Perioperative management of disease modifying anti-rheumatic drugs: Recommendations based on a meta-analysis

Abstract

The objective of this paper is to make recommendations for the perioperative management of antirheumatic treatment based on the best available evidence. A systematic review was performed including studies in which patients with rheumatic diseases treated with biological and non-biological disease-modifying antirheumatic drugs (DMARDs) had undergone surgery. A total of 5285 studies were recorded, of which 27 were finally included. These contained information on 5268 patients and 7933 surgeries. The majority were women (mean age 55 years) diagnosed with rheumatoid arthritis, and the most studied drug was methotrexate (MTX). The final recommendations include: maintaining treatment with MTX or leflunomide in the perioperative period in the absence of other risk factors for postoperative complications (Level of Evidence 1c, Grade D recommendation). Treatment with DMARDs should be temporarily suspended, or the surgery scheduled as far as possible from the last dose, and if there were other risk factors a space at least two doses (Level of Evidence 2c; Grade D recommendation).

Resumen

Con el objetivo de proponer recomendaciones para el manejo perioperatorio de los fármacos modificadores de la enfermedad (FAME) en pacientes con enfermedades reumáticas que van a ser sometidos a cirugía, se ha realizado una revisión sistemática de la literatura. Se realizó una búsqueda de todos los estudios publicados y de los resúmenes de congresos, recopilando 5.285 documentos, de los que finalmente se incluyeron 27 estudios que proporcionan información de 5.268 pacientes y 7.933 cirugías. La mayoría eran mujeres (edad media: 55 años), estaban diagnosticados de artritis reumatoide y el fármaco más estudiado fue el metotrexate (MTX). Las recomendaciones finales son las siguientes: mantener el tratamiento con MTX o leflunomida en el período perioperatorio en ausencia de otros factores de riesgo de complicaciones posquirúrgicas (Nivel de evidencia 1c; Grado de recomendación D) y con respecto a los FAME biológicos, suspenderlos momentáneamente o programar la cirugía lo más alejada posible a la última dosis, espaciando al menos 2 dosis si existieran otros factores de riesgo (Nivel de evidencia 2c; Grado de recomendación D).

Introduction

A high number of patients with inflammatory rheumatic disease are submitted to surgical operations, specifically orthopaedic surgeries, throughout the course of their illness. In Spain, just for rheumatoid arthritis (RA), it is estimated that 26% of patients will be subject to some orthopaedic procedure.¹ Surgical complications can vary, the rate of major complications in orthopaedic surgery for RA being 3.4 in every 100 patients per year.² The development of postoperative infections is of particular concern; they occur in around 2% or more² of interventions,^3,4 according to the series.

Discontinuing both biological and synthetic disease-modifying anti-rheumatic drugs (DMARDs) is habitual practice before surgical operations on patients with inflammatory rheumatic disease. The objective of this procedure lies in the immunosuppressive characteristic of these drugs, which theoretically increase the probability of postoperative infection. It is also based on the unverified belief that these drugs can affect surgical wound healing. On the other hand, discontinuing the primary medication for an inflammatory disease can lead to its reactivation, a situation associated with all kinds of complications, including an increased risk of infection. Consequently, other rheumatologists are reluctant toward said discontinuation. Data that support 1 practice or another are scarce, leaving many clinical questions unanswered, including whether discontinuation is really necessary, how long before the surgery treatment should be discontinued and how long after DMARD treatment should start again, etc.

The objective of this document was to develop recommendations for perioperative management of both biological and synthetic DMARDs that are used in treating rheumatic diseases. In addition, these recommendations were to be based on the best evidence available.

Materials and methods

A systematic literature review was performed, following the Cochrane⁵ methodology, of all studies where patients diagnosed with any rheumatic disease treated with biological or synthetic DMARDs were to have surgical interventions.

Search strategy

The search for references was done by 2 reviewers (BH and LC) using the following electronic databases: Medline (from 1950 to 14 June 2010), the Cochrane Library (from 1972 to 14 June 2010) and EMBASE (from January 1961 to 14 June 2010). The initial search was broadened with a manual search of summaries from the last 5 European League Against Rheumatism (EULAR) conferences (2007–2011) and American College of Rheumatology (ACR) conferences (2006–2010). In addition, all the bibliographical citations from the studies included were actively searched. The search results were processed by a reference manager in order to eliminate duplicates and select those that complied with selection criteria, based on titles and summaries. Articles with titles related to the subject, but without a summary, were included for closer reading.

