: 593 Selected : 4 Specific drug : Ustekinumab : 1 Already included with previous MeSH terms : 3 Visceral surgery : 23 Review : 5 Case report : 1 Unrelated : 1 “ anti tumor necrosis factor AND orthopedic surgery ” : Total : 271 Selected : 1 Already included with previous MeSH terms : 6 Review : 4 Ca

Disease-modifying antirheumatic drugs (DMARDs) have become essential treatments in the management of patients with inflammatory rheumatic diseases. Their use has resulted in a marked improvement of disease control and a limitation of joint damage, although some patients still require subsequent corrective or joint replacement surgery. Due to their immunosuppressive effects, some DMARDs are associated with an increased risk of infection. The aim of this review is to discuss the available literature on the management of DMARDs during the perioperative period, particularly in the case of orthopaedic surgery. Conventional synthetic DMARDs such as methotrexate, sulfasalazine, leflunomide, hydroxychloroquine appear to be safe during the perioperative period. Conflicting results on biological DMARDs, mainly tumour necrosis factor antagonists, are reported in the literature, including both increased and unchanged risk of superimposed infections after surgery. Taking into account the available literature, we included some propositions for the management of patients who will undergo surgical interventions.


Introduction
The first-line treatment of rheumatoid arthritis (RA) comprises various conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) [1].Methotrexate is the cornerstone csDMARD that is used either as monotherapy or in combination with other DMARDs.If low disease activity or remission is not achieved with csDMARDs, treatment is escalated with the use of biological DMARDs (bD-MARDs), including monoclonal antibodies or soluble receptors acting on specific cytokines, co-stimulatory signals and B cells.More recently, targeted synthetic DMARDs (tsDMARDs) blocking cell signalling such as Janus kinase (JAK) inhibitors have been introduced for the management of RA.In addition to their pathogenic role in immunemediated inflammatory diseases, the pathways targeted by bDMARDs and tsDMARDs contribute to hosting defence against pathogens [2].As an example, the use of tumour necrosis factor (TNF) inhibitors is associated with an enhanced risk of infectious complications including various bacterial and viral infections as well as opportunistic infections [3].Despite recent treatment progresses, some RA patients experience end-stage joint disease with structural damage requiring surgery.One critical question is how to manage immunosuppressive agents in patients who will require surgery, in order to avoid the occurrence of complications after surgery, but also to limit perioperative disease flares.This article aims to review the evidence regarding the use of DMARDs during the perioperative period and provide some guidance to the clinicians caring for patients with RA and other auto-immune diseases.To do this, the Medline database was screened to address the question of drug discontinuation and occurrence of perioperative complications (infection and delayed wound healing) in RA patients treated with cs-or bDMARDs undergoing orthopaedic surgery.

Rheumatoid arthritis as a risk factor for perioperative complications
Several studies have attempted to examine surgical site complications (wound infection, delayed healing, and deep infection) among RA patients, unfortunately without always taking into account on-going DMARDs treatment.Even so, one would have to deal with factors such as disease activity that cannot be dissociated from DMARDs use, immunity impairment inherent to the disease, and concomitant glucocorticoid use.Table 1 summarises the conclusions of studies that have tried to give an overview of infection risks in RA patients, and aims to reflect the postoperative complications seen in various centres when no drug interruption guidelines are involved.The largest study on total knee arthroplasties performed between 2006 and 2010 compared the number of infections after surgery in 11 755 RA and 339 348 osteoarthritis (OA) patients and showed that the risk was higher in RA patients (4.5% vs 3.8% infections, p <0.001) [12].The results from the Nordic Arthroplasty Register Association included 13 384 total hip arthroplasties in RA and in 377 287 OA patients [7].The relative risk (RR) of infectious complications was higher in RA patients who underwent surgery between 2002 and 2010 (RR 1.4, 95% confidence interval [CI] 1.0-1.8;p = 0.05) as compared with those who had surgery between 1995 and 2001 (RR 1.1, 95% CI 0.8-1.5;p = 0.5).Although not formally proven, the authors raise the fact that this slight increase could be related to the introduction of TNF inhibitors in Scandinavian countries.The Norwegian Arthroplasty Registry [19] reported a significantly increased RR of 1.6 (95% CI 1.06-2.38;p = 0.027) of infectious complications in RA versus OA patients undergoing total hip or knee arthroplasty.In contrast, the risk of infections within 90 days following total hip arthroplasty was not significantly higher in 5565 RA compared with 132 834 non-RA patients (odds ratio [OR] for infection risk 1.47, 95% CI 0.90-2.41;p = 0.12) in another study [18].Despite these contradictory data, the larger studies consistently show a small but significant increase in risk of postoperative infections in RA compared to OA patients.

