Methods

Data source and study population

Data from a large cohort of patients with RA or PsA receiving care in community and academic settings across the USA enrolled in the Consortium of Rheumatology Researchers of North America (CORRONA) were analysed. Details of this cohort have been previously published.9 Subjects eligible for analysis were required to have a rheumatologist-confirmed diagnosis of RA or PsA and to have initiated treatment with a non-biological or anti-tumour necrosis factor (TNF) DMARD; normal ALT/AST levels were also required at baseline. Individuals with concomitant diagnoses of RA and PsA were excluded. The observation period extended through 1 April 2007.

Drug exposures

The four categories of DMARD exposure were: (1) MTX, (2) LEF, (3) MTX+LEF, or (4) other non-biological DMARD combinations excluding MTX and/or LEF which served as the referent exposure category. The drug exposure categories were mutually exclusive; patients were allowed to change DMARDs over time. Observation time began after initiation of any new DMARD, and patients could contribute exposure time to multiple categories. The use of other non-biological DMARDs (eg, hydroxychloroquine, sulfasalazine, gold compounds, azathioprine or minocycline) and TNF-α inhibitors was permitted in conjunction with each of the four DMARD exposure groups. ALT/AST determinations while patients received biological DMARDs other than TNF inhibitors (eg, anakinra, abatacept, rituximab) were excluded from analysis, as were tests when patients were not receiving any DMARDs.

Exposure to one of the four DMARD categories at the time of ALT/AST determinations was defined as the independent variable of interest; those initiating treatment on the day of liver enzyme tests were included in the analysis of the previous DMARD regimen. Given the prolonged metabolic effects of MTX and the long circulating time for LEF, patients were included in these categories until they changed exposure categories; a separate sensitivity analysis censored observations at the time of treatment discontinuation. Administered MTX and LEF doses in CORRONA are recorded by patients and treating physicians at each clinical encounter. Although 10 and 20 mg daily doses of LEF were initially evaluated, in view of a relatively small number of individuals receiving 10 mg, the results for both doses were combined.

Outcomes of interest

The primary outcome was any elevation of ALT/AST above upper limits of normal (ULN) using normal cut-off values from a reference laboratory (Covance): ALT (women 34 (age 18–69 years) and 32 (age ⩾70 years); men 43 and 35, respectively); AST, irrespective of age (women 34 and men 36). A secondary analysis examined AST and/or ALT elevations >2× ULN. In both primary and secondary analyses the observation time was censored after the first elevated transaminase level. After identifying patients with new liver function test abnormalities, we then determined the proportion of these with persistent abnormalities at the next physician visit.

A separate set of sensitivity analyses evaluated patients with at least two and at least three follow-up visits with laboratory tests, and additional analyses evaluated the incidence rates of abnormal liver function tests using person-time (rather than people) as the denominator to account for varying lengths of follow-up.

Statistical analysis

Descriptive statistics were used for all comparisons. Following a determination of per-person incidence of ALT/AST elevations by DMARD exposure category, additional risk factors were included in multivariable analyses. Logistic regression examined relationships between use and dose of MTX, LEF or MTX+LEF and the risk of elevated ALT/AST levels among patients with RA. Those receiving non-biological DMARDs other than MTX and/or LEF served as the reference group. In bivariate analyses, additional factors associated with elevated ALT/AST values with p values <0.05 were included in multivariable models; covariates of high clinical interest were forced in. A parallel model was created for patients with PsA and included as many of the same covariates as possible to facilitate a qualitative comparison with patients with RA. A few covariates, such as LEF use, had to be excluded in models that examined patients with PsA due to a very low prevalence of use. Generalised estimating equations were used to account for within-person clustering of study visits. All analyses were performed using STATA Version 10 (StataCorp, College Station, Texas, USA).

