Background

Some material throughout this article has been reproduced with permission from Nakafero et al.,1 Leaviss et al.,2 Abhishek et al.,3 Abhishek et al.4 and Fuller et al.5 These are Open Access articles distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Rheumatoid arthritis (RA), inflammatory bowel disease (IBD), psoriasis (PsO) +/− arthritis (PsA), ankylosing spondylitis (AS) and systemic lupus erythematosus (SLE) affect > 4% of adults and are treated with immune-suppressing drugs such as methotrexate.1,6–11 These medicines can cause adverse events (AEs), such as acute liver injury, acute kidney injury (AKI) and cytopenia. Fortnightly-to-monthly blood tests are recommended when starting these treatments and 3-monthly testing is recommended once a stable dose has been maintained for a few months.12–15 At this time, the responsibility for prescribing and monitoring, including with blood tests, is usually handed over to the general practitioner (GP).12,13 The purpose of monitoring is to facilitate the early detection of an asymptomatic AE, allowing treatment to be stopped before clinically significant damage occurs. However, recommendations on the necessary frequency of testing within the various patient groups are inconsistent.12–15 The evidence on the optimum monitoring strategy in terms of the frequency and identity of tests for different patient populations is unclear. Many guidelines recommend fixed blood-testing intervals, for example 3-monthly, while others recommend more frequent testing in the presence of adverse prognostic factors.12–15 However, these prognostic factors are either not specified or narratively described in the recommendations.13–15

The overall purpose of these reviews is to inform the development of the optimum strategy of blood testing for adults with common inflammatory conditions treated with immune-suppressing drugs after the first 3 months of treatment. Review 1 will determine the type and frequency of those AEs that would be identified by blood monitoring at reported intervals (e.g. every 3 months). Review 2 will determine which prognostic factors have been found to predict the likelihood of one or more of these specified AEs.

Objectives

• Review 1: To identify the frequency of liver-, blood- and kidney-related AEs identified by regular monitoring of blood tests or otherwise, and any related dose adjustments/discontinuations of therapies in adults aged ≥ 18 years with inflammatory conditions treated with immune-suppressing drugs for > 3 months.
• Review 2: To identify patient and treatment factors that predict specified liver, blood and kidney AEs in adults aged ≥ 18 years with inflammatory conditions treated with immune-suppressing drugs.

Methods

Review 1

This was a conventional systematic review of AEs. The review was conducted in accordance with the York Centre for Reviews and Dissemination (CRD) guidance, reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Harms guidelines,16 and registered with the CRD PROSPERO database (CRD42020208042).

Protocol amendment: The published protocol was amended to include studies published in the year 1995 onwards (to allow contemporaneous cohorts) and with ≥ 200 participants. The latter approach was chosen because the AEs of interest were relatively uncommon, meaning that the likelihood of their being found and reported was much reduced in smaller studies.

Inclusion and exclusion criteria

Search strategy

A single search strategy for both reviews was developed in consultation with an information specialist (see Appendix 1). Thesaurus and free-text terms for the relevant populations were combined with terms for the interventions and validated study design filters for AEs and prognostic studies.17 The following bibliographic databases were interrogated in September and December 2020 to identify relevant citations: MEDLINE, Web of Science, EMBASE and the Cochrane library, from 1970 to date, without any language restrictions. Bibliographies of included studies and relevant systematic reviews and meta-analyses were reviewed manually to identify any additional relevant studies. We originally excluded studies from before 1970 as inflammatory conditions were almost exclusively treated with corticosteroids or non-steroidal anti-inflammatory drugs (NSAIDs) before that date.

Study selection

All citations and abstracts were screened according to the selection criteria (Table 1). Three reviewers (ME, CC, JL) independently screened 10% of the sample of the titles and abstracts of citations retrieved by the search to compare results for accuracy and clarity of the application of the inclusion criteria. Each reviewer then screened 30% of the remaining titles and abstracts to identify articles that satisfied the inclusion criteria. Two reviewers independently made a judgement on inclusion of the full papers (ME, CC); any disagreements on inclusion were resolved by discussion and, where necessary, consultation with a third reviewer (AA). If an article satisfied all the eligibility criteria except for information on frequency of monitoring, then the corresponding author was contacted to ascertain this information and enable the inclusion of that study in the review.

Data extraction and critical appraisal

The following data were extracted: location (country/countries), indication, drug(s), mean [standard deviation (SD)] or median dose, route of administration (oral, subcutaneous, intravenous, other), data collection time point(s) and details of blood tests; reported levels of blood-test result abnormality that required dose change or discontinuation; number or rates of specified AEs, rates of discontinuation or dose adjustment or interruption due to specified AEs. These data were extracted by one reviewer and validated by a second reviewer. Any disagreements were resolved by consensus or referral to a third team member (AA). Data on time points and details of blood tests were requested from authors if not reported.

Data on the following AEs were extracted:

1. Liver: acute/drug-induced liver injury (DILI), hepatotoxicity, hepatitis, liver failure, hepatic encephalopathy or coma, decompensation of liver disease, jaundice, chronic liver disease, granuloma, liver fibrosis, cirrhosis, fatty liver, nodular regenerative hyperplasia, sclerosing cholangitis, cholestasis; increased alanine transferase (ALT), aspartate transferase (AST), bilirubin or alkaline phosphatase.
2. Blood: anaemia, thrombocytopenia, leucopenia, lymphopenia, neutropenia +/− sepsis, blood or platelet transfusion.
3. Kidney: AKI, chronic kidney disease (CKD), nephritis, glomerulonephritis, haematuria, proteinuria, albuminuria, crystalluria, dialysis, end-stage renal disease, decline in glomerular filtration rate, elevated creatinine.

The Cochrane Risk of Bias tool 2.0 (RoB 2)18 was used to appraise risk of bias in randomised controlled trials (RCTs) and the Cochrane Risk Of Bias In Non-randomised Studies – of Interventions (ROBINS-I)19 was used for observational studies. These tools were modified to include additional questions on rigour of method used to detect AEs, and the quality of AE reporting, as recommended by Cochrane AEs Methods group. Judgements of high, low or unclear risk of bias for each domain was made by the lead reviewer (CC). A second reviewer independently quality assessed 10% of studies as this has been shown to be a cost-effective use of resources for very large reviews.20 Any disagreements were resolved by consensus or referral to a third team member (AA). No studies were excluded based on quality assessment.

Synthesis

Studies and outcome data were not sufficiently homogeneous [different populations, therapies (combinations and doses), follow-ups, outcomes and definitions of outcomes] to conduct a meta-analysis, so study characteristics and outcome data were tabulated and presented in a narrative synthesis. Studies and outcomes were subgrouped by therapy and follow-up, where appropriate.

Review 2

This was a systematic review of prognostic factor studies, that is of variables associated with the risk of a subsequent health outcome or event (e.g. AE) among people with a particular condition, for example, RA or exposed to a particular therapy, for example, methotrexate. It was conducted in accordance with PROGnosis RESearch Strategy (PROGRESS) framework (focusing on prognostic factor research)21 and Cochrane prognosis methods group,22 reported in accordance with the PRISMA guidelines,23 and registered with CRD PROSPERO database (CRD42020208049).

Protocol amendment: The revision applied to the protocol for review 1, that is only 1995 onwards and only ≥ 200 total study participants, was also applied to this review.

Inclusion and exclusion criteria were applied using the PICOTS24 system (Table 2).

Search strategy

A single approach was conducted for reviews 1 and 2 (see Appendix 1). To ensure the review was as current as possible, an update search restricted to 1 January 2020 onwards was conducted in January 2023.

Study selection

A single approach was conducted for reviews 1 and 2. For the update search, the screening was conducted by two reviewers (CC, JL), with disagreements or uncertainty on inclusion of full text articles resolved by discussion among them, and where required with a third reviewer (AA).

Data extraction and quality assessment

The following data were extracted based on a modified version of the CHARMS-PF checklist:22 location; population and sample size; outcomes to be predicted; start and end of follow-up period; index and comparator prognostic factors; missing data; analysis; and results [estimates and corresponding standard errors/SD or 95% confidence interval (CI)]. The effect sizes of interest [e.g. hazard ratios (HRs)], cut points and adjustment factors were also extracted. HRs were prioritised over rate ratios and odds ratios. We did not transform from one reporting scale to another.

Crude (unadjusted) and estimates additionally adjusted for other patient characteristics were extracted with the latter estimates prioritised for the evidence synthesis. All data were extracted by the lead reviewers (JL, CC) and validated by at least one other reviewer (AA, TC, DvdW, MG). Any disagreements were resolved by consensus or referral to AA. No attempt was made to contact the authors of included studies to enquire about missing or incomplete data. Where estimates and 95% CI could be calculated from raw data, this was calculated by one reviewer using Stata^® (StataCorp LP, College Station, TX, USA), and where only p-values and group sample sizes were available the Campbell collaboration effect size calculator was used.25

The QUality In Prognosis Studies (QUIPS) tool26 was used to appraise risk of bias. Judgements of high, low or unclear risk of bias for each domain were independently made by two reviewers (AA and JL). Any disagreements were resolved by consensus or referral to DvdW. Review findings were synthesised using an approach informed by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework to assess the quality of the evidence (certainty in the evidence) for each prognostic factor–outcome combination.27 Evidence from RCTs were considered to be high quality and those from observational studies were considered to be low quality. The quality of evidence was upgraded for large effect size (up one or two levels depending on the magnitude of the effect size), dose–response (up one level), and downgraded one level each for high risk of bias, imprecision, inconsistency and for single study. Data synthesis was performed independently by JL and AA, with disagreements resolved through discussion. Any uncertainties were discussed with DvdW.

Data synthesis

Given the range of conditions, treatments and time points being reviewed, the studies were heterogeneous. Study characteristics and outcome data were presented in a narrative synthesis. Inadequate reporting of prognostic studies, the limitations of indirect estimation methods and the uncertainties occasioned by conversion of different estimates of effect (hazard, risk and odds ratios) indicated that the pooling of the data in a quantitative meta-analysis was not appropriate.

Small study bias

Visual assessment of potential small-study bias with funnel plots was planned to be performed if effect estimates from > 10 studies for a prognostic factor were identified for a drug and AE–outcome pair. This was not the case in any instance.

Results and discussion

Numbers of included articles reviews 1 and 2

The search of bibliographic databases generated a total of 13,721 citations after deduplication processes (Figure 1). Most papers excluded at title and abstract stage were excluded for: no relevant outcomes reported; no discrete outcome data for relevant therapies; non-relevant combinations of therapies; follow-up < 3 months (review 1 only); and published as conference abstract or reviews. Two-hundred and eight papers satisfied almost all the criteria for inclusion, but 200 were excluded because they did not report information on the frequency of monitoring blood tests. The corresponding authors of all these papers were contacted for this information, which was provided in eight cases.28–35 The remaining 200 papers for which this information was unavailable were therefore excluded. For Review 1, the final number of included articles was 166.

FIGURE 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 flow diagram.23 Reproduced with permission from Leaviss et al.2 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

For review 2, two additional articles were identified by clinical experts,36,37 which were published after the first searches had been conducted. Both were also retrieved and identified in the update search for review 2. The number of included papers for review 2 from the original searches was 40. Seventeen articles reported relevant data for both reviews 1 and 2 and were therefore included in both reviews.31,38–53 Update searches (January 2023) for review 2 retrieved 2679 citations after deduplication; full texts were checked for 252 citations. A total of 234 articles were excluded for the same reasons as above. Eighteen new articles satisfied the inclusion criteria. The final total for review 2 was 56 studies (58 articles). The results of the two reviews are now reported in order (review 1, followed by review 2). The quality assessments are included in Appendix 2, Tables 37–44 for review 1 and Appendix 2, Tables 45–51 for review 2.

Results review 1

Characteristics of included studies. These are described in detail in Appendix 3, Tables 52–65 and Appendix 4, Tables 66–77.

Methotrexate was the most evaluated therapy assessed in 73 articles. The mean age of participants in the RA articles fell between 40 and 60 years, with only 1 exception in the 60 studies, which had a much younger populatiion;54 the mean age of PsO participants fell in the range of 41.9–47.6 years in all 7 studies, while the 2 IBD studies reported on participants with quite different mean ages: 2555 and 45.6 years.56

Methotrexate was assessed as monotherapy in at least 1 arm in 47 articles. One multicentre RCT evaluated named drugs (85–90% either methotrexate or sulfasalazine in each arm), with or without prednisolone, but only reported combined findings data.57

Anti-tumour necrosis factor-alpha drugs

Adalimumab

Adalimumab was assessed in 43 studies (43 articles). Evaluations were undertaken in populations with PsO (n = 15); RA (n = 14 articles); IBD (n = 6); AS (n = 5); inflammatory arthritis (n = 1); spondyloarthritis (n = 1) and a population including RA, PsO and AS patients.39 Twenty-five articles reported the findings of a multicentre RCT, or an open label extension (OLE); 2 reported a RCT, but the number of centres was not reported58,59 and there was 1 multicentre controlled clinical trial;60 12 were multicentre cohort studies, and 3 were single-centre cohort studies.

The mean age of RA patients fell between 50 and 60 years, AS patients between 29 and 46 years and IBD patients between 40 and 46 years; for PsO patients, the mean age of participants was between 39 and 51 years, except for one study that focused on an older population (range 65–85 years) and had a higher mean age (72.3 years).61 Thirty-five of the 43 studies had at least 1 arm of adalimumab (or a biosimilar), evaluated as monotherapy; only 8 articles evaluated combination therapy for adalimumab, always with methotrexate.62–69

Certolizumab

Certolizumab was assessed in seven articles of trials and cohort studies. Evaluations were undertaken almost exclusively in RA populations (n = 5 articles); one article assessed certolizumab in spondyloarthritis patients29 and one in PsO patients.61 It should be noted that only one patient in the PsO study received certolizumab: the principal therapies were adalimumab and etanercept.61 Five articles reported the findings of a multicentre RCT, and two cohort studies.61,70

The mean age of participants in four of the five articles studying RA populations fell between 50 and 56 years; the fifth study focused on older populations, which mean ages of > 80 years.71 The mean age of the AS patients was 37 years29 and the PsO patients was 72 years.61 A majority of patients in each study was female, except for the PsO study (38.3%).61 Three of the six studies assessed certolizumab in combination therapy, principally with methotrexate,71,72 but also with unspecified non-biologic background medication;29 only three studies evaluated certolizumab as monotherapy in any arm.61,70,73

Etanercept

Etanercept was assessed in 24 studies (25 articles). Evaluations were undertaken in populations with RA (n = 11 articles); PsO (n = 9); AS (n = 2); inflammatory arthritis (n = 1); and AS (n = 1). Twelve articles reported the findings of a multicentre RCT, or an OLE; one was a single-centre RCT;74 nine were multicentre cohort studies and one a pooled analysis of related trials.75

The mean age of participants and the proportion of female patients depended on the condition studied, although the mean age of patients in 23 of the 24 studies fell between 41 and 63 years; 1 study focused on an older population (range 65–85 years) and had a higher mean age (72.3 years).61 Where reported, the percentage of female participants was > 70% in the 11 studies of RA patients and 20–46% in the 9 studies of PsO patients. Twenty of the 24 studies had at least 1 arm of etanercept (or its biosimilar), evaluated as monotherapy; only 4 studies evaluated combination therapy of etanercept, with either methotrexate76–78 or sulfasalazine.79,80

Golimumab

This therapy was assessed in seven articles of trials and cohort studies. The included populations were diverse: one study each in RA, AS, IBD, PsO and spondyloarthritis, and two in mixed populations including AS, PsO and PsA patients.39,81

Where reported, the mean age of the populations in the included studies ranged from 38 to 72 years, and the percentage of female patients from 28% to 66%. Six and seven studies assessed golimumab as monotherapy in at least one arm, although only three reported separate data for this therapy,81–83 while three reported combined findings for this therapy and other anti-tumour necrosis factors (TNFs), the studies often including only small numbers of golimumab patients.39,61,84 Only one study evaluated golimumab and methotrexate as combination therapy.34,85

Infliximab

Infliximab was assessed in 30 studies (32 articles). Evaluations were undertaken in populations with RA (n = 12 articles); IBD (n = 6); PsO (n = 6); AS (n = 4); inflammatory arthritis (n = 1); spondyloarthritis (n = 1) and a population including RA, PsO and AS patients (n = 1). Eleven articles reported the findings of a multicentre RCT, or an OLE; 1 was a single-centre RCT;86 14 were multicentre cohort studies and 3 were single-centre cohort studies.

