The efficacy of TNFα blocker is now well established in patients with rheumatoid arthritis (RA) 1 , ankylosing spondylitis (AS) 2 and psoriatic arthritis (PsA) 3 . Consequently, the prescription of these drugs becomes more and more frequent. Their use in patients with rheumatic disorders has led to less joint destruction and patients' functional prognosis has been greatly improved 4 5 6 . The requirement for total joint arthroplasty (TJA) tended to decrease in rheumatic patients before the use of TNFα blockers in rheumatology, thanks to strategies of earlier and more intensive management of recent rheumatic disorders 7 8 . The use of biologic treatments, and in particular of TNFα blockers, in rheumatic disorders will probably increase this downward trend. However, the need for TJA remains frequent, particularly due to joint destructions occurring before the introduction of TNFα blockers. Moreover, many patients already have one or more TJA at the time of TNFα blockers introduction.

The increased risk of tuberculosis and other opportunistic infections in patients receiving TNFα blockers is now well known 9 10 . An increased risk of serious bacterial infections in RA patients receiving TNFα blockers has also been established through two meta-analyses of randomized controlled trials 11 12 and retrospective cohort studies 13 14 , although other studies gave contradictory results 15 16 .

One of the most severe complications of TJA is surgical site infection, leading to long and expensive hospitalizations, complicated additional surgical procedures, increased mortality rates and severe functional disability. Despite systematic preventive measures, the risk of TJA infection persists and has been estimated at 1% for total hip arthroplasty and 2% for total knee arthroplasty 17 18 19 . Moreover, a two- to four-fold increased risk has been reported in RA 20 21 , although not found in other studies 22 . The role of treatments and particularly TNFα blockers in this increased risk remains unclear 23 . Some studies concluded a similar risk of postoperative infection after orthopedic surgery whether the patients were exposed or not to TNFα blockers 24 25 26 , whereas other studies highlighted a higher risk with TNFα blockers 27 28 reaching a two-fold increase 28 . However, these data remain controversial.

Other identified risk factors of TJA infections are systemic malignancy 29 , previous prosthetic joint infection of the index joint and of any joint 21 , arthroplasty revision 21 29 , increased operative time 21 and postoperative surgical site infection not involving the arthroplasty 29 . Nothing is known about the relevance of these risk factors in patients exposed to TNFα blockers.

The objectives of the present study were to evaluate the microbial agents, natural history and risk factors of TJA infections in patients receiving TNFα blockers, through a case-control study.

In the present study, TJA infection appears as a rare but potentially severe complication of TNFα blockers. Main risk factors are primary TJA or TJA revision for the joint subsequently infected within the past year and steroid intake.

Microbial agents identified in the present study were similar to those usually observed in TJA infections in patients having or not a rheumatic disorder 20 29 36 37 and in RA patients not exposed to TNFα blockers 38 . Nevertheless, most of the infections were related to MSSA. This was more often involved than in previous studies concerning TJA infections in patients having or not a rheumatic disorder (22 to 45%) 20 29 36 37 , and in RA patients not exposed to TNFα blockers (37%) 38 . No opportunistic agent was observed, but the low number of cases does not allow conclusions about the risk of opportunistic TJA infections in this population. TNFα blockers are known to compromise local wound healing 39 , so that an increased rate of polymicrobial infections could have been expected; however, the rate of polymicrobial infections (n = 2, 10%) was also consistent with data from the literature concerning TJA infections in patients having or not a rheumatic disorder (11 to 25%) 29 36 37 , and in RA patients not exposed to TNFα blockers (15%) 38 .

This study suggests that the outcome of TJA infections is particularly severe in rheumatic patients exposed to TNFα blockers, leading to hospitalization in an ICU in 20% of the cases, to death in 10% and to moderate to severe functional disability in about 40%. A high rate of bacteremia (60%) was observed in this study compared with 44% in a retrospective study assessing the natural history of TJA infections in RA patients not exposed to TNFα blockers 38 . Thus, TNFα blockers could increase the severity of the sepsis, but further controlled studies are needed to assess this hypothesis.

If TJA infections of large joints are often severe, their incidence appears rare in rheumatic patients exposed to TNFα blockers. The French national RATIO registry identified only 13 cases over two years. However, this study was not designed to estimate the incidence of TJA infections in these patients. The main objective of RATIO was to collect in an exhaustive way all over the country two rare side effects of TNFα blockers, which are opportunistic infections (including tuberculosis) and lymphomas. Severe documented bacterial infections (except pneumonias) were also collected, knowing that it was impossible to be exhaustive. Thus, we focused on specific types of severe infections such as TJA infections. Taking into account the fact that septic arthritis is frequent in RA, we were convinced that all cases of TJA infection have not been declared in the RATIO registry. But whatever it is, this study reporting 20 cases of TJA infection in patients treated with TNFα blockers is to our knowledge the largest of the literature concerning this peculiar complication.

Actually, the herein reported case-control study was designed to assess risk factors of TJA infection in patients exposed to TNFα blockers. Steroids are a classic risk factor of infection in RA patients exposed 13 16 or not 40 41 to TNFα blockers. This study confirms this data specifically for TJA infections in patients exposed to TNFα blockers. The fact that patients with TJA infection had more steroids than those without infection could also be an argument in favor of a more severe rheumatic disorder. No threshold of steroid intake was identified through this study, probably because of the small sample size.

