4)<0.0001
 Total moderate/severe15.73 (12.6–19.7)20.59 (16.2–26.2)5.08 (3.5–7.3)<0.0001
Serious adverse events    
 Lower respiratory tract infections, total1.86 (1.0–3.6)3.38 (1.9–6.1)0.70 (0.2–1.8)0.014
 Bacterial skin and subcutaneous tissue infections1.45 (0.7–3.0)1.23 (0.3–3.1)0.35 (0.04–1.3)0.12
 Skin and subcutaneous tissue infections, total2.07 (1.1–3.9)1.84 (0.8–4.1)0.70 (0.2–1.8)0.12
 Bone and joint infections1.03 (0.4–2.5)0.31 (0.01–1.7)0.18 (0.004–1.0)0.14
 Other1.45 (0.7–3.0)0.61 (0.1–2.2)0.70 (0.2–1.8)0.51
 Total6.42 (4.5–9.1)6.15 (4.0–9.5)2.28 (1.3–3.9)0.0016
Table 4. Relative risk of infection in patients receiving etanercept or infliximab compared with controls*
  • *

    Values are the relative risk (95% confidence interval).

All adverse events  
 Pneumonia2.36 (0.6–9.5)4.68 (1.2–17.6)
 Lower respiratory tract infections, total2.81 (1.2–7.4)3.51 (1.4–9.5)
 Herpesvirus infections, total1.33 (0.5–3.4)2.63 (1.1–6.4)
 Bacterial skin and subcutaneous tissue infections2.53 (1.0–6.2)2.26 (0.8–6.1)
 Skin and subcutaneous tissue infections, total2.36 (1.3–4.4)2.93 (1.5–5.5)
 Gastrointestinal and oral soft tissue infections4.14 (0.9–19.9)6.14 (1.3–29.6)
 Bone and joint infections5.91 (0.7–50.7)3.51 (0.3–38.7)
 Total3.30 (2.3–4.8)4.14 (2.8–6.0)
 Total moderate/severe3.10 (2.0–4.7)4.05 (2.6–6.3)
Serious adverse events  
 Lower respiratory tract infection, total2.66 (0.7–11.8)4.82 (1.4–20.8)
 Bacterial skin and subcutaneous tissue infections4.14 (0.9–19.9)3.51 (0.6–19.2)
 Skin and subcutaneous tissue infection, total2.95 (0.9–9.4)2.63 (0.7–9.3)
 Bone and joint infections5.91 (0.7–50.7)1.75 (0.1–28.0)
 Total2.82 (1.4–5.9)2.70 (1.3–5.9)
Table 5. Characteristics of patients with a propensity score >0.4*
CharacteristicPropensity score 0.4–0.7Propensity score >0.7
Etanercept (n = 124)Infliximab (n = 89)Control (n = 152)Etanercept (n = 323)Infliximab (n = 208)Control (n = 81)
  • *

    Full function was assessed by means of the Hannover Functional Status Questionnaire, which measures the ability to perform activities of daily living. RF = rheumatoid factor; CRP = C-reactive protein; DAS28 = Disease Activity Score in 28 joints; DMARDs = disease-modifying antirheumatic drugs.

Patient-years of followup120.287.1142.9303.9194.979.6
Age, mean ± SD years54.7 ± 12.152.8 ± 11.555.7 ± 11.052.5 ± 12.653.9 ± 12.953.6 ± 12.6
Female, %75.868.584.980.274.588.9
RF positive, %78.277.577.084.281.781.5
Swollen joint count, mean ± SD9.3 ± 5.58.5 ± 5.58.6 ± 5.911.3 ± 6.411.9 ± 6.510.7 ± 5.8
CRP level, mean ± SD mg/liter24.8 ± 29.928.8 ± 49.823.7 ± 30.637.0 ± 46.744.5 ± 66.936.3 ± 68.2
DAS28, mean ± SD5.9 ± 1.05.6 ± 1.15.7 ± 1.16.3 ± 1.16.4 ± 1.16.3 ± 1.1
Full function, mean ± SD %58.3 ± 23.457.2 ± 20.058.6 ± 22.648.6 ± 22.850.2 ± 21.649.5 ± 23.7
No. of previous DMARDs, mean ± SD2.5 ± 0.62.5 ± 0.72.5 ± 0.74.4 ± 1.34.5 ± 1.33.8 ± 1.3
Respiratory tract infections.

