Prevalence and predictors of adverse events with methotrexate mono- and combination-therapy for rheumatoid arthritis: a systematic review

Abstract

Objectives

This systematic review aims to summarize rates of adverse events (AEs) in patients with RA or inflammatory arthritis starting MTX as monotherapy or in combination with other csDMARDs, and to identify reported predictors of AEs.

Methods

Three databases were searched for studies reporting AEs in MTX-naïve patients with RA. Randomized controlled trials (RCTs) and observational cohort studies were included. Prevalence rates of AEs were pooled using random effects meta-analysis, stratified by study design.

Results

Forty-six articles (34 RCTs and 12 observational studies) were identified. The pooled prevalence of total AEs was 80.1% in RCTs (95% CI: 73.5, 85.9), compared with 23.1% in observational studies (95% CI: 12.3, 36.0). The pooled prevalence of serious AEs was 9.5% in RCTs (95% CI: 7.4, 11.7), and 2.1% in observational studies (95% CI: 1.0, 3.4). MTX discontinuation due to AEs was higher in observational studies (15.5%, 95% CI: 9.6, 22.3) compared with RCTs (6.7%, 95% CI: 4.7, 8.9). Gastrointestinal events were the most commonly reported AEs (pooled prevalence: 32.7%, 95% CI: 18.5, 48.7). Five studies examined predictors of AEs. RF status, BMI and HAQ score were associated with MTX discontinuation due to AEs; ACPA negativity, smoking and elevated creatinine were associated with increased risk of elevated liver enzymes.

Conclusion

The review provides an up-to-date overview of the prevalence of AEs associated with MTX in patients with RA. The findings should be communicated to patients to help them make informed choices prior to commencing MTX.

Rheumatology key messages

• Gastrointestinal events are the most commonly reported AEs in patients with RA commencing MTX (32.7%).

• The pooled prevalence of alopecia was low (6.3%), potentially reducing patients’ concerns when starting MTX.

• Limited number of predictors have been examined to date, other possible predictors should be explored.

Introduction

In the last three decades, MTX has become the anchor treatment in the management of patients with RA, mainly due to its effectiveness in achieving clinical remission or low disease activity in a large proportion of patients [1, 2]. MTX is usually the conventional synthetic DMARD (csDMARD) of first choice for the treatment of newly diagnosed patients with RA, and is central to the recent treatment guidelines published by the EULAR and the ACR [3, 4].

MTX, as any treatment, is associated with a number of potential adverse events (AEs). These AEs range from mild (but often distressing for the patient) gastrointestinal (GI) symptoms to more serious events, such as pneumonitis. GI events, such as nausea, diarrhoea or abdominal pain were the most frequently occurring AEs (30.8%) in a systematic review published in 2009 about the long-term safety (>2-years follow-up) of MTX [5]. Moreover, GI AEs are one of the main reasons why a significant proportion of patients discontinue or stop adhering to MTX treatment. A study of 762 patients with RA found that GI events accounted for 32.7% of the AEs leading to MTX discontinuation [6]. In another study, nausea was reported in 74% of patients who were non-adherent to MTX due to AEs [7].

Liver enzyme abnormalities are also a common AE associated with MTX treatment, mainly elevated alanine transferase (ALT) and aspartate transferase (AST) enzymes, thus necessitating that all patients undergo regular blood tests [8]. In a review of randomized controlled trials (RCTs) comparing MTX to placebo, 15% of the patients who received MTX developed elevated liver enzymes compared with 3% in the placebo group [9]. Mucocutaneous AEs of MTX are diverse and include alopecia, skin rash, stomatitis or oral ulcers. Although alopecia is not common (6%), it can be a major concern affecting patients’ decision to start MTX therapy [9]. The reported prevalence of neurological AEs associated with MTX varied in previous reviews and ranged between 5.5% and 35% [5, 10]. Pulmonary AEs are not uncommon and range from a mild cough to more severe pneumonitis. Pneumonitis is a rare but severe idiosyncratic reaction that is documented in many studies, and usually occurs early after starting MTX therapy and is not related to the dose [11, 12].

To date, there have been many studies reporting on MTX-associated AEs; nevertheless, they have reported inconsistent results. Furthermore, previous reviews of the incidence and prevalence rates of AEs have been mainly limited to RCTs or have included both incident and prevalent MTX users. Therefore, a review compiling up-to-date incidence and prevalence rates of the AEs associated with starting MTX in patients with RA is needed.

In addition, identifying predictors of AEs can help individualize therapy, promote adherence, and subsequently improve effectiveness. Various demographic, clinical and drug-related factors have been examined as predictors of MTX-related AE occurrence, with folic acid supplementation among the most recognized factors [13, 14]. A review of RCTs of folate supplements vs placebo with MTX found a significant decline in the incidence of GI and hepatic AEs in the folate group, with relative risk (RR) of 0.74 (95% CI: 0.59, 0.92) and 0.23 (95% CI: 0.15, 0.34), respectively [15]. However, the evidence of association is not always conclusive or consistent across studies and therefore, uncertainty exists regarding the predictors of theses AEs. A few reviews have covered potential predictors of MTX-related AEs [16, 17], but there have been no recent systematic reviews bringing together the literature on predictors of AEs.

Therefore, this review aimed to (i) summarize the prevalence and incidence rates of AEs (i.e. total AEs, serious AEs, discontinuation due to AEs, GI, elevated liver enzymes, mucocutaneous, neurological and pulmonary AEs) in patients treated with MTX for RA and to (ii) identify the reported demographic, clinical and treatment related predictors of these AEs.