Selection criteria, data collection and analysis

The selected studies included inflammatory rheumatic disease patients being treated with classic and biological DMARDs whose objectives were: (1) to compare perioperative strategies of DMARD treatment (discontinue vs. continue treatment), (2) to measure risk of using DMARDs in relation to surgical complications, and/or (3) to measure the frequency of complications. Studies concerning isolated clinical cases were excluded. Two independent reviewers (AB and LO) selected the articles, according to title and summary, and a third reviewer (LC) compared the selected articles. Four reviewers (AB, LO, MG and BH) performed a detailed analysis of the selected articles, gathering data (Table 1) independently on paper, and a fifth (DT) included these data in an Excel^® file. Quality was evaluated using the New Castle-Ottawa⁶ scale for risk of bias in observational studies and the Jadad⁷ scale for clinical trials. A meta-analysis would be performed (LC) if homogeneity existed in at least 3 studies (in study type, population and result measurement). It would also be performed in observational studies, as the few clinical trials existing were of low quality. The final level of evidence in supporting the recommendations was established based on the levels of evidence from the Oxford Centre for Evidence-Based Medicine.⁸

Table 1.

Data gathered from the articles.

Heading	Data	Details
Publication data	Author
	Journal and year of publication
Study characteristics	Study type
	Intervention	Controlled/open/randomised
	Observational	Cohort (prospective longitudinal observational), retrospective, case-control, multiple cases, other
Patient characteristics	Total no., mean age, sex distribution
	Base disease
	Diagnosis (indicate total no.)	Rheumatoid arthritis, spondylitis, psoriatic arthritis, connective tissue disease, etc.
	Activity	Is there a baseline measurement of activity before surgery? (yes/no), measurement of activity (DAS28, DAS, BASDAI, etc.), measurement of baseline activity
	Confusion factors
	Comorbidity (indicate no.)	Diabetes mellitus, concomitant malignancies, previous infections, kidney failure, peripheral vascular disease, heart disease
	Corticosteroid use	Indicate the no. of patients using corticosteroids, mean period of use, mean accumulated dosage, mean dosage in surgery
Perioperative DMARD	Indicate no. of patients in treatment	Methotrexate, leflunomide, sulfasalazine, anti-malarial drugs, combination DMARD therapy, anti-TNF and specifications, rituximab, abatacept, tocilizumab, DMARD + biological drug
	Indicate how many DMARDs or biological drugs were discontinued
	Mean period of discontinuation before surgery
	Mean period before restarting after surgery
Surgery	Type	Orthopaedic (arthroscopy of any joint, knee, hip, other OTS), dental surgery, digestive surgery, other surgeries
	Characteristic	Planned or emergency surgery
Result	Infection
	Local infection
	Systemic infection
	Abnormal scarring
	Other complications
	Death
	Reactivation of disease

Results

A total of 5285 documents were collected, of which, after eliminating duplicates and performing the first title and summary screening, 82 were selected and evaluated in detail. Of these, 56 were excluded due to causes indicated in Annex 1. Ultimately, 27 studies (Annex 2) published between 1991 and 2011 and the abstracts of 5 conference communications were included,^{11,23,79,82,84} in addition to 5 clinical trials,^{67,70,75,84,87} which were of questionable quality (all with a Jadad⁷ score of 2 or lower). The rest consisted of retrospective longitudinal studies, case series and 2 case-controls. Even though 2 cases were called case-controls, in reality they were retrospective comparative studies with different types of patients, not divided by result but by drug.^39,77 The authors were from the United States (n = 10), Japan (n = 7), the United Kingdom (n = 4) and France (n = 3).

In total, these documents provided information about 5226 patients on whom 6327 surgeries were performed. In this sample, 4128 (79%) patients were female and the mean age was 56.8 years (minimum: 17; maximum: 94). All patients were diagnosed with RA in accordance with ACR criteria,⁸⁹ with the exception of 3 studies: 20% of those in the Ruyssen-Witrand⁵⁷ study were diagnosed with spondylitis, and small percentages of psoriatic arthritis (4%) or juvenile idiopathic arthritis (JIA) (2%) were found in other studies.^73,85 The postoperative follow-up duration was obtained in 19 studies and had a mean of 6 months (0.5–24). Surgery type was difficult to specify, as this information was incomplete. From the data analysed, surgeries were typified as follows: 957 (15%) knee arthroplasties, 774 (12%) hip arthroplasties, 412 (6%) ankle and foot surgeries, 135 (2%) hand and wrist surgeries, 124 (2%) elbow surgeries, 114 (2%) shoulder surgeries, 64 (1%) arthroscopies and 1935 (30%) orthopaedic surgeries of another kind. In 1587 (25%) cases, the type of trauma surgery was not specified. Furthermore, 614 (9%) surgical procedures were included that were not trauma surgeries, but predominantly digestive surgery. Emergency surgery occurred in only 10 cases.