Serious and nonserious respiratory tract infections were significantly more common in patients receiving infliximab or etanercept than in the control group (Table 3). In the anti-TNF groups, 14 cases of pneumonia were observed. This incidence was 2.4–4.7 times higher than that in the control group (Table 4). Furthermore, there was 1 case of pulmonary tuberculosis in the infliximab group, 1 case of lung abscess in the etanercept group, and 19 bronchitides, leading to a significantly increased risk of lower respiratory tract infections, in the etanercept and infliximab groups (Table 4). Large differences in the frequency of upper respiratory tract infections and influenza-like illness were observed. However, half of these infections (34 of 63) were graded as mild.

Skin and subcutaneous tissue infections.

There were more episodes of erysipelas, furuncle, abscess, and paronychia in the TNF groups compared with the control group (Table 2). This led to a significantly increased risk of bacterial skin infections (Tables 3 and 4). Herpes simplex, the milder form of herpesvirus infections, was also more frequently observed in patients receiving etanercept or infliximab, whereas the event per patient-year rate was comparable with that for herpes zoster. Because 5 of 11 episodes of herpes simplex infection were graded as mild and 6 were graded as moderate, this refers especially to an increase in the number of mild-to-moderate viral skin infections. Furthermore, there were 3 fungal skin infections. In total, a significantly increased risk of skin and subcutaneous tissue infections in patients receiving TNF inhibitors was observed (Tables 3 and 4).

Bone and joint infections.

Seven cases of serious infective arthritis were reported (Table 2). The data suggest a higher risk of bone and joint infections in patients treated with biologic agents, especially in those receiving etanercept. However, because there were only 8 bone and joint infections in total, the corresponding confidence boundaries of the RRs are very wide (Table 4).

Septic shock.

In 5 patients (3 assigned to etanercept and 2 controls), septic shock developed following severe or recurrent infections. Four of these patients died, of whom 1, who had been treated with a combination of etanercept and MTX followed by leflunomide alone, had septic shock after recurrent urinary tract infections, 2 had septic multiple organ failure after pneumonia (both patients had chronic obstructive lung disease and were treated with etanercept alone and leflunomide, respectively), and 1, who had been treated with a combination of etanercept and leflunomide, had septic shock following bacterial peritonitis.

Other infections.

Increased rates of gastrointestinal and oral soft tissue infections, urogenital tract infections, and ear and throat infections were also observed (Tables 2, 3, and 4).

Adverse event/serious adverse event total.

The rates of total adverse events and serious adverse events per 100 patient-years were significantly higher in the TNF groups than in the control group (Table 3). The corresponding RRs were 2.7–2.8 for serious adverse event and 3.3–4.1 for adverse events in general, and 3.1–4.1 for moderate-to-severe adverse events (Table 4). A significant increase was also observed when upper respiratory tract infections and influenza-like illness were excluded. However, the higher risks were only partly attributable to the drugs etanercept and infliximab. The different predispositions of the patients must also be taken into account.

We used propensity score methods to differentiate the patients according to their risk of being treated with TNF inhibitors. The risk factors included in this model were also risk factors for infection (age, number of DMARD failures, DAS28, CRP, the presence of RF, and disability). The propensity score model fit quite well. The Hosmer-Lemeshow goodness-of-fit statistic did not show a significant difference between observed and predicted frequencies within deciles of risk (P = 0.24). Patients with a propensity score of <0.4 were not representative of those treated with biologic agents. These patients had a mean DAS28 of 5.1, a mean of 1.4 DMARD failures, and good functional capacity (68% of full function). They were, therefore, excluded from the following analysis. Within the selected groupings, the patients in the TNF groups did not differ significantly from the selected control patients with respect to age, DAS28, disability, ESR, CRP, and RF (Table 5). However, the controls in the high-risk group had a lower number of DMARD failures than did patients receiving biologic agents. Using these subgroups of patients, adjusted RRs of 2.1–3.0 for adverse events and serious adverse events, respectively, were found (Table 6). These adjusted RRs indicate that nearly one-third of the increase found in the unadjusted RRs is likely to be attributable to differences in the case mix.

Table 6. Adjusted relative risks of adverse events in patients versus controls*
  • *

    Values are the relative risk (95% confidence interval). Propensity score methods were applied to stratify patients according to risk factors, and Poisson regression was used to calculate the adjusted relative risks.

