Clinical Rheumatology

, Volume 33, Issue 2, pp 227–236

Prednisolone combined with adjunctive immunosuppression is not superior to prednisolone alone in terms of efficacy and safety in giant cell arteritis: meta-analysis

Authors

    • Department of RheumatologyNorfolk and Norwich University Hospital
    • University of East Anglia
  • Y. K. Loke
    • Norwich Medical SchoolUniversity of East Anglia
  • R. A. Watts
    • Norwich Medical SchoolUniversity of East Anglia
  • A. J. MacGregor
    • Department of RheumatologyNorfolk and Norwich University Hospital
    • University of East Anglia
    • Norwich Medical SchoolUniversity of East Anglia
Original Article

DOI: 10.1007/s10067-013-2384-2

Cite this article as:
Yates, M., Loke, Y.K., Watts, R.A. et al. Clin Rheumatol (2014) 33: 227. doi:10.1007/s10067-013-2384-2

Abstract

To conduct a meta-analysis of published data of the effectiveness of drug treatment in giant cell arteritis (GCA) to provide evidence to support the optimal use of glucocorticoids (GCs) and adjunct therapy. MEDLINE, CENTRAL and EMBASE searches were used to identify randomised control trials on the treatment of GCA. Studies included were trials in which: (1) the participants were classified as having GCA by the 1990 ACR criteria or biopsy proven disease; (2) parallel-group randomised control of at least 16 weeks duration had been conducted with at least 20 participants; (3) the design included either alternative adjunct immunosuppressant regimens, alternative GCs dosing or routes of administration; and (4) outcome data was included on either relapse rates or rates of infection. One thousand eight hundred thirty-six articles were retrieved, of which only 37 met the primary inclusion criteria. Sixteen of these studies reported some information about the GCs or adjuvant regimen used. Only ten studies were of sufficient quality to be included in the meta-analysis. Together these comprised 638 participants of which 72 % were female. Three studies compared various GCs regimens, with two comparing IV GCs, the latter showing a marginal benefit with respect to relapse (risk ratio (RR) = 0.78, 95 % CI = 0.54 to 1.12) but a greater risk of infection (RR = 1.58, 95 % CI = 0.90 to 2.78). Another three used methotrexate as an adjunctive agent and showed marginal benefit with respect to relapse (RR = 0.85, 95 % CI = 0.66 to 1.11). The remaining four trials compared prednisolone to dapsone, infliximab, adalimumab and hydroxychloroquine, respectively. There are various clinical trials of varying quality. The results from this meta-analysis show that the use of adjunct agents is not associated with improved outcome.

Keywords

Giant cell arteritisGlucocorticoidsImmunosuppressionMeta-analysisTreatmentVasculitis

Introduction

Giant cell arteritis (GCA) is a large vessel vasculitis of unknown aetiology that typically occurs in patients above the age of 50 years. It is the most common vasculitis in older persons with an annual incidence in the UK of 22 per 100,000 [1]. High-dose glucocorticoids (GCs) are the mainstay of treatment, with evidence of benefit demonstrated since the 1950s [2, 3]. Untreated GCA can lead to blindness and can occur in up to 20 % of cases treated with GCs [4, 5].

Despite the severe consequences, there is no consensus on the optimal treatment for GCA, and there is considerable variation in clinical practice with regard to the initial (starting) dose and subsequent rate of dose reduction of GCs [6]. Both European League Against Rheumatism (EULAR) and the British Society for Rheumatology (BSR) have produced guidelines on GCs dosing. The guidelines were developed through an extensive literature search using a modified Delphi technique. Recommendations were based on data from clinical trials and expert opinion of the panel members [7, 8]. The EULAR guidelines advise treatment with prednisolone 1 mg/kg/day (maximum, 60 mg/day) continued for 1 month [7]. The BSR recommends prednisolone 40 to 60 mg (not <0.75 mg/kg) daily in patients without visual loss until the resolution of symptoms and laboratory abnormalities, 500 mg to 1 g of intravenous methylprednisolone in those with evolving visual loss, and at least 60 mg a day for those patients with established visual loss [8].