Comorbidity data and possible risk factors for postoperative infection were limited. Comorbidity was reported in 7% of the cases: ischaemic cardiopathy in 210 (4%) patients, DM in 120 (2%) patients and less than 1% for the group of high blood pressure (n = 21), kidney failure (n = 12) bronchiectasis (n = 9) and malignancies (n = 4). Previous report of infection was not found in any of the cases. In general, the description of the patients and the comparability of the study groups were not very satisfactory, except for noted exceptions (Table 2). Dixon²³ did not provide data on comorbidity or predisposing factors for developing infection, but he adjusted for them in multivariate reference models. In the Giles⁷⁴ study, comorbidity was analysed, but no numerical data were given. Furthermore, there were only references to results that had no significant influence. In the Fuerst⁷³ study, the author referred to the collection of comorbidity data, but did not provide them nor adjust for them in reference models of different drugs. The Alarcón⁶⁷ study did not specify characteristics of the population studied, despite being a clinical trial. The author simply referred to the comparison groups as homogenous.

Table 2.

Quality (absence of bias) of the studies included, in descending order.

Study	Selection	Comparability	Result	Comments
Grennan, 2001^75	^b	^a	^b	Jadad = 2
Alarcón, 1996^67	^b	^a	^a	Jadad = 2
Tanaka, 2003^87	^b	^a	^a	Jadad = 2
Sany, 1993^84	^b	^a	–	Jadad = 1
Carpenter, 1996^70	^b	–	–	Jadad 0^e
den Broeder, 2007^71	^c	^a	^c	Retrospective observational
Giles, 2006^74	^c	^a	^b,e	Infections case-control
Bongartz, 2008^28	^b	^a	^b	Retrospective observational
Dixon, 2007^23	^b	^a	^a	Prospective observational
Escalante, 1995^72	^b	^a	^a	Ambispective observational
Kawakami, 2010^39	^b	^a	–	Retrospective observational
Perhala, 1991^82	^b	–	^a	Retrospective observational
Hirano, 2010^28	^b	–	^a	Retrospective observational
Ruyssen-Witrand, 2007^57	^b	–	–	Retrospective observational
Fuerst, 2006^73	^b	–	–	Prospective observational
Bridges, 1991^69	^b	–	–	Retrospective observational
Jain, 2002^79	^b	–	–	Retrospective observational
Hirao, 2009^77	^a	^a	–	Series of cases
Murata, 2006^81	^a	–	–	Retrospective observational
Kanazawa, 2011^80	^a	–	–	Retrospective observational^d
Bibbo, 2003^68	–	–	–^e	Infections case-control
Hiroshima, 2011^78	–	–	–	Series of cases
Wendling, 2005^88	–	–	–	Series of cases
Talwakar, 2005^86	–	–	–	Series of cases
Arkfeld, 2007^11	–	–	–	Series of cases^d
Saech, 2009^83	–	–	–	Series of cases^d
Shergy, 2005^85	–	–	–	Series of cases^d

^a,b,cAccording to the New Castle-Ottawa bias scale for cohorts (adapted in EC) or case-controls studies.

^dOnly abstract available.

^eThis is a case-control study. This table refers to risk exposure, and not to the outcome, which would be the criteria for cases selection.

Data regarding previous rheumatic disease activity and treatment were not reported systematically. A report was found on some measure of activity in 21 articles. However, in only 1 case was this measurement an activity index (DAS28). In the rest, acute phase reactants or non-validated semiqualitative measures were used, such as the doctor's opinion, or 20% deterioration in the inflamed joint count. Regarding concomitant treatments, 1287 of the 2230 patients (58%) reported consuming corticosteroids with a mean dosage of 7.5 mg (5–10). Furthermore, in some studies, the use of corticosteroids was the only RA treatment before surgery. Synthetic DMARDs that some patients used before surgery were methotrexate (MTX) (11%) and leflunomide (LEF) (2%). The drug most studied was MTX, found in 9 studies, including clinical trials. Leflunomide was analysed in 2 studies, 1 being a clinical trial.⁸⁷ In 1399 (22%) cases, consumption of synthetic DMARDs was noted without type specification, thus making it impossible to obtain clear information regarding combinations. Biological DMARDs were used with 2033 (32%) patients, of which anti-TNFs were the most used. Tocilizumab and rituximab treatments were reported in isolated cases. Biological DMARDs were included in 8 observational studies but no clinical trials. The evidence found for drug treatment is provided in Annex 2. In several studies, no specific DMARD was studied, but rather all were studied as risk factors.