Methods

Search strategy

A systematic search of published literature was carried out using three electronic databases: Embase, Medline and the Cochrane Central Register of Controlled Trials (CENTRAL). A search strategy combining free text and MeSH terms was developed with the help of the centre librarian to identify relevant studies and implemented in the three databases (Supplementary Table S1, available at Rheumatology online). To capture the AEs rates in the current era, only manuscripts published from 1 January 2005 (allowing sufficient time for a study after the year 2000 to be completed and published) up to 12 February 2019 were included in this review. PRISMA reporting guidelines were followed in this systematic review [18]. Studies with multiple publications were reviewed to choose a representative report/s which included the outcomes of interest.

Both RCTs and longitudinal observational studies were included in this review. Only studies including adult (≥18 years old) patients with RA, inflammatory arthritis or undifferentiated arthritis were assessed. Studies examining genetic predictors, or the effect of natural products and supplements were excluded. Case reports, case series, reviews and conference abstracts were also excluded.

Patients had to be MTX-naïve before starting the study; retrospective studies were included if MTX treatment start date was defined and patients had not received MTX before the start date. Patients could have started MTX monotherapy or in combination with other csDMARDs or Corticosteroids (CSs), but not in combination with biologic DMARDs (bDMARDs) or targeted synthetic DMARDs (tsDMARDs). Only studies in English or translated into English were included.

Outcomes and data extraction

The outcomes of interest were the prevalence and incidence rates of the total of ‘any’ AEs, serious AEs, discontinuation of MTX due to AEs, elevated liver enzymes (including ALT and AST enzymes), GI, mucocutaneous, neurological and pulmonary AEs. The AE rates were only collected from the study arms with MTX monotherapy or MTX combined with GCs or csDMARDs, but not from other comparator arms. Some AEs were grouped together due to similarity, such as stomatitis and oral ulcers, and pneumonitis and interstitial lung disease (ILD).

The incidence rates of AEs were retrieved directly from studies reporting the rate in person-years. No attempts were made to estimate the incidence rates due to insufficient data on the number of events and attrition rate in most studies. The prevalence estimates were calculated as the number of patients who developed each AE divided by the study population at baseline; therefore, a patient could have more than one event during a study period but would only be counted once in the prevalence estimates.

For studies assessing predictors of MTX-related AEs, data on the possible predictors including the adjusted effect estimates and 95% CIs were retrieved from these studies.

Risk of bias assessment

The risk of bias in individual studies was assessed using a checklist adapted from previously published tools [19, 20]. The checklist examined the risk of bias related to the review outcomes and was divided into three main domains for all studies (selection, attrition and outcome measurement). Additional assessment of analysis and reporting was performed for studies examining predictors of AEs. Publication bias across studies was examined graphically by funnel plots for total AEs, serious AEs and discontinuation of AEs in RCTs.

Data analysis

Descriptive statistics were used to summarize the included studies. Pooled prevalences of reported AEs were estimated using random effects meta-analysis and stratified by the study design (RCTs and observational studies). Where studies had several eligible MTX arms, the arms were combined (i.e., the number of AEs/baseline sample sizes were summed) and the study included only once in each meta-analysis. Freeman–Tukey double arcsine transformations were performed for all meta-analyses to stabilize the variance and to include prevalence estimates of 0 or 100% [21]. Overall and subgroup heterogeneity was quantified using the I² statistics if more than three studies reported the AE. Meta-analyses were conducted using the metaprop package in Stata 14 (College Station, TX, USA) [22].

Results

A total of 5098 records (1529 found in Medline, 2168 in Embase and 1401 in CENTRAL) were identified. Of those, 1958 references were duplicates. Two authors (A.A.S., S.D.S.) screened 3140 titles and abstracts independently, and 2790 references were excluded for various reasons (Fig. 1). The full texts of 350 articles were then reviewed by the same two authors. Any discrepancies during screening were discussed between the two authors, disagreements were referred to a third author. Of the 350 full texts screened, 46 fulfilled the inclusion criteria.

Fig. 1.

PRISMA flow diagram capturing the different phases of the systematic review

AE: adverse event.

The 46 manuscripts comprised 34 clinical trials and 12 observational studies (Table 1). Twenty-two studies were based in Europe, nine in Asia, two in Australia, one in North America and 12 were multi-national studies. The total number of participants included was 10 122; the mean proportion of women was 73.3% and ranged between 53.4% and 86.1%. The mean reported follow-up duration of all studies was 16 months with an s.d. of 8 months [15 months for RCTs (s.d. = 8), 19 months for observational cohorts (s.d. = 8)].

Table 1.

Description of included studies including demographic and clinical characteristics of the study populations