The mean age of RA patients fell between 46 and 62 years, and AS patients between 38 and 42 years; for PsO patients, the mean age of participants was between 41 and 47 years, except for one study that focused on an older population (range 65–85 years) and had a higher mean age (72.3 years).61 Where reported, the percentage of female participants was between 70% and 92% in the 12 studies of RA patients, 40% and 52% in IBD patients, 29% and 38% in the studies of PsO patients, and 19% and 35% in AS patients. Twenty-seven of the 29 studies had at least 1 arm of infliximab (or a biosimilar), evaluated as monotherapy; only 2 studies evaluated combination therapy of infliximab, both with methotrexate.86–88 Several studies evaluated multiple anti-TNF-alpha drugs.

Leflunomide

Leflunomide was assessed as monotherapy or in combination with another relevant therapy in one or more arms or groups in 16 articles of trials and cohort studies. Evaluations were undertaken almost exclusively in RA populations (n = 15 articles); one article included both RA and PsO patients.40 Six articles reported the findings of a multicentre RCT, two were single-centre RCTs, and eight were multicentre cohort studies; seven of which were prospective studies, and one a retrospective database study.89 The mean age of participants in 11 of the 15 articles fell between 50 and 61.3 years; the mean age of participants was not reported in 2 studies,90,91 and in 2 others was 42.6–44.6 years.92,93 Where reported, the percentage of female participants was generally > 70% in all studies.

Fourteen of the 15 studies had at least 1 arm of leflunomide monotherapy; 2 studies evaluated a combination of methotrexate and leflunomide.40,94 All studies evaluated the therapy at a dose of 10–20 mg/day, except one study that assessed the efficacy and safety of a single dose of 50 mg/week in one arm, compared with the standard dose of 10 mg/day in a second arm.93

Thiopurines

Azathioprine and mercaptopurine were assessed in 19 articles of trials and cohort studies. Evaluations were principally undertaken in IBD (n = 14 articles), but also in RA (n = 3), and SLE (n = 1). It should be noted that the number of patients receiving a thiopurine in 2 of the 3 RA population studies was relatively small: 4 out of > 100031 and 85 out of > 2000.95 There were only 3 RCTs; the remaining 15 articles all reported on cohort studies.

The mean age of participants in the RA articles fell between 54 and 58 years, with 63–81% female; the patients in the SLE study had a mean age of 40–42 years, and were > 90% female,96 while the IBD participants had a mean age that fell in the range of 25–45 years, and between 28% and 56% were female.

Azathioprine was assessed as monotherapy in at least one arm in 14 of 18 studies; 1 study reported findings that did not discriminate between either azathioprine or mercaptopurine.42 Only three studies evaluated azathioprine in combination with mercaptopurine and did not also evaluate either as monotherapy.47,50,97 Four studies (five articles) evaluated mercaptopurine as monotherapy.33,38,48,56

Mycophenolate

Mycophenolate mofetil (MMF) and mycophenolate sodium were assessed in four articles of trials and cohort studies. Evaluations were undertaken in SLE (n = 3 articles) and RA (1 article). Two articles reported the findings of multicentre RCTs, and one each a multicentre cohort study and a single-centre cohort study. The mean age of SLE participants fell between 30 and 42.1 years, and RA participants 55–56 years; and the percentage of female participants was > 85% in all three SLE studies, but < 78% in the RA study.

5-aminosalicylates and sulfasalazine

These therapies were assessed in 12 studies (13 articles). Seven studies were RCTs, four were multicentre cohorts and one was a single-centre cohort study. Evaluations were undertaken in a mixed population of PsA, reactive arthritis (ReA) and AS patients in one study98 and a SLE population in one study.99 Five studies evaluated sulfasalazine in RA populations, and five studies evaluated sulfasalazine and/or 5-aminosalicylate (5-ASA) in IBD populations, most commonly ulcerative colitis (UC). The UC populations tended to be younger, with mean or median ages ranging from 39 to 54 years, compared with the age range of 50–59 years in the RA population studies. The RA study populations always had > 60% female participants, and usually > 70%; a majority of the seven PsO studies had only 30–40% female participants; and the two IBD populations had 52–59% female participants. The UC studies also tended to have a lower percentage of female participants (43–54%) compared with the RA studies (63–81%), and the SLE study (> 82%).

Sulfasalazine was evaluated as monotherapy in at least one arm in nine studies,31,79,80,95,99–103 and in combination with etanercept in one study,79,80 methotrexate in one study.94 Four studies had at least one mesalazine monotherapy arm;101,103–105 5-ASA therapy was not evaluated in combination therapy. Non-specified and multiple drugs were evaluated in some studies.

Quality assessment review 1

The results of the risk-of-bias assessments for each included study are reported within the narrative synthesis for each therapy or combination therapy, but some of the key points are summarising briefly here. The full details of the assessments are reported in Appendix 2, Tables 37–44.

Randomised controlled trials

These studies were assessed using the RoB 2.18 Most trials were assessed as having some concerns regarding bias due to deviation from the intervention, either due to the absence of blinding or any information on blinding, as well as potential concomitant medication usage that was not balanced across arms and could affect outcomes.

Safety data sets usually include all patients, so the risk of bias from missing data was low across almost all these trials. Further, in terms of outcome measurement, given that this review only assessed outcomes measured using objective methods, for example, blood tests, the risk of bias in outcome measurement was also often low. However, even studies with protocols rarely detailed the safety outcomes, so that the selective reporting domain was frequently also categorised as ‘Some concern’.

In terms of the overall risk-of-bias assessment, in accordance with the Cochrane guidance, if any domain was assessed as being at high risk of bias, then the overall risk was ‘High’. If all domains were assessed as at low risk of bias, then the overall categorisation was ‘Low’. If only one domain was judged to have some concerns, and the remainder were judged to be low, then the overall risk of bias was categorised as ‘Low’. If two domains were judged to have some concerns, and the remainder were judged to be low, then the overall risk of bias was categorised as ‘Some concern’. If at least three domains were judged to have some concerns, then the overall risk of bias was categorised as ‘High’.

Non-RCTs and cohort studies

These studies were assessed using the ROBINS-I tool.19

Generally, there were no issues with confounding. It was only present in some single-cohort or extension studies where patients had all received different interventions before the start of the single-cohort period (potential ‘time-varying confounding’)66,81,91,106 or where the indication for treatment with the relevant therapy differed among participants.42,107 Populations were usually very clearly defined, but in the case of retrospective cohort studies there was a risk of selection bias,39,108 although such study designs also tended to have a low risk of missing data bias as all participants with data were identified by the design.48,95

Bias in terms of intervention classification was common because full details of the interventions being received by the participants were often lacking: the drug might be listed, but no dose or frequency information given, only numbers of participants exposed to an intervention (serious risk of bias), or dose and frequency were reported, but precise numbers of participants receiving each relevant intervention were not reported (moderate risk of bias), or none of this information was reported: dose, frequency and numbers (critical risk of bias).

Bias due to deviations from the interventions being evaluated (pre-specified or unplanned cointerventions that might affect an outcome) occurred where some participants were exposed concomitant medications, but others not (especially if this was not balanced between groups109) or participants in a single cohort were exposed to different combinations of interventions, as well as being without discrete outcome data for individual interventions.42,47,48,61,83,84,110

As with the RCTs, safety data sets usually include all patients, so the risk of bias from missing data was usually low across these studies. However, sometimes the incidence rates of AEs were reported, without reference to the percentage of patients experiencing such events, so that it was unclear if all patients were included in the safety data set.111,112

In terms of outcome measurement, the AEs of interest were also often not defined, but might be listed only as ‘haematological condition’.113 The risk of bias was categorised as ‘No Information’ for such studies. However, given that this review only assessed outcomes measured using objective methods, for example, blood tests, the risk of bias in outcome measurement was often low. In the same way, even studies with protocols rarely detailed the safety outcomes, so that the selective reporting domain was frequently also categorised as ‘No information’. However, some studies were only interested in a single outcome, for example neutropenia, and so the risk of selective reporting bias for outcomes was low.31,40

Synthesis review 1

The full details of the outcome data for each therapy are presented in the tables in Appendix 3, Tables 52–65.

Methotrexate

Fifty-two studies evaluated relevant AEs for patients receiving methotrexate monotherapy (Tables 3–7). Forty-three of the 52 studies were conducted in populations with RA; 3 studies were conducted in RA and PsO populations;32,40,51 4 in PsO populations alone;135–138 and 1 study each in populations with inflammatory arthritis114 and IBD.55 Thirteen studies had a follow-up of approximately 6 months and performed monthly monitoring.67,71,72,76,114,115,120,122,126,130,135,139,140 Eight studies employed 1- to 2-month monitoring for as long as 1 year138,141,123 or 2 years,45,116,124 and even for as long as approximately 5 years or more.95,107,131

Twenty studies performed approximately 3-monthly monitoring for a variety of follow-up periods: 4 months,136 6 months,63,125,128,137,143 1 year,34,55,74,94,142,144 and up to 2 years32,118,119,121 or longer.44,51,127,132 Small numbers of studies employed different approaches, such as 6-monthly monitoring for either 1 year,55,86,117,144 18 months31 or 2 years.90

For the studies with RA populations, 12 of the 30 RCTs were assessed as having a low risk of bias,34,67,115,116,117,118,120,122,123,140,141,144 6 were assessed as having some concerns71,94,119,121,125,126 and 10 as being at high risk of bias.43,72,74,76,90,124,128,133,139,142 Of the 14 cohort studies, 1 was assessed as being at low risk of bias,86 7 at moderate risk of bias,44,45,107,127,130,131,132 4 at serious risk of bias31,56,63,143 and 1 at critical risk of bias.95 For the three cohort studies conducted in combined RA and PsO populations, two were assessed at moderate risk of bias40,51 and one at critical risk of bias;32 for the four studies in PsO populations alone, the RCTs were assessed at being at low risk of bias137 or having some concerns;135,138 and low risk of bias for the single cohort study.136 The cohort studies in populations with inflammatory arthritis114 and IBD55 were both assessed as being at moderate risk of bias.

Where full details of dosage were reported, patients in most studies were exposed to a methotrexate dose of between 7.5 mg and 20 mg per week. The only exceptions to this was one study with a lower threshold end of the dose: 5–25 mg/week,32 and six studies with a higher threshold for the upper end of the dose: 7.5–25 mg/week in RA patients44,123 and inflammatory arthritis patients;114 10–25 mg/week130 or 15–25 mg/week in either RA patients107 or IBD patients;55 5–25 mg/week in PsO and 5–30 mg/week in RA patients;32 and 10–30 mg/week in RA patients.116

Blood

The incidence of most reported AEs, such as anaemia, neutropenia, leucopenia and thrombocytopenia, was < 2% where reported. Eight studies reported discrete data for anaemia as an outcome: four reported an incidence < 2%, two slightly higher (< 4%) and two between 5.5% and 7.5% at the < 10 g/dl threshold, with higher rates for patients with early RA.

Ten studies reported discrete data for neutropenia as an outcome: three recording no cases at all; five reported an incidence < 2.3%; two studies reported higher incidence rates; these were for transient neutropenia74 and included grade 1 neutropenia: 5.9% of the 8.1% recorded.31

Sixteen studies reported discrete data for leucopenia as an outcome. Twelve of the 16 studies, with follow-ups ranging from 6 to 24 months, and monitoring monthly to 3-monthly, reported percentages of patients experiencing leucopenia as < 1% (6 studies) and 1.6–3.8% (6 studies), including 1 study that reported an incidence of 1.6% in the first 12 months and 3.1% in the second 12 months.124 A 9-year study also reported an incidence rate of 2.5%.107 The studies with the highest incidence (7.3% and 10.6%, respectively) had follow-ups of 3–4 years127 or almost 5 years,45 or had only a small number of patients in the methotrexate arm (3/34 patients, 8.8%).126

Eight of 11 studies with follow-up ranging from 6 to 58 months, and monitoring monthly to 6-monthly, reported that < 2% of patients experienced thrombocytopenia; the 3 studies with higher incidence either reported only mild thrombocytopenia (< 3%),123 had only a small number of patients in the methotrexate arm (2/34 patients, 5.9%)126 or had a follow-up of 3–4 years.127 The incidence of pancytopenia and pancreatitis was < 1% across a total of six studies with up to 2-year follow-up and 3-monthly monitoring.