These guidelines have been based, only partially, on published data from randomised control trials. The risk of blindness in untreated patients precludes the use of placebo-controlled studies. Therefore, these studies have used GC regimens in both the intervention and control arms. Both the disease and its treatment with high-dose GCs have significant morbidity since high dosages of GCs are associated with substantial toxicity. Although mortality is not increased in cohorts of patients with GCA [9, 10], GCs usage is associated with a sevenfold increase in the risk of severe opportunistic infections compared to the rest of the population [11]. Patients treated with greater GC dosages are at the greatest risk of morbidity [6]. Various immunosuppressive agents have been trialled in an effort to reduce toxicity from GCs and to improve efficacy of treatment.

Existing studies of treatments in GCA have focused on design, including randomised control trials which compare either different GC regimens or the use of adjunct agents in an attempt to reduce overall GC dosages (steroid sparing). The outcomes of interest have been the rate of relapse and the toxicity associated with GCs. Currently, there is only one meta-analysis comparing the outcomes from trials using methotrexate as an adjunct agent in the treatment of GCA which showed a benefit in those patients receiving methotrexate [12]. However, there have been many other trials using several other adjunct agents. In this report, we synthesise the published evidence, with the use of meta-analysis, with the aim of identifying whether there is benefit from the use of adjunct agents or IV GCs containing regimens in terms of the outcomes of relapse rate, infection rate and total GC dose.

Methods

Search strategy

The search attempted to capture all clinical drug trials that included patients with a clinical diagnosis of GCA. Searches were conducted using MEDLINE (Ovid), EMBASE and CENTRAL databases. There was no language restriction. The dates of the searches were from 1946 to 19 August 2013. The bibliographies of trials which were identified in the initial search were hand searched for additional studies.

Inclusion criteria

Studies included were trials in which:
  1. (1)

    the participants were classified as having GCA by the 1990 ACR criteria or biopsy proven disease;

     
  2. (2)

    parallel-group randomised control trials of at least 16 weeks duration with at least 20 participants;

     
  3. (3)

    the design included either alternative adjunct immunosuppressant regimens, alternative GCs dosing or routes of administration;

     
  4. (4)

    outcome data was included either on relapse rates or rates of infection.

     

When an article could not be rejected based on the information contained within the abstract, the full text version was retrieved and reviewed. The searches were carried out independently by two members of the research team (MY and YKL). Any disagreement on whether a study should be included or excluded was resolved by consensus discussion between MY and YKL.

Data extraction and validity assessment

Data extraction and quality assessment judgements were taken independently by the two reviewers (MY and YKL) and recorded onto a data sheet. The main outcomes of interest were poor treatment response (as assessed by relapse rate or failure to reach a lower dose of GCs), infection-related adverse events and total dose of oral GCs. Quality judgements were made using the Critical Appraisal Skills Programme (CASP) tool for randomised control trials [13]. Authors of papers were contacted to clarify or provide missing data necessary for the meta-analysis and asked whether they knew of any existing or ongoing studies.

Analytical approach

The analysis focused on adverse outcomes of therapy, namely relapses and infections and total dose outcomes in the included trials. Occurrences of relapse were considered as relapse as defined by the authors of the primary study, or alternatively, outcomes related to unfavourable response to treatment such as failure to reach a lower dose of GC. Infection was defined as in the original source article and included all patients who developed an infection regardless of the need for inpatient admission. The efficacy of different GC regimens was further evaluated in a Poisson regression model in which relapse was assessed as a function of the patient characteristics and summary measures that characterised the individual regimen (initial dose, total dose, dose at day 60, total dose and duration of treatment).