References	Country	Study design, setting	Follow-up length	MTX arm (n)	Disease duration mean (s.d.) or median (IQR)	Age, mean (s.d.)	Female sex (%)	MTX starting dose (mg/week)	MTX dose mean (s.d.) (mg/week)	Comparator arm (n)
Clinical trials
Stouten et al. [23]	Belgium	RCT Multicentre	2 years	High risk patients:				15	NR	5 MTX arms
				MTX+SSZ+GCs (98)	1.8 (3.1) weeks	53.2 (11.9)	64 (65)
				MTX+GCs (98)	2.6 ( 3.3) weeks	51.8 ( 13.1)	63 (64)
				MTX+LEF+GCs (93)	3.1 (6.4) weeks	51.1 (12.8)	64 (69)
				Low risk patients:	3.2 (6.6) weeks	51.0 (14.0)	38 (81)
				MTX (47)	1.9 (2.7) weeks	51.4 (14.4)	33 (77)
				MTX+GCs (43)
Stamm et al.[24]	Austria	RCT	2 years	MTX (36)	9.4 (2.3) weeks	52.9 (14.0)	28 (78)	10	NR	IFX+MTX (38)
		Multicentre								PBO (16)
Atsumi et al.[25]	Japan	RCT Multicentre	2 years	MTX+PBO (157)	4.3 (2.8) months	49.0 (10.3)	127 (81)	8	11.5 (2.8	CZP+MTX (159)
Emery et al.[26]	Multi-national	RCT	1 year	MTX+PBO (213)	2.9 (2.9) months	51.2 (13.0)	170 (80)	10	22 mg at week 8	CZP+MTX (655)
		Multicentre
Bijlsma et al.[27]	The Netherlands	RCT	2 years	MTX+PBO (108)	Median 27 (15–46) days	Median 53.5 (44.5–62)	69 (64)	10	NR	TCZ+MTX (106)
		Multicentre								TCZ+PBO (103)
Burmester et al.[28]	Multi-national	RCT	2 year	MTX+PBO (287)	0.4 (0.48) years	49.6 (13.1)	229 (80)	7.5	16.2 (6.2) at week 104	TCZ+MTX (578)
		Multicentre								TCZ+PBO (292)
Scott et al.[29]	The UK	RCT	2 years	MTX (75)	0.14 (0.19 months	54 (13.0)	54 (72)	7.5	NR	Anakinra+MTX (79)
		Multicentre
Emery et al.[30]	Multi-national	RCT	1 year	MTX+PBO (116)	0.50 (0.49 year	49.1 (12.4)	89 (77)	7.5	NR	ABA+MTX (119)
		Multicentre	(+1 year no treatment)							ABA+PBO (116)
Lee et al.[31]	Multi-national	RCT	2 years	MTX+PBO (186)	2.7 years	48.8 (13.3)	145 (78)	10	18.5 mg at month 3	TOF (770)
		Multicentre
Nam et al.[32]	The UK	RCT	78 weeks	MTX+IV GCs (57)	Median 6.9 (5–10) months	52.9 (12.8)	41 (72)	10	NR	IFX+MTX (55)
		Multicentre
Nam et al.[33]	The UK	RCT	78 weeks	MTX+PBO (55)	Median 8 (6–11) months	48.4 (13.3)	40 (73)	10	median 25 (20–25)	ETN+MTX (55)
		Multicentre
Takeuchi et al.[34]	Japan	RCT	26 weeks	MTX+PBO (163)	0.3 (0.4) year	54.0 (13.2)	128 (76)	4–6	6.6 (0.6 mg	ADA+MTX (171)
		Multicentre
Chen et al.[35]	China	RCT	24 weeks	MTX+PBO (132)	32 (20) months	47.7 (11.4)	113 (86)	10	NR	Anbainuo (132)
		Multicentre								Anabainuo+MTX (132)
de Jong et al. [36]	The Netherlands	RCT	3 months	MTX+SSZ+HCQ +IM GCs (91)	162 (97) days	53 (15)	55 (60)	NR	25 (1)	3 MTX arms
		Multicentre		MTX+SSZ+HCQ +oral GCs (93)	184 (92) days	54 (14)	67 (72)		At week 3
				MTX+oral GCs (97)	154 (83) days	54 (14)	68 (70)
de Rotte et al.-a[37]	The Netherlands	RCT (de Jong 2013)	9 months	MTX (285)	NR	54 (14)	199 (70)	NR	25 (1)	Two cohorts combined
		+ Prospective cohort		MTX (102)		52 (16)	72 (71)		15 (2)
Detert et al.[38]	Multi-national	RCT	48 weeks	MTX+PBO (85)	1.6 (1.7) months	52.5 (14.3)	57 (67)	15b	15 mg stable dose	ADA+MTX (87)
		Multicentre
Kavanaugh et al.[39]	Multi-national	RCT	26 weeks	MTX+PBO (517)	4.5 (7.2) months	50.4 (13.6)	382 (74)	7.5	NR	ADA+MTX (515)
		Multicentre
Leirisalo-Repo et al.[40]	Finland	RCT	2 years	PBO+MTX+SSZ	Median 4 (3–6) months	46 (11)	31 (63)	10b	Median 20 (15– 25)	IFX+MTX+SSZ
		Multicentre		+HCQ+GCs (49)						+HCQ+GCs (50)
Bakker et al.[41]	The Netherlands	RCT	2 years	MTX+PBO (119)	NR	53 (13)	72 (61)	10b	NR	2 MTX arms
		Multicentre		MTX+oral GCs (117)		53 (14)	70 (60)
Hobl et al.[42]	Austria	RCT	16 weeks	Standard MTX (10)	NR	51 (15)	7 (70)	15	NR	2 MTX arms
		Single centre		accelerated dose MTX (9)		62 (7)	7 (78)	25
Stohl et al.[43]	Multi-national	RCT	1 year	MTX+PBO (207)	1.2 years	49.2	153 (74)	7.5	NR	OCR+MTX (398)
		Multicentre
Wevers-de Boer et al.[44]	The Netherlands	Non-RCT (induction phase)	4 months	MTX+oral GCs (610)	NR	NR	NR	15	NR	ADA+MTX
		Multicentre
Tak et al.[45]	Multi-national	RCT	1 year	MTX+PBO (249)	0.9 (1.1) years	48.1 (12.7)	192 (77)	7.5	NR	RTX+MTX (499)
		Multicentre							>18 at week 8