Discontinuations due to blood AEs were still less common: from no reported cases at all107,122,124 or < 1%31,51,131 to 1.9%.132 For specific AEs, discontinuation rates were: zero;124 < 1% for anaemia;114 < 2.1%45,123 and < 3%126 for leucopenia; 2.3–5.5% for neutropenia,32,114 and from 5.5% to 6.8% for thrombocytopenia.32

Liver

The incidence of hepatic function abnormalities ranged from 1%135 to 12.8%,131 across 15 studies with follow-ups ranging from 6 months to approximately 12 years, and with monthly, 3- or 6-monthly monitoring. The only incidence higher than 12.8% was 27% for PsO patients (the same study reported 12% for RA patients).32

Two studies reported elevated plasma liver enzyme levels,124,142 with an identical rate of 16.3% for the first year, and one study a reduced rate of 5.9% in the second year,124 and discontinuation rates of 3.2% and 1.6% for those time periods, respectively.124

Otherwise, elevated liver enzyme levels were reported either as unspecified or for ALT or AST. Only four studies provided unspecified enzyme levels, all with 1-year follow-up and 6-monthly monitoring: two reported elevations affecting between 13% and 53% of methotrexate-exposed patients, depending on therapy (53% for methotrexate alone, 13% for methotrexate + folinic acid123) and condition: 22% for RA and 35% for PsO.40 Where levels over the upper limit of normal (ULN) were reported, the rates were ≤ 1% for > 2 × ULN (Curtis 2010).40 Specifically for ALT, incidence rates for any elevation over the ULN ranged from 1.6% to 44% across 8 studies, all in RA populations, with 2 falling in the range of 1.6–4.3% at 6 months,72,140 5.9% at 2 years118 and 5 in the range of 20.6–44% between 6 months and > 4 years.44,74,116,122,126 In the case of one of these studies, only 3% of patients had therapy discontinued, although the incidence rate was 39%.44

Seven studies reported data for 2 × ULN, with an incidence rate ranging from 1.3%143 to 15%;44 the rate was 6% in the one study reporting 3–4 × ULN. Discontinuations were rare, for example ≤ 3%,44,120,122,126,130 with the exception of one study that had much higher rates.123

Incidence rates for any elevation over the ULN ranged from 3.8% to 32% across six studies, though a seventh study with approximately 5 years follow-up reported a rate as high as 63% (only 3.5% discontinued), and 52% for 1–2 × ULN.45 Three studies reported data for 2 × ULN, with an incidence rate ranging from < 1% to 7.9%; the rate was 1.9% in the one study reporting 3–4 × ULN, which also reported < 1% for higher levels.45

It should be noted that one study found that 21% of patients experienced grade 1 elevated ALT in months 1–6, but only 5% in months 6–12, and grade 2 ALT was experienced by 6% in months 1–6 and 2% in months 6–12.117 Lower levels of AST were reported (≤ 2%) and there was less difference between the first and second 6 months.117

Discontinuation rates due to liver function tests (LFTs): ≤ 1%,127,135 1.9%,132 2.3%,131 5.3–5.5%;114 and 40.1% over 5–6 years in one study with monthly monitoring.95 One study reported rates of 12–27.3%, with higher for PsA patients compared with RA patients.32

Otherwise, serious AEs (SAEs) were rare: there were no cases of serious ALT elevation, abnormal hepatic function or hepatobiliary disorders in one 1-year study with 6-monthly monitoring,117 and rates of ALT SAEs were < 1% in one study122 and < 4.5% in another.115

Kidney

Five studies reported renal function abnormalities; four of the five studies reported an incidence rate of < 1% up to 1 year, with monthly or 6-monthly monitoring; the study with the highest incidence had only a small number of patients in the methotrexate arm (2/46 patients, 4.3%).94 Five studies reported abnormal elevation of creatinine; three of the five studies reported an incidence rate of < 1% up to 2 years with monthly or 2-monthly monitoring; two studies reported a higher incidence of up to 3%.117,118 Discontinuations due to any renal AEs were also rare. Only one included study reported any cases: an incidence rate of 1.5% in RA patients and 0% in PsO patients at 18–21 months with 3-monthly monitoring.32 Four studies reported renal function abnormalities, and only one of these studies reported an AE rate of > 1% with 3-monthly monitoring and 1-year follow-up.94

Methotrexate in combination therapy with anti-tumour necrosis factor-alpha

Methotrexate and adalimumab

Eight studies evaluated relevant AEs for patients receiving a combination of methotrexate and adalimumab (or a biosimilar). Seven of the eight studies were conducted in populations with RA; one study had an IBD population.56 Four of the six RCTs were assessed as having a low risk of bias,62,65,67,69 and two assessed as having some concerns.64,68 Both cohort studies were assessed as being at serious risk of bias.56,63 Where reported, the final dose of methotrexate in any arm ranged from 7.5 mg/week to 25 mg/week, except in two trials which also had arms with doses as low as 2.5 mg/week and 5 mg/week.62,64 Adalimumab dose was 40 mg/every other week. Of the seven studies with outcome data, five had 6–month follow-ups, and performed monitoring every 4–6 weeks;62,64,65,68 one study had a 48-week follow-up, again with monitoring performed at every visit;67 and one study had a 1-year follow-up, with monitoring every 3 months;69 and one study a median duration of approximately 1 year, but monitoring was performed monthly.56

The incidence of most reported AEs, such as anaemia, neutropenia, thrombocytopenia, hepatitis, and hepatic and kidney events, was < 2% and was not reported by many studies. Discontinuation due to any of blood or liver AEs was still less common. Only elevated ALT and AST was reported more frequently, with a much higher incidence among patients, depending on details of therapy and how far the elevation exceeded the ULN, but treatment discontinuation, where reported, was rare (< 1%).

Blood AEs:

• Anaemia data were reported by four studies.56,62,63,68 The incidence was ≤ 3.6%, with the highest percentages being reported for patients with early RA (3.6%)63 or exposed to a combination with the adalimumab biosimilar PF-06410293 (3%).68 Discontinuation for anaemia was only reported by one study, which occurred in one out of eight patients.56
• Neutropenia was reported as an AE by three studies.64,67,69 There was no case reported in one study;64 < 1% incidence of grade 3 and 4 neutropenia in another study, with 6-week monitoring,67 and only 0.5% discontinued due to neutropenia in a third study, with 3-month monitoring.69
• One study reported each of the following: leucopenia at ≤ 1%;62 thrombocytopenia at < 1%;64 haematological reactions at < 2%;65,66 and in one study, grade 3 and 4 haemoglobin AEs of 3.1% and 1%, and grade 3 and 4 lymphocytes AEs of 7.6% and 0.3%, respectively.67

Liver AEs:

• Hepatotoxicity [abnormal liver tests, e.g. gamma glutamyl transferase (GGT), ALT and/or AST above the ULN] was reported by seven of the eight studies – all in RA populations.62,64–69 Hepatic disorder with an incidence rate of 3.9% at 26 weeks was reported in one study.67 Three studies reported an incidence of elevated ALT ranging from 2.7% to 7%;65,66,68,69 and one study an incidence of elevated AST of 2.4–2.3%, depending on whether adalimumab or its biosimilar was used.68 Two studies reported categories of elevation:62,69 the incidence of ALT and AST ≥ 3 × ULN was ≤ 2% in one study of 6 months duration, depending on therapy62 and 7% and 4% in a second study of 1 year duration;69 and the incidence of ALT and AST ≥ 1.5 or > 2 × ULN was ≤ 4.1% in one study of 6 months duration, depending on therapy62 and 16% and 10% in a second study of 1 year duration.69 One study reported incidence of grade 3 and 4 AEs for ALT and AST, which were all ≤ 1.6%.67 Discontinuations due to elevated liver enzymes was < 1% in the two studies where this was reported.64,69
• Hepatitis events were reported for two patients in one study;63 1/466 patients experienced hepatobiliary disorder over 72 weeks;66 and the proportion of patients with gastric/hepatic events ranged from 3.7% to 4.7% in one study.68 In one study, 1/385 patients discontinued treatment due to each of the following AEs: hepatitis, liver injury, chronic cholecystitis.69

Kidney AEs:

• A single case of grade 3 creatinine elevation (> 3.0 to 6.0 × ULN) was reported in one study after 48 weeks.67 No other relevant kidney AEs were reported by any included study.

Methotrexate and infliximab

Five studies (six articles) evaluated relevant AEs for patients receiving a combination of methotrexate and infliximab (or a biosimilar).41,56,86–88,145 Three of the five studies were multicentre RCTs in populations with RA, and one a single-centre controlled trial of RA patients; one study was a single-centre cohort study of an IBD population.56 One of the three RCTs was assessed as having a low risk of bias,145 and two assessed as having some concerns.41,87,88 One cohort study was assessed as being at low risk of bias86 and one at serious risk of bias.56 Where reported, the final dose of methotrexate in any arm ranged from 10 mg/week to 25 mg/week, except in one trial which used a dose of 5 mg/week.41 Infliximab or biosimilar dose was 3 mg/kg. Of the six studies with outcome data, four had a 1-year follow-up, and performed monitoring either monthly,56 every 3 months88 or at 6 months and 1 year;86,145 one had a follow-up of 2 years and performed monitoring every 3 months87 and one 5 years also performed monitoring 3-monthly.41 This last study did not provide discrete data for the methotrexate combinations with sulfasalazine or infliximab, so the total data are reported here.

The incidence of most reported AEs, such as anaemia, neutropenia, hepatic and kidney events, was < 2% and was not reported by many studies. Only elevated ALT was reported more frequently, with a much higher incidence among patients, depending on details of therapy and how far the elevation exceeded the ULN, but treatment discontinuation, where reported, was uncommon and temporary.

Blood AEs:

• Only one study reported any blood AEs.145 This 12-month study with 6-month monitoring reported an incidence of < 1% for grade 3 or 4 events for anaemia and leucopenia, but 1.4% and 2.2% respectively for neutropenia.

Liver AEs:

• Two studies reported ALT elevations.41,86 The percentages of patients experiencing ALT of > 2 × ULN were reported as 14% and 17.9%, and > 3 × ULN of 1.9% and 8.9% in the 1-year86 and 5-year41 studies, respectively. The longer study had more than four times the number of patients and only reported data for either infliximab or sulfasalazine in combination with methotrexate. This study reported that 90% of those affected had normal levels within 3 months and 10% of patients with elevated ALT discontinued therapy. Of this figure, 7% had normal levels in the following 3 months; only 3% had persistently elevated levels.
• Grade 3 ALT and AST were only reported for infliximab biosimilar BCD-055 in combination with methotrexate (2.1% and 0.4%, respectively) and not for infliximab and methotrexate. The incidence of abnormal blood test was reported as < 1.6% at 12 months and at 24 months in one study, and liver AEs at 4% and 5% at 12 months and at 24 months, respectively.87,88

Kidney AEs: None reported.

Methotrexate and etanercept

Eight studies evaluated relevant AEs for patients receiving a combination of methotrexate and etanercept or a biosimilar.76–78,86,139,141,146,147 Six of the eight studies were in populations with RA, of which five were multicentre RCTs76,139,141,146,147 and one a single-centre controlled trial of RA patients;86 two studies were RCTs of PsO patients.77,78 Three of the seven RCTs were assessed as having a low risk of bias,78,141,147 two as having some concerns77,146 and two at high risk of bias.76,139 There was one cohort study, which was assessed as being at low risk of bias.86 Where reported, the final dose of methotrexate in any arm ranged from 5 mg/week to 20 mg/week, except in one trial which used a dose of 5 mg/week.41 Etanercept or biosimilar dose was 25 mg twice per week, or 50 mg once per week. Of the eight studies with outcome data, four had follow-up of approximately 6 months and monitoring frequency of 4–6 weeks,76–78,139 and four had a follow-up of approximately 1 year, with monitoring performed every 6 weeks,139,141 3 months,146 3 months then 6 months, or 6 months.86,147

The incidence of most reported AEs, such as anaemia, leucopenia and haemoglobin, were generally < 2.3% and was not reported by many studies and could be worse for certain etanercept biosimilars. Elevated liver enzymes affected < 3.2% of patients on most reported instances, but could be higher than 20% in some studies, and again the incidence tended to be worse for biosimilars. Treatment discontinuation, where reported, always affected < 2.5% of patients.

Blood AEs:

• Anaemia was reported by one 48-week study,147 with an incidence of 0% on the etanercept + methotrexate but 6.8% in arm of the methotrexate and etanercept biosimilar HD203 combination. A 6-month study assessing the methotrexate and etanercept biosimilar ANB combination76 reported both decreased platelet count (4.5%) and decreased haemoglobin level (1.5%).
• Only two studies reported leucopenia [white blood cell (WBC) < 3 × 10⁹/L],76,139 both for a combination of methotrexate and the etanercept biosimilar ANB: in the first study 21.2% of patients had a decreased WBC, but only 2.27% of < 3 × 10⁹/L (all discontinued),76 while in the second study the incidence was only 0.17% after the first 3 months up to 6 months.139

Liver AEs:

• Only one study reported details of ALT elevations.86 The percentage of patients experiencing ALT of 2–3 × ULN and 3–4 × ULN were both reported as 2.2% at 1 year.86
• Four 6-month studies reported abnormal LFTs (elevated liver enzymes considered to be an AE) ranging from 2.9%77 to 3.2% (Chen 2016) and 20.5%76 to 21.8%.78 The three studies with the higher frequencies employed a combination of methotrexate and the etanercept biosimilars ANB76,139 and rhTNFR-Fc.78
• One 1-year study of RA patients with 3-month monitoring reported incidence rates of patients with an increase of AST between 0% and 6.6%, and of ALT between 0% and 7.9%, depending on the details of the regimen.146 And a 6-month study of PsO patients reported elevated ALT of 2.5% and AST of 2.9% (discontinuation was 1.7% for both): all events were considered mild or moderate in severity.77

Kidney AEs:

• A 6-month study assessing the methotrexate and etanercept biosimilar ANB combination reported no instances of an abnormal renal function test.76

Methotrexate and certolizumab

Two studies (two articles) evaluated relevant AEs for patients receiving a combination of methotrexate and certolizumab.71,72 All three studies were multicentre RCTs in populations with RA, and all three had approximately 6 months (24-week) follow-up and approximately monthly monitoring after. One of the three RCTs was assessed as having some concerns,71 and one as being at high risk of bias.72 Where reported, the final dose of methotrexate in any arm ranged from 5 mg/week to 10 mg/week, except in one trial which used a dose of 15 mg/week.94 The certolizumab dose, where reported, ranged from 100 mg to 400 mg every 2 weeks.71,72 Many AEs, such as anaemia, neutropenia and kidney events, were not reported, or there was explicit reporting of no such events. Only hepatic elevations were reported more frequently; incidence was low and there was a slight trend towards higher incidence with lower doses of certolizumab.

Blood AEs: No study reported any relevant outcome.

Liver AEs:

• Only one 6-month study reported ALT and AST elevations.72 This study reported low levels, with the following proportions of patients experiencing elevated ALT and AST, respectively, based on dosing regimen: certolizumab 200 mg: 0.4% and 0.8% and certolizumab 400 mg: 3.3% and 2.4%.72
• One 6-month study reported the following percentages of patients with abnormal hepatic function, based on therapy received: certolizumab 100 mg: 11.1%; certolizumab 200 mg: 3.7%; certolizumab 400 mg: 5.9%; placebo: 5.2%.71 A second 6-month study reported similar low levels, with hepatic enzyme increases in the following proportions of patients: certolizumab 200 mg: 1.2% and certolizumab 400 mg: 1.2%.72

Kidney AEs: No study reported any relevant outcome.

Methotrexate and golimumab

One study (the GO-FURTHER trial) evaluated relevant AEs for patients receiving a combination of methotrexate and golimumab.34,85 This was a multicentre RCT in a population with RA (groups aged < and ≥ 65 years), which had approximately 2 years follow-up and applied approximately 4-monthly monitoring, except for LFTs, which were monitored every 8 weeks. This RCT was assessed as having a low risk of bias. The final dose of methotrexate was not reported; the golimumab dose was 2 mg/kg. No liver AEs were reported by the study. The study reported a single case of septic arthritis and a single case of acute pancreatitis.34

Methotrexate and leflunomide

Four studies (four articles) evaluated relevant AEs for patients receiving a combination of methotrexate and leflunomide.40,92,94,120,133 Three of the four studies were multicentre RCTs in populations with RA,94,120,133 and one study was a multicentre cohort study of RA and PsO population.40 One of the two RCTs was assessed as having a low risk of bias for the short-term follow-up,120 but a high risk of bias for the OLE,133 while two trials were assessed as having some concerns.92,94 The cohort study of RA and PsO patients was assessed as being at moderate risk of bias.40 Where reported, the final dose of methotrexate in any arm ranged from 10 mg/week to 20 mg/week. Leflunomide dose was 10–20 mg/day. One study had a 6-month follow-up with 6 weekly monitoring;92 one had a 6-month and 1-year follow-up, with monthly monitoring,120,133 one study had a 1 year follow-up with 3-month monitoring,94,133 and the length of follow-up was unclear in the fourth study, but monitoring was conducted approximately every 5 months.40

Many AEs, such as anaemia, neutropenia, leucopenia and kidney events, were not reported. Only elevated ALT was reported more frequently, with a higher incidence at lower levels and in the shorter term – incidence was generally temporary, decreased over time and discontinuation was rare.