Data synthesis

The data were pooled using Review Manager (RevMan) version 5.1.7 (Nordic Cochrane Centre, Copenhagen, Denmark). A fixed effect model was used to estimate pooled relative risks from dichotomized outcome events in the trials. For continuous variables (such as GC dose), a meta-analysis of mean differences between groups was conducted. Missing standard deviation values were imputed from similar included trials where necessary, as recommended by Furukawa et al. [14]. Statistical heterogeneity was assessed using the I2 statistic with I2 > 50 % indicating a substantial level of heterogeneity. If substantial heterogeneity was detected, the random effects model was used to incorporate heterogeneity into the pooled estimate. Asymmetry testing for publication bias was used only if there were more than ten included trials and if there was no evidence of significant statistical heterogeneity.

Results

Study characteristics

In all, 1,836 citations were identified; of which, only 37 met the initial inclusion criteria (see PRISMA statement—Fig. 1). Of these 37 studies, 16 reported some information on the GC regimen used and included relevant outcome data. Six of the 16 were excluded due to the lack of information or because they involved a trial of patients with established GC side effects. These studies were assessed qualitatively but not included in the meta-analysis (Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10067-013-2384-2/MediaObjects/10067_2013_2384_Fig1_HTML.gif
Fig. 1

PRISMA Statement

Table 1

Studies excluded from meta-analysis

Study

Trial design

Reason for exclusion

Bengtsson et al. [15]

Alternate day GCs

(27 participants)

Not RCT

Participants had side effects from pretreated GCA

DeSilva and Hazelman [16]

RCT of azathioprine vs GCs

(31 participants)

Participants had side effects from pretreated GCA

Liozon et al. [17]

RCT of dapsone

(48 participants)

Same data in 1993 trial (see Table 2)

Mariette et al. [18]

RCT of adalimumab

(70 participants)

Participants had side effects from pretreated GCA

Martinez-Taboada et al. [19]

RCT of etanercept

(17 Participants)

Participants had side effects from pretreated GCA

Schaufelberger et al. [20]

Open label RCT of

cyclosporin

(60 participants)

No efficacy data

The remaining ten studies were randomised trials that compared either different GC regimens or GCs with adjunct agents. Together these comprised 638 participants with a mean age of 73 years of which 72 % were female. Eighty-three percent had a positive temporal artery biopsy (Table 2).
Table 2

Study characteristics

Study

Trial design

Age (years)

Female (%)

TAB + ve (%)

Duration (I/C weeks)

Relapse: markers and/or Sx?

Infection

Spiera et al. [24]

RCT of MTX vs GCs

(21 participants)

74

56

78

68/60

Either

Any

Jover et al. [25]

RCT of MTX vs GCs

(42 participants)

77.6

71.4

100

29/94

Sx

Any

Hoffman et al. [26]

RCT of MTX vs GCs

(98 participants)

75

61.7

87

23/24

Both (ESR >40 mm/h)

SAE

Hunder et al. [21]

RCT of GC regimens

(60 participants)

68.7

80

100

Sx

Chevalet et al. [22]

RCT of GC regimens

(164 participants)

73.3

70.7

78

36a

Either

Any

Mazlumzadeh et al. [23]

RCT of GC regimens

(27 participants)

74

70.4

100

Either

Any

Sailler et al. [30]

RCT of HCQ vs GCs

(64 participants)

100

Either

Liozon et al. [27]

RCT of dapsone vs GCs

(48 participants)

75

63.8

100

62/59

Either

Any

Hoffman et al. [28]

RCT of infliximab vs GCs

(44 participants)

69.5

68.7

83

Stopped earlyb

Both (ESR >40 mm/h)

Any

Seror et al. [29]

RCT of adalimumab vs GCs

(70 participants)

74

74.3

71.4

Either

Any

Duration of treatment is recorded in the above table as median values for the intervention then control (I/C). Infection as defined in the original article is severe adverse event (requiring hospitalisation and IV antibiotics)

RCT Randomised control trial, TAB +ve (%) percentage of participants with a positive temporal artery biopsy, MTX methotrexate, GCs glucocorticoids, HCQ hydroxychloroquine, SAE severe adverse event, Sx symptoms

aMean time to <7 mg/day of oral GC

bOne study was stopped early once the initial analysis at 22 weeks had shown infliximab to be performing worse than the control group relapse—as defined in the original article either based on symptoms, rise in inflammatory markers or both

Three of the ten included studies (251 participants) compared different GC regimens [2123]. Only two of these three studies were blinded. The studies were of varying quality and size. The small number of trials precluded the use of a detailed assessment of publication bias.