Verstappen et al.-a[46]	The Netherlands	RCT	2 years	Intensive MTX (149)	NR	54 (14)	102 (69)	7.5b	16.1 (CI = 14.8–17.3)	2 MTX arms
		Multicentre		Standard MTX (140)		52 (15)	91 (65)		14 (CI: 13.1–14.8)
Emery et al.[47]	Multi-national	RCT	24 weeks	MTX+PBO (160)	2.9 (4.8) years	48.6 (12.9)	134 (84)	10	19.1 (2.73) at week 23	GOL+MTX (477)
		Multicentre
Westhovens et al.[48]	Multi-national	RCT	1 year	MTX+PBO (253)	6.7 (7.1) months	49.7 (13.0)	199 (79)	7.5	19 (2.3) at year 1	ABA+MTX (256)
		Multicentre
Bejarano et al.[49]	The UK	RCT	56 weeks	MTX+PBO (73)	7.9 (5.4) months	47 (9)	39 (53)	7.5	NR	ADA+MTX (75)
		Multicentre
Braun et al.[50]	Germany	RCT	24 weeks	SC MTX (188)	median	median		15b	NR	2 MTX arms
		Multicentre		oral MTX (187)	2.5 (range: 0–535)	58 (20–75)	149 (79)
					2.1 (range: 0–293) months	59 (22–75)	138 (74)
Emery et al. [51]	Multi-national	RCT	1 year	MTX+PBO (263)	9.3 (SE = 0.4) months	52.3 (SE = 0.8)	191 (73)	7.5	Median 19.6 at week 8	ETN+MTX (265)
		Multicentre
Capell et al. [52]	The UK	RCT	1 year			Median		7.5	Median	SSZ (55)
		Multicentre		SSZ+MTX (56)	1.9 year s	56 (30–78)	42 (75)		12.5 (0–20)
				MTX (54)	1.8 years	53 (34–79)	43(79)		15 (0–20)
Breedveld et al.[53]	Multi-national	RCT	2 years	MTX+PBO (257)	0.8 (0.9 years	52.0 (13.1)	190 (74)	7.5	16.9	ADA+MTX (268)
		Multicentre								ADA (274)
Dirven et al.-a[54]	The Netherlands	RCT Multicentre	5 years	MTX+ other DMARDs (498)	Median 24 (14–53) weeks	54 (14)	336 (68)	NR	NR	Combined arms of Goekoop- Ruiterman 2005
Goekoop- Ruiterman et al.[55]	The Netherlands	RCT	1 year		Median				NR	IFX+MTX (128)
		Multicentre		sequential monotherapy (126)	23 (14–54)	54 (13)	86 (68)	15
				MTX step-up combination (121)	26 (14–56)	54 (13)	86 (71)	15
				MTX+SSZ+ GCs (133)	23 (15–53) weeks	55 (14)	86 (65)	7.5
Ichikawa et al.[56]	Japan	RCT	96 weeks	MTX (23)	8.2 (4.8	52.7 (9.3)	16 (70)	4	NR	Bucillamine (24)
		Multicentre		MTX+bucillamine (24)	10.6 (6.6) months	49.1 (12.9)	17 (71)
Observational studies
Bluett et al.-a[57]	The UK	Prospective cohort	At least 1-year	MTX (431)	Median 7 (4–12) months	Median 58 (48–68)	272 (63)	7.5–10	NR	NA
		Multicentre
Gaujoux-Viala et al.[58]	France	Prospective cohort	2 years	Optimal MTX dose (76)	8.8 (9.7) months	44.5 (12.6)	56 (74)	14.9 (4.9)	15.1 (4.4)	NA
		Multicentre		Non-optimal MTX dose (212)	7.7 (9.2) months	50.3 (11.3)	163 (77)	11.2 (3.1)	11.9 (3.1) at 6 months
Takahashi et al.[59]	Japan	Prospective cohort	76 weeks	MTX (79)	2.2 (5.7) months	56.7 (14.7)	68 (86)	8	NR	NA
		Multicentre
Cummins et al.[60]	Australia	Prospective cohort	Median = 2 years	MTX+SSZ +HCQ (119)	Median 6 (4–10.5) months	52.8 (13.1)	80 (67)	10	NR	NA
		Single centre
Kudo-Tanaka et al.[61]	Japan	Prospective cohort	1 year	MTX (29)	156.8 (107.6) days	57.9 (13.1)	22 (76)	4–10	Median 6.3 (range: 4–10)	No MTX (19)
		Multicentre
Schipper et al.[62]	The Netherlands	Prospective cohort	1 year	MTX tight control (126)	Median 15 (8–26) weeks	56 (13)	78 (62)	15	NR	Usual care (126)
									25 mg target
Hider et al.-a[63]	The UK	Prospective cohort	at least	MTX (309)	NR	59.3 (48.3–69)	204 (66)	7.5–10	NR	NA
		Multicentre	2 years						25 mg target
Ideguchi et al.[64]	Japan	Retrospective cohort	437 person-years	MTX (273)	9.9 (10.6) years	57.6 (11.4)	229 (84)	mean 3.8	5.5 (1.9 mg	NA
Proudman et al.[65]	Australia	Prospective cohort	3 years	MTX+SSZ+HCQ (61)	Median 12 (6–104) weeks	56 (14)	46 (76)	10	median 10 mg (0–15)	NA
		Multicentre
Shoda et al.[66]	Japan	Prospective cohort	mean 39.7 (12.0) months	MTX (69)	NR	61.0 (9.0)	58 (84)	2	7.2 (3.4) mg	NA
		Single centre
Weaver et al.[67]	USA	Prospective cohort	1 year	MTX (941)	3.5 (7.1) years	56.8 (14.0)	705 (75)	median	median at 1 year	bDMARDs
		Multicentre		MTX+HCQ (325)	4.6 (6.9) years	53.8 (14.5)	260 (80)	10	12.5 mg
				MTX+LEF (191)	7.4 (8.4) years	55.5 (12.6)	145 (78)	10	12.5 mg
				MTX+HCQ+SSZ (42)	7.2 (7.5) years	47.8 (13.9)	33 (79)	15	15 mg
								12.5	15 mg
Singal et al.[68]	India	Prospective cohort	20 months	MTX+CQ (24)	48 months	38	18 (75)	NR	NR	NA
		Single centre