Blood AEs:

• One study reported that no clinically relevant decreases in leucocytes or neutrophils were observed in the first 6-month follow-up, but that haemoglobin levels < 10 g/dl affected 5.2% of patients in the 6- to 12-month period.120,133

Liver AEs:

• Three studies reported ALT and AST elevations.40,120,133 The percentages of patients experiencing ALT or AST of > 2 × ULN was 5% in one study with 5-monthly monitoring.40 The percentages of patients experiencing elevated ALT 2–3 × ULN was 6.2% at 6 months and 2.1% at 6–12 months, and > 3 × ULN was 3.8% at 6 months (2.3% discontinued) and 2.1% at 6–12 months.120,133 The percentages of patients experiencing elevated AST 2–3 × ULN was 3.1% at 6 months (1.5% discontinued) and 1% at 6–12 months, and > 3 × ULN was 1.5% (0.8% discontinued) at 6 months and 1% at 6–12 months.120,133 The levels were reported to normalise in all of these patients in the first 6 months,120 while during the 6- to 12-month period half or more of affected patients normalised and only one patient discontinued treatment.133 Normalisation to < 1.2 × ULN was reported for almost all of these patients.
• Two reported LFT abnormalities: in one study the percentage of patients with this outcome was 3.8% in the first 6 months120 and 1% in the second 6 months;133 in the second study, with 1-year follow-up and 3-month monitoring, there were 17 events among 93 patients.94

Kidney AEs: None reported.

Methotrexate and sulfasalazine

Four studies (five articles) evaluated relevant AEs for patients receiving a combination of methotrexate and sulfasalazine.41,43,94,134 Four of the five studies were multicentre RCTs in populations with RA.41,43,94,134 One of the five RCTs was assessed as having a low risk of bias,134 two were assessed as having some concerns,41,94 and one was assessed as being at high risk of bias.43 Where reported, the final dose of methotrexate in any arm ranged from 5 mg/week to 10 mg/week, except in one trial which used a dose of 15 mg/week.94 The sulfasalazine dose, where reported, was 2–4 g/day. Two studies had a 6-month follow-up, and performed monitoring monthly;43,134 one had a follow-up of 1 year and performed monitoring every 3 months;94 and one 5 years, but also performed monitoring 3-monthly.41 This last study did not provide discrete data for the methotrexate combinations with sulfasalazine or leflunomide, so the entire data are reported here.

Three studies explicitly reported no blood AEs, and two studies no kidney-related AEs. Three of the included studies reported relatively high rates of ALT elevation (e.g. up to > 20%), but very few or no SAEs or discontinuations due to any hepatic abnormalities or events.

Blood AEs:

• Two 6-month studies explicitly reported that there were no SAEs because of treatment.43,134
• One 6-month study reported that there were no cases of blood platelet elevation or normal blood routine examination.134
• One 6-month study reported that there was a single SAE of amenorrhoea in patients receiving prednisolone as well as the combination therapy; and 2/51 (3.92%) patients not receiving prednisolone experienced leucopenia (but therapy was not discontinued).134

Liver AEs:

• Three studies reported ALT elevations.41,43,134 The percentages of patients experiencing any ALT elevation by 6 months were reported as between 11% and 21.5%134 and 9.8% and 11.3% for AST.43,134 Proportions of patients with ALT of > 2 × ULN were reported as 17.9%, and > 3 × ULN as 8.9% in the 5-year study.41 This study only reported data for either infliximab or sulfasalazine in combination with methotrexate. This study also reported that 90% of those affected had normal levels within 3 months and 10% of patients with elevated ALT discontinued therapy. Of this figure, 7% had normal levels in the following 3 months; only 3% had persistently elevated levels.41
• Hepatic or liver function abnormalities were reported by two studies.43,134 In one 6-month study, hepatic adverse drug reactions (ADRs) were experienced by 8.76% of patients, and abnormalities by 21.5%; there were no cases of jaundice or any SAEs.43 Hepatic dysfunction was reported for between 2% and 7.6% of patients in one study, with the higher incidence among those exposed to prednisolone; ≤ 2% were deemed to be SAEs.134 There was a single case of acute hepatic function damage that led to discontinuation.134

Kidney AEs: One study explicitly reported no relevant AEs or SAEs.43

Anti-tumour necrosis factor-alpha (and biosimilars)

Adalimumab

Twenty-three studies (27 articles) evaluated relevant AEs for patients receiving adalimumab or a biosimilar such as BI 695501,148 GP2017149 or IBI303 (Xu). Eleven of the studies (13 articles) were conducted in populations with PsO, 5 in populations with RA, 5 with IBD (6 articles) and 3 with AS (4 articles). Eighteen were multicentre RCTs; there were 6 multicentre cohort studies66,83,109,136,150,151 and 1 single centre cohort study.108

For the studies with PsO or psoriatic arthritis populations, four of the eight RCTs were assessed as having a low risk of bias149,152–155 and four were assessed as being at high risk of bias.58,156–159 Of the three cohort studies, one was assessed as being at low risk of bias,136 one at moderate risk of bias,108 and one at serious risk of bias.151 For the studies with RA populations, the two RCTs were assessed as having a low risk of bias28 and a high risk of bias.160 Of the two cohort studies, both were assessed as being at moderate risk of bias.109,148 For the studies with IBD populations, the three RCTs were assessed as either having some concerns161 or a high risk of bias.60,162,163 Of the two cohort studies, one was assessed as being at moderate risk of bias150 and one at a serious risk of bias.83 For the four studies with AS populations, the three RCTs were assessed as having either a low risk of bias164 or a high risk of bias.165 One trial has some concerns for the main part of the trial35 and a high risk of bias for the OLE.146

Where full details of dosage were reported, the final dose of adalimumab in every trial was 40 mg every other week, except for two Japanese trials in which several patients also received a dose of 80 mg every other week.156,162

Most studies had a follow-up of 1 year or less (27 studies): 5 studies only had a follow-up of 16 or 20 weeks, 8 of approximately 6 months, 9 of 1 year, and 5 studies of 1–2 years. Three studies had a follow-up of more than 2 years.108,109,162

Twenty studies performed monitoring at intervals of 4–8 weeks for at least part of the study. One study performed monitoring every 2 weeks for a period after the first 3 months.160,164 Seven studies performed monitoring at 3-month intervals,60,66,108,109,146,150 and four studies with intervals of 6 months or more.58,108,150,152,162 A study might increase the interval over time, for example 3 months up to 6 months, then 6 monthly thereafter.58,60,61,150

Adverse events, such as anaemia, neutropenia and leucopenia, were reported by few studies. The incidence of most haematological AEs was rare (< 1% or < 2%), regardless of length of follow-up or monitoring frequency, with the exception of single instances of rates of > 3% for thrombocytopenia for an adalimumab biosimilar149 and cytopenia.58 One study reported numbers of haematological events: 16 events in 266 patients over almost 4 years.162 Discontinuations due to haematological events were reported by four studies as being ≤ 1.3%.84,108,158,159

The frequency of hepatic AEs ranged from 12.9% to 39.5% across three studies, but were also as low as 1.2% in one study, probably due to outcome definition.154 Serious hepatic events were reported by one study and with an incidence rate of < 1% for adalimumab both with and without concomitant drugs.109 The incidence of hepatobiliary disorders ranged from 7% to 11.9%, but this was reported only by one study.156 ALT elevation: five of six studies reporting an incidence of ≤ 4.8% across one of more arms, while one study reported rates of 9–17%164 and another higher rates in two arms of: 6/38 (15.8%) and 7/43 (16.3%).156 Both studies used biosimilars, though the biosimilar was not used in the arms with higher rates. There were lower reported rates of AST elevation, < 2–9%, with the highest rates in the same studies as for ALT,156 and the highest rates also for GGT.156 The lower end of the scales are consistent with two studies reporting simply liver enzyme elevation of 0.3–5.9%,152,157 and discontinuation rates at ≤ 1%.108,152,155,159 The majority of the results for specific ALT and AST levels fell within the lower thresholds, especially for ALT and AST > 2 × ULN: ≤ 9.3% or ≤ 3.2% respectively. All SAEs where reported were < 1%.

No kidney AEs were reported in any studies, including by studies that otherwise reported specific AEs such as hypercreatininaemia.136

Certolizumab monotherapy

One study, a 6-month, multicentre RCT of RA patients, at low risk of bias, evaluated relevant AEs for more than one patient with certolizumab monotherapy at a dose of 400 mg/every 4 weeks.73 This study only reported a single case of chronic renal failure (1/109, < 1%), and < 1% (1/111) and a single incidence of discontinuation of therapy due to elevated blood creatinine/increased blood urea. One multicentre cohort study at low risk of bias also evaluated this dose of certolizumab pegol, but for 2 years, with monitoring performed at 6 months, 9 months, 1 year and 2 years: the incidence of hepatic AEs was < 1%.70

Etanercept

Eighteen studies (19 articles) evaluated relevant AEs for patients receiving etanercept or a biosimilar such as GP2015166 or rhTNFR-Fc78 or ANB.76 One study reported both 6-month and 2-year follow-ups.79,80 Nine of the studies were conducted in populations with RA, eight with PsO and one with AS.167 Eleven were multicentre RCTs; there was one single-centre RCT,74 five were multicentre cohort studies and one a pooled analysis of multicentre studies.75

For the 9 studies (10 articles) with RA populations, 2 RCTs were assessed as having a low risk of bias,141,168 1 having some concerns169 and 3 as being at a high risk of bias.74,76,79,80 Of the three cohort studies, one assessed as being at low,144 one at moderate75 and one at serious risk of bias.112 For the eight studies with PsO or PsA populations, two of the five RCTs were assessed as having a low risk of bias,78,166 two having some concerns77,170 and one as being at high risk of bias.171 Of the three cohort studies, one was assessed as being at low risk of bias,136 one at moderate risk of bias,172 and one at serious risk of bias.173 The RCT of AS patients was assessed as having some concerns (Davis 2003).

Where full details of dosage were reported, the final dose of etanercept in all but one study was 25 mg twice weekly, except for the addition of a 10 mg twice weekly arm in two trials, both in RA.74,144 The exception was a RCT OLE study only assessing 50 mg twice weekly in PsO patients.171

The majority of studies had a follow-up of 1 year or less (14 studies): 2 of 4 months,136,169 7 of approximately 6 months76–78,80,172 (Combe 2006; Davis 2003) and 5 of 1 year.74,141,144,166,170,173 Three studies had a follow-up of approximately 2 years;79,112,171 1 of 5 years168 and 1 of 7 years.75 Studies with approximately 6 months duration conducted monitoring at a frequency of every 1–3 months, except for one study, which monitored every 6 months.76 After 6 months, almost all studies conducted monitoring every 3 or 6 months, except for one RCT that employed 1–2 monthly monitoring.141

Adverse events, such as anaemia, neutropenia and leucopenia, were reported by few studies, but the incidence of most haematological AEs was rare (< 1%) regardless of length of follow-up or monitoring frequency (Table 8). There were only two studies with higher rates: 2.3% for anaemia and 12.1% for leucopenia for the biosimilar anbainuo76 and 16% neutropenia in one of the oldest studies.74 Septic arthritis was reported by three studies with follow-ups ranging from 6 months to 5 years, but the incidence was never higher than 1.9%.75,80 Three studies specifically recorded discontinuations due to blood AEs,79,80,168,169 but only one recorded any relevant cases, with an incidence of < 1%.168

Only two studies, both with only 6 months follow-up, reported the outcome ‘abnormal LFT’, affecting < 7% of patients.76,78 ALT and AST elevations were reported as either being ≤ 6.3% and ≤ 4.4%, respectively,144,166,168 by up to three studies, or as high as 24–25.5% and 12.1–16% by other studies.74,77 Incidence for ALT or AST > 2 × ULN was ≤ 5.5% and ≤ 3.6%, respectively.136 Grades 3 and 4 ALT or AST elevations were always < 1%.77,168,170,171 Other reported outcomes had an incidence of ≤ 2%. Again, where any liver-related discontinuations were reported, they were always < 1%.77,168,169,172,173

No kidney AEs were reported in any studies, except < 1% incidence of elevated serum creatinine in a 4-month study of PsO patients.136

Golimumab

This was a clinically heterogeneous set of studies. One multicentre RCT82 and two multicentre cohort studies81,83 reported relevant AE data. The RCT was at low risk of bias,82 and both cohort studies were judged to be at serious risk of bias.81,83 All three studies had a follow-up of 12–24 months, but monitoring ranged from monthly to 3 months and 6 months, and the populations were all different: AS,82 RA, PsO, AS81 and IBD.83 One study reported no cases of anaemia, leucopenia, or neutropenia,83 while a second reported only single cases of neutropenia and thrombocytopenia leading to treatment discontinuation.81 Only one 2-year study with 3-monthly monitoring reported any liver AEs. The incidence of ALT elevation was ≤ 2.9% in all treatment arms, except for the arm with the highest dose (100 mg) of golimumab (GOL), when it was 5%; the incidence of AST elevation was ≤ 2.9% in all treatment arms;82 total bilirubin was also < 2.4% in any arm.82 None of these studies reported any relevant kidney AEs.