Results

Of 10 863 CORRONA participants with RA (n = 9755) or PsA (n = 1108) with at least one follow-up visit, 5858 follow-up visits among 2104 unique patients (n = 1953 with RA, n = 151 with PsA) were identified after initiation of a new DMARD and when ALT/AST determinations were performed. The mean time between study visits at which liver function tests were performed was approximately 5 months. The characteristics of this study population are presented in table 1. As shown in the third row, between 2% and 12% of the treatment periods were contributed by patients with PsA; the remainder were from patients with RA. Patients receiving MTX+LEF or MTX monotherapy were less likely to have a history of liver disorder or to use daily alcohol, and were more likely to receive folate supplements.

Table 2 describes the relationship between per-person incidence of ALT/AST elevations and DMARD exposure. The proportion of patients with RA with transaminase elevations was greatest in those receiving MTX+LEF: 31% for any AST or ALT >1× ULN and 5% for >2× ULN. In the subgroup of patients with PsA using MTX, this proportion was 35% and was numerically higher than in patients with PsA receiving neither MTX nor LEF (28%, p = 0.18). Among the patients with PsA, there was not enough exposure to LEF (alone or in combination with MTX) to assess this exposure independently. Sensitivity analyses that evaluated patients that had at least two or three follow-up visits with laboratory tests, and those analyses that compared incidence rates by drug exposure yielded results similar to those in table 2 (data not shown). The proportion of persons with AST or ALT >3× ULN was exceedingly small (<1%).

After identifying persons with new onset liver function test abnormalities, we then evaluated their persistence. Among those with new liver function test abnormalities initially >1× ULN, 37% had abnormalities at the next physician visit. For those with new abnormalities >2× ULN, 48% had abnormalities (>1× ULN) at the next visit.

Factors potentially associated with ALT/AST elevations among patients with RA are shown in table 3. After multivariable adjustment, there was a synergistic effect of the MTX+LEF combination, especially with higher doses of MTX. Among several other factors, a past history of liver disorder and daily alcohol use were associated with ALT/AST elevations. Of interest, there was no significant protective effect of folate supplementation (adjusted odds ratio (OR) 0.97, p = 0.88). Although not significant, trends from the parallel model for patients with PsA presented in table 4 suggested that these patients had a greater risk for ALT/AST elevations associated with the use of MTX and with higher BMI. When the data represented in tables 3 and 4 were pooled together, patients with PsA had a 2.76-fold (95% CI 1.84 to 4.15) greater likelihood of ALT/AST abnormalities after multivariable adjustment. Results from the sensitivity analysis that censored participants at the exact time of MTX, LEF or other DMARD discontinuation resulted in similar but less precise risk estimates.

Discussion

In a large cohort of patients with RA and PsA receiving MTX, LEF, both in combination or other non-biological or TNF inhibitor DMARDs, we observed that increased ALT/AST laboratory tests were common. In patients with RA the incidence of these abnormalities ranged from 14% to 22% for patients on monotherapy alone. With combination MTX+LEF, this proportion was greater (31%) with a significant dose-response relationship with MTX. Compared with RA, patients with PsA appeared to be at increased risk, as were individuals with a history of pre-existing liver disorder and daily alcohol use.

The incidence of ALT/AST elevations we observed with the use of MTX or LEF is comparable to those reported in randomised controlled trials where incidence rates between 5.4% and 16.3% have been reported in the first year after their initiation.10 11 12 In addition, we found that combination therapy with MTX+LEF increased the incidence of ALT/AST elevations approximately 2–5-fold (depending on MTX dose) compared with those receiving either monotherapy. This is consistent with data from a clinical trial of combination MTX+LEF therapy in which a >4-fold increase in liver transaminase elevation was observed with MTX and LEF compared with MTX alone.4 However, in that trial the MTX dose was constrained by the protocol (to either 15 mg or 20 mg weekly), precluding evaluation across a wide range of MTX doses. To facilitate comparison between our results and a less select population, a recent report using data from the National Health and Nutrition Evaluation Examination Survey (NHANES) assessed the prevalence of liver enzyme abnormalities in the general US adult population.13 In contrast to the relatively high proportions that we found and have previously been described for patients with RA and PsA, the NHANES found that 6% of the population had liver enzyme tests >1× ULN and 1% had elevations >2× ULN.

