The treatment of psoriatic arthritis (PsA) has moved beyond symptomatic therapy with non-steroidal anti-inflammatory agents (NSAIDs) and analgesics. Many patients now receive disease-modifying anti-rheumatic drugs (DMARDs) and biologics. Tumour necrosis factor (TNF) inhibitors are well established as effective biologics for treating PsA and co-stimulatory blockade agents such as Alefacept have also been evaluated. Clinicians are therefore interested in understanding the relative efficacy and toxicity of DMARDs and biologics for PsA. Our primary aim in this systematic review is to provide comparable data about established and more modern treatments for PsA.

Therapeutic advances in treating PsA have been accompanied by developments in outcome assessments. Modern randomised controlled trials (RCTs) use summated indices such as the American College of Rheumatology (ACR) response criteria and psoriatic arthritis response criteria (PsARC).1 2 However, such modern outcome assessments cannot be applied retrospectively to earlier RCTs. Consequently, treatments evaluated in older trials may appear less effective than modern therapies; this might lead to potential bias against older treatments. We therefore based our systematic review on the simplest criterion of benefit—whether treatment was stopped for inefficacy or adverse effects. These outcome measures are routinely reported in clinical trials under the CONSORT guidelines.3 They have face validity and are used to assess effectiveness in clinical practice. We have previously established their validity compared with ACR response criteria in a systematic review of treatments in rheumatoid arthritis.4

METHODS

Search strategy

We searched the Medline, PubMed and EmBase databases from 1966 to June 2006 for articles with the headline arthritis, psoriasis/psoriatic, psoriatic arthritis therapy and/or treatment, and randomised controlled trial. Cochrane clinical trials register and Cochrane database for systematic reviews were also searched. This was supplemented by manually searching bibliographies of these articles and of previously published reviews.

Criteria for considering studies for this review

Using a predefined protocol we selected studies for evaluation using the following criteria: (a) randomised controlled trials; (b) enrolled patients with PsA; (c) at least one of the treatment group was placebo; (d) DMARDs or biologicals were currently used in clinical practice; (e) publication was in English. Trials were only included if they had adequate allocation concealment (studies were excluded if an open allocation schedule or unsealed or open envelopes were used) and double-blinded assessment. Studies involving experimental and non-licensed treatments were excluded.

Quality of trials

The Jadad score assessed the quality of the trials.5

Types of outcome measures

The primary outcome for efficacy was the number of patients withdrawn because of lack of efficacy. Sensitivity analysis was carried to compare efficacy assessed by patient withdrawal with the number of patients who achieved ACR20 response1 and the number of patients who achieved PsARC response.2 The primary outcome for toxicity was the number of patients withdrawn due to adverse events.

Meta-analysis

We used Review Manager (RevMan version 4.2 for Windows, Oxford, UK: The Cochrane Collaboration, 2002) software. Results were expressed as risk ratios (RR) with 95% confidence intervals (CI) for dichotomous outcomes based on the random effects model. Sensitivity analyses were carried out based on agents used and outcome measured. The validity of patient withdrawal due to lack of efficacy as an outcome measure was evaluated by comparing it with ACR20 and PsARC responses. Ratio of numbers needed to treat (NNT) to numbers needed to harm (NNH) was calculated to assess the benefit versus risk of each treatment.

RESULTS

Description of studies

A preliminary search identified 32 potentially relevant RCTs. Nine studies were excluded6^–15 that used experimental treatments6^–9 or interventions not used in clinical practice.10^–14 These included studies of intravenous methotrexate13 and fumaric acid.14 The 23 remaining articles were reviewed in detail; a further five were excluded (one that was excluded was an open label study,15 three did not have a placebo arm16^–18 and one also included patients with spondyloarthopathy and information on number of patients with PsA who withdrew because of toxicity and lack of efficacy was not available).19 The remaining 18 RCTs2 20^–36 fulfilled the criteria for inclusion in this review and are listed in table 1.