Infliximab

Eighteen studies (22 articles) evaluated relevant AEs for patients receiving infliximab or 1 of the biosimilars: SB2,178,180,181 PF-06438179/GP1111106 or CT-P13.175,177,186 The included studies covered a range of conditions: RA (five studies); PsO (six studies); IBD (four studies) and AS (three studies); and a sample including RA, PsO and AS patients.176 The majority of studies were multicentre cohort studies (n = 11); otherwise, there were eight multicentre RCTs87,88,135,174,175,177–182 and one single-centre cohort.176

For the five studies with RA populations, two RCTs were assessed as having some concerns in terms of risk of bias;87,88,178,180,181 both cohort studies were assessed as being at moderate risk of bias.106,111 For the six studies with PsO or PsA populations, all three RCTs were assessed as having some concerns in terms of risk of bias;135,174,179 of the three cohort studies, one was assessed as being at low risk of bias,136 one at moderate risk of bias113 and one at serious risk of bias.61 The two RCTs of AS patients were both assessed as being at high risk of bias,175,177,182 and the one cohort study as being at a serious risk of bias.187 The four IBD population studies were all cohorts: two were assessed as being at moderate risk of bias150,186 and two at serious risk of bias.49,83 The remaining cohort study concerned patients with RA, PsO and AS patients – and was at moderate risk of bias.176

Where full details of dosage were reported, the final dose of infliximab or its biosimilar was 3 mg/kg, except for seven studies with a 5 mg/kg dose,49,113,135,174,175,177,182 including one trial comparing the two doses.179

Relatively few studies (n = 5) had a follow-up of < 1 year, and mostly within the range of 16–30 weeks.135,136,174,175,178,179 The majority of studies had a follow-up of 1–2 years.6,61,83,87,88,106,111–113,150,177,180–182,186,187 Two studies had follow-ups of 3–5 years.49,176 Frequency of monitoring fell within two time intervals: 1–2 months61,106,112,135,136,174,176,178–180,182 and 3–6 months.49,83,87,88,111,113,150,175,177,181,186,187

Blood AEs were very rare. Two studies reported data on leucopenia: incidence rates were 1.6% or lower, regardless of follow-up (ranging from 30 weeks to 3 years) or monitoring frequency (ranging from 3 to 6 months).112,177 Three studies, all with follow-ups of 2 years or longer, reported < 1% incidence of pancytopenia113,176 and/or thrombocytopenia.112,176 Haematological disorders and abnormal blood tests were all reported by single studies and had an incidence of ≤ 2%, despite follow-ups of 1–3 years.87,88,111 Elevated serum creatinine phosphokinase was reported by one study with incidence rates of 1.6% and 3.1% – the higher rates reported for the biosimilar CT-P13 compared to infliximab.175,177

Liver enzyme elevations were the most frequent AEs, with incidence rates for ALT up to 30 weeks ranging from 2.7%178 to 10.9%,175 and for AST from 2.7%178 to 9.4%.175 Another study reported very high rates at 30 weeks: 34% for ALT and 24% for AST.179 Still lower rates (1.0–7.9% for ALT and 2.0–4.3% for AST) were reported for 12- and 18-month follow-up in a single trial.180,181 The incidence rates of ALT and AST were reported as 24.5% and 11.3%, respectively, in one 2-year study.182 In the studies with the lowest rates, the rates at the higher end of the range were always for biosimilars SB2 and CT-P13.175,178,180,181 Only one study reported actual levels of liver enzyme elevation (> 2 × ULN), with incidence rates of 9.8% (ALT) and 6.4% (AST) at 16 weeks.136 The incidence of discontinuations due to ALT or AST elevation was reported as < 1% by one study for 1-year follow-up,177 while the rate of discontinuations due to LFT abnormalities was reported to be 2% at 6 months in another study.135 Abnormal LFTs had an incidence rate of ≤ 2% in two studies with 6 months or 2 years follow-up,135,182 but the rate was as high as 12.5% in a third study.177 All other reported liver-related AEs had an incidence rate of < 1%. There were no discontinuations reported due to liver toxicities by one study,111 but a discontinuation rate of 3.3% at 2 years for unspecified AEs of the liver in another study.187 The data from one study are not reported here because it was unclear what proportion of elevations had resolved within the first 3 months.49

In terms of kidney AEs, there was a single reported case (< 1%) of sepsis and pyelonephritis in a 30-week study with monthly monitoring.179

Anti-tumour necrosis factor drugs (unspecified)

Eight studies reported combined data for three relevant therapies (adalimumab, etanercept and golimumab or infliximab) or all four of these therapies,84 or unspecified anti-TNFs plus either methotrexate or leflunomide,30 for a variety of populations and a variety of outcomes. One study was a 2-year retrospective cohort study of a mixed population (PsO, RA and AS) with 6-monthly, then 3-monthly testing.39 The study reported an overall incidence of 8.9% for ALT of > 2 × ULN. The second was a cohort of PsO patients followed for 1 year and with 3–6 months monitoring.61 This study reported data on all AEs, but no AEs relevant to this review were reported. Six of the seven cohort studies were assessed as being at serious risk of bias; only one was assessed as being at moderate risk of bias.184 A 1-year multicentre RCT at low risk of bias and with monthly monitoring evaluated different anti-TNF drugs.188 This trial reported haematological events: 10 events in more than 100 patients, including 1 SAE, over 1 year, and 16 events of elevated ALT.189 Discontinuations for blood and liver function disorders were reported by a multicentre cohort study of patients with spondyloarthritis: the incidence as ≤ 1.7%.84 A large retrospective cohort study of RA patients reported an incidence rate of 2% for hepatotoxicity by 17 months, with 2–3 monthly monitoring.30

Etanercept and sulfasalazine

This combination was evaluated by a single multicentre RCT (two articles) with 6 months and 2 years follow-up, and monthly monitoring to 6 months, 2-monthly monitoring to 1 year, and 3-monthly monitoring to 2 years.79,80 The RCT was assessed as being at a high risk of bias. Etanercept was given at a dose of 25 mg twice a week and sulfasalazine at a dose of 2–3 mg/day.

Reported relevant AEs were very rare. No relevant liver or kidney AEs were reported for this combination. The study only reported mild leucopenia, with a total incidence of 5% at 6 months and 7.9% at 2 years, and no discontinuation of therapy due to this AE.79,80

Infliximab and thiopurines

This combination was evaluated by one single-centre cohort study in IBD patients with a mean follow-up of 12.8 months, and monthly monitoring.56 This cohort study, at moderate risk of bias, evaluated a range of therapy combinations, but numbers of patients exposed to particular combinations were small: 19 patients received infliximab with either thiopurines or mycophenolate. As a result, reported relevant AEs were rare and only discontinuations were reported: 1/5 (20%) had infliximab + mercaptopurine therapy discontinued due to neutropenia, while 2/46 (4.3%) discontinued therapy due to acute hepatitis.

Thiopurines

Azathioprine

Nine studies evaluated relevant AEs for patients receiving azathioprine monotherapy. Seven studies were conducted in populations with IBD, all of which performed frequency of monitoring at approximately 3-month intervals; two studies were conducted in RA populations and performed monitoring at 4–6 weeks95 or every 6 months.31 One study of SLE patients conducted monitoring every 3 months.96 One cohort study did not report discrete data for azathioprine and included two patients receiving mercaptopurine and 349 patients only received azathioprine, so the findings are reported here.33 In the studies of IBD patients, there was one RCT with some concerns regarding risk of bias,38 and five cohort studies at moderate risk of bias.33,46,190–192 The studies of RA patients were two cohort studies which were both assessed as having a serious or critical risk of bias.95 The RCT of SLE patients was judged to be at a high risk of bias.96 Where reported, almost all studies used a dose of 2.0–2.5 mg/kg for azathioprine; one study reported a range of 50–200 mg/day.190

The follow-up in these studies ranged from < 6 months38 (Wong) to 1–< 2 years31,95,191 or ≥ 2 years.46,96,190 Leucopenia was reported by four studies, with a frequency after the first 3 months of therapy ranging from 4.0–4.9%38,96 to 16.6%,191 and 39.6%.46 Discontinuation rates ranged from 0.9%38 to 5.3%.191 Myelotoxicity was reported by one study affecting 4.3% of patients,190 and thrombocytopenia affected 0.3% of patients in another study, all of which led to discontinuations.191 The reported rate of neutropenia (25%) was skewed by this being grade 1 only and one out of only four patients.31 Hepatotoxicity appeared to occur in approximately 12–14% of patients in the short term,38 but was less frequent longer term.190 Liver toxicity was rarer still, with an incidence of < 2.5% across two studies, except for the highest dose of azathioprine, which produced a rate of 5% at 2 years.193 One study reported renal toxicity, with an incidence and discontinuation rate of < 1%.191

Mercaptopurine

Only one study evaluated relevant AEs for patients receiving mercaptopurine monotherapy (the TOPIC trial). This was conducted in populations with IBD and performed frequency of monitoring at approximately 3-month intervals. One article was the full RCT with some concerns regarding risk of bias,38 and one a cohort study at moderate risk of bias,53 which evaluated a cohort of patients from the trial. This study used a dose of 1.0–1.5 mg/kg for mercaptopurine and had 6 months follow-up. Only leucopenia, hepatotoxicity, myelotoxicity and pancreatitis were reported as relevant AEs. Leucopenia was reported by tone study, with a frequency range of 11–12.5% in one study.38 In this study, discontinuation was reported: 3.2% of patients receiving 6-mercaptopurine (6-MP) had treatment discontinued due to leucopenia. Hepatotoxicity was reported by two articles from this one study, with a frequency of 22.7–25.3%. Acute pancreatitis was reported in this RCT part of the study, with a frequency of 1.1% in a study of mercaptopurine (all cases led to treatment discontinuation).

Thiopurine

Seven studies evaluated relevant AEs for patients receiving either azathioprine or mercaptopurine. One study was a RCT of IBD patients with some concerns in terms of risk of bias (Coenen 2015),194 and the remaining six studies were cohort studies conducted in populations with IBD. All studies performed frequency of monitoring at approximately 3-month intervals.

One of the cohort studies was judged to be at low risk of bias,97 while the remaining five were all assessed as being at a moderate risk of bias.42,46–48,50 The majority of studies employed azathioprine doses of 2.0–2.5 mg/kg and 0.75–1.5 m/kg for mercaptopurine.42,46,47,50,97 The exception reported using median doses of 0.9 mg/kg (azathioprine) and 0.8 mg/kg (mercaptopurine).48 The follow-up in these studies ranged from 6 months194 or 1 year42 to a median follow-up duration of approximately 5 years.50 Leucopenia was reported by four of the seven studies with follow-ups ranging from 3 months to 5 years: the incidence was reported to decrease over time (Table 9). For example, one study reported generally declining rates for the period from 3–6 months (26.7%) to after 24 months (14.7%), of which only 12% were designated as grade 3 or 4.46 In one study, 2.4% of patients discontinued due to leucopenia;97 in a second, leucocyte counts recovered a median of 27 days after cessation of thiopurine treatment [interquartile range (IQR) 26 days, range 1–153 days]; in another, the median time from initiation to cessation due to leucopenia was 347.5 days [(IQR 159.0–866.0) days (p < 0.0001)].46 Myelotoxicity (thrombocytopenia and/or leucopenia) was reported by one study48 and thrombocytopenia (platelet count < 100,000) by one study,47 resolving after withdrawal of treatment or dose reduction in 11/210 (5.2%) patients, and no severe cytopenia; the same study reported neutropenia affecting 11/210 patients (5.2%) and lymphopenia 19/210 (9%). Haematological ADRs were reported in 7.5% of patients in the RCT (Coenen 2015). Hepatotoxicity (abnormal liver tests, e.g. GGT, ALT and/or AST above the ULN) was reported by three studies, with the proportion of patients affected reported as ranging from 0%42 to 3.8%47 across two studies, and as high as 24.3% in one study, of which < 1% was categorised as ‘severe’.48 Percentages of patients receiving thiopurines who had treatment discontinued due to hepatotoxicity ranged from 0.008%48 and 1.9%47 to 5.5%.97 There were no relevant kidney AEs reported.

Leflunomide alone

Fourteen studies (14 articles), exclusively of RA populations, except for one study which also included PsO patients,40 evaluated relevant AEs for patients receiving a leflunomide. The majority of studies were multicentre cohort studies (n = 9); otherwise, there were three multicentre RCTs93,124,121 and two single-centre RCT.90,142 Four of the five RCTs were judged to be at a high risk of bias90,93,124,142 and only one was assessed as having some concerns.121 Of the cohort studies, two were at low risk of bias,102,195 three at moderate risk of bias40,196,197 and four at serious risk of bias.31,89,91,112

Four studies had a 6-month follow-up, and performed monitoring every 1–3 months;93,102,195–197 two had a follow-up of 1 year and performed monitoring every 3 months;91,142 and the remainder had follow-ups of 18 months or 2 years, with 3–6 months monitoring,90,112,121 except for one study with monthly monitoring,124 and two studies with a follow-up of either 4 or 5 years.31,89,196

Six studies reported data on leucopenia: incidence rates were highest in the first 6 months or year ≤ 6.4%; one study explicitly reported rates of 4% for the first year, and 2.7% for the second year (Table 10).124 Discontinuation rates of ≤ 1% were reported by two studies.102,124 Four studies reported anaemia and/or neutropenia, and the incidence ranged from 2.2% to 4.3%; discontinuations due to neutropenia were reported as < 1% by one study.102 Otherwise, there were specifically no reported cases or only a single case of either pancytopenia,102,121 thrombocytopenia.121,196

Liver enzyme elevations were the most frequent AEs, with incidence rates at 31.9% or 20% (> 1 × ULN) for ALT and AST, respectively; reducing markedly at higher levels: > 2 × ULN: 6.7–12.6% ALT (two studies) and 2.6–8.9% AST (two studies); and > 3 × ULN: 0–2.8% ALT (five studies) and 0–2.2% AST (three studies). AEs specified only as abnormal LFT/increased hepatic enzymes were reported by five studies with a range of 6 months to 4 years follow-up, and monthly to 6-monthly monitoring: the incidence rate ranged from 2.7% to6.4%. The incidence of SAEs or discontinuations due to ALT or AST elevations within 1 year was ≤ 1.6%,91,102,124,195 with slightly higher rates for periods > 2 years (1.6–2.2%).124,196 Liver biopsies were rare (< 1.1% across three studies), and liver damage195 or dysfunction89 had an incidence of 4–4.5%, but only in one study each.

In terms of kidney AEs, there was a single reported case of acute renal failure (SAE) in a 6-month study.195

Combinations of classes of drug

One study reported relevant AE data for methotrexate and sulfasalazine57 This study was a 2-year trial of RA patients with 6-monthly monitoring.57 The study was judged to be at a high risk of bias. It only reported data for discontinuations due to relevant AEs, rather than the incidence of AEs experienced by patients that did not lead to discontinuation. Discontinuations due to leucopenia and thrombocytopenia occurred in < 3% and < 1% of patients, respectively, while discontinuations due to elevated liver enzymes occurred in < 4% of patients.

Mycophenolate

This was a largely clinically homogeneous set of studies: all three studies were in populations with SLE with 86–93% female participants. One was a multicentre RCT,96 one a multicentre198 and one a single centre199 cohort study. The RCT was at high risk of bias,96 while one cohort study was at low risk of bias198 and the second at a serious risk of bias.199 Dosages were not reported except 1440 mg/day enteric-coated mycophenolate sodium (EC-MPS).96 Follow-up ranged from 6 months to 3 years. Monitoring was monthly up to 6 months in two studies,96,198 and thereafter was 3–4 months for 2 or 3 years.96,199 There was a separate and distinct study: a pooled set of RCTs at low risk of bias in a RA populations, exposed to MMF at 1 g/twice a day, with monthly follow-up for 6 months.