In multivariable models the adjusted likelihood of ALT/AST elevations in patients with PsA was approximately 2–3-fold greater than in patients with RA. This is consistent with concerns suggesting that patients with psoriasis may be particularly sensitive to MTX.8 In one randomised controlled trial examining LEF treatment in PsA, the incidence of ALT/AST abnormalities (10.4% over 6 months of treatment) was relatively low and comparable to those reported in randomised controlled trials in RA.14 In contrast, in a long-term retrospective cohort study of patients with psoriasis and PsA treated with MTX (78% with PsA), ALT/AST elevations were reported at least once in 57%.15 Although dissimilarities in study populations and/or follow-up may account for some differences, these observations may also reflect channelling of high-risk patients away from more hepatotoxic therapies. This hypothesis is supported by table 1, which shows that patients receiving MTX and MTX+LEF were less likely to have a history of liver disorders and regular alcohol use and were more likely to use folate supplementation.

The strengths of this study include evaluation of a large number of patients with RA and PsA receiving care in routine clinical settings, which allowed focus on interactions between various doses of MTX and LEF. Although previous randomised controlled trials in RA have reported relationships between DMARD exposures and ALT/AST elevations, risks associated with varying DMARD doses typically have been limited by allowance of only two doses of MTX. Thus, a focus on MTX dose contributes to the novelty of our report. Moreover, the safety profile of these DMARDs in patients with PsA has been less well studied outside clinical trials. The results from such trials may have limited generalisability to the much broader spectrum of patients treated with these agents in clinical practice.

As a potential limitation of these analyses, a definition of any ALT/AST elevation >ULN is very sensitive. For this reason, elevations >2× ULN were also evaluated. Moreover, any persistent ALT/AST elevations into the abnormal range have been shown to correlate with histopathological changes on liver biopsy in patients with RA.2 3 These data served as the basis for published guidelines for monitoring of MTX treatment in RA,16 and any elevation may prompt a change in therapy or reductions in dose. It has also been shown that patients starting MTX increase the magnitude of their initial normal AST measurements by twofold from baseline, even while remaining in the normal range.16 In light of these separate lines of published evidence, we believe that it is somewhat arbitrary to label elevations into the abnormal range of between 1× and 2× normal as uniformly inconsequential. Additionally, monitoring laboratory tests are not presently mandated by CORRONA, and it may be that patients on potentially hepatotoxic DMARDs have had more follow-up laboratory tests, resulting in a detection bias. However, our sensitivity analyses evaluating patients with a minimum number of follow-up laboratory tests did not change the results. Furthermore, although we considered patients who changed drug treatment groups as remaining exposed until they started a new drug, we recognise the possibility that there was a “carryover” effect from the previous drug therapy. If this were this case, it would have increased the proportion of patients in the non-MTX, non-LEF exposure group with ALT/AST elevations; however, this effect would serve only to attenuate our results towards the null. Finally, we acknowledge that the results from this US cohort may not be generalisable to individuals receiving care in other settings where, for example, the prevalence of alcohol use and other hepatotoxic agents may differ.

In conclusion, the proportion of patients with RA with ALT/AST elevations >ULN cared for in routine clinical practice in the US CORRONA registry receiving MTX or LEF was approximately 14–22%, and greater for those receiving combination MTX+LEF, particularly at doses of MTX ⩾10 mg/week. We have documented what appears to be a selection bias of high-risk patients away from therapeutic regimens which may be associated with more hepatotoxicity. We have also shown that liver enzyme test abnormalities were numerically more frequent in patients with PsA. These data support the need for continued careful monitoring of transaminase enzymes in patients with both RA and PsA who receive treatment with MTX, LEF or a combination of the two drugs.