Methotrexate (MTX) is a cost-effective immunosuppressant. Although anecdotal reports of its use in myasthenia gravis (MG) date back several decades, there are no published randomized or quasi-randomized trials [1]. Azathioprine (AZA) is considered by most to be the first-line steroid-sparing immunosuppressant in MG [2] and data show that it has significant steroid-sparing activity compared with prednisone alone after 18 months of treatment [3]. However, AZA is considered an expensive therapy in the developing world.

The incidence of MG in sub-Saharan Africa is expected to be similar to that elsewhere in the world were patients able to access appropriate health care [4]. Despite the need for cheaper MG therapies in developing countries, developed countries are investigating more costly alternatives [5]. As MTX is substantially cheaper than AZA it has the potential for wider availability in Africa. Efficacy data for MTX in MG is therefore of critical importance.

MTX is used effectively as an immunosuppressant in target-organ autoimmune disease such as Crohn's disease and psoriasis and it is considered the disease modifying agent of choice in rheumatoid arthritis (RA) [6]. Mertens et al. [7] alluded to the efficacy of MTX in generalized MG decades ago. However, many recent reviews on MG therapies either do not include MTX as an option [5, 8, 9] or only regard this agent as an alternative if other drugs have failed [10].

MTX is a structural analog of folic acid and exerts an anti-proliferative effect by metabolic interference with DNA synthesis. The safety and efficacy of MTX in RA is well established [6] and studies suggest that MTX may show earlier efficacy compared with AZA [11]. In RA, substantial data show that a weekly MTX dose of 17-20 mg produces an optimal efficacy/toxicity ratio [6, 12, 13].

The main objective of this study was to determine the efficacy of MTX as a steroid-sparing agent in subjects who had recently been diagnosed with generalized MG, and compare this with the efficacy of AZA. In MG, AZA has been shown to be an effective steroid-sparer compared to placebo after a treatment period of 18 months [3]. To achieve this we performed a 24-month parallel study with a design similar to that of Palace et al. [3].

A randomized, single-blind study of AZA and prednisone as compared to MTX and prednisone was planned to recruit 60 patients who had recently been diagnosed with generalized MG. This sample size calculation was based on the proportion of patients with MG who were steroid-independent after receiving AZA vs. placebo for a period of 36 months (AZA-Placebo Δ = 40%) [3] and AZA vs. MTX at six months in inflammatory bowel disease (AZA-MTX Δ = 7%) [14], as well as taking into consideration anticipated feasibility and recommendations by Barohn et al. [15].

To meet the objective of comparing the steroid-sparing effectiveness between AZA and MTX, the primary outcome was the average prednisone dose required in each group to achieve and maintain minimal manifestation status (MMS) over 24 months. The clinical MG measurements were assessed by blinded assessors and used to adjust prednisone dosing per protocol, aiming for MMS as defined by the MGFA (i.e. no symptoms of limitation of functioning during everyday living even if some fatiguable weakness was noted on examination) [16]. The MG activities of daily living (MG-ADL) questionnaire was used to objectively assess functional status and subjects in MMS were required to have no MG symptoms (MG-ADL = 0) [17].

Secondary outcomes included the improvement in quantitative MG score compared to baseline, frequency of adverse events and treatment failures in each group. Treatment failures were defined as: withdrawal from the study whether due to intolerance of study medication or uncontrolled MG disease requiring alternative medications; hospitalization for MG relapse or intravenous antibiotic therapy; and death (all-cause mortality).

The study was conducted between 2005 and 2010 and due to slow recruitment it was halted after five years. Eight eligible subjects with AChR-Ab-positive generalized MG were either recruited to a thymectomy trial (n = 5) [19] or declined study participation (n = 3) as they lived too far from the study site. Thirty-one subjects entered the study. Twenty-four subjects were randomized to either MTX (n = 11) or AZA (n = 13). The residual 7 subjects were willing to follow study protocol but were not randomized either because they could not afford AZA if potentially randomized to that group (n = 5) and therefore included into the MTX-group, or because they opted for standard of care therapy with AZA (n = 2)(Figure 1).