^aStudies that included predictors of adverse events;

^bStudies in patients used s.c. MTX or both oral and s.c. MTX.

ABA: abatacept;

ADA: adalimumab;

CQ: chloroqine;

CZP: certolizumab pegol;

ETN: etanercept;

FA: folic acid;

GC: glucocorticoids;

GOL: golimumab;

IFX: infliximab;

IQR: inter-quartile range;

NA: not applicable;

NR: not reported;

OCR: ocrelizumab;

PBO: placebo;

RCT: randomized clinical trial;

RTX: rituximab;

TCZ: tocilizumab;

TOF: tofacitinib.

The outcome-based risk of bias was moderate in 11 studies, low in 29 studies and unknown in five studies (Supplementary Table S2, available at Rheumatology online). Moderate risk of bias was graded to studies mainly because of issues related to selection or reporting bias.

Incidence and prevalence of AEs

Of the included studies, six reported IRs of total AEs with a mean of 412.4/100 person-years (95% CI: 275.4, 549.3) [25–28, 32, 33], whereas the mean IR of serious AEs from eight studies was 10.0/100 person-years (95% CI: 7.1, 12.9) [25–28, 32, 33, 44, 53]. The IR of MTX discontinuation due to AEs was only reported in two studies, with IRs of 10.6 and 6.5/100 person-years [26, 28].

The pooled prevalence of total ‘any’ AEs in RCTs (80.1%; 95% CI: 73.5, 85.9) was more than two-fold the prevalence in observational studies (23.1%; 95% CI: 12.3, 36.0) (Table 2). In contrast, a greater proportion of patients discontinued MTX due to AEs in observational studies (15.5%; 95% CI: 9.6, 22.3) compared with RCTs (6.7%; 95% CI: 4.7, 8.9) (Fig. 2).

Table 2.

Pooled prevalence estimates of reported adverse events

Type of AEs		Number of studies	Overall pooled prevalence (%, 95% CI)	Number of studies	RCTs pooled prevalence	Number of studies	Observational studies pooled prevalence (%, 95% CI)
					(%, 95% CI)
Total ‘any’ AEs		31	73.8 (65.5, 81.4)	26	80.1 (73.5, 85.9)	5	23.1 (12.3, 36.0)
Serious AEs		28	8.6 (6.6, 10.8)	25	9.5 (7.4, 11.7)	3	2.1 (1.0, 3.4)
Discontinuation due to AEs		25	8.4 (6.1, 11.1)	19	6.7 (4.7, 8.9)	6	15.5 (9.6, 22.3)
Gastrointestinal	Any gastrointestinal event	8	32.7 (18.5, 48.7)	6	39.7 (22.2, 58.7)	2	18.0 (11.1, 26.0)
	Nausea	20	19.0 (14.6, 23.9)	19	19.2 (14.6, 24.1)	1	16.7 (6.7, 35.9)
	Abdominal pain/ discomfort	11	9.3 (5.5, 13.9)	11	9.3 (5.5, 13.9)	0	NR
	Diarrhoea	16	8.0 (5.8, 10.4)	16	8.0 (5.8, 10.4)	0	NR
Liver	Any elevated liver enzymea	9	15.1 (10., 20.9)	6	15.3 (9.3, 22.4)	3	14.7 (5.5, 26.7)
	Elevated ALT > 1× ULN	13	8.9 (4.6, 14.3)	13	8.9 (4.6, 14.3)	0	NR
	Elevated ALT > 2× ULN	2	4.1 (2.4, 6.1)	2	4.1 (2.4, 6.1)	0	NR
	Elevated AST > 1× ULN	8	8.0 (2.8, 15.5)	8	8.0 (2.8, 15.5)	0	NR
Mucocutaneous	Any mucocutaneous event	5	24.7 (9.8, 43.6)	5	24.7 (9.8, 43.6)	0	NR
	Alopecia	10	6.3 (3., 9.4)	9	6.9 (4.3, 10.1)	1	1.3 (0.2, 6.8)
	Oral ulcer/stomatitis	10	7.1 (2.9, 12.6)	8	7.8 (3.1, 14.2)	2	3.7 (0.0, 11.4)
Neurological	Any neurological event	4	24.7 (6.7, 48.9)	4	24.7 (6.7, 48.9)	0	NR
	Headache	20	8.5 (6.1, 11.3)	20	8.5 (6.1, 11.3)	0	NR
	Dizziness	10	6.9 (3.5, 11.2)	9	7.2 (3.6, 11.8)	1	3.4 (0.6, 17.2)
Pulmonary	Any pulmonary event	4	30.7 (11.7, 53.8)	4	30.7 (11.7, 53.8)	0	NR
	Pneumonitis/ILD	11	0.09 (0.0, 0.3)	11	0.09 (0.0, 0.3)	0	NR

^aElevated liver enzymes include any abnormal liver function above the ULN.

AE: adverse event;

ALT: alanine transaminase;

ILD: interstitial lung disease;

NR: not reported;

RCTs: randomized clinical trials;

ULN: upper limit of normal.

Fig. 2.

Forest plots for the pooled prevalence estimates of (A) total adverse events and (B) discontinuation of MTX due to adverse events

ES: estimated proportion/prevalence.

The pooled prevalence of any GI AEs from eight studies was 32.7% (95% CI: 18.5, 48.7) (Fig. 3). Nausea was the most commonly reported event of the GI system with a pooled prevalence of 19.0% (95% CI: 14.6, 24.1), followed by abdominal pain (9.3%; 95% CI: 5.5, 13.9) and diarrhoea (8.0%; 95% CI: 5.8, 10.4). With regard to liver AEs, the pooled prevalence was 15.1% (95% CI: 10.0, 20.9) for any elevated liver enzyme, and 8.9% (95% CI: 4.6, 14.3) for elevated ALT above the upper limit of normal (ULN), reported in 9 and 13 studies, respectively.