Method
We used a combination of the following MeSH terms: surgery, arthroplasty, infection, wound healing, surgical wound, rheumatoid arthritis, postoperative complications, perioperative period, orthopaedic surgery, anti-tumour necrosis factor, tocilizumab, abatacept, rituximab, belimumab, anakinra, tofacitinib, apremilast, methotrexate, leflunomide, sulfasalazine, detailed in appendix 1.We included 37 articles based on the exclusion criteria summarised in figure 1.Based on the titles and the abstracts, our first selection process allowed us to exclude all the studies that did not discuss the link between perioperative complications and immunosuppressive drugs.We did not find any publications on the perioperative use of some DMARDs, such as apremilast, tofacitinib, belimumab and anakinra.

Methotrexate
A study examining the perioperative use of methotrexate (median dose 10 mg, 2.5-25 mg) included 388 RA patients who were prospectively randomised into three groups: 88 patients in whom methotrexate was continued, 72 patients in whom methotrexate was discontinued two weeks prior to surgery and resumed two weeks after surgery, and a control group of 228 patients without methotrexate treatment [31].After a follow-up of one year, there was a significant reduction in the incident rate of infections in patients in whom methotrexate was maintained (2%), as compared to those in whom methotrexate was discontinued (15%, p <0.003) and the control group (10.5%, p = 0.026).With- in six weeks of surgery, RA flares were numerically more frequent in the methotrexate discontinuation group (8%) and the control group (4%) than in the methotrexate maintenance group (0%).A further logistic regression analysis showed a significant risk of infection with use of D-penicillamine, ciclosporin, hydroxychloroquine, chloroquine and prednisone.Sixty-five patients included in this study were reviewed ten years later, including 23 patients from the methotrexate continuation, 8 from the methotrexate interruption, and 34 from the control groups.This long-term follow-up did not reveal any case of delayed infections [32].
A retrospective study included 122 RA patients undergoing 201 orthopaedic procedures [33].Methotrexate was maintained in 77 cases and discontinued for more than one week prior to surgery in 21 cases.The remaining patients (103 surgeries) had not been treated with methotrexate.
There was no significant difference in terms of postoperative infection (3.9, 4.8 and 3.9, respectively) or occurrence of delayed wound healing (1.3, 9.5 and 7.8%, respectively).There was also no significant difference regarding the incidence of disease flares in the three groups (3.9, 14.3 and 6.8%, respectively).Eighty RA patients undergoing 129 hand surgical procedures without perioperative drug interruption were divided into various treatment groups, including methotrexate monotherapy, glucocorticoids only, methotrexate plus glucocorticoid, and no therapy [34].The authors observed four wound infections in the methotrexate groups (5%), and two in the group without therapy (4%).
A randomised non-blind prospective study examined the frequency of postoperative complications in 64 RA patients separated into two groups.In group A (32 patients, 50 surgical procedures) methotrexate was discontinued seven days before surgery; in group B (32 patients, 39 procedures), methotrexate was maintained [35].There was no postoperative infection but prolonged wound healing in six and four cases in groups A and B, respectively (not significant).
A prospective non-randomised study included 32 RA patients undergoing elective total joint arthroplasty who were divided into two groups.In group one, methotrexate was discontinued one week prior to and after surgery (n = 19 with 26 procedures), whereas in group two, methotrexate was maintained throughout the perioperative period (n = 13 with 16 procedures) [36].Four postoperative infectious complications occurred in group two, including two prosthetic infections, one infected joint fusion, and one deep wound infection, but none was observed in group one (p = 0.03).No disease flare was observed in either group.A systematic review including the studies described above concluded that low dose methotrexate is safe during the perioperative period, and associated with significantly fewer disease flares [37].