Baseline values between the AZA- and MTX-group were similar including the duration of MG symptoms prior to study entry or MG diagnosis (results not shown, p = 0.51) and the proportion on prednisone, as well as the cumulative prednisone exposure prior to baseline (Table 1). The baseline bodyweight, MGFA severity grade, quantitative MG scores, and proportions requiring pre-study IVIg or P/E were similar (Table 1). Seven patients (AZA = 3; MTX = 4) required hospitalization at the time of diagnosis.

Prednisone doses were not significantly different between the AZA and MTX groups throughout the study except at months 10 (p = 0.047) and month 12 (p = 0.039) when the average daily doses by month were significantly lower (2-sided) in the MTX-group; mean difference (Δ) between MTX and AZA (MTX-AZA) at month 10 = 10.21 mg; 95% confidence interval (CI) 0.12; 20.21; mean Δ month 12 = 10.89 mg; 95% CI 0.58; 21.21. Months 11, 13 and 14 showed a trend (p = 0.078) towards lower doses of prednisone in the MTX-group (Figure 2). The average daily prednisone dose per kilogram bodyweight at month 12 was also lower in the MTX-group being on average 0.15 mg/kg prednisone (SD ± 0.08) daily as compared with that in the AZA-group of 0.31 mg/kg (SD ± 0.22)(mean Δ 0.16; 95% CI 0.29; 0.03; p = 0.019).

Intolerable diarrhoea (AZA n = 1; MTX n = 1), hearing loss and acne (AZA n = 1) and loss of appetite (MTX n = 1) resulted in study withdrawals during the first six months (Table 2). MG relapses requiring hospitalization (with or without additional IVIg or P/E) occurred in both groups within six months and in two AZA-patients between 7-12 months. The time to first treatment failure was similar in the two groups (p = 0.99)(data not shown).

One patient (70-year old male) suffered a fatal myocardial infarction, an event judged to be unrelated to either MG or study medications (Table 2). This patient had several pre-morbid risk factors for ischemic heart disease and had been randomized to AZA. He died at 12.5 months into the study during which time his MG had improved consistently.

At the end of the study, incorporating dose adjustments due to adverse events (Table 4), the final average AZA dose was 2.23 mg/kg daily (SD ± 0.74; n = 11) and the final MTX dose was 16.73 mg weekly (SD ± 2.14; n = 14). The final mean pyridostigmine dose did not differ between the AZA-group (176 mg; SD ± 136) and MTX-group (150 mg; SD ± 103), p = 0.56. By the end of the study, three AZA- and four MTX-subjects had completely weaned pyridostigmine and four in each group had completely weaned prednisone.

We present the results of the first controlled trial using methotrexate in myasthenia gravis. In this single-blinded study involving recently diagnosed generalized MG subjects, we show that MTX has an onset of steroid-sparing efficacy from 10 to 12 months in subjects using an average dose of 17 mg weekly which is earlier than that of AZA. Two previous trials studying AZA, the comparator, found the efficacy as steroid-sparing immunosuppressant to become apparent between 15 to 18 months [1, 3].

MG has a fluctuating disease course and it has been suggested that a more robust measure of treatment efficacy would be sustained minimal manifestation status (MMS) as opposed to transient MMS that may be followed by deterioration [1]. Both treatment groups in this study achieved similar proportions of participants with sustained MMS after similar treatment periods (Table 2). The responsiveness to treatment was also similar in both groups as measured by the improvement in quantitative MG score compared to baseline [15]. These outcomes are to be expected with adherence to the treatment protocol in this study as MMS was achieved by adjusting doses of prednisone in addition to the steroid-sparing effects of AZA or MTX. Steroid-sparer efficacy is regarded as being inversely related to the dose of prednisone required to maintain MMS in each group. From Table 2 it is evident that most improvement occurred in the first 6 months and was likely attributable to a steroid effect.

Longitudinal observation of MG subjects has indicated that approximately 10-15% of patients may go into spontaneous remission, usually after several years [23]. In our study the proportion of subjects in MMS (i.e. those without MG symptoms) was at least 25% in both groups at 12 months, escalating to more than 50% in both groups by 24 months (Table 2). Also, the majority of subjects had either moderate or severe disease at study entry with similar disease severity and duration in both treatment groups. It is therefore unlikely that spontaneous remissions contributed significantly to the outcomes investigated.