Fig. 3.

Forest plots for pooled prevalence estimates of (A) gastrointestinal adverse events (AEs), (B) elevated liver enzymes, (C) mucocutaneous AEs and (D) neurological AEs

The pooled prevalence of any mucocutaneous AEs estimated from five RCTs was 24.7% (95% CI: 9.8, 43.6). Alopecia and oral ulcers were reported in ten studies with a pooled prevalence of 6.3% (95% CI: 3.9, 9.4) and 7.1% (95% CI: 2.9, 12.6). The estimated prevalence of neurological AEs was 24.7% (95% CI: 6.7, 48.9) pooled from four RCTs. The pooled prevalence from four RCTs reporting pulmonary AEs was 30.7% (95% CI: 11.7, 53.8), but the pooled prevalence of pneumonitis/ILD was only 0.09% (95% CI: 0.0, 0.3) estimated from 11 RCTs. Funnel plots and additional forest plots are provided in the supplementary material (Supplementary Figs 1–16, available at Rheumatology online).

The findings on the prevalence of AEs were pooled from studies with MTX monotherapy and in combination with other csDMARDs. A sensitivity analysis for studies in which patients with RA started with MTX monotherapy did not show a meaningful change in the results (Supplementary Table S3, available at Rheumatology online). Note that many of these studies allowed adding or switching MTX to other DMARDs for inadequate response during the study period.

Predictors of AEs

A total of five studies (three RCTs and two observational studies) assessed possible predictors of AEs [37, 46, 54, 57, 63]. For the development of any AE, some of the examined predictors were associated with increased risk of AE occurrence but none of these predictors were statistically significant as shown in Table 3 [37, 46].

Table 3.

Baseline predictors of adverse events in included studies

Adverse events	Possible predictor	Study	Effect size (95% CI)
Development of AEs	Age	Verstappen et al. [46]	aOR=1.00 (0.97, 1.03) per year increase in age
	Female sex		aOR=1.25 (0.49, 3.17) vs male sex
	BMI		aOR =1.08 (0.95, 1.21) per unit increase in BMI
	Swollen joint count		aOR=0.96 (0.89, 1.03) per number of swollen joint
	Tender joint count		aOR=0.99 (0.93, 1.05) per number of tender joint
	VAS pain		aOR=1.01 (0.99, 1.03) per unit mm increase
	VAS general well-being		aOR=1.01 (0.99, 1.03) per unit mm increase
	HAQ score		aOR=1.14 (0.54, 2.38) per unit increase in HAQ
	RF positive		aOR=2.46 (0.77, 7.89) vs RF−
	ESR		aOR=0.99 (0.97, 1.01) per mm/hour increase in ESR
	Creatinine		aOR=1.01 (0.98, 1.04) per µmol/l increase in creatinine
	Creatinine clearance		aOR= 1.00 (0.98, 1.02) per ml/min increase in clearance
	AST		aOR=0.99 (0.94, 1.05) per U/l increase in AST
	ALT		aOR=1.00 (0.97, 1.03) per U/l increase in AST
	MTX-polyglutamate	de Rotte et al. [37]	OR=1.00 (0.99, 1.01) for total MTX-PG concentration after 3 months. In individual MTX-PGs 1–5, there were no significant association at 3, 6 and 9 months with AEs
MTX discontinuation due to AEs	Female gender	Bluett et al. [57]	HR=1.39 (0.82, 2.35) vs male gender
	RF positive		HR =0.37 (0.21, 0.64) vs RF−
	HAQ score		HR=1.39 (0.98, 1.96) per unit increase in HAQ
		Hider et al. [63]	at year 1: OR =1.87 (1.11, 3.15) per unit increase in HAQ
			at year 2: OR=1.42 (0.89, 2.26) per unit increase in HAQ
	BMI	Verstappen et al. [46]	OR =1.21 (1.02, 1.44) per unit increase in BMI
	Creatinine clearance		OR=0.97 (0.94, 1.00) per unit increase in creatinine clearance
Elevated ALT >2× ULN	Smoking	Dirven et al. [54]	OR =1.6 (0.98, 2.7) smokers vs non-smokers
	ACPA positivity		OR =1.8 (1.1, 3.1) vs ACPA−
	Baseline ALT >1 × ULN		OR =3.1 (1.6, 6.2) vs normal ALT at baseline
	Age		aOR = 0.99 (0.98, 1.01) per year increase in age
	Female gender		aOR=1.21 (0.73, 1.99) vs male gender
	BMI		aOR=1.01 (0.95, 1.06) per unit increase in BMI
	Alcohol		aOR=1.29 (0.81, 2.03) drinking alcohol compared to non-drinkers
	Symptoms duration		aOR=1.00 (1.00, 1.00) per unit increase in duration
	DAS		aOR=0.89 (0.68, 1.17) per unit increase in DAS score
	HAQ		aOR=0.74 (0.52, 1.04) per unit increase in HAQ
	RF positivity		b OR =1.68 (1.00, 2.81) vs RF−
	ESR		aOR=1.00 (0.99, 1.01) per unit increase in ESR
	Liver disease history		aOR=0.57 (0.07, 4.61) vs no history of liver disease
Elevated ALT >1× ULN	Age	Verstappen et al. [46]	OR=1.02 (0.99, 1.04) in I-group, OR=1.01 (0.98, 1.03) in C-group
	Female sex		OR=1.16 (0.58, 2.35) in I-group, OR=0.60 (0.28, 1.29) in C-group
	BMI		OR=1.08 (0.97, 1.21) in I-group, bOR =1.12 (1.00, 1.25) in C-group
	Swollen joint count		OR=1.04 (0.99, 1.10) in I-group, OR=1.03 (0.96, 1.09) in C-group
	Tender joint count		OR=1.02 (0.97, 1.07) in I-group, OR=1.00 (0.95, 1.06) in C-group
	VAS pain		OR=1.00 (0.98, 1.01) in I-group, OR=1.00 (0.98, 1.01) in C-group
	VAS general well-being		OR=1.00 (0.99, 1.02) in I-group, OR=1.01 (0.99, 1.03) in C-group
	HAQ		OR=0.88 (0.51, 1.15) in I-group, OR=1.21 (0.67, 2.20) in C-group
	RF positivity		OR=1.20 (0.58, 2.49) in I-group, OR=1.04 (0.46, 2.34) in C-group
	ESR		OR=1.01 (1.00, 1.02) in I-group, OR=1.00 (0.99, 1.02) in C-group
	Creatinine		OR=0.99 (0.97, 1.02) in I-group, bOR =1.03 (1.00, 1.07) in C-group
	Baseline ALT		OR =1.03 (1.00, 1.06) in I-group, OR =1.08 (1.04, 1.12) in C-group