Reported relevant AEs were very rare (Table 11). Leucopenia was reported by two SLE studies with an incidence of 7.5% at 12 weeks and 5.1% at 6 months in one study,198 but with no cases in a second study in 2 years,96 and a discontinuation rate of < 1% in the third study (Conti). The shortest duration study also reported similar, albeit slightly lower, rates of thrombocytopenia, for example ≤ 5.6%.198 Only the 2-year study reported on relevant liver AEs, and found no cases of liver toxicity.96 The 6-month study reported a significant reduction in serum creatinine levels at 24 weeks,198 and one study a single case of pyelonephritis as a SAE (incidence rate of < 1%).96 Only serum creatinine clearance was reported as an AE for the RA study, with an incidence of ≤ 1.1% in patients.200

Sulfasalazine and 5-aminosalicylates

Sulfasalazine

There were 9 studies (10 articles) evaluating sulfasalazine as monotherapy: 4 multicentre RCTs and 4 multicentre cohort studies31,95,100,101 and 1 single-centre cohort study.103 Four studies were conducted in RA populations, three in IBD populations, one in SLE99 and one in a mixed population of PsO, ReA and AS.98 The RA studies included two RCTs, both at high risk of bias,79,80,100 one cohort study at serious risk of bias31 and one at critical risk of bias.95 For the IBD patients, there was one RCT at high risk of bias,201 and two cohort studies at moderate risk of bias.101,103 The dose was generally 2–3 g/day of sulfasalazine in these studies, except for three studies that reported slightly different regimens: 0.5–4 g/day,98,99 and 0.25–6 g/day.101

Three studies only had 4–6 months follow-up,80,98,101 two had an 18-month follow-up,31,201 but the remaining five studies (six articles) had follow-ups ranging from 2 years79 to 12 years.100 For the RCTs in the other populations, one was at low risk of bias98 and the other at a high risk of bias.99

Across four studies, the incidence of leucopenia ranged from 0% – for a definition of ‘mild leukopenia’79,80 or ‘within normal ranges’101 – to 3.4–6%.98,99 Discontinuations due to leucopenia were explicitly reported as zero in one study at both 6 months and 2 years,79,80 and as ≤ 1.7% in two other studies.98,100 Neutropenia was reported in two studies: one 18-month study reported an incidence rate of 8.8%, of which 6.1% were grade 1, and 0.1% led to discontinuation of therapy,31 while a 5-year study reported 40 events in 1321 sulfasalazine patients.95 Liver abnormalities were only reported as clinically meaningful by two studies, both with an incidence of ≤ 2%,98,99 although an increase in hepatic enzymes led to discontinuation in 1.1% of patients in one 6-month study with monthly monitoring98 and liver function abnormalities led to 20 discontinuations in 1321 patients in 5 years in another study.95 The only recorded kidney AE reported by any of the included studies was a single instance of nephrotoxicity, which led to discontinuation in one study.103

5-aminosalicylates

This was a clinically homogeneous set of studies: three studies were in populations with UC,103–105 and the fourth in a population of mixed IBD.101 One multicentre RCT with some concerns regarding risk of bias,105 one RCT (number of sites unknown) at high risk of bias,104 one multicentre101 and one single-centre cohort study,103 both at moderate risk of bias. Only one study reported dose: 4 g/day.104 Follow-up ranged from 4 to 12 months in three studies, all with 3-monthly monitoring, and monitoring of at least once a year in the 5-year study.103 Reported relevant AEs were very rare (Table 12). Only the 3–4-month study reported any haematology AEs, noting a significant decrease in platelets and leucocytes (percentage of patients affected not reported).101 The same study also reported no significant change in ALT or AST levels. The only other study to report any liver AEs found a single case of acute hepatitis that led to treatment discontinuation.104 Two studies reported no significant changes in creatinine levels;101,105 one study reported impaired renal function, with an incidence of 12% (< 1% discontinuation rate).103

Principal summary of findings by outcome

Blood

Anaemia

The cumulative incidence of anaemia in patients with RA and other inflammatory arthritis treated with methotrexate ranged between 1.9% and 7.5% over 6 months to up to 2 years of follow-up, in studies that monitored full blood count (FBC) monthly, 3- or 6-monthly intervals with no relationship between interval of monitoring blood tests and incidence of anaemia.63,114,120,115,116,117

Similarly, the cumulative incidence of anaemia in patients with RA, inflammatory arthritis, PsO and IBD treated with anti-TNF alpha monotherapy ranged between 0% and 2.3% over 6 months to up to 5 years of follow-up in studies that monitored blood tests 6-weekly, monthly, 2-, 3- or 6-monthly with no relationship between interval of monitoring blood tests and incidence of anaemia.76,83,144,149,168,174,183

In patients with RA treated with methotrexate and anti-TNF alpha drugs, the cumulative incidence of anaemia ranged between < 1% and 6.8% over 6–12 months in studies that monitored blood tests at 6-weekly to 6-monthly intervals with no relationship between interval of monitoring blood tests and incidence of anaemia.62,63,68,76,145

Leucopenia

The cumulative incidence of leucopenia in patients with RA and other inflammatory arthritis treated with methotrexate ranged between < 1% and 3.8% over 6 months to up to 2 years; and 7.3–10.3% in studies of between 3-year and 5-year duration, that monitored FBC monthly to 3-monthly with no relationship between interval of monitoring blood tests and incidence of leucopenia.31,45,55,76,116,119,120,121,122,123,124,125,126,127

The cumulative incidence of leucopenia in patients with RA, IBD, PsO, AS and other inflammatory arthritis treated with anti-TNF alpha ranged between < 1% and 1.9% over 6 months to up to 2 years; and < 1–1% in studies of between 3-year and 5-year duration, that monitored FBC monthly, 3-monthly, 6-monthly with no relationship between interval of monitoring blood tests and incidence of leucopenia.76,79,80,83,112,152,161,183,184

The cumulative incidence of leucopenia in patients with RA treated with methotrexate and anti-TNF alpha ranged between 0.17% and 2.27% over 6–12 months in studies that monitored with a FBC at 6-weekly to 6-monthly intervals with no relationship between interval of monitoring blood tests and incidence of leucopenia.62,76,145

The cumulative incidence of leucopenia in patients with RA treated with leflunomide ranged between 0% and 6.8% over 6 months–2 years in studies that monitored with a FBC at 4-weekly, 6-weekly, to 3-monthly intervals with no relationship between interval of monitoring blood tests and incidence of leucopenia.102,112,124,121,195

The cumulative incidence of leucopenia was higher in patients prescribed thiopurines (higher for mercaptopurine than for azathioprine) over 6 months to up to 5 years compared to methotrexate, but there was no relationship between interval between testing that varied from 2 monthly to 3 monthly and cumulative incidence of leucopenia.33,38,96,191

The cumulative incidence of leucopenia in patients with SLE treated with mycophenolate was up to 7.5% over 2 years in studies that monitored with a FBC at 4-weekly, 3-monthly, intervals with no relationship between interval of monitoring blood tests and incidence of leucopenia.96,198

The cumulative incidence of leucopenia in patients with RA, PsO, AS, ReA, SLE treated with sulfasalazine ranged between 0% and 6% over 6 months to 5 years in studies that monitored with a FBC at 4-weekly, 3-monthly and annual intervals with no relationship between interval of monitoring blood tests and incidence of leucopenia.79,80,98,99

Neutropenia

The cumulative incidence of neutropenia in patients with RA treated with methotrexate ranged between 0% and 8.1% over 6 months to up to 2 years in studies that monitored FBC monthly, two monthly, 3-monthly or 6-monthly with no relationship between interval of monitoring blood tests and incidence of neutropenia.31,67,74,120,117,118,128

The cumulative incidence of neutropenia in patients with RA, PsO, IBD treated with anti-TNF alpha ranged between 1.1% and 3.1% over 6 months to up to 2 years in studies that monitored FBC monthly, or 3-monthly with no relationship between interval of monitoring blood tests and incidence of neutropenia.28,74,83,149,152,153,166,176

The cumulative incidence of neutropenia in patients with RA treated with methotrexate and anti-TNF alpha ranged between < 1.0% and 2.2% over 6 months to up to 1 year in studies that monitored FBC 6-weekly to 6-monthly with no relationship between interval of monitoring blood tests and incidence of neutropenia.67,69,145

The cumulative incidence of neutropenia in patients with RA treated with leflunomide ranged between 2.2% and 3% over 6 months to up to 4 years in studies that monitored FBC monthly, 3-monthly, or 6 monthly with no relationship between interval of monitoring blood tests and incidence of neutropenia.31,102

Thrombocytopenia

The cumulative incidence of thrombocytopenia in patients with RA treated with methotrexate ranged between 0% and 3.0% in studies with up to 5 years follow-up and was 6.9% in a study that followed up patients up to 9 years; these studies undertook monthly, 3- to 4-monthly and 6-monthly monitoring FBC.31,45,107,116,119,120,121,122,125,126,127

The cumulative incidence of thrombocytopenia in patients with RA, AS, PSO, and inflammatory arthritis treated with anti-TNF alpha ranged between < 0.0% and 3.2% in studies with between 1 year and 5 years follow-up that undertook monthly, 6-weekly or 6-monthly monitoring FBC.112,149,176,183

The cumulative incidence of thrombocytopenia in patients with IBD treated with thiopurines ranged between < 1.0% and 5.2% in studies with up to 2 years follow-up that undertook monthly or 2–3-monthly monitoring FBC.47,191

Liver

There was substantial variation in reporting of AEs affecting the liver with some studies reporting elevated liver enzymes (unspecified), abnormal LFT (which could include bilirubin and albumin alongside liver enzymes), different liver enzymes separately or using terminologies such as hepatitis, hepatotoxicity, liver AEs without any further qualification. Several reported the rates for any elevated liver enzyme, while others reported rates according to severity of liver enzyme elevation. The cumulative incidence of any liver enzyme elevation was very high – sometimes in the range of 30–40% while meaningful elevations in liver enzymes, for example, two- to threefold upper level of normal, were present in X% to X% participants. This made synthesising evidence on this outcome challenging.

Any abnormal LFT or liver enzyme elevation occurred in 1–53% patients with RA, PsO, PsA, treated with methotrexate over 6 months to 12 years, with monitoring blood tests conducted monthly, 3- or 6- monthly.32,55,67,71,76,94,115,120,125,131,132,135,137,139 The particularly high rates were observed in people with PsO and those not prescribed concurrent folic acid.40,123 When these studies were excluded, the highest cumulative incidence was 35%. In Chinese patients with RA treated with methotrexate and leflunomide for 6 months and undergoing monitoring blood tests every 6 weekly, 9.5% and 34.8%92 patients receiving total glucosides of paeony (TGP) and not receiving TGP experienced hepatotoxicity.

Any abnormal LFT or liver enzyme elevation or hepatobiliary dysfunction without any further details provided occurred in 0.4–21.8% patients with RA, PsO, treated with methotrexate and anti-TNF alpha over 6 months, with monitoring blood tests conducted monthly, 6 weekly, or 3 monthly with no relationship between frequency of testing and blood test abnormalities.65,66,68,69,71,72,76–78

Any abnormal LFT or liver enzyme elevation occurred in 1–3.8% patients with RA, treated with methotrexate and leflunomide over 6 months to 1 year, with monitoring blood tests conducted monthly, with no relationship between frequency of testing and blood test abnormalities.120,133

Any abnormal LFT occurred on 17 occasions (n = 93) in one study where patients with RA treated with methotrexate and sulfasalazine over 1 year, underwent monitoring blood tests conducted 3 monthly.94

Any abnormal LFT or liver enzyme elevation or hepatobiliary dysfunction not otherwise detailed occurred in 1–12.5% patients with RA, PsO, treated with anti-TNF alpha over 6 months–2 years, with monitoring blood tests conducted monthly, 6 weekly, 2 monthly, 3 monthly or 6 monthly with no relationship between frequency of testing and blood test abnormalities.70,76,77,78,135,152,156,175,177 In infliximab-treated patients with RA, < 1% were reported to have liver AEs over 18 months with monitoring blood test done 6 monthly.111

Any abnormal LFT or liver enzyme elevation occurred in 2.9–5.4% patients with RA, treated with leflunomide over 6 months–2 years, with monitoring blood tests conducted monthly or 3-monthly with no relationship between frequency of testing and blood test abnormalities.90,102,124,142 In leflunomide-treated patients with RA, 4% were reported to have hepatic dysfunction (Schultz 2017) over 4 years with monitoring blood tests done 6 monthly; and 4.5% were reported to have hepatic damage195 over 6 months with monitoring tests done 3 monthly.

Any abnormal LFT or liver enzyme elevation occurred in 1.1–2% patients with PsO, ReA, AS, or SLE, treated with sulfasalazine over 6 months–5 years, with monitoring blood tests conducted 3 monthly or annually.98,99

Information on severity of any liver enzyme elevation was only reported in a handful of studies, but these indicated that clinically relevant liver enzyme elevation that usually require treatment change based on the current guidance, that is at least twofold above the ULN or meet the definition of DILI, that is at least fivefold above the ULN were uncommon.

For methotrexate, 1% and 2.2% patients with RA or PsA had fivefold or greater increase in liver enzyme level or at least grade 3 liver enzyme elevation over 1-year follow-up with no relationship between frequency of testing that ranged between monthly to 6 monthly and rate of side effect.40,86,141

For anti-TNF-alpha, < 1% patients with RA or AS had at least grade 3 liver enzyme elevation or SAE due to liver enzyme elevation over 1-year follow-up with no relationship between frequency of testing that ranged between monthly to 2 monthly and rate of side effects.35,141,157

For methotrexate and anti-TNF combination therapy, 1–7% patients with RA had at least threefold increase in liver enzymes or at least grade 3 liver enzyme elevation over 1-year follow-up with no relationship between frequency of testing that ranged between monthly to 3 monthly and rate of side effect.62,69,141

For leflunomide, 1% and 6.4% patients with RA had threefold or greater increase in liver enzyme level over 2-year follow-up with no relationship between frequency of testing that ranged between monthly to 3 monthly and rate of side effect.40,91,124

For methotrexate and sulfasalazine, 8.9% patients with RA had threefold or greater increase in liver enzyme level over 5-year follow-up with 3-monthly testing.41

For methotrexate and leflunomide, 1.0–5.0% patients with RA and PsO had threefold or greater increase in liver enzyme level over 1-year follow-up with monthly to 3-monthly testing with no relationship with frequency of testing.40,92,133

Hepatotoxicity was reported in 3.8–13.7% patients treated with thiopurines (mostly azathioprine) in studies of IBD and SLE patients that lasted between 6 months and over 5 years with monitoring done 3-monthly.33,38,47,48,96,190,193 The cumulative incidence was higher (at 22.7–25.3%) in those treated with mercaptopurine, even though the follow-up period was considerably small at 6 months.38,53

Hepatitis

Drug-induced hepatitis due to methotrexate or methotrexate plus anti-TNF alpha occurred in < 1% patients63,66,118 with RA followed up over 6 months–2 years with monitoring blood tests done 6 weekly, 3- or 6-monthly. The severity was sufficient to be classified as SAE in < 1% patients in one study that was restricted to patients with PsO treated with anti-TNF alpha alone over 5 months and monitored with 2-monthly blood tests.151 There was an instance of viral hepatitis reported in one study restricted to AS patients monitored monthly.165 Drug-induced hepatitis due to thiopurine occurred in 3.3% patients63,66,118 with IBD followed up over 2 years with monitoring blood tests done 2- to 3-monthly.191

Alanine aminotransferase/aspartate aminotransferase elevation Alanine aminotransferase elevation due to methotrexate was evaluated in a large number of studies in patients with RA72,74,107,115,116,117,118,120,121,122,126,140 and data on occurrence of ALT elevation were reported as > ULN, unspecified, and using a variety of cut-offs > ULN (e.g. ≥ 1–2 × ULN, > 1.5 × ULN, > 2 × ULN, 2–3 × ULN, > 3 × ULN, > 2.5 < 5 × ULN and > 5 × ULN). The follow-up period varied from 6 months to 9 years. The monitoring blood tests were undertaken monthly, 2, 3 or 6 monthly. It was clear that any ALT elevation was common occurring in 26–44% patients, the occurrence of abnormal blood-test results reduced as the higher thresholds were used, for example, < 3.5% for > 3–5-fold ULN. At each level of abnormality, there were insufficient number of studies to draw any inference on the frequency of testing and relationship between side-effect occurrences.