Leflunomide
Two studies provided conflicting results regarding the perioperative safety of leflunomide.A case-control study including 41 RA patients treated with leflunomide (82 orthopaedic procedures) versus 41 patients in whom leflunomide was discontinued four weeks prior to surgery (79 orthopaedic procedures) showed no difference in the infec-tion percentage (6.1 vs 6.3% local infections, respectively) [38].
A more recent prospective study included 201 patients with RA and psoriatic arthritis who underwent orthopaedic surgery and were treated with various combinations of methotrexate, leflunomide, etanercept, infliximab and glucocorticoids [39].An increased number of postoperative complications was observed, including infections, impaired wound healing and necrotic scars in patients treated with leflunomide compared to methotrexate (40.6 vs 13.6%).A logistic regression analysis confirmed that the complication risk was significantly higher for leflunomide (OR 3.48, 95% CI 1.31-9.24;p = 0.01), but not for the other drugs.Based on these results, the authors recommended stopping leflunomide without providing any specific indication regarding the timing and duration of treatment discontinuation.They also suggested that preoperative cholestyramine administration could help to reduce circulating levels of leflunomide.

Sulfasalazine
There is no specific study on the safety of sulfasalazine in patients with inflammatory arthritis undergoing orthopaedic surgery.However, a retrospective cohort study compared the number of surgical site infections following 1219 orthopaedic procedures in RA patients, in whom TNF inhibitors were maintained or discontinued four drug half-lives before surgery (infliximab 39 d, etanercept 12 d, adalimumab 56 d) [40].There was no significant effect of TNF inhibitors on the risk for infection (OR 1.5, 95% CI 0.45-5.2;p = 0.53).Interestingly, the risk of infection was lower among those treated concomitantly with sulfasalazine (OR 0.21, 95% CI 0.05-0.89;p = 0.035).The authors attributed the protective effect of sulfasalazine to its action on bacterial folic acid synthesis.

Hydroxychloroquine
Hydroxychloroquine did not show any statistically significant association with either delayed healing or postoperative infections [41].

Biologic disease-modifying antirheumatic drugs TNF inhibitors
The effect of TNF inhibitors on the risk of infectious complications in patients undergoing orthopaedic surgery has been examined in several studies.Most of these are retrospective and very heterogeneous regarding their design and conclusions.They include mainly RA patients and occasionally patients with other rheumatic diseases.Most of these studies compared the rates of infections in patients treated with TNF inhibitors versus csDMARDs without specifying whether the treatment was discontinued during the perioperative period.There is only one prospective randomised clinical trial that specifically compared the perioperative discontinuation versus maintenance of TNF inhibitors [42].We have arbitrarily divided the presentation of the results into two parts: the first includes all the studies in which the use of TNF inhibitors was associated with an increased risk of infections, and the second describes those that did not find any difference in the number of infectious complications.All these studies are summarised in table 2.