^aUnivariable regression analysis.

^bEffect size in bold is statistically significant.

Data on predictors of adverse events were retrieved directly from these studies without any amendments or attempt to combine the results.

ALT: alanine transferase;

AST: aspartate transferase;

AUC: area under concentration curve;

C-group: conventional strategy arm; HR: hazard ratio;

I-group: intensive strategy arm; MTX-glu3: MTX-polyglutamate 3;

MTX-PG: MTX polyglutamate;

OR: odds ratio;

ULN: upper limit of normal;

VAS: visual analogue scale.

Three studies examined predictors of MTX discontinuation due to AEs. In the Norfolk Arthritis Register (NOAR) observational study, Bluett et al. reported 67 discontinuations over 1608 person-years of follow-up. RF positivity was associated with lower risk of MTX discontinuation due to AEs [RF+: (HR)= 0.37 vs RF−, 95% CI: 0.21, 0.64)], whereas HAQ score and female gender were associated with increased risk of MTX discontinuation due to AEs (HAQ: HR= 1.39 per unit increase in HAQ, 95% CI: 0.98, 1.96; female gender: HR = 1.39 vs male gender, 95% CI: 0.82, 2.35) [57]. In another NOAR study (Hider et al.), 34 (11%) patients stopped MTX due to AEs in the first year, and 46 (15%) after 2 years of treatment. They also reported that a high baseline HAQ score (OR= 1.87 per unit increase in HAQ, 95% CI: 1.11, 3.15) was associated with increased risk of MTX discontinuation due to AEs after the first year of follow-up; however, the effect was not statistically significant at year two (OR= 1.42, 95% CI 0.89, 2.26) [63]. A third study by Verstappen et al. also examined predictors of MTX withdrawal due to AEs (24 discontinuations over 2 years). They reported that an increased BMI and decreased creatinine clearance were associated with higher risk of MTX discontinuation due to AEs (BMI: OR= 1.21 per unit increase in BMI, 95% CI: 1.02, 1.44; creatinine clearance: OR= 0.97 per unit increase in clearance, 95% CI: 0.94, 1.00) [46].

With regard to elevated liver enzymes, baseline ACPA positivity and high baseline ALT were associated with an increased risk of elevated ALT above two times ULN in the paper by Dirven et al. (ACPA+: OR= 1.8 vs ACPA−, 95% CI: 1.1, 3.1; baseline ALT >ULN: OR= 3.1 vs normal ALT, 95% CI: 1.6, 6.2) [54]. Higher BMI, creatinine and ALT at baseline were associated with increased risk of elevated ALT above ULN in the conventional treatment arm of the CAMERA-I study (BMI: OR= 1.12 per unit increase in BMI, 95% CI: 1.00, 1.25; creatinine: OR= 1.03 per unit increase in baseline creatinine, 95% CI: 1.00, 1.07; ALT: OR= 1.08 per unit increase in baseline ALT, 95% CI 1.04, 1.12) [46].

Discussion

This comprehensive systematic review summarizes the incidence rates and provides precise pooled prevalence estimates of AEs in patients with RA starting MTX for the first time. This review affirms previous studies that GI events are the most commonly reported MTX-related AEs, but in addition has shown that pulmonary, mucocutaneous and neurological events are also commonly reported. Furthermore, this review identified various baseline predictors of certain AEs, such as BMI, HAQ score, RF and ACPA.

The main objective of this review was to summarize the knowledge on the AEs associated with MTX, the most commonly used csDMARD for RA. Patients often worry about AEs related to their treatment, including MTX, and this could potentially affect their adherence to treatment [69, 70]. Patients report that they do not receive enough information regarding AEs from their rheumatologists [71], and the anxiety surrounding starting treatment might be lessened if patients had more information about the AEs associated with MTX. However, prior to this review, an extensive literature search would have been required by each rheumatologist to acquire the information necessary to inform patients about the potential AEs associated with MTX. This review offers valuable insight into the occurrence of MTX-related AEs, which can facilitate discussion between physicians and patients regarding any concerns prior to commencing MTX.

Although our review included studies with MTX-naïve patients in both RCTs and observational studies, prior reviews mainly including RCTs have reported comparable results. The pooled prevalence of elevated liver enzymes in this review was 15.1% compared with 15% in a previous review of RCTs of MTX treatment, and 20.2% in a review including non-RCTS with ≥2 years follow-up [5, 9]. Although liver enzyme abnormalities are not the most common AEs, it is usually a concern that requires frequent monitoring, and the treating physician often halts or discontinues MTX treatment if liver enzymes are elevated.