Risk of complications between strategies: discontinuation vs. continuation

Table 3 shows data regarding risk of complications in studies that directly compared strategies. We performed a meta-analysis of the surgical infections and abnormal scarring: the pooled odds ratio (OR) for presenting infection complications with any DMARD whose use was not discontinued in the perioperative period was 0.8 (95% confidence interval [CI], 0.6–1.4). There were no apparent variations between DMARD types (Fig. 1). The meta-analysis of abnormal scarring did not show a defined grouped estimator for any strategy; OR 1.4, 95% CI 0.2–7.7 (Fig. 2).

Table 3.

Compared risk of complications between discontinuing and continuing treatment in the perioperative period.

Study	Drug	Risk of complications if not discontinued^a
		Infections	Other	Resurgence
Alarcón, 1996^67	MTX	0.22 (0.01–5.41)	Abnormal scarring: 0.5 (0.01–15.7)	1.0 (0.02–54.5)
Bridges, 1991^69	MTX	12.6 (0.6–265.9)	Abnormal scarring: 12.6 (0.6–265.9)	–
Carpenter, 1996^70	MTX	16.3 (0.8–334.7)	–	1.6 (0.03–83.9)
Grennan, 2001^74	MTX	0.8 (0.2–41.7)	Any complication: 0.09 (0.02–0.40)	0.06 (0.0–1.15)
Murata, 2006^81	MTX	1.4 (0.2–12.6)	Abnormal scarring: 0.13 (0.01–1.5)	0.24 (0.05–1.3)
Sany, 1993^84	MTX	–	Abnormal scarring: 0.6 (0.2–2.4)	–
Tanaka, 2003^87	LEF	0.9 (0.3–3.5)	–	–
Dixon, 2007^22	Anti-TNF	0.6 (0.3–1.04)	–	–
Den Broeder, 2007^71	Anti-TNF	1.5 (0.4–5.2)	Abnormal scarring: 11.2 (1.4–90)	–
Ruyssen-Witrand, 2007^57	Anti-TNF	–	Any complication: 2.0 (0.5–8.6)	–
Talwakar, 2005^87	Anti-TNF	3.3 (0.05–197)	Any complication: 11.0 (0.7–187)	1.6 (0.04–57.6)
Wendling, 2005^88	Anti-TNF	0.6 (0.01–29.2)	–	0.03 (0.0–0.6)
Bongartz, 2008^28	Any	0.7^b (0.1–5.0)	–	–

a

Odds ratio and 95% CI calculated using crude data from the study if the study itself did not provide them.

b

Hazard ratio.

Figure 1.

Result of meta-analysis of infection complications risk when treatment was not discontinued during the perioperative period.

Figure 2.

Result of meta-analysis of scarring complications risk when treatment was not discontinued during the perioperative period.

Discontinuing methotrexate

In the Grennan⁷⁵ randomised clinical trial (RCT), it was evident that the incidence rate for complications was lower in the group that continued with MTX (2%) than in the group that discontinued its use (15%). The OR for complications after continuing treatment is 0.09 (95% CI: 0.02–0.40). Six months after surgery, no patient who had continued using MTX showed disease reactivation, compared to 6 (8%) patients who had discontinued. There were no variations of activity over the long term.

In the Alarcón⁶⁷ study, a RCT of strategies was not performed. The sample size was not large enough due to doctors abandoning the study because they did not consider it very ethical to continue or discontinue treatment. Patients (26) were randomised to receive MTX or a placebo before and after surgery (13 in each group) with an observation period of 12 weeks after surgery. There were 5 (38%) complications with MTX and 2 (15%) with the placebo. In each group, 9 (70%) knee surgeries were performed on only the patients who showed complications. There was no resurgence of the disease in any of the groups, and physical functioning after 12 weeks was comparable. However, the sample size prevented a reliable conclusion from being reached, as shown in Table 3 by the wide confidence intervals calculated. An observational study with the same group⁶⁹ analysed data for 38 patients undergoing MTX treatment, who underwent planned surgery. There were 8 complications among the 19 patients who continued with MTX until at least 2 weeks before surgery, compared to no complications in the 34 patients who discontinued MTX treatment 4 weeks or more before surgery. There were groups with other similar risk factors. The percentage of knee replacements was greater among patients who continued, as well as the percentage of diabetic patients. The evaluation was not performed independently and the confidence intervals were exceedingly wide.