The use of TNF inhibitors was associated with an increased risk of infection
In summary, there are six retrospective [43,44,[46][47][48][49] and one prospective study [45] showing an increased risk for infection.Most of them do not provide information on whether TNF inhibitors treatment was discontinued before surgery.First of all, we describe the prospective study and then the retrospective ones.
A prospective observational non-randomised study evaluated the risk of septic arthritis among patients treated with TNF inhibitors (11 881) or csDMARDs (3673) [45].Patients treated with TNF inhibitors were more likely to develop septic arthritis than patients on csDMARDs (HR 2.3, 95% CI 1.2-4.4),with 20 events in the csDMARD group and 179 events in the TNF inhibitor group.A subsequent analysis showed that previous arthroplasty was associated with an increased risk of septic arthritis (HR 2.45, 95% CI 1.90-3.17) in both groups.
In a retrospective nonrandomised multicentre study, a significant increase in infectious complications was observed in RA patients treated by bDMARDs (TNF inhibitors, abatacept, tocilizumab) undergoing arthroplasties (OR 2.12, 95% CI 1.48-3.03;p <0.0001) [47].The prevalence of infections was 2.1% for bDMARDs (out of a total of 1626 arthroplasties) versus 1.0% in patients treated with csDMARDs (on a total of 29 903 arthroplasties).In contrast, the prevalence of infections was the same for other surgical procedures in the bDMARD and csDMARD groups (1.6 vs 1.0%This latter result must be interpreted with caution, as csD-MARDs were not interrupted perioperatively.Finally, the authors did not observe any delay in wound healing. A retrospective study analysed all consecutive total hip and total knee arthroplasties.5.7% developed a superficial wound infection and 0.7% suffered a serious infection leading to removal of the prosthesis [46].The use of bD-MARDs (OR 5.  [49].Within the inflammatory rheumatic disease group, the risk of infection was higher in TNF in-hibitor than in csDMARD users, with an OR of 2.54 (95% CI 1.08-5.97;p = 0.032) with TNF inhibitors and an OR of 2.49 (95% CI 1.06-5.84;p = 0.036) for csDMARDs.This risk increased when the TNF inhibitors had been given within one administration interval (i.e., without stop) with an OR of 10.05 (95% CI 1.17-86.29;p = 0.035).Furthermore, the highest rate of infections was observed in cases of elbow and foot surgery.Goodman et al. performed a meta-analysis in 2016 including eight observational and three case-control studies, covering a total of 3681 patients exposed to TNF inhibitors and 4310 with no recent exposure [66].Results showed a higher risk of infection in the TNF inhibitor group with an OR of 2.47 (95% CI 1.66-3.68;p<0.0001).They did not analyse the continuation or stop of the TNF inhibitors before the operation.