The pooled prevalence of oral ulcers and stomatitis was 7.1% in our review compared with 9% and 5.7%–8% in two other reviews of MTX monotherapy in RCTs [9, 72]. Lastly, the pooled prevalence of alopecia was 6.3% compared with 6% and 1%–4.9% in the aforementioned reviews. Although alopecia is one of the most worrying AE for people starting MTX, the low prevalence of this AE could be of some reassurance to patients.

Another important finding is the low prevalence of pneumonitis or ILD in the included studies (pooled estimate =0.09%). The prevalence was 0.43% in a previous review and an even higher incidence of pneumonitis was documented in an older literature review of retrospective studies published between 1970 and 2000 [5, 73]. This earlier review of observed high occurrence of MTX-pneumonitis could be due to refractory RA disease with pre-existing lung involvement [74], in addition, study design precludes drawing any conclusions on causation. Yet, pneumonitis remains a serious and threatening AE that requires monitoring and should be differentiated from other common pulmonary AEs such as mild chest pain and cough.

Although MTX treatment arms of RCTs were treated as observational studies in this review, there was significant variation in the pooled prevalence estimates of the AEs between RCTs and observational studies. For example, the pooled prevalence of total AEs was 80.1% in RCTs, while in observational studies the estimate was 23.1%. This could be explained by closer monitoring of AEs in RCTs and by higher adherence to therapy following stringent protocols. This is in contrast to what is observed for discontinuation of MTX due to AEs in which RCTs reported low withdrawal rates (pooled estimate: 6.7%), compared with observational studies (pooled estimate: 15.5%). The difference could also be attributed to stricter inclusion criteria in RCTs, or to the length of follow-up in the included studies, with observational studies having a relatively longer follow-up time (19 vs 15 months for RCTs).

Few studies examined potential predictors of the AEs in this review. Bluett et al. showed that RF positive patients had lower odds of discontinuing MTX due to AEs [57]. A plausible explanation for the association between RF positivity and withdrawal due to AEs might be explained by the variation in treatment response and disease activity. The same study reported that RF positivity and DAS-28 were associated with earlier discontinuation of MTX due to inefficacy, and that more patients stopped MTX due to inefficacy (33%) compared with AEs (16%). Thus, fewer patients with RF positivity and high DAS were included for the analysis of MTX withdrawal due to AEs. In addition, patients who are RF positive or with high HAQ score at baseline are less likely to have a good treatment response [75], thus may be less tolerant to AEs and more likely to stop MTX [76].

Dirven et al. [54] reported that ACPA and RF positive patients had higher odds of liver AEs. They also reported that high ALT at baseline was associated with increased risk of elevated ALT, conforming to the results in the CAMERA-I study. Higher ALT at baseline is an expected and a strong predictor for the enzyme elevation during subsequent follow-ups, which suggests pre-existing liver pathology or lower enzyme threshold before starting MTX. The two studies have examined multiple predictors of liver AEs, and showed contrasting results for RF status and BMI. Finding predictors of elevated liver enzymes could be of great benefit in clinical practice; patients at increased risk can be monitored more frequently than the recommended intervals by the 2015 ACR guidelines [4].

Overall, a limited number of predictors were identified. Although our risk of bias rating was low to moderate in most of the studies that examined predictors, they still had potential for bias. It is important to note the possible bias due to inadequate adjustment for confounders or selective reporting of results. Identification of these predictors of MTX AEs may help to optimize drug therapy in patients likely to experience these AEs early in the course of treatment. This also may lead to better treatment adherence, and subsequently improved effectiveness [7]. Furthermore, knowledge of risk factors of MTX AEs may guide physicians’ treatment decisions, such as dosing of MTX, frequency of monitoring or stopping MTX therapy.

A key strength of this review was not excluding studies with MTX in combination with other csDMARDs or GCs. This comes close to the variation seen in clinical practice as MTX could be started alone, or in combination with other csDMARDs and GCs [77]. However, the inclusion of combination therapy could lead to a higher level of heterogeneity among studies. These treatments also share many of the AEs of MTX safety profile [78], which could result in overestimation of the prevalence rates of some AEs associated with MTX. Another strength is the inclusion of studies of MTX with both short and long-term follow-up, which ranged between 4 months and 5 years. Moreover, the current review included manuscripts published after 2005, to be representative of current clinical practice since the implementation of the treat-to-target strategy in the early 2000s.

One of the limitations of our review was the inability to perform a meta-analysis of IRs. Unfortunately, the lack of detailed data reported on attrition rates and number of events in some studies prevented pooling the IR of different AEs. Moreover, a few RCTs allowed switching from MTX to comparators, such as bDMARDs. This could affect the precision of the pooled prevalence of AEs. The inclusion of studies in which MTX was combined with other csDMARDs could also have affected the estimates of the outcome. An inevitable weakness of this review is the exclusion of elevated liver enzymes results from a few studies that do not fit with the presented categories. The variation in reporting and assessment is a common methodological challenge faced when reviewing certain outcomes such as AEs. Another issue that was not addressed in our review was accounting for duration of follow-up in the meta-analysis of AEs. The prevalence of AEs might vary in studies with short observation and exposure time compared with studies with longer observation; however, direct weighting of observation time is limited by the uncertainty regarding the relationship between time since MTX onset and risk of AEs.

In conclusion, the review provides a thorough and updated summary on AEs associated with MTX treatment in patients with RA. Furthermore, identified predictors of MTX-related AEs may help to optimize drug therapy in patients likely to experience these AEs early in the course of treatment, but the results of these predictors should be interpreted with caution and further research is required to investigate the predictors of AEs.