Carpenter et al.⁷⁰ performed an open clinical trial, in which the surgeons decided, according to their preferences, whether to discontinue MTX 2 weeks before surgery or not. This was done without a blind assessment and thus had many biases. Even though the sample size was small and the confidence intervals were vague, a greater tendency toward infection was detected in the group that continued MTX treatment, as found in previous studies.

Sany⁸⁴ carried out a similar study, although with randomised assignment. Of the 32 patients who did not discontinue MTX use, 13% presented some kind of complication, compared to 19% who discontinued treatment at least 1 week beforehand. No group reported infection.

Murata⁸¹ performed a retrospective study of complications resulting from surgical procedures in RA in patients who had continued using MTX, compared to procedures whose patients had discontinued treatment at least 2 weeks before surgery. The groups were quite comparable, except that all patients in each group were referred by a different source. There were 4 complications in the group that continued with MTX and 3 in the group that discontinued its use. There were 3 cases of the disease reactivating in each group (5% of those who continued and 14% of those who discontinued).

In the Loza⁴³ review, the Sany and Grennan studies were meta-analysed, and no variations regarding morbidity related to surgical scarring were found between those who discontinued MTX treatment and those who did not (OR, 0.69; 95% CI, 0.23–2.02).

Discontinuing leflunomide

The Tanaka⁸⁷ RCT analysed the effect of discontinuing leflunomide over 4 weeks (2 before and 2 after surgery) on the rate of postoperative infections. Patients could also be treated with other DMARDs, but all treatments were discontinued before intervention in the 2 groups. The groups were quite comparable, both with more than 80% of patients using corticosteroids, although at low doses. The rate of infection was practically the same in both groups. The study did not provide data regarding resurgence of RA activity.

Discontinuing anti-TNF

In the British registry of biologists, Dixon²³ examined the risk of severe postoperative infection (30 days) associated with discontinuing or continuing anti-TNF treatment (28 safety days). Adjusting for age, sex, activity, diabetes and steroids, the OR for severe postoperative infection when anti-TNF use was discontinued was 0.56 (95% CI, 0.30–1.04). This corresponded to an infection rate of 7.3% continuing treatment and of 4.8% discontinuing (n = 1694). Van den Broeder,⁷¹ in a retrospective study, examined the combined risk of early infection (less than 30 days) and delayed infection in patients being treated with anti-TNFs, those who discontinued its use and those who did not, depending on whether or not the period up until the surgery represented 4 half-lives. The rate of surgical infection was 4% in patients not exposed, 5.8% in those exposed who discontinued use and 8.7% in those exposed who continued use. Perioperative use of anti-TNF drugs was not significantly associated with an increase in infection (OR, 1.5; 95% CI, 0.4–5.2), but was significantly associated with abnormal scarring (OR, 11.2; 95% CI, 1.4–90).

In the Ruyssen-Wytrand⁵⁷ study, the rate of complications among patients who discontinued anti-TNF use more than 5 half-lives before the surgery (36 surgeries) was 19.4% compared to 18.4% among those who discontinued use later or did not discontinue at all (P = .48). If use was discontinued more than 2 half-lives before surgery, the rate of complications was 17.6% compared to 30% among those who discontinued use later or not at all (P = .24).

In the 16 surgeries exposed to anti-TNF drugs in the Talwakar⁸⁶ study, no infection was found in the group that discontinued or the group that maintained anti-TNF use. One patient in the discontinuation group experienced resurgence (using etanercept). In the Wendling⁸⁸ study, no serious complications occurred, and no infection was found in any of the groups. There were 6 cases (12%) of moderate reactivation with each anti-TNF, and orthopaedic surgery was significantly correlated with discontinuation.

Discontinuing any disease-modifying anti-rheumatic drug

In the Bongartz¹⁵ study, discontinuing any DMARD at the moment of surgery was associated with decreased risk. However, its relationship with prosthetic infection was not statistically significant (OR, 0.65; 95% CI, 0.09–4.95).²⁸

Resurgence of activity

Reference to baseline activity of the disease before surgery existed in only 8 studies and only 2 provided numeric data.^75,77 In the postoperative period, an increase in phase reactants is typically produced and many habitual measures of disease activity thus remain altered. In the case of reactivation, if treatment was discontinued in the perioperative period, the pooled OR was 0.2 (95% CI, 0.05–0.7), in favour of not discontinuing. However, this meta-analysis had high heterogeneity (I² = 29.8%), mainly in the results of the anti-TNF studies (Fig. 3). In the Kawakami³⁹ study, which compared biological drugs to non-biological, the presence of arthralgia was used as a criterion for disease recurrence. Eleven cases of recurrence were found among those who discontinued anti-TNF use. Percentages of discontinuation were not given for any group, thus we cannot know what associations existed with discontinuation itself.