The use of TNF inhibitors was not associated with increased risk of infection
We found three prospective studies -two studies in which TNF inhibitors were maintained [42,58] and another that compared TNF inhibitor discontinuation with a group of patients not receiving a TNF inhibitor [60] -and 11 ret-  [58].No significant association was observed between infection rates (total deep and superficial infections) and medical treatment, although 3% infections in the group A compared with 5.3% in the group B were reported.
A study divided RA patients undergoing orthopaedic surgery (total joint arthroplasty) into two groups; one biologic (267 joints) with drug interruption between two and four weeks before surgery according to the Japanese guidelines [65] (principally TNF inhibitors), and one non-biologic (300 joints) [60].No significant differences in the incidence of complications were observed between the two groups (TNF inhibitor stopped vs csDMARDs continued) in terms of superficial and deep infections, which were observed in 0.37 and 1%, respectively in the biologic group vs 0.67 and 0% in the non-biologic group.
A retrospective study reported 11 patients (10 RA and one psoriatic arthritis) who underwent 16 operations.The TNF inhibitor was discontinued in 12 patients and maintained in four during the perioperative period.There were no serious systemic or wound infections.There was one episode of disease flare in a patient who stopped therapy [50].
In a retrospective study that included 50 surgical procedures in 30 RA patients, TNF inhibitors were maintained in 32 cases and withheld prior to surgery in 18 cases [51].
There was no major complication, but disease flares occurred more frequently in patients who discontinued therapy (p = 0.02).
A retrospective parallel cohort included 768 RA patients with 1219 orthopaedic procedures who were divided as follows: group 1, TNF inhibitor-naïve; group 2A, TNF inhibitors stopped (four half-lives), and group 2B, TNF inhibitors continued [40].There was no significant difference in the occurrence of infections between groups (OR A retrospective study including 39 RA patients on TNF inhibitors and 74 patients on csDMARDs undergoing orthopaedic surgery (ankle arthrodesis; total hip, knee, elbow or shoulder arthroplasty) found no differences in overall adverse events in wounds (csDMARDs: 6.8%, TNF inhibitors 5.1%, p = 1.0), nor in duration of wound healing and fever.This study, however, does not provide sufficient data on discontinuation timing [54].
A case-control study performed within the French national registry RATIO (Research Axed on Tolerance of bIOtherapies) included 20 patients with various rheumatic diseases (RA 90%, ankylosing spondylitis, psoriatic arthritis) treated with TNF inhibitors [55].Patients who presented with infection after a total arthroplasty of a large joint were compared with 40 controls.A rate of 67% of flares was observed.Increased steroid intake was associated with infection (OR 5, 95% CI 1.1-21.6;p = 0.03) as was previous arthroplasty infection (OR 88.3, 95% CI 1.1-7071; p = 0.04).Neither TNF inhibitor use nor duration of exposure showed a statistically significant increase of infection rate, but infections had a worse outcome in patients on TNF inhibitors.
A retrospective study including RA patients who stopped infliximab four weeks prior to 52 orthopaedic procedures (total hip, knee, elbow or shoulder arthroplasty; foot/hand surgery, arthroscopic synovectomy and fractures) showed that infliximab did not increase the risk of superficial and deep infections [56].
In a study on hand surgery, authors reviewed 35 Swanson's metacarpophalangeal joint arthroplasties performed on RA patients [57].They only interrupted their biologic treatment (TNF inhibitor, etanercept) two or three weeks prior to surgery and resumed after wound healing, while maintaining the other synthetic DMARDs, steroids or nonsteroidal anti-inflammatory drugs (NSAIDs).In addition, Review article: Biomedical Intelligence antibiotic prophylaxis was used.They observed no deep infection, only one superficial infection, one late suture granuloma, and one presumed rheumatoid flare.
A multicentre retrospective study reviewed 381 hip and knee arthroplasties performed on RA patients [59].No further details were provided on the timing protocol for withholding biological agents (94% TNF inhibitors) or synthetic DMARDs except that 67.5% of patients were being treated at the time of surgery.Based on the small numbers of surgical site infections [7], the authors did not find any significant association between treatments and infection, except for prednisone use exceeding 15 mg/day (OR 21, 95% CI 3.5-127.2;p <0.001).
A large retrospective study examined use of DMARDs and biological agents in 6024 RA patients undergoing surgery from the US database of the Veteran Health Administration [61].To determine the drug-withhold timing, they developed an algorithm.The authors did not report any significant increase in risk of surgical site infection or wound infection between perioperative drug continuation and interruption.However, they found an increased risk of generalised infection with bDMARD interruption after surgery (OR 9.2, 95% CI 1.99-42.60;p = 0.005), which may be explained by drug interruption because of an on-going postoperative complication, as well as a significantly higher risk of postoperative wound infection with an OR of 14.15 (95% CI 1.76-113.76;p = 0.013).Risk factors for surgical site infection and delayed wound healing were retrospectively analysed among 332 elective orthopaedic procedures performed on RA patients in various Japanese centres [62].Their bDMARDs (etanercept, adalimumab, infliximab, tocilizumab, abatacept) and csD-MARDs were interrupted prior surgery as per the Japanese College of Rheumatology guidelines [65].The authors did not find any association of infection or delayed wound healing with bDMARD or csDMARD use.Unfortunately, the small number of patients on new biological drugs (abatacept or tocilizumab) does not allow for any firm conclusions.
Another recent retrospective study of a larger sample examined risk factors for the same parameters (infection and wound healing) after 1036 various orthopaedic procedures on RA patients [63].TNF inhibitors (etanercept, adalimumab, infliximab, golimumab) were interrupted two to four weeks before surgery, again according to the Japanese guidelines.There was no significant correlation between infection or delayed wound healing and bDMARDs.A higher risk of infection with foot surgery (OR 3.167, 95% CI 1.256-7.986;p = 0.015) and delayed wound healing with total knee arthroplasty (OR 4.044, 95% CI 1.436-11.389;p = 0.008) was observed.

Tocilizumab: anti interleukin-6 receptor monoclonal antibody
A retrospective case-control study compared wound healing and inflammatory state in 44 RA patients treated with either tocilizumab or csDMARDs [67] who underwent upper and lower limb total arthroplasties.The mean time from last infusion of tocilizumab was 16.1 days (range 3-27).There were no differences in terms of C-reactive protein (CRP) levels (p<0.001),white blood cell counts, temperature, superficial or deep infections and no wound healing delays.
More recently, a multicentre retrospective study collected 161 cases of orthopaedic surgery (mainly arthroplasties: 89/161 cases) among 122 RA patients treated with tocilizumab [68].According to the Japanese guidelines [69], treatment was interrupted prior to surgery but with some variability among patients.The mean time between the last infusion and surgery was 23.5 ± 12.7 days, and the time for resuming therapy was 24.8 ± 12.7 days after surgery.The authors reported a small number of surgical site infections -two superficial (1.9%) and one deep (0.6%) infection -as well as a few cases of wound healing delays (14%) and disease flares (22.4%).In a multiple logistic regression analysis, no correlation was found between the duration of treatment interruption and disease flares (OR 1.033, 95% CI 1.013-1.055;p = 0.0015).Delays in wound healing were mostly related to concomitant use of prednisolone (OR 5.494, 95% CI 1.026-101.926;p = 0.0459).There were also significant correlations between foot and spinal surgery and delayed wound healing with ORs of 3.066 (95% CI 1.009-9.327;p = 0.0483) and 158.750 (95% CI 10.521-6993.007;p = 0.0002).