A similar pattern of reporting and results were observed in studies that reported AST elevation. These studies were in RA populations and followed up patients for up to 9 years with monitoring bloods done monthly to 6-monthly.45,72,74,107,116,117,118,120,121,126,144

Several studies evaluated anti-TNF drugs in patients with RA, AS, IBD, PsO and follow-up ranged from 4 months to 2 years.28,39,74,77,82,136,144,153,155,156,159–161,164,166,168,174,175,178,179,181,182 Monitoring blood tests were done 2 weekly, monthly, every 2 months, every 3 months and 6 monthly. Just as for methotrexate, the abnormalities were reported as unspecified, >ULN, > 1–1.5 × ULN, > 1.5–3 × ULN, > 2 × ULN, > 3 × ULN, > 2.5 × ULN, > 3–5 × ULN, > 5 × ULN, markedly abnormal. For the cumulative incidence varied from 25% to 2% or lower depending on the severity of blood test abnormality. At each level of abnormality, there were insufficient number of studies to draw any inference on the frequency of testing and relationship between side-effect occurrences. A similar pattern was seen for AST elevations in this population.74,77,136,144,148,153,156,159,164,166,174,175,178–182

Two studies evaluated ALT elevation in those with RA treated with methotrexate and anti-TNF for 1–5 years. Patients underwent monitoring blood tests 3 or 6 monthly. The incidence of ALT elevation greater than or equal to twofold ULN did not vary according to frequency of testing but was higher in infliximab-treated populations than in etanercept-treated populations.41,86

Five studies102,121,195,196,197 reported on ALT elevation in RA patients treated with leflunomide for between 6 months and 5 years, monitored with monthly or 3-monthly blood tests and reported that ALT elevation above ULN was common, present in 32% cases in one study,102 but ALT elevations more than threefold ULN was present in 1–2.8% patients.102,195–197 Four of these studies reported on AST elevation in patients treated with leflunomide for 6 months to 5 years, monitored with monthly to 3-monthly blood tests and had similar results.102,121,196,197

Bilirubin and other liver enzymes

Elevated GGT was reported in 1.1% and 4.2% patients with RA treated with methotrexate and followed up over 2 years with monthly to 3-monthly monitoring blood tests.116,118 Elevated GGT was reported in 2% and 13.2% patients with AS or PsO treated with anti-TNF and followed up over 6 months–1 year with monthly to 4-monthly monitoring blood tests and no relation between frequency of testing and outcomes.156,166,175 Elevated ALP was reported in 1.7% and 16% patients with RA that were followed up for 2 years,116 and 9 years,107 respectively, with monthly monitoring blood tests. No such instances were reported with anti-TNF alpha.174

Elevated bilirubin was reported in 0–7% patients with AS or PsO treated with anti-TNF alpha over 6 months–2 years with blood testing done 2–8 weekly across studies with evidence of relation between frequency of testing and detection of abnormalities.164,174,182

Kidney adverse events

The cumulative incidence of nephrotoxicity often defined using increase in creatinine in patients with RA treated with methotrexate (one study with concurrent anti-TNF-alpha) ranged between < 1.0% and 4.3% in studies with 4 months to up to 2 years of follow-up that undertook monthly, 2-, 3- or 6-monthly monitoring blood tests.67,76,94,116,117,118,123,132 The cumulative incidence of nephrotoxicity often defined using increase in creatinine in patients with RA or PsO treated with anti-TNF-alpha ranged between < 1.0% and 1.6% in studies of between 4 months and 6 months duration that monitored blood tests either monthly or 2-monthly monitoring.73,136

Discussion review 1

Liver

The most frequent hepatic AEs with incidence rates of > 10% can be summarised as:

• Any abnormal LFT (ALT, AST, or elevated enzymes) in RA patients exposed to methotrexate alone or certain combinations with methotrexate, principally sulfasalazine, anti-TNF-alpha, and, to an extent, the following monotherapies: leflunomide, infliximab and etanercept.

Otherwise, most studies reported rates ≤ 10% for any hepatic AE for these and other therapies and indications.

Abnormal LFTs (studies, n = 24)

The incidence of abnormal LFTs was generally < 10%, but with higher reported rates occurring in studies evaluating the following therapies: methotrexate and methotrexate and anti-TNF-alpha drugs (1–21.8%), principally in RA patients, and infliximab (2–12.5%) in PsO and AS patients. Consistently lower rates than 10% were reported for methotrexate and leflunomide, leflunomide, methotrexate and sulfasalazine, sulfasalazine and etanercept. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Unspecified liver enzyme elevations (studies, n = 21)

The incidence of unspecified liver enzyme elevations was generally < 10%, but with higher reported rates occurring in studies evaluating the following therapies: methotrexate (< 1–28%), methotrexate and sulfasalazine (8.9–17.9%), methotrexate and adalimumab (1–16%), methotrexate and leflunomide (1–12.4%), and leflunomide alone (< 1–17%), almost exclusively in RA patients. Consistently lower rates than 10% were reported for methotrexate and certolizumab, certolizumab, sulfasalazine and etanercept. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Alanine aminotransferase enzyme elevations (≥ 2 ULN) (studies, n = 15)

The incidence of elevated ALT was generally < 10%, but with higher reported rates occurring in studies evaluating the following therapies: methotrexate (< 1–21%), methotrexate and sulfasalazine (11–21.5%), methotrexate and infliximab (1.9–17.9%), leflunomide alone (< 1–14.8%), almost exclusively in RA patients. consistently lower rates than 10% were reported for methotrexate and etanercept, and other anti-TNF-alpha agents. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Aspartate aminotransferase enzyme elevations (≥ 1 ULN) (studies, n = 27)

The incidence of elevated AST was generally < 10%, but with higher reported rates occurring in studies evaluating the following therapies: methotrexate (< 1–32%), methotrexate and sulfasalazine (9.8–11.3%), infliximab alone (2–24%) and etanercept (< 2.1–16%), exclusively in RA patients. Consistently lower rates than 10% were reported for adalimumab and leflunomide alone. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Hepatotoxicity (studies, n = 8)

The incidence of hepatotoxicity was reported as low as < 4%, but with higher rates frequently occurring in studies evaluating the following therapies: thiopurines (< 1–25.3%) in IBD patients or in a single study of methotrexate and leflunomide in RA patients (9.5–34.8%). No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Gamma glutamyl transferase, ALT, AST, ALP and bilirubin elevations and hepatobiliary disorders (studies, n = 13)

The range of the incidence of elevations in these markers could be extremely wide across a small number of studies, so they are not summarised here.

Hepatitis (studies, n = 6)

The incidence of hepatitis was < 1%, but with slightly higher rate (3.3%) occurring in a single study evaluating azathioprine for IBD patients. The lower reported rates were for methotrexate, methotrexate and adalimumab and adalimumab. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Liver injury, dysfunction, or damage (studies, n = 5)

The incidence of these outcomes was ≤ 1.1%, but with slightly higher rates (≤ 4.5%) for dysfunction or damage occurring in two studies evaluating leflunomide for RA patients. The lower reported rates were for adalimumab and methotrexate and leflunomide. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Blood

The most frequent haematological AEs identified were:

• Leucopenia in IBD patients exposed to thiopurines (< 1% to more than 30%), and RA patients exposed to etanercept and sulfasalazine (5–7.9%) and methotrexate (< 1–10.6%).
• Neutropenia in multiple indications exposed to methotrexate (< 1–8.1), leflunomide (2.2–4.3%), thiopurine (5.2%) and sulfasalazine (8.8%).
• Thrombocytopenia in RA patients exposed to methotrexate (< 2–6.9%), IBD patients exposed to thiopurine (5.2%) and SLE patients to mycophenolate (< 5.6%).

However, most studies reported rates < 3% for any haematological event.

We included studies in people with SLE. Some of the haematological manifestations in these populations could be due to disease activity rather than due to drugs.

Anaemia (studies, n = 17)

The incidence of anaemia was low (< 3%) across most studies with highest reported rates occurring in studies evaluating the following therapy: methotrexate and its combinations (0–7.5%), as well as leflunomide (4.6%) in RA patients. Consistently lower rates than these were reported for anti-TNFs. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Leucopenia (n = 38)

The incidence of leucopenia was low (< 3%) across most studies, but with some much higher reported frequencies reported in studies evaluating the following therapies: principally thiopurines (in IBD patients) (< 1% to more than 30%), and to a lesser extent etanercept and sulfasalazine (5–7.9%) and methotrexate (< 1–10.6%) (RA patients) and mycophenolate (≤ 7.5%) (SLE patients). Consistently lower rates than these were reported for methotrexate combinations, leflunomide, sulfasalazine and anti-TNFs. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Neutropenia (n = 18)

The incidence of neutropenia was low (< 3%) across most studies. The highest reported rates more consistently occurred in studies evaluating the following therapies: methotrexate (< 1–8.1%), leflunomide (2.2–4.3%), thiopurine (5.2%) and sulfasalazine (8.8%) (multiple indications). consistently lower rates than these were reported for methotrexate combinations, and anti-TNF. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Thrombocytopenia (n = 18)

The incidence of thrombocytopenia was low (≤ 1%) across most studies. The highest reported rates more consistently occurred in studies evaluating the following therapies: methotrexate (< 2–6.9%) (RA patients), thiopurine (5.2%) (IBD patients) and mycophenolate (< 5.6%) (SLE patients). Consistently lower rates than these were reported for methotrexate and adalimumab, anti-TNFs, azathioprine alone and sulfasalazine. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Pancytopenia (n = 7)

The incidence of pancytopenia was low (generally < 1%), with highest reported rates occurring in studies evaluating the following therapy: adalimumab in PsO patients. Consistently lower rates than these were reported for methotrexate and infliximab across different indications. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Kidney

No frequent kidney AEs were reported.

Renal function disorders or toxicity (studies, n = 6)

The incidence of renal function disorders was consistently low (< 1%) across all studies, except for one study reporting a rate of 4.3% for RA patients exposed to methotrexate; otherwise, consistently low rates were reported for methotrexate, certolizumab and azathioprine. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Creatinine elevation (studies, n = 8)

The incidence of renal function disorders was consistently low (≤ 1.6%) across all studies, except for two studies reporting rates of 2.7–3% for RA patients exposed to methotrexate; otherwise, consistently low rates were reported for methotrexate, methotrexate and adalimumab, anti-TNF-alpha and mycophenolate. No data on this outcome from other studies satisfying the inclusion criteria were identified for any of the remaining specified therapies.

Limitations

• The search was sensitive and extensive but, given the large number of therapies and populations, some relevant studies might have been missed.
• Despite only including studies with at least 200 participants overall, some included studies still had small numbers of participants in trial arms with relevant therapies.
• Only studies that reported at least one of our AEs or noted that all AEs were reported (even if none of our relevant AEs were then reported) were included: if a study did not report any relevant AEs and only reported AEs with a frequency of ≥ 2% or ≥ 5%, or reported only AEs generally or only certain types, for example infections, then participants might have experienced relevant AEs, but because they were not explicitly reported, the study was excluded;
• Two hundred studies with relevant data but without monitoring information were identified but not extracted because assumptions would have to be made about monitoring and testing regimes.

Results review 2

Characteristics of included studies

After deduplication, 16,400 titles and abstracts were reviewed. Finally, 56 studies, reported in 58 manuscripts, were included (see Figure 1).

Thiopurines

Thiopurines were evaluated as a predictor of specific blood-tested AEs in relevant populations in 18 studies.38,42,46–48,50,53,194,202–211 Almost all studies evaluated the use of thiopurines in populations with IBD; however, one study also included patients with SLE or RA.207 Of the 18 studies, 6 were case-control studies,38,47,50,53,209,211 2 of which were nested within a RCT38,53 and 3 of which were nested within a retrospective cohort study.47,50,211 In terms of the remaining study designs, there were nine retrospective cohort studies,46,48,202,204–208,210 two prospective cohort studies42,203 and one RCT.194 Most of the studies evaluated both mercaptopurine and azathioprine.38,42,46–48,50,53,194,204,205,208 Four studies evaluated only azathioprine203,207,210,211 and two studies also evaluated 5-ASA.202,206 Most studies did not report drug naivety; however, four studies reported that all included patients were thiopurine naive.38,194,204,207 Nine studies reported participants’ concomitant medications, and these included 5-ASA;38,46,48,50,53,203 systemic or corticosteroids;38,42,47,50,53,194,203 biologics38,194,203 or anti-TNF-alpha.50 In one study, the concomitant medication was allopurinol.208 Mean age of participants ranged from 25.7 years46 to 59 years;208 one study included an elderly group, the mean age of which was 66 years.204

Anti-TNF-alpha

Anti-TNF-alpha was evaluated in 10 studies.39,49,212–219 Four of these studies evaluated the use of anti-TNFs in populations with one of the following conditions: RA, AS, PsO or PsA;39,213,215,217 one study evaluated AS only,214 two evaluated PsA only216,219 and three studies evaluated populations with IBD.49,212,218 Of the 10 studies, one study used a RCT design,216 one a prospective cohort design,219 and the remaining eight were retrospective cohort studies.39,49,212–215,217,218 Two studies only evaluated infliximab,49,216 while the remaining studies evaluated two or more of the following anti-TNF drugs: entanercept,39,213–215,217,219 adalimumab,39,212–215,217,218 golimumab,39,212–214 infliximab212,217 and certolizumab.212

All studies apart from one39 reported participants’ concomitant medications. Where this was reported, methotrexate was the most common concomitant drug among the study populations,213,215,216 apart from one study,218 in which the population of IBD patients were most commonly taking thiopurines as concomitant medication with only 3% taking methotrexate. Other commonly reported concomitant medications were corticosteroids,213,214,216 NSAIDs,219 thiopurines212,213,216,218 and other anti-TNFs.217 The mean age of participants ranged from 31 years218 to 57.5 years.215

Methotrexate

Methotrexate was evaluated in 22 studies (22 articles).31,36,37,40,41,43–45,51,52,220–236 Seventeen of the studies evaluated the use of methotrexate in populations with RA,31,37,40,41,43–45,52,220–224,226–228,230,232–234,236 five studies evaluated disease-modifying antirheumatic drugs (DMARDs) in mixed populations of either RA, AS or Ps/PsA and SLE36,51,225,229,231 and one study in a PsO only population.235

Of the 22 studies, 11 were retrospective cohort studies.36,37,40,44,45,220,222,223,225,226,228,229,231,233,234,236 One case-control study was nested within a retrospective cohort,230 There were a further seven prospective cohort studies.51,52,221,227,232,235 One RCT was reanalysed as a cohort,41 and there were two additional RCTs.43,224

Of those studies that reported concomitant medication, all participants were taking folic acid in six studies,41,51,52,222,225,232,233 while some, but not all, participants were taking folic acid in four further studies.220,221,224,231 Other commonly reported concomitant drugs were prednisone,44,221,226 NSAIDs,40,44,221,224–226,232,234 sulfasalazine,43,44,225,230 corticosteroids232 and other DMARDs.223,227–229

Most studies did not report drug naivety; however, eight studies reported that all included patients were methotrexate naive.31,41,226,227,230–234 The mean age of participants ranged from 47.1 years43 to 68.8 years221 (although in this last-cited study, the investigators specifically compared groups of younger vs. older participants).