Adverse events, total2.31 (1.4–3.9)3.01 (1.8–5.1)
Moderate/severe adverse events1.96 (1.1–3.6)2.93 (1.6–5.3)
Serious adverse events, total2.16 (0.9–5.4)2.13 (0.8–5.5)

Similar results were found when considering the relative increase in the risk of infection during followup while patients were receiving the initial treatment (completer analysis). By Cox regression analysis, the unadjusted HRs of an adverse event in the etanercept group and the infliximab group, respectively, compared with the control group, were 2.6 (95% CI 1.7–3.9) and 3.9 (95% CI 2.6–5.9). The corresponding adjusted HRs in the subsample of patients with a propensity score of >0.4 were, again, nearly one-third lower (for etanercept, 1.7 [95% CI 0.9–3.0]; for infliximab, 2.9 [95% CI 1.6–5.2]). Adjustment was made for the propensity score (HR 3.2 [95% CI 1.0–10.3]) and chronic lung disease (yes/no; HR 2.1 [95% CI 1.2–3.5]). Other possible risk factors were not significantly associated. Because of the limited number of patients with serious infections, no significant associations between any of the risk factors and the outcome of serious adverse event were found.

Infections in the anakinra group.

Eleven adverse events corresponding to 17.5 events per 100 patient-years (95% CI 8.8–31.2) were observed in patients receiving anakinra. Among these were 2 serious infections (infective arthritis/acute osteomyelitis), corresponding to a serious adverse event rate of 3.2 per 100 patient-years (95% CI 0.4–11.5). However, the size of the anakinra group was small (n = 70; 62.8 patient-years). Therefore, the confidence intervals of the infection rates are wide, and possible differences between patients receiving etanercept or infliximab and those receiving anakinra did not reach statistical significance.


  1. Top of page
  2. Abstract
  6. Acknowledgements

In this study, we compared the rates of infections in patients treated with biologic agents with the rates in patients receiving conventional DMARDs. We found that in their practices, treating rheumatologists observed significantly more infections in patients in whom treatment with etanercept or infliximab was started. In comparison with the control group, the incidence of serious infections was 2.7–2.8 times higher in patients receiving biologic agents, and the incidence of adverse events in general was 3.3–4.1 times higher. Obviously, the increase is partially caused by differences in patient characteristics. However, our data suggest that the main part of the increase is likely to be caused by TNF inhibitors, and that susceptibility to all types of infections is higher. Special concern exists for lower respiratory tract infections (especially pneumonia), bacterial skin infections (erysipelas), and bone and joint infections. For anakinra, lower RRs were observed. However, because of the small number of patients receiving anakinra treatment, the results need to be further validated.

The rates of serious infections in our study were similar to those reported by Klareskog et al (28) in the Trial of Etanercept and Methotrexate with Radiographic Patient Outcomes trial (4% for the first year for etanercept, compared with 4.5% in our study), Maini et al (29) and Lipsky et al (2) in the Anti–Tumor Necrosis Factor Trial in Rheumatoid Arthritis with Concomitant Therapy trial (6% for the first year for infliximab, compared with 5.2% in our study), St Clair et al (30) in early RA, or Moreland et al (31) in the long-term followup of patients receiving etanercept. Lower rates were reported by Genovese et al (4).

The rates of nonserious infections as well as the infection rates in patients treated with conventional DMARDs vary widely. These rates are difficult to compare due to inconsistent reporting standards. Boerbooms et al (32) observed infection rates of 18–27% among patients receiving MTX, in a review of 3 randomized controlled trials involving small numbers of patients (≤33), and a rate of ≤4% in 1 randomized controlled trial and 3 open prospective studies involving >100 patient-years. In a retrospective cohort study of patients with RA, Doran et al (14) reported infection rates of 19.6 and 9.6 per 100 patient-years for objectively confirmed infections and for infections requiring hospitalization, respectively. Nevertheless, the treatment applied in the 1960s or 1970s was different from the treatment used today. Furthermore, nearly one-fourth of the patients in the study described by Doran et al were older than age 70 years at study entry (33), compared with only 6% of the patients in our study. This may explain the large difference in our study, because age is a risk factor for infection, especially in the elderly (25, 34).

Our results differ from those of randomized controlled trials in which similar infection rates were reported for the placebo arms and the etanercept (3, 28, 31) or infliximab arms (1, 2, 29). Nevertheless, recent findings by St Clair et al (30) support our results. In a randomized controlled trial of 1,049 patients receiving infliximab and MTX or placebo and MTX, those investigators observed significantly higher infection rates in the 2 infliximab groups (5.0% and 5.6% at 1 year) compared with the placebo group (2.1%). These rates are comparable with our findings.