Figure 3.

Result of meta-analysis for risk of disease recurrence when not discontinuing treatment in the perioperative period.

In the Alarcón⁶⁷ and Carpenter et al.⁷⁰ studies, activity resurgence was not found in any group (neither among those who discontinued nor those who continued MTX treatment). However, they did not explain how they defined resurgence, nor did they provide baseline disease activity. In the Sany⁸⁴ study, all patients who discontinued MTX more than 4 weeks showed a resurgence, but the authors did not clarify how many discontinued for that length of time nor how resurgence was defined.

Kanazawa⁸⁰ demonstrated that, in 68 operations, activity resurgence (without definition) occurred in 3 patients who discontinued etanercept treatment more than 12 days. In addition, activity resurgence appeared in all patients who discontinued more than 21 days. The study concluded that the preoperative use of biological drugs did not constitute an independent risk factor for infection.

Other result measurements

Two Japanese studies were centred on the appearance of fever or increase in C-reactive protein (CRP) in patients with RA who underwent surgery. In the Hirao⁷⁷ study, body temperature and CRP were analysed in 22 surgeries exposed to tocilizumab and 22 exposed to a non-biological DMARD. At first, no complications were observed in either group. However, patients using tocilizumab did not show fever or elevated CRP. Hiroshima⁷⁸ analysed 8 surgeries in 5 patients using tocilizumab, comparing them to 16 using anti-TNFs and 16 using classic DMARDs. All patients using tocilizumab discontinued its use 4 weeks before surgery and restarted treatment 4 weeks after it. Temperature and CRP levels increased in both the anti-TNF and the DMARD groups, but not in the tocilizumab group. There was no comment as to whether there were complications.

Kawakami³⁹ found a greater difference (significant) between the pre- and postoperative CRP in those exposed and those not exposed to biological drugs. In addition, an association between the use of biological drugs and deep vein thrombosis (DVT) was found (OR, 3.0; 95% CI, 1.1–7.8), while MTX was not associated with DVT (OR, 1.2; 95% CI, 0.4–3.4).

Hirano⁷⁶ examined the time up until total recovery of the surgical wound and did not find variations between those exposed to anti-TNFs and those not. Nor were there differences in postoperative fever or anaemia between the 2 groups.

Comparison between drugs

Several studies did not provide comparative data between strategies, but they provided data on the risk of complications associated with drugs.

Regarding MTX, in the Grennan⁷⁵ article, a group of patients not using MTX was included. In comparing this group to those exposed to MTX, whether they had discontinued use or not in the operative period, the number of complications did not generally differ between the groups (OR, 0.75; 95% CI, 0.37–1.53). Furthermore, no group showed more reactivation (OR, 0.95; 95% CI, 0.33–2.72). Likewise, a group not using MTX was included in the Murata⁸¹ study. In comparing the rates of infections, no variations were found (OR, 1.05; 95% CI, 0.26–4.33), nor were they found in reactivation (OR, 0.89; 95% CI, 0.29–2.76). Perhala⁸² retrospectively compared the proportion of complications among RA patients exposed to MTX to those among patients not exposed, the figures being 9% and 6%, respectively (OR for infections, 1.5; 95% CI, 0.4–5.9). Jain⁷⁹ compared several postoperative results between 4 groups, divided according to the drugs being used at the moment of surgery. The perioperative guidelines were not modified in any group: 48 used only MTX, 30 used only prednisolone, 30 used both and 21 did not use any drug. There was a 5% rate of infection in the surgical wound among those using MTX and 4% among those who not using it (P > .05, but without adjustment for other risk factors).

In the case of leflunomide, Fuerst⁷³ used logistic regression to see the effect of continued treatment during the perioperative period with MTX, leflunomide, etanercept, infliximab or corticosteroids on postoperative infections. No association was found for any of the drugs, not even corticosteroids, except for leflunomide (OR, 3.5; 95% CI, 1.3–9.2). It was not clear whether the data were adjusted for risk factors of infection.

Regarding biological DMARDs, Arkfeld^10,11 compared the number of infections after elbow surgery in 11 patients exposed to anti-TNFs and 11 patients not exposed. Four (36%) of the exposed elbows were infected compared to 1 (9%) of those not exposed (calculated OR for anti-TNF, 5.7; 95% CI, 0.5–62.7).