A further three studies compared prednisolone to methotrexate (starting dose range, 7.5 mg to 10 mg/week; maximum dose range, 10 mg to 20 mg/week) [2426]. The remaining four trials compared prednisolone to immunomodulatory adjuncts of dapsone (an antibiotic) [27], two monoclonal antibodies that inhibit tumour necrosis factor alpha of infliximab [28] and adalimumab [29], respectively and hydroxychloroquine (an antimalarial agent) [30].

Adverse events—relapse, or failure to reach lower dose of prednisolone

Ten studies were included in the assessment of adverse events, three of which compared different GC regimens. There was a suggestion that those trials that used dapsone and intravenous (IV) GCs were associated with better outcome, although they comprised relatively small numbers. Figure 2 summarises their efficacy and risk of unfavourable outcome (relapse or failure to reach lower dose of prednisolone).
https://static-content.springer.com/image/art%3A10.1007%2Fs10067-013-2384-2/MediaObjects/10067_2013_2384_Fig2_HTML.gif
Fig. 2

Relapse or failure to reach lower dose of prednisolone

No GC regimen with an adjunct agent was shown to have consistent advantage over oral GCs alone. The results of the three trials that included methotrexate were heterogeneous, with one trial showing significant benefit, while the other two did not. Dapsone showed a significant advantage, but the trial was limited by its small sample size and risk of bias. The adjunct treatments that were ineffective included hydroxychloroquine, adalimumab and alternate day GCs (which appeared to be significantly less beneficial than standard care).

Of the adjunct trial agents that had been examined in more than one study, methotrexate was associated with a risk ratio (RR) of 0.85 (95 % CI = 0.66 to 1.10) and IV GCs with an RR of 0.78 (95 % CI = 0.54 to 1.12). The heterogeneity values for these studies were 43 and 0 %, respectively, and therefore a fixed effects model was used. These results suggest a marginal benefit in terms of reduced relapse risk for these regimens which was not statistically significant.

Adverse events—infection

Eight of the ten studies reported infection outcomes (outcome (Fig. 3) data was missing for the alternate day GCs trial and hydroxychloroquine study). Methotrexate as an adjunct showed no increased risk for infection with a risk ratio of 0.85 (95 % CI = 0.49 to 1.47). The infliximab trial was stopped early due to an increased risk of infection in those individuals receiving infliximab (RR = 1.27, 95 % CI = 0.78 to 2.08). The other anti-TNF trial involving adalimumab revealed a significantly increased number of patients with overall infections in the intervention arm as compared to control, and this reached statistical significance (RR = 1.93, 95 % CI = 1.09 to 3.39), although there was no clear difference between arms if the analysis was restricted to serious infections only. An increased risk of infection was also seen with the IV GC trial (RR 1.58, 95 % CI = 0.90 to 2.78).
https://static-content.springer.com/image/art%3A10.1007%2Fs10067-013-2384-2/MediaObjects/10067_2013_2384_Fig3_HTML.gif
Fig. 3

The risk of infection for the various trial agents

Figure 4 illustrates the oral GCs’ sparing effects exhibited by the various agents. IV GC regimens reduce the need for oral prednisolone, but only one of the methotrexate trials showed any steroid sparing effect (Jover et. al [25]; mean difference (grammes), 1.12 g (CI = −1.99 to −0.24 g).
https://static-content.springer.com/image/art%3A10.1007%2Fs10067-013-2384-2/MediaObjects/10067_2013_2384_Fig4_HTML.gif
Fig. 4