One limitation of our study is a possible underestimation of milder infections (especially in the control group), which might not have been reported to the treating rheumatologist. Complete on-site monitoring is beyond the means of large cohort studies such as RABBIT. Studies such as ours have, by far, more complete coverage of adverse events and serious adverse events than do databases of spontaneous reporting systems such as MedWatch (35, 36). Nevertheless, studies such as ours cannot claim to have captured all adverse events.

The infection that causes the greatest concern is tuberculosis (6–9, 37, 38). This was already known at the start of RABBIT. By the end of the study, only 1 case of tuberculosis was observed among patients receiving infliximab. Other rare granulomatous opportunistic infections such as histoplasmosis, listeriosis, aspergilliosis, and nocardiosis were not observed. Nevertheless, the incidence rate of other severe lower respiratory tract infections, such as pneumonia, is higher in patients with RA (14, 29). The rate of pneumonia in RA patients receiving infliximab and MTX was higher than that in patients receiving treatment with MTX alone (29, 30). This difference reached statistical significance in a study of early RA (29). Because patients with RA and pneumonia have a higher risk of mortality (39), an increased risk of pneumonia, as shown by our data, is another special concern of anti-TNF treatment. Thus, infective arthritis or bacterial skin infections are likely to represent further safety problems associated with this treatment.

Cases of serious cellulitis were observed in randomized controlled trials of infliximab (29) and etanercept (4). However, these trials were not powered to detect significant differences in the incidence of this infectious disease. We observed no significant difference between the groups in the incidence of cellulitis but did observe a tendency toward a higher incidence of the more rapidly progressive infection, erysipelas, among patients receiving etanercept. Although the event rates for single diagnoses were too low to allow conclusions to be drawn, the total rate of bacterial skin infections was significantly higher in the TNF groups. In contrast to the evidence of an increased risk of bacterial skin infections in patients receiving anti-TNF treatment, the evidence in the case of viral infections is less convincing (38). Our data suggest an increase in the incidence of herpesvirus infections in general. However, the increase was especially attributable to more mild infections in the anti-TNF groups.

Sepsis is another major concern. However, in our study as well others (28, 29, 31, 40), the number of cases is too low to make comparisons between the different drugs. Furthermore, our data suggest that bone and joint infections are another concern in patients receiving etanercept. This is supported by single reports of septic arthritis by other investigators. Nevertheless, further confirmation is needed, because the number of cases is rather low.

The main objective of this study was to describe the frequency of serious and nonserious infections in patients with RA who receive TNF inhibitors compared with the frequency in patients receiving conventional DMARD therapy. However, it must be kept in mind that patients treated with biologic agents are likely to have a higher predisposition for infectious diseases. Therefore, it is obvious that the differences between patients and controls cannot be attributed solely to the new drugs. According to our estimation, nearly one-third of the increase in the rate of adverse events must be attributed to differences in patient characteristics. This estimate should be further validated by including more patients and other important risk factors (e.g., exposure to other immunsuppressive drugs, comorbid conditions, disease activity at the onset of infection) in the analysis.


  1. Top of page
  2. Abstract
  6. Acknowledgements

We thank the following rheumatologists, each of whom enrolled at least 25 patients: S. Wassenberg, G. Herborn (Ratingen); K. Babinsky (Halle); A. Kapelle (Hoyerswerda); T. Klopsch (Neubrandenburg); W. Demary (Hildesheim); G-R Burmester, R. Haux (Berlin); K. Rockwitz (Goslar); A. Bussman (Geilenkirchen); H-E Schroeder (Dresden); E. Edelmann (Bad Aibling); E. Wilden, T. Karger (Cologne); R. Dockhorn (Weener); A. Graessler (Pirna); B. Krummel-Lorenz (Frankfurt/Main); K. Krueger (Munich); W. Liman (Hagen); H. Soerensen (Berlin); W. L. Gross (Luebeck/Bad Bramstedt); C. Richter (Stuttgart); W. Ochs (Bayreuth); S. Schewe, T. Dexel (Munich); L. Meier (Hofheim); H. Tremel, V. Petersen (Hamburg); C. Kneitz (Würzburg); M. Zaenker (Eberswalde); M. Bohl-Bühler (Brandenburg); A. Gause (Elmshorn); T. Grebe (Attendorn); K. Alliger (Zwiesel); U. Lange (Bad Nauheim); D. Pick (Grafschaft Holzweiler). We thank P. Ramlau for statistical analyses and U. Kamenz and C. Bungartz for their careful monitoring of the study.


  1. Top of page
  2. Abstract
  6. Acknowledgements