Giles⁷⁴ performed a case-control to see the effect that anti-TNF treatment would have on postoperative infections. Anti-TNF therapy was significantly correlated with the development of postoperative infections in the bivariate analysis (OR, 4.4; 95% CI, 1.1–18.4) and after adjustment for age, sex, use of corticosteroids, diabetes and rheumatoid factor (OR, 5.3; 95% CI, 1.1–24.9).

Dixon²³ examined the risk of severe postoperative infection (30 days) associated with exposure to anti-TNFs (discontinued or not) compared to non-biological DMARDs during the surgical period. Adjusting for age, sex, activity, diabetes and steroids, the OR for severe postoperative infection using DMARDs compared to anti-TNFs was 0.75 (95% CI, 0.44–1.28). In the DMARD group, the rate of infection was 5.9% compared to 7.1% in the anti-TNF group.

Den Broeder⁷¹ compared those exposed and not exposed to anti-TNFs. The OR for surgical infections in those exposed compared to those not exposed was 0.8 (95% CI, 0.3–2.0). Furthermore, sulfasalazine was identified as a protective factor against infection, with an OR of 0.21.

Kawakami³⁹ compared the rate of infections among patients exposed and not exposed to anti-TNFs. These patients were grouped by age, sex and type of surgery. The rate of infection was clearly greater among those exposed (adjusted OR, 21.8; 95% CI, 1.2–386.1). In the Hirano⁷⁶ study, the total number of complications in the group using anti-TNFs (5%) was not different from that in the unexposed group (7%), with an OR of 0.7 (95% CI, 0.1–4.0).

Ruyssen-Witrand⁵⁷ provided the rate of postoperative complications with anti-TNF treatment at approximately 19% (24/127), including infections (9%), thrombosis (<1%) and scarring complications (5%). In the Shergy⁸⁵ study, the rate of infection with infliximab was 3% and the rate of complications in general was 9%.

In the Saech⁸³ study, 13 patients using rituximab who underwent orthopaedic surgery experienced a soft tissue infection and another experienced a urinary tract infection, although neither was severe. There were also 3 cases of abnormal scarring.

For other DMARDs, Bibbo⁶⁸ did not find any association between postoperative infection in RA patients who underwent foot or ankle surgery and those who were exposed to a non-biological DMARD. Escalante⁷² studied risk factors for complications and did not find any association with DMARDs, just with azathioprine (RR = 2.13; 95% CI, 1.04–4.4). The risk of complications was the same among surgical procedures in patients exposed to and not exposed to DMARDs. Furthermore, there was no difference for prednisone (RR = 1.3; 95% CI, 0.9–1.8).

Other factors associated with postoperative risk of complications

In addition to DMARDs, other factors (for both the patient and the surgery) important when selecting patients at greater risk were studied. Among the patient factors, the use of steroids^4,23, diabetes⁷⁹ and hypertension⁸⁰ stand out. Bongartz²⁸ demonstrated that RA is a risk factor for surgical complications. However, age was not identified as an important risk factor. Regarding the disease, no study found any association with duration, functional class or CRP levels before the operation.^28,81

Regarding factors of the surgery itself, Ruyssen-Witrand⁵⁷ found a rate of complications of 12% for orthopaedic procedures, and of 6% for infections, while 50% of abdominal procedures had complications, all of them infections. Furthermore, complications in emergency orthopaedic procedures had a rate of 20%. Den Broeder⁷¹ found a greater risk of complications in elbow, foot and hand surgeries. Kanazawa⁸⁰ found greater risk in knee surgeries. Bongartz²⁸ identified the presence of infections from previous operations as a clear risk factor.

Discussion

Upon initiating this review, we decided to include studies of any quality, since a previous search alerted us to the lack of clinical trials. Conclusions and recommendations should be prudent and based, if possible, on high quality studies.

Performing clinical trials in the perioperative context is complicated, as Alarcón⁶⁷ revealed in a clinical trial on perioperative strategies, which ultimately did not achieve the planned sample size. The main obstacles in the Alarcón study—besides budget cuts and little cooperation from surgeons in recruitment—were the preconceived ideas, from both rheumatologists and traumatologists, regarding how immunosuppression should be managed during this period. Strangely enough, the proportion of physicians who did not consider discontinuation to be very ethical was similar to that of those who thought the same of continuation. In both cases, these physicians abandoned the study. This polarisation in opinion also became evident in the Steuer⁶⁰ study, a survey of rheumatologists in which 35% of rheumatologists and 46% of traumatologists considered MTX to be clearly correlated with postoperative complications. Even in the same centre, it was difficult to predict which patients’ treatment would be discontinued, as the decision was not often based on the patient's age, the severity of their underlying disease or their comorbidities.²⁸ It was not even homogenous within the same centre.²⁷