Primary TJA or TJA revision within the past year was identified as an important risk factor of subsequent infection of this TJA. This could be due to bacterial perioperative contamination through a hematogenous way or a closed infected site, which is probably more difficult to control in the case of immunomodulation by TNFα -blockers. An infliximab-induced blood neutrophil deactivation has previously been demonstrated indeed 42 . Previous studies assessing orthopedic surgeries failed to demonstrate that the perioperative withdrawal of TNFα blockers reduced the risk of infection 24 28 43 . In the present study, TNFα blockers were less often withdrawn (stopped for more than five half-lives) before arthroplastic surgery in cases than in controls. This difference tended to reach statistical significance (P = 0.08), suggesting that TNFα blocker withdrawal could limit the infectious risk. No strong evidence exists but prospective controlled studies are not conceivable. In this context, withdrawal of TNFα blockers before surgery is now highly recommended by all societies of rheumatology and experts 14 44 45 . In this study, we considered five half-lives since the last administration of TNFα blockers for classifying patients as no longer exposed to the drug at the time of surgery. Recommended delays between last TNFα blocker administration and orthopedic surgery depends on each society of rheumatology. For example, the Dutch Society for Rheumatology recommends four half-lives for each TNFα blocker 45 , whereas the French guidelines established in 2005 recommend two weeks for etanercept, and four weeks for infliximab and adalimumab 44 . However, it is well specified in the latter guidelines that it is a minimum and that this time has to be increased in the case of high infectious risk surgery such as TJA setup.

The present study suggests that a previous TJA infection before the introduction of TNFα blockers could be a risk factor of reoccurence, although statistical significance was not reached. In case of previous TJA infection, high reoccurence rates of TJA (10% at 3 years and 26% at 10 years) 46 have previously been reported. In RA patients exposed or not to TNFα blockers, previous surgical site infection is an identified risk factor of postoperative infection after orthopedic surgery 43 and after TJA 21 . Interestingly, the British Society of Rheumatology recommended that previous sepsis on a TJA that remains in situ is a definitive contraindication of TNFα blocker use 47 .

Diabetes mellitus has previously been associated with an increased risk of infection after orthopedic surgery in RA 48 . The low number of cases (including only two diabetic cases) in this study did not permit to confirm this hypothesis concerning TJA infections in case of TNFα-blocker therapy. Based on data from the literature, methotrexate does not seem to modify the risk of postoperative infections after orthopedic surgery 48 , and particularly after TJA 43 , but previous results were heterogeneous 49 50 . Our study supports the absence of increased risk of TJA infection if methotrexate is added to anti-TNFα treatment. Regarding the type of TNFα blocker, there was no significant difference between cases and controls in our study.

The present study has several limitations and strengths. Some potential risk factors (recent skin infection, number of previous DMARDs, cumulative steroid intake, non-steroidal anti-inflammatory drugs) were not assessed because of missing data in several files. Even if it was prospective, the possible recruitment bias in the RATIO registry has already been detailed. There were only two Staphylococcus epidermidis infections although we would have expected more 51 . The case selection method may have led to underestimate the early less aggressive operatively-induced infections and to recruit the more severe hematogenous infections. The low number of cases has limited the statistical power of the study, and the retrospective recruitment of controls in a single center could have led to confounding factors. At the beginning of the study, no data published in the literature suggested a different risk of prosthetic infection between females and males, so we decided that controls would not be required to be of the same gender as the cases. However, a recent study highlighted a higher risk of revision due to deep infection on hip arthroplasty in males than in females 42 , and gender could be a confounding factor in our study. Controls were not selected in the same way as cases (i.e., through a national study) but in one center, which may induce bias. However, the tertiary care center where controls were selected is a referral center, therefore receiving patients having the most severe rheumatic disorders, which we believe selects controls close to the cases. Furthermore, the presence of a control group including two controls per case, even if it is imperfect, increases the validity of the results. To our knowledge, this study is the first case-control study assessing risk factors of TJA infection in rheumatic patients exposed to TNFα blockers.

TJA infection is a rare but severe complication in patients receiving TNFα blockers. Microbial agents do not differ from those usually identified in TJA infections, with Staphylococcus species involved in most cases. Two important modifiable risk factors have been identified: recent TJA setup in the previous year, that is primary TJA or TJA revision, in particular if TNFα blockers are not withdrawn before surgery, and steroid intake.

Practical implications of this study are the following. It may be preferable to perform arthroplasty, if needed, before the introduction of TNFα blockers. In cases of prosthetic surgery after the introduction of TNFα blockers, their withdrawal during the perioperative period is highly recommended. Finally, steroid intake should be reduced as low as possible in patients with both TJA and TNFα blockers.

AS: ankylosing spondylitis; CRP: C-reactive protein; DMARDs: disease-modifying anti-rheumatic drugs; IQR: interquartile range; MRSA: methicillin-resistant Staphylococcus aureus; MSSA: methicillin-susceptible Staphylococcus aureus; OR: odds ratio; PsA: psoriatic arthritis; RA: rheumatoid arthritis; RATIO: Research Axed on Tolerance of bIOtherapies; TJA: total joint arthroplasty; TNF: tumor necrosis factor.

RATIO has been supported by a research grant from the INSERM (Réseau de Recherche en Santé des Populations 2003 and 2006), and by an unrestricted grant from Abbott, Shering Plough and Wyeth, but these commercial sources played no further role in the herein reported work. The authors declare that they have no competing interests.

MG, LG and DS designed the study. MG, DG, MHG, JMB, DW, CM, and PD participated in data collection. LG and FT analyzed and interpreted the data. MG and LG drafted the manuscript. XM, MD and DS were involved in revising the manuscript critically for important intellectual content. All authors read and approved the final manuscript.