Search strategies were generated by an expert documentalist (MPR) with MeSH terms and free text in the following bibliographic databases: Medline (via OVID), Embase, and Cochrane Library, all from their inception until June 2019. Articles about studies in human subjects and in English or Spanish were used as limits. Subsequently, a secondary search of the bibliography of the articles finally included was performed.

Annex 1 shows the Medline search strategy. Based on this, Embase and Cochrane were generated. All citations resulting from the searches were entered into EndNote® libraries to facilitate their handling.

Two reviewers, rheumatologists with extensive experience in systematic reviews (TO, EL), independently screened the articles resulting from the search strategy in the different bibliographic databases. In case of discrepancies, a third researcher (LC) was included. The search results were first screened by title and abstract, or by full article, for those lacking an abstract, in sessions lasting a maximum of 60min. Following this process, the full text of the selected articles was retrieved and analysed in detail. The articles retrieved were classified according to the topics of interest they covered (efficacy, safety, adherence, satisfaction, etc.) and the information was collected directly in tables prepared for this purpose. The Oxford Centre for Evidence-Based Medicine guidelines were used to homogeneously evaluate the studies, assigning a level of evidence to each study according to the question it answered.6

Despite the fact that most of the studies included examined the efficacy of parenteral MTX, most of them did not have a comparator group. Moreover, as older studies were included, currently used efficacy endpoints, such as remission or EULAR response, could only be found in the most recent studies. The domains assessed include data on activity (including remission), structural damage, pain, stiffness, function, fatigue, hand strength, physician and patient global ratings, quality of life, acute phase reactants, corticosteroid consumption and, as a possible efficacy variable, saving/ delayed prescription of biologic therapies, MTX survival, and treatment changes.

Almost two thirds of the studies included in this review looked at the safety and tolerability of parenteral MTX, some comparing it with orally administered MTX, others with different doses of parenteral MTX, but most did not have a comparator group (appendix 2). As with efficacy, the comparison between oral and parenteral formulation has not been done with parallel designs, but sequentially. Regardless of the pharmacokinetics of MTX, this cannot entirely avoid the existence of residual effects that bias the results obtained with the injected route. It should be noted that the type of adverse event (AE) was not always well defined, nor are the heterogenous ways of grading their severity, and there is tremendous variability in how they are defined and recorded.

In general, in both RCTs,15,48 and observational studies (regardless of how treatments were compared), parenteral MTX does not increase the rate, type, or severity of AEs described for oral MTX. Nor have changes been observed with different doses of parenteral MTX or when comparing different concentrations of the sc formulation.

It has been suggested that injected administration of MTX could possibly have a favourable effect on the rate of gastrointestinal AEs relative to the oral formulation. One of the included RCTs notes this,48 but not the RCT by Braun et al.15 In fact, the authors commented in the Discussion on this finding as unexpected. There is one observational study (of low quality) that specifically compared the intensity of gastrointestinal AEs between doses of 7.5 and 15mg/week in the two formulations. In general (although it depends on the type of AE), the intensity was higher with the oral formulation.

Similarly, it has been suggested that the injected route may improve the findings on transaminase levels reported with the oral route. The RCT by Braun et al.15 found elevated transaminases in 4.3% of patients on MTX per os and in 1.6% on the parenteral formulation. Comparative studies are lacking to definitively confirm this point.

Four observational studies of moderate-good quality and one CE of low level of evidence stand out, in which adherence to parenteral treatment was directly assessed (albeit using different definitions). However, they lack a comparator group, and in the first one described, the inclusion criterion was that the subject was a “potential adherent” (Appendix 3). It should be noted that these are all studies with few patients, so the validity of the results is very limited.

The first of these studies (apparently short-term, although of unclear duration) examined self-reported adherence to parenteral drugs.8 Forty patients with RA and psoriatic arthritis (PsA) were included, 20 of whom initiated im MTX. These participants were provided with education and training for self-injection, and reported 92.5% adherence, defined as ≥80% compliance with the prescribed schedule.

Another study54 found that in 12 patients with rheumatic disease who switched from MTX per os to im at 6months, there was no difference in adherence to MTX im, given the discontinuation of oral MTX due to inefficacy or AE. Stamp et al.,80 after analysing 30 RA patients who switched from MTX per os to sc MTX, found only one non-adherent patient at 6 months.

The study by Flipo et al.31 (low quality), which measured compliance using the Morisky index, revealed moderate adherence. In this work, physicians considered their patients’ adherence to be higher than the patient-reported compliance.

A few studies evaluated the persistence of parenteral MTX,10,54 one of which was only found in abstract form at congress.42 Drug persistence has been seen to decrease over time and that injected MTX persists longer than MTX per os.

It is important to take into account the favourable role that education and training for parenteral administration, the use of prefilled syringes, and the smaller volume of syringes may have on adherence.8,62,81

Studies assessing satisfaction are highly heterogeneous in both form and content and of medium-low quality (Annex 4). In general, all the studies report good or very good satisfaction, above 70%, although in each of them, the analysis of satisfaction was approached from a different perspective; for example, satisfaction with the change of drug delivery route,59 the type of device used for parenteral administration,25,81 or even pain at the site of injection,32 among others. Given the characteristics of the design and quality [of the studies], it is difficult to draw definitive conclusions in this category.

The results found with respect to pharmacokinetics are shown in Annex 5. The evidence on differences in bioavailability between oral and parenteral formulations is based on one 8-week, good quality, crossover RCT, as well as multiple small-sample, short duration, moderate quality, observational studies.41,46,49,51,66 The pharmacokinetics of both sc, im, and injected MTX at different doses have been evaluated.

With MTX per os, it has been observed that as the dose increases, its bioavailability decreases, reaching a plateau at 15mg/week, a phenomenon that is not observed with parenteral MTX. In other words, from 15mg/week onwards, the bioavailability of parenteral MTX would be higher than that of the oral route, and is similar at lower doses. Also based on these findings, one of the studies46 suggested that to achieve greater efficacy in individuals with MTX25mg/week, they should switch to the parenteral route.

Schiff and Sadowski78 suggest a formula based on pharmacokinetics for dose escalation from oral to parenteral (y=0.6101x+2.9274), which could have clinical applicability.

The systematic review included a quality study on the cost-effectiveness of parenteral MTX in RA patients naïve to MTX, adjusted to the peculiarities (including costs) of our national health system.22

Other studies have performed other drug-economic analyses, specifically cost minimisation.29,30 Using UK costs, the use of sc MTX in patients refractory to oral MTX can save £7,197 per patient in the first year and £9.3million per year in new patients. Other studies56 have estimated that, for every 1,000 RA patients, 40 are being treated with sc MTX, and, taking into account the fact that 36 of these patients have a response equivalent to that of an anti-TNFα, the savings (from using sc MTX instead of biologic therapy) are estimated to be £306,000, or £300 per patient-year. All pharmacokinetic studies found can be found in Annex 6. The difference in health systems and the difference in the variables used in each study, as well as their quality, make it difficult to draw adequate conclusions.