Rituximab: anti-CD20 antibody
The French Auto-Immunity and Rituximab (AIR) registry analysed data that included 94 orthopaedic and 23 abdominal procedures in RA patients [70].The mean time between the last rituximab infusion and surgery was 6.4 months (4.3-8.7).There were 12 postoperative complications in nine patients: eight surgical site infections, and one fatal case of septic shock.In a univariate analysis, spine surgery was significantly associated with postoperative complications (p = 0.048), but no association was reported regarding the timeframe between surgery and the last rituximab infusion.

Abatacept: CTLA4-IgG fusion protein
A retrospective study included seven RA patients undergoing eight orthopaedic procedures (knee, hand, foot and vertebral surgery) [71].Abatacept was discontinued on average 15.9 days prior to surgery (mean of 33.1 days in the perioperative period) but concomitant medications (prednisolone, methotrexate, sulfasalazine) were maintained.There was no case of postoperative surgical site infection, delayed wound healing or disease flare.

Ustekinumab: anti-p40 antibody targeting both interleukin-12 and interleukin-23
A retrospective study was performed on 131 psoriasis patients who underwent various surgical procedures (orthopaedic, cardiovascular, dermatological, dental extraction and neurosurgery) [72].Only 13 patients were treated with ustekinumab, and 118 patients were on TNF inhibitors (infliximab, adalimumab or etanercept).The treatment (TNF inhibitor or ustekinumab) was maintained in 87 patients and discontinued in 44 patients according to the guidelines of the British Association of Dermatologists, which advises that ustekinumab be interrupted 12 weeks prior to surgery (four half-lives) [73].Three wound infections were observed without knowing the type of treatment: two in the discontinuation group (knee arthroplasty), and one in the maintenance group (dental implant).Given its small sample, this study seems reassuring as far as the perioperative safety of TNF inhibitors is concerned, but

Review article: Biomedical Intelligence
does not really provide much information on perioperative ustekinumab use.