Leflunomide and mycophenolate

One study evaluated leflunomide228 and one study mycophenolate.229 Both studies were retrospective cohort/database studies; one study was conducted in a RA population with a mean age of 57 years228 and one in a mixed indication population with a mean age of 52 years.229

Combinations of classes of drug

Four studies evaluated populations taking two or more different classes of drugs relevant to the review.31,237–239 One study evaluated a number of drugs (methotrexate, sulfasalazine, leflunomide and azathioprine) in RA;31 one study, azathioprine and methotrexate in populations with IBD and RA, respectively;237 one study the use of mesalamine and infliximab and azathioprine in a population with IBD;238 and the fourth study239 evaluated non-biologic DMARDs or anti-TNFs in patients with PsA. Three studies were prospective cohorts with nested case-control studies,31,238,239 and one study had a retrospective cohort design.237 No details of dosing regime, drug naivety or concomitant medications were reported for any study. The mean age of participants ranged from 39 years238 to 58.8 years.31,237

Characteristics of outcomes

Appendix 3, Tables 52–65 present the outcome characteristics of included studies, including how the outcome was defined and the timing of tests and follow-up if reported. In some studies, outcomes were grouped under broad headings of ‘myelotoxicity’ or ‘hepatotoxicity’, while in others, outcomes were more specific, for example ‘leucopenia’, or ‘elevation of transaminase’. Differences were also found in how the outcome was recorded. Most broadly, studies reported any incidence of the outcome. In some studies, discontinuation of treatment or dose adjustment due to the outcome was reported; however, it was sometimes unclear where incidence necessarily led to discontinuation or dose adjustment as this may not have been explicitly stated.

Thiopurines

The most commonly studied outcome in the 18 studies evaluating thiopurines was leucopenia, with 9 studies measuring this outcome.38,42,46,50,205,208–211 Leucopenia was sometimes graded by severity, for example ‘mild’, ‘moderate’ or ‘severe’, and most studies reported the WBC threshold used to record incidence of the outcome, even where the outcome was not graded by severity. In addition to these studies that only recorded leucopenia, a further five studies included leucopenia under the broader outcome of myelotoxicity, along with other haematological AEs such as bone marrow suppression, lymphopenia or neutropenia.47,48,204,207,208 Similarly, one study reported ‘haematological AEs’ as the outcome in their study.194 Nine studies measured hepatotoxicity.42,47,48,53,203–205,208,209 This was reported to be measured in most studies by either elevation of ALT and/or AST, and this was also sometimes graded by severity but sometimes not. One study reported AKI as the outcome.202

Eleven of the studies reported ‘incidence’ of the outcome,38,47,50,53,194,202–206,208 three of which also reported discontinuation due to the AE.38,203,205 The remaining seven studies explicitly recorded the outcome only if it had caused discontinuation or dose adjustment.42,46,48,207,209–211 Follow-up of participants varied widely, ranging from 20 weeks38 to 23 years.50

Anti-TNFs

Six of the 10 studies evaluating anti-TNFs reported liver enzyme elevation as the outcome.39,49,213,214,216,218 As with the studies evaluating thiopurines, individual studies varied in whether they recorded the outcome by severity of liver enzyme elevation, for example mild, moderate and severe, or whether participants were simply recorded as having the outcome if they met the reported threshold, most commonly > 2 × ULN. Two studies recorded neutropenia as the outcome, as defined by neutrophil count.212,215 Two studies assessed predictors of abnormal laboratory results217 and decline in renal function.219 Nine studies reported ‘incidence’ of the outcome; the exception reported scale of decline in renal function.219 Follow-up of participants ranged from 12 months213,216 to 14 years.214

Methotrexate

The most studied outcome in the 22 studies was hepatotoxicity, recorded in 12 studies (although terminology varied widely between studies). This outcome was variously defined as hepatotoxicity,40,220,224,227,235 elevation of transaminases,221,222,231 elevated liver enzymes,41,44,232 transaminitis,37,230 hepatic AEs,43 hepatic abnormalities,45 persistent transaminitis,226 liver/hepatic toxicity,51,52,234 abnormal hepatic function235 and liver disease.36 Severity of hepatotoxicity was sometimes graded, for example into ‘mild’, ‘moderate’ and ‘severe’. Most studies reported AST and/or ALT at various levels > ULN to record incidence of the outcome, even where the outcome was not graded by severity.

Five studies recorded haematological AEs such as leucopenia,221,236 cytopenia,225,234 or simply haematological ‘events’45 or ‘toxicity’.52 One study reported change in estimated glomerular filtration rate (eGFR) in the previous year and the dependent variable.223 Twenty-one of the 22 studies reported ‘incidence’ of the outcome; 6 of these studies also recorded whether participants discontinued with the drug due to the recorded AEs.31,44,45,51,224,228 Follow-up of participants varied widely, ranging from within the first 6 months233 to 18 years.36

Leflunomide and mycophenolate

The two studies evaluating these therapies both reported as the outcome of interest as discontinuation due to abnormal blood test results.228,229

Combinations of classes of drug

Four studies evaluated populations taking a combination of different classes of drugs relevant to the review237–239 and studied neutropenia,31 liver injury, graded by severity238 and liver abnormalities.237,239 All four studies reported ‘incidence’ of the outcome.

Quality assessment review 2

The results of the quality assessments of all 58 included articles (56 studies) using the QUIPS tool is presented in Table 13. Full details are in Appendix 2, Tables 45–51.

Study participation: Participants in most studies met the criteria for the specific inflammatory diseases of interest with baseline characteristics adequately described. However, where retrospective analysis of a cohort using data sets (e.g. from one institution) were used, it was often unclear how many eligible participants were screened for inclusion. In these cases, it was often not possible to tell whether only participants who met the criteria for inclusion and had data for all relevant outcomes were included in the analyses, and if so, how many potentially eligible participants were excluded due to a lack of data. This presents a risk of selection bias across these studies.

Study attrition: A lack of clear reporting of attrition and missing data in the retrospective cohort studies meant that it was not possible to judge whether there was any, and if so, how much data were missing for each variable of interest. This level of detail was rarely reported, leading to the risk of bias from potentially missing data. Furthermore, in these cases, it was not possible to judge whether patients whose data were missing differed in characteristics compared to those without missing data.

Prognostic factor measurement: Definitions of key prognostic factors were in the main well reported; however, some studies had many potential prognostic factors (any demographic or clinical characteristic entered in a multivariable analysis), and in these cases there was generally a lack of detail for how these were defined. Given the lack of reporting of missing data, any methods of imputation of missing data were rarely reported.

Outcome measurement: Most of the included studies used clear definitions of each outcome measured, with details of laboratory tests and diagnostic thresholds clearly described. The timing of occurrence of the outcome was mostly not reported, or only reported descriptively, precluding evidence synthesis on the timing of the outcomes.

Adjustment for other prognostic factors: The following list of other prognostic factors (adjustment variables) was considered by the review authors to require inclusion in multivariable analyses for these studies – age, sex, body mass index (BMI), alcohol intake, concomitant immune-suppressing drugs, disease duration, CKD, and other comorbidities. While the inclusion of all these variables was not an expectation, in 42 studies were considered to include either no or only a selective limited choice of adjustment (existing prognostic) factors in the multivariable model. Some studies only included variables in their multivariable analysis when the unadjusted effect of those variables was significant. It was also often unclear whether there was missing adjustment factor data and/or how these were handled with multiple imputation only reported in four studies.36,194,228,229

Statistical analysis: Selective reporting of only significant results occurred in 27 studies and where this was the case, prognostic effect estimates were consequently not available for all variables, leading to potential reporting bias.

Synthesis

A narrative synthesis of the results prioritising at least low-quality evidence is presented here and details outlined in Tables 14–18. Population characteristics are described in Appendix 4, Table 78. Full details and evidence found to be very low quality is only reported in Appendix 4, Tables 79–114.

Hepatotoxicity

Methotrexate

There was high-quality evidence that baseline elevated liver enzymes41,52,222,227,231,232 are associated with hepatotoxicity, and folate supplementation45,224,231 is associated with reduced risk of hepatotoxicity. There was moderate-quality evidence that increased risk of hepatotoxicity is associated with excessive alcohol consumption,36,37,40,41,222,232 diabetes,36,222,226,231,232 pre-existing liver disease,40,41,226,227 PsO +/− PsA (compared to RA)36,51,220,231 and concurrent leflunomide40,41,230 treatment. There was low-quality evidence that Charlson comorbidity index36,231 was associated with hepatotoxicity. There was low-quality-evidence that age,41,44,52,220–222,226,227,230,232,235 smoking,36,40,41,222,226,232,235 auto-antibodies associated with RA,41,52,222,226,230,232 disease severity/activity,41,52,226,230 inflammatory markers,41,52,226 disease duration,41,226,230 methotrexate dose40,41,220,222,226,230,231 and concomitant anti-TNF-alpha drugs40,41,226,231 were not associated with hepatotoxicity.

Anti-TNF

There was high-quality evidence of baseline liver enzyme elevations39,213,214 being associated with hepatotoxicity. There was low-quality and moderate-quality evidence from two studies of an increased risk of hepatotoxicity with increased BMI and comorbidities, respectively.214,239 There was low-quality evidence of no increased risk of hepatotoxicity with positive anti-nuclear antibody (ANA) or rheumatoid factor (RF).214,239 Prescription of other steroid sparing drugs were not associated with hepatotoxicity, low-quality evidence from three studies.39,214,239

Thiopurines

There was moderate-quality evidence that pre-existing ALT elevation increased the risk of hepatotoxicity,49 and high-quality evidence that use of mercaptopurine rather than azathioprine48,53,204 increased the risk of hepatotoxicity. There was moderate-quality evidence that age47,53,204,206 and male sex47,48,53,204–206 increased and smoking47,48,206 did not increase the risk of hepatotoxicity.

There was low-quality evidence that disease activity,48,204,238 concomitant biologic and non-biologic immune-suppressing drug therapy,47,48,53,206 azathioprine dose47,53 were not associated with hepatotoxicity.

There was moderate-quality evidence from one study that included IBD patients treated with different drugs (5-ASA, corticosteroids, azathioprine, anti-TNF-alpha, or none) that liver steatosis was associated with hepatotoxicity.238

Cytopenia (including neutropenia)

Methotrexate

There was low-quality evidence that chronic liver disease225 increased cytopenia risk.

Anti-TNF

There was low-quality evidence for no association between increasing age, sex or inflammatory disease type and neutropenia.212,215 There was low-quality evidence of increased risk of cytopenia from previous neutropenia, and reduced risk with increased baseline neutrophil count.215 There was moderate-quality evidence that leucopenia or low neutrophil count at baseline were associated with increased risk of neutropenia in those prescribed conventional DMARDs +/− biologics.31

Thiopurines

There was high-quality evidence that mercaptopurine use was associated with an increased risk of cytopenia compared to azathioprine.38,48,204,205 There was high-quality evidence that poor thiopurine metabolisers (based on TPMT/NUDT genotype +/− enzyme intermediate or low activity42,203,207,210) had increased risk of cytopenia. There was low-quality evidence that female sex38,47,48,50,203–205,210 was associated, and current smoking38,47,48,203 and disease activity38,47,48,210 were not associated with cytopenia.

There was moderate-quality evidence from one study that included autoimmune-rheumatic disease patients treated with different drugs that low baseline leucopenia and low neutrophil count were associated with neutropenia.

Nephrotoxicity

There was moderate-quality evidence that concomitant use of NSAIDs234 and risk factors for renal function decline219 were prognostic for nephrotoxicity in patients prescribed methotrexate and anti-TNF-alpha, respectively.

Composite toxicity (treatment discontinuation with cytopenia, AKI, or elevated liver enzymes). There was moderate-quality evidence that epilepsy and blood-test abnormalities in the first few months of shared-care prescription228 were associated with composite toxicity in patients prescribed leflunomide and moderate-quality evidence that CKD stage 3229 was associated with composite toxicity in patients prescribed MMF.

Discussion review 2

The review retrieved 56 studies published from 1995 to January 2023 that reported potential prognostic factors for prespecified AEs in patients with a range of conditions who were prescribed immune-supressing drugs. Most of these were designed as retrospective cohort studies. The most consistent finding was that, across drug types, baseline elevated liver enzymes were associated with increased risk of subsequent hepatotoxicity after adjusting for many other prognostic factors. The largest quantity of evidence related to prognostic factors associated with an increased risk of hepatotoxicity, with much of this from low- or moderate-quality evidence. The main reasons for downgrading evidence were single study, imprecision and inconsistency. Factors shown to increase risk included BMI, age, comorbidities and the specific drug prescribed or use of concomitant drugs. These findings varied by drug type (anti-TNF-alpha, methotrexate or thiopurines). Conversely, there was strong evidence that supplementation of folates was shown to reduce risk in patients prescribed methotrexate. Several factors were shown to predict an increased risk of cytopenia. These included previous neutropenia, comorbidities and poor metaboliser based on TPMT/NUDT genotype +/− enzyme intermediate or low activity. Little evidence was identified for prognostic factors for nephrotoxicity, and the quality was low, but included concomitant use of NSAIDs in those prescribed methotrexate.

The key findings suggest that patients prescribed immune-supressing drugs are, in general, at higher risk of increased liver, blood and kidney AEs if they have a prior history of or baseline elevated markers of toxicity, if they have comorbidities, or if they have tested positive for indicators of poor metaboliser activity. Identifying patients at the earliest opportunity who are at increased risk due to these factors could potentially help to reduce the means that they may avoid risk of AEs and ensure benefit from appropriate adjustments to treatment are being made.

The review was broad, with a focus on identifying risk factors for liver, blood and kidney AEs. The strengths of this systematic review are its comprehensive inclusion of evidence that spans all relevant immune-supressing drugs prescribed to patients with a range of conditions and up-to-date searches which retrieved evidence as recently as January 2023. Furthermore, it was conducted and reported following international guidelines by a team of highly experienced reviewers and clinicians. The limitations are that, while the search was sensitive and extensive, some relevant studies might still have been missed given the large number of therapies and populations; and despite only including studies with at least 200 participants overall, some included studies still had small numbers of participants in trial arms with relevant therapies resulting in low event rates for the outcomes of interest. Studies of certain populations may have been disproportionately excluded due to generally lower sample sizes, for example SLE. The evidence base itself was extensive and the risk of bias was generally low or moderate according to QUIPS assessments. However, the included studies’ available data and analyses for the outcomes of interest were relatively limited, with the result that quality of the evidence was assessed as low or very low according to the GRADE criteria, except for findings for prognostic factors for elevated liver enzymes. Prognostic factor findings were mainly assessed as very low quality, and this was in the main due to data being derived from small single studies, or where this was from multiple studies there was heterogeneity in outcomes, study designs, cut points used to describe prognostic factors and populations studied which also prevented meta-analysis.

Patients prescribed immune-supressing drugs are, in general, at higher risk of increased liver, blood and kidney AEs if they have a prior history of or baseline blood-test abnormalities, if they have comorbidities, or if they have tested positive for indicators of poor metaboliser activity. Identifying patients at the earliest opportunity who are at increased risk due to these factors could potentially help to reduce the risk of AEs and ensure appropriate adjustments to monitoring are being made.