Study characteristics

All studies recruited patients with established PsA who had peripheral arthritis. Fifteen studies used parallel designs and three used crossover designs. For all three-crossover trials we included data only from the initial randomised, double-blind parallel phase.23 28 34 Eleven RCTs assessed DMARD monotherapy and one study DMARD combination. Five assessed TNF inhibitors and one Alefacept. The average Jadad score was 3, and 13 (72%) of the RCTs had a Jadad score of 3. Seven studies used PsARC as an outcome measure2 30^–35 (one with sulfasalazine, one leflunomide, two etanercept, two infliximab and one adalimumab) and seven used ACR20 response criteria (one with leflunomide, one adalumimab, two etanercept, two infliximab and one alefacept).30^–36

Efficacy

Treatment was better than placebo (RR = 0.35; 95% CI 0.25, 0.49; p = 0.00001) (fig 1). The RCTs did not show significant heterogeneity (χ² = 18.38; p = 0.10). Five RCTs assessed TNF inhibitors (table 2), which were substantially more effective than placebo (RR = 0.25; 95% CI 0.13, 0.48; p = 0.0001). The effect size of treatment with gold was similar (RR = 0.25; 95% CI 0.11, 0.53; p = 0.0003). Five RCTs evaluated sulfasalazine (table 2); its benefits were moderate (RR = 0.45; 95% CI 0.23, 0.89; p = 0.02). One RCT evaluated leflunomide (RR = 0.44; 95% CI 0.23, 0.83; p = 0.01); its benefit was similar to sulfasalazine.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Overall efficacy based on withdrawals due to lack of efficacy.

Only one RCT evaluated methotrexate. This small 12-week placebo-controlled trial of low-dose methotrexate was terminated early after recruiting only 37 patients.21 Methotrexate significantly improved physician global assessment; there were no significant effects on other outcomes (eg, tender and swollen joint scores, patient global assessment, ESR).

Toxicity

Treatment resulted in more withdrawals due to toxicity than placebo (RR = 2.33; 95% CI 1.61, 3.37; p = 0.00001) (fig 2). Toxicity leading to withdrawal was most common with leflunomide (RR = 3.86; 95% CI 1.2, 12.39; p = 0.02) and gold salts (RR = 2.34; 95% CI 1.1, 4.97; P = 0.03) (table 2). Withdrawal as a result of toxicity due to sulfasalazine was relatively high (RR = 1.76; 95% CI 0.98, 3.14; p = 0.06). With TNF inhibitors withdrawal due to toxicity was more than with placebo (RR = 2.2; 95% CI 0.82, 5.91; p = 0.12) but this was not statistically significant.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Overall toxicity based on withdrawals due to side-effects.

Benefit versus risk

TNF inhibitors had highest ratio of numbers needed to treat (NNT) to numbers needed to harm (NNH) (table 2) followed by leflunomide, gold salts and sulfasalazine.

Sensitivity analysis

The validity of using the withdrawal rate as the principal outcome measure of efficacy was examined by comparing results with meta-analysis using ACR20 and PsARC as outcome measures in seven studies that reported these (table 3). These showed comparable results, indicating the validity of using withdrawals for lack of effect.

DISCUSSION

Compared with rheumatoid arthritis there have been relatively few RCTs in PsA. Consequently, the treatment strategy for PsA is largely based on those used in rheumatoid arthritis, even though the pattern of synovial inflammation, the severity of joint destruction and the presence of skin involvement that makes important contributions to the overall quality of life all differ from rheumatoid arthritis. With the advent of TNF inhibitors, it is an appropriate moment to systematically review treatments aimed to modify the course of PsA. Methodologically in this meta-analysis we used patient withdrawals due to lack of efficacy and toxicity as primary outcome measures. Patient withdrawal is complex and may be the result of a combination of both adverse events and poor efficacy. The reason for withdrawal, which is determined arbitrarily by the investigator, is a potential limitation. Hence we validated this method against ACR response criteria in a previous meta-analysis of combined DMARDs in rheumatoid arthritis.4 It overcomes the difficulties from using a variety of different outcome measures in historic trials. Here we have confirmed the validity of this approach again by showing it gave comparable results to ACR response criteria and PsARC. We have also explored the possibility of validating this methodology against improvement in Psoriasis Area and Severity Index. However, in seven trials where Psoriasis Area and Severity Index was assessed, only those patients with >3% of skin involvement were included in the analysis. The percentage of eligible patient and baseline Psoriasis Area and Severity Index scores are different in different trials, hence we have not included this formally as sensitivity analysis. Ideally, we would also like to perform a meta-analysis of the effect of treatment on structural damage but technically it is complicated. Hitherto, the effect of treatment on structural damage has only been reported in trials of TNF antagonists. All of the trials showed statistical significant retardation in joint damage. Undoubtedly meta-analysis would have shown the same result.