Discussion
Our review of the current literature shows that the number of clinical studies on the effect of DMARDs during the perioperative period is scarce.Furthermore, the limited numbers of patients included in most studies, as well as the sometimes poor study design, are major limitations to making recommendations with a good level of evidence.Finally, as several studies were observational, confounding factors may have introduced major bias leading to conflicting results and sometimes, wrong conclusions.Methotrexate is by far the best studied among all csD-MARDs, in particular in prospective randomised clinical trials addressing the question of treatment maintenance during the perioperative period.Based on the results of these studies, it seems reasonable to continue methotrexate administration during the perioperative period.This recommendation is not only supported by the absence of major safety signals, but also aims to avoid disease flares that may ultimately lead to use corticosteroids or an increase their dosage, thus promoting a higher risk of complications.However, one should also bear in mind that low median doses of methotrexate (10 mg/week) were used in most of the studies, and that the safety of larger doses, as currently used, has not been formally examined.The evidence currently suggests that other csDMARDs (HCQ, sulfasalazine) can also continue to be safely administered during the perioperative period.National recommendations on the use of bDMARDs in patients undergoing surgical procedures have been based primarily on limited evidence and mostly on clinical experience and expert opinions.Some of the recommendations, which vary greatly between countries, are based essentially on bDMARDs half-lives (table 3).In France, the CRI (Club Rhumatisme et Inflammation) adjusts its recommendations to discontinue TNF inhibitors according to the surgical site: two half-lives for interventions in a sterile mi-lieu such as cataracts, to four to five half-lives for high-risk procedures such as abdominal surgery [74].In the UK and in Japan, guidelines advise stopping TNF inhibitors between three and five half-lives prior to any surgical procedure [69,75].In Spain [76], like in Switzerland (Commission des médicaments des pharmaciens suisses) [77], guidelines recommend stopping administration without any further detail regarding the timing of discontinuation or the type of surgery.The German Rheumatology Society gives further recommendations on synthetic DMARDs, mainly based on expert opinion and drug half-lives.Preoperative interruption is only recommended for ciclosporin (1-2 days before surgery), mycophenolate mofetil (1-2 days before surgery) and azathioprine (1-2 days before surgery) [78].In 2017, The American College of Rheumatology published new recommendations with low and moderate level of evidence: they suggest not stopping cs-DMARDs and stopping bDMARDs based on the timing of administration rather than the half-life, i.e., schedule surgery one week after normal administration, for example 9 weeks for infliximab, and 3 weeks for adalimumab [79].
Various case reports [80,81] and a case-control study comparing postoperative inflammation [82] reported a lack of inflammatory response (CRP and fever) in patients treated with tocilizumab.These findings are consistent with the well-described role of interleukin (IL)-6 in the control of acute-phase protein production by hepatocytes and body temperature.Thus, physicians should be aware of the specificity of tocilizumab's mode of action when assessing a patient with a postoperative complication.Notably, they should not rely on CRP as a marker of infection since CRP elevation is blunted by IL-6 inhibition.Furthermore, and regardless of surgery, severe infections can occur in tocilizumab-treated patients with an OR of 1.78 (95% CI 0.98-3.23),including mainly skin and subcutaneous infections, when tocilizumab is used in combination with methotrexate.Adverse events were also noticed with tocilizumab 8 mg/kg in combination with csDMARDs, as compared with controls treated with csDMARDs, with an OR of 1.53 (95% CI 1.26-1.86)[83].
Regarding bDMARDs, the preoperative withdrawal will depend on several parameters.Decisions should ideally be made in a multidisciplinary setting involving the rheumatologist, surgeon, and specialist in infectious diseases in certain cases.Points to consider should include individual patient and disease characteristics, including the presence of comorbidities, previous infectious complications, the type of immunosuppressive therapy and the severity of the disease.High disease activity is a risk factor for infection [84], as is the presence of prosthetic joint [45].The current literature does not clearly address whether bDMARDs should be discontinued according to their half-life or administration interval, and it seems difficult to completely dissociate one from the other.However, the administration interval has the advantage of being easy to use in routine practice.The intrinsic risk of infectious complications associated with the type and site of surgical procedures should be assessed.Particular attention should be given to an articular pathology that is known to be a risk factor for subsequent postoperative complications [85].The same applies to the type and severity of the autoimmune disease.Indeed, uncontrolled systemic lupus erythematosus may have major consequences on morbidity or even mortality that may preclude the complete cessation of immunosuppressive therapy.The decision of bDMARD discontinuation will also depend on whether bDMARDs are used as monotherapy or in combination with methotrexate or other csDMARDs, since in the latter case the disease may still be controlled by the administration of csDMARDs, whereas patients left without therapy in the former situation are at risk of flares.Finally, it is common practice to withhold bDMARDs before surgery because of the increased risk (or fear) of infection and delayed wound healing.However, the benefits of such an approach should be carefully weighed against the side effects of increasing glucocorticoid doses to treat flares.

Key messages
-Owing to scarce and limited literature, there are major limitations to recommending the discontinuation of disease-modifying antirheumatic drugs (DMARDs).-Methotrexate is by far the best studied conventional synthetic DMARD (csDMARD) and it seems reasonable to continue administering methotrexate during the perioperative period.-National recommendations on the use of biological DMARDs (bDMARDs) vary greatly between countries, but they all propose stopping bDMARDs during the perioperative period.-Clinicians should be aware of a possible lack of temperature and acute-phase marker elevation in patients treated with the IL-6 inhibitor, tocilizumab.-Several points should be taken into consideration prior to bDMARD withdrawal, including the presence of comorbidities, previous infectious complications, the use of a concomitant csDMARD, and disease severity.

Table 1 :
Rheumatoid arthritis as a risk factor for postoperative infection.

Table 2 :
Overview of studies with TNF inhibitors.