Regarding the nuances that should be considered when accepting study conclusions as valid, we were able to show that the definition of discontinuation varied from 1 study to the next. In some studies, the definition was very sophisticated, with timetables, etc. Furthermore, it especially concerned us that the definition of a half-life should vary so much between the studies. To say a patient had discontinued medication before an operation was especially complicated in observational studies, since they were based on collecting the dates of the last dose before surgery and of the surgery itself. Consequently, there was a lack of supporting evidence, not for discontinuing or continuing medication, but for how long it should be discontinued.

On the other hand, the definition of complications was fairly constant, mainly regarding postoperative infections and abnormal scarring. That allowed us to perform a meta-analysis. Even though the definition for disease reactivation was unclear and not homogenous, our meta-analysis resulted in favour of continuing medication.

At the moment of deciding a perioperative strategy, it is important to consider other factors, primarily those that increase risk of infection, such as age, diabetes, kidney failure or the use of corticosteroids in medium-high doses.^3,22,23 There does not seem to be a firm relationship between clinical factors related to disease expression and complications. In a case-control concerning risk factors for developing infections—not included as it did not provide data regarding drugs—Hämäläinen³ did not find any association, not even with the previous disease duration, nor with the Steinbrocker criteria, ESR or the rheumatoid factor. However, factors related to surgery or admission did seem associated: hospitalisation period,³ day of hospitalisation (greater risk on Monday),³ ischaemia period³ and type of operation (greater risk in prosthetic knee and hand synovectomy^2,4). In addition, Hämäläinen³ highlighted another variable (collected in only the Bongartz²⁸ study) plausibly and clearly correlated with greater risk: the presence of infection in previous operations. In general, as use of corticosteroids was a constant risk factor among the studies,^23,25 it seems reasonable to not discontinue immunosuppressive treatment if discontinuation makes an increase in corticosteroid dose obligatory.

It is important to indicate that in this review we found studies primarily about RA and in planned orthopaedic surgery above all. While this could be the most frequent situation, it is not possible to generalise from this. Thus, we concluded that the data was in favour of continuing treatment in order to avoid reactivation, and also that no data supported discontinuation to avoid complications. However, perhaps it is more important to consider other factors besides drugs when deciding whether or not to discontinue medication during the perioperative period.

From this review it may be inevitably concluded that, in addition to the need to perform quality studies, comparisons should be made between discontinuation and continuation strategies of immunosuppression. Studies should include control of confusion factors and objective measures of results, including primary disease activity and complications.

In conclusion, it is recommended that when a patient with an inflammatory rheumatic disease undergoes surgery, risk of infection should be considered, in accordance with perioperative risk factors and DMARD type. The following are risk factors to be considered: age, diabetes, kidney failure or use of corticosteroids in medium-high doses, hospitalisation period, ischaemia period, type of operation (greater risk in prosthetic knee and hand synovectomy) and the presence of infections in previous operations.

For the patient treated with synthetic DMARDs who does not present other risk factors of postoperative complications—such as old age, diabetes, corticosteroid treatment, kidney failure or certain surgeries—maintaining treatment with MTX or leflunomide is recommended during the perioperative period (Level of evidence 1c; Grade of recommendation D). The evidence analysis did not support a specific strategy of discontinuing or continuing the use of immunosuppressive drugs, but there were data that identified diabetes, corticosteroids and some kinds of surgery as greater risks of complications; consequently, we thought it simpler to not make any modifications in treatment regarding the surgery. Furthermore, simple strategies are easier to achieve and expose patients to fewer safety problems. Maintaining treatment is thus the desirable option in most cases.

For the patient treated with biological DMARDs without other associated risk factors for postoperative complications, discontinuing treatment momentarily, or planning the surgery as far ahead as possible from the last dose, is recommended. In the presence of other risk factors for postoperative complications, such as diabetes or corticosteroid treatment, surgery should be put off for the period of at least 2 dosages (Level of evidence 2c; Grade of recommendation D).

Level of evidence

Level of evidence 3.

Ethical responsibilities

Human and animal protection

The authors declare that no experiments were performed with humans or animals for this study.

Data confidentiality

The authors declare that no patient data appear in this article.

Right to privacy and informed consent

The authors declare that no patient data appear in this article.

Conflict of interest

The authors have no conflict of interest to declare.