Study withdrawal as a result of drug intolerance was as frequent amongst those taking MTX as AZA (13%), and comparable to that observed in a previous study (11%) when these two agents were compared in patients with Crohn's disease, although only for six months [14]. A 48-week comparison between the two agents in RA found more frequent AZA withdrawals [11]. In our study, MTX was well tolerated with few patients requiring dose adjustments to manage raised transaminase levels, which did not necessarily occur in individuals with an increased body mass index as previously reported [24]. The low incidence of liver toxicity may have been due to folate supplementation [25, 26]. In this study 25 mg of folate was supplemented every week although previous reports show (a) no differences between 5 mg folate weekly and higher doses in reducing MTX-toxicity and (b) that higher doses of folate may result in an individual requiring higher MTX doses to achieve efficacy [25–27]. Although we advised against consuming more than one alcoholic drink per week, others have allowed moderate alcohol intake without this apparently influencing transaminase levels in their subjects [25].

MTX has anti-inflammatory, anti-proliferative and immune-modulatory activities some of which are due to the cellular induction of adenosine and consequent suppression of NF-_KB activation, a key transcription factor (reviewed in [28]). These effects may explain the early (within 3-6 weeks) radiographic improvements observed in MTX-treated RA subjects [12]. MTX has also been shown to have immunosuppressive effects including apoptosis and clonal deletion of activated T cells, an inhibitory effect on proliferating cells [27] including B cells, a reduction in pro-inflammatory cytokine production by T cells and macrophages [29] and CD95-dependant apoptosis of activated memory T cells [27]. In RA, accumulated data suggest that strategies to use MTX earlier with faster dose escalations and intensive regimens of 17-20 mg/week showed control in disease activity within three to four months (reviewed in [6, 30]). The optimal initiation dosing strategy recommends starting at 15 mg weekly unless the patient is elderly or bodyweight is below 50 kg in which case some would start at 7.5 mg weekly [6, 21].

A major limitation of our study is its small size with only 24 of 31 participants undergoing randomization. As we recruited substantially fewer subjects than we anticipated, the possibility exists that we have falsely accepted MTX treatment as equivalent to AZA when it is inferior (type 1 error). Nevertheless, our results do suggest that MTX shows an earlier onset of steroid-sparing efficacy than AZA. This study represents the "real-life" situation in that a spectrum of patients with generalized MG were recruited including AChR-Ab-negative and thymoma-associated variants: two less prevalent MG subgroups often associated with severe disease requiring immunosuppression [7, 9, 10, 31, 32]. Another limitation of our study is that we did not systematically examine MuSK antibody status of all our AChR-Ab-negative subjects. MuSK-positive MG can be difficult to manage with immunesuppressants [10]. We did identify one MuSK-positive subject who was randomized to MTX, required and responded to plasma exchange at study entry, and maintained a "responder" status throughout the 2-year of study.

Due to a concurrent study assessing the role of thymectomy in MG [19] in which AChR-Ab-positive MG subjects amenable to possible thymectomy were recruited to that study, all the patients recruited into this study had refused thymectomy and consequently this was not performed. In this manner bias might have been introduced. However, < 20% of this study population had mild generalized (MGFA Grade 2) disease. Another source of potential bias was the fact that socioeconomic status prevented five subjects from being randomly assigned to either treatment group. Other limitations, due to limited funding, included patients being unblinded to the study medications and prednisone dosing being captured by intended script dosing and verified by patient recall at the following visit. The latter may have resulted in minor inaccuracies in daily prednisone dosing.

In a study population that is representative of that seen in a MG clinic, we present class III evidence [33] that an average methotrexate dose of 17 mg weekly is an effective steroid-sparer in generalized MG evident from 10-months of treatment initiation. Our data suggests that in addition to an earlier onset of action, methotrexate has similar efficacy and tolerability to azathioprine in recently diagnosed generalized MG and may be the drug of choice in financially constrained health systems.