Oral GCs Sparing (total dosage in grammes)

GC starting dose and taper

The control arms for the trials which compared GCs to GCs plus adjunct immunosuppression allow for comparison of outcome between the different GC regimens used. The relapse rates in the GC control arms ranged from 27.9 to 92.3 %. As expected, there was a positive correlation between length of the study and relapse rate. The relapse rate for studies of over 18 months of follow-up ranged from 71.0 to 92.3 % and for those with follow-up under 12 months ranged from 27.9 to 92.0 %. To account for different lengths of follow-up, the number of relapses reported in each study was modelled in a Poisson regression. This showed no clear association between relapse rate and starting GC dose, dose at day 60 or total GCs dose.

Discussion

The results from this meta-analysis show that there is no benefit from using adjunct therapy with GCs for the treatment of GCA. GCs have been the gold standard treatment for GCA since their first use for the condition in the 1950s [2, 3]. Many patients become dependent on long-term GCs, and there is need to achieve a balance between treating the disease effectively and exposing patients to GC’s toxicity. Whilst adjunct agents are frequently used in rheumatic diseases for their steroid sparing effect, this meta-analysis shows that there is no benefit from adjunct agents in GCA either in terms of efficacy or toxicity. Indeed, one of the trials had to be stopped early due to an increased infection rate in those receiving infliximab. The other anti-TNF trial using adalimumab was recently published. This study was an RCT of adalimumab given for 10 weeks in newly diagnosed participants with GCA. Unfortunately, due to slow recruitment rate, the study was underpowered for its primary outcome of increasing the proportion of participants in disease remission at 6 months from 40 to 70 % in the intervention group. This trial also revealed an increased risk for infection in those participants receiving adalimumab. Two IV steroid trials showed a small benefit over oral prednisolone but had a greater infection rate. No starting dose or rate of taper for GCs was shown to be superior.

There were three trials that used methotrexate as an adjunct agent. These trials used methotrexate at treatment induction, rather than at a later point in the treatment of individuals with difficult to treat GCA, as in common clinical practice. The three trials used a relatively low dose of methotrexate (starting dosage range of 7.5 to 10 mg/week, increasing to a maximum dose range of 10 to 20 mg/week). These trials have been the subject of a previous meta-analysis [12]. The authors of that study concluded that adjunctive methotrexate lowers the risk of relapse and reduces the exposure of corticosteroids. However, only one of the methotrexate trials showed any steroid sparing effect (Jover 2001; mean difference (grammes), 1.12 g (CI = −1.99 to −0.24 g). The trial by Jover et al. also showed a benefit for reduction in relapse, but the other two studies did not. It is this trial that drives the positive outcome in the previous meta-analysis. The dosage of methotrexate across these trials was relatively low, but similar across the studies. The trial by Jover et al. used 10 mg/week of oral methotrexate throughout the study period (24 months). In addition, the use of GCs was similar in the two largest studies involving methotrexate. Both the trial by Hoffman et al. and Jover et al. used an oral dose of prednisolone at a maximum of 60 mg per day. Therefore, the results of the Jover trial are difficult to explain in relation to the other two methotrexate adjunct trials. However, the previous meta-analysis did contain individualised patient data and therefore was able to draw survival curves, which we were not able to do in our study.

Most of the trials had a relatively short follow-up period of between 6 months to 1 year. Current guidance from the BSR and EULAR recommends that patients should be taking GCs for around 18 months [7, 8]. Due to the heterogeneity of the studies, it is difficult to draw conclusions regarding the optimum GCs starting dose and tapering regimen over longer time periods, and had these been reported, the conclusions drawn might have been different.

Limitations of the present analysis include the small number of studies using the various adjunct agents and the fact that studies reported a range of different end points. There were differences in the definition of relapse. Some studies required symptoms and the presence of rising inflammatory markers (either CRP or erythrocyte sedimentation rate