Some trials have included patients receiving two active agents; particularly methotrexate and a TNF inhibitor. However, subgroup analysis of one of these trials, evaluating adalimumab, suggested no added benefit from receiving two agents over a single agent.37 One RCT assessed the effect of efalizumab in combination with methotrexate or sulfasalazine in 107 patient with PsA38 The result was negative suggesting efalizumab is ineffective in PsA. Unfortunately this was published after the deadline of our search strategy. Efalizumab is an anti-T cell agent; therefore, in a different class compared with other agents included in this review. Including efalizumab in this result would not affect our conclusion.

The Group for Research and Assessment of psoriasis and psoriatic arthritis (GRAPPA) in conjunction with the Outcome Measures in Rheumatology (OMERACT) group is refining and developing outcome measures for a variety of disease domains for PsA.39 This includes assessment for spine involvement, enthesitis, dactylitis and fatigue. Other outcomes such as quality of life, radiographic and skin scores are also important and relevant in PsA that we did not evaluate in this meta-analysis, as these have not been reported in many clinical trials. This collaboration has resulted in a number of systematic reviews that have evaluated specific aspects of the treatment of PsA, including the assessments involved39 and peripheral arthritis, axial disease, enthesopathy, dactylitis and psoriatic nail disease.40^–45 These have all highlighted the benefits of treatment with DMARDs and biologics, though there are differences in these various clinical settings.

It is perhaps surprising to find that TNF inhibitors have the best ratio of NNT:NNH. A recent systematic review of therapy for PSA has also found positive evidence for usefulness of TNF inhibitors in symptom control, physical function, quality of life and control of radiographic progression with low toxicity profile.40 The lack of statistical significant difference in withdrawal due to toxicity with TNF inhibitors could not be explained by the small sample size as the number of patients enrolled in trials of these agents is larger than other DMARDs. Sulfasalazine is considered to be generally well tolerated but only has a moderate effect in PsA.46 Our meta-analysis confirms moderate efficacy and good tolerability of sulfasalazine in PsA. The effect size of treatment with oral and intramuscular gold salts was among the highest compared with placebo but it was poorly tolerated. This is in contrast to another meta-analysis where it appeared that gold salts were not statistically better than placebo for the treatment of PsA;47 48 however, this study used a pooled index as the primary outcome measure. The sensitivity of this instrument to detect change is unknown. The difference between our finding and this study could be explained by the difference in our chosen outcome measure to detect change especially as there were only two studies with a total of 320 patients. Leflunomide had one of the best ratios for efficacy to tolerability and the effect size for treatment was also better than sulfasalazine. Only one randomised control trial is available for this drug so far and its effect in terms of disease modification is yet to be established in PsA. The impact of methotrexate in combination with TNF inhibitors is a further complexity in comparing trials in which some patients were treated with placebos while others were treated with methotrexate. It is surprising that methotrexate, which is often used to treat PsA has not been studied in more detail; the study by Willkens et al21 was too small to produce a definitive result and was discontinued early due to slow recruitment. A definitive RCT of methotrexate is needed to complete the evidence base.

We conclude that gold, sulfasalazine, leflunomide and TNF inhibitors all are effective treatment for PsA. The effect size of treatment was highest with gold and TNF inhibitors, and previous systematic reviews of TNF inhibitors have already highlighted their value.49 50 Tolerability was worst with gold and leflunomide; more patients on active treatment had to be withdrawn from treatment due to side-effects when compared with placebo. The NNT:NNH ratio was best with TNF inhibitors followed by leflunomide, gold and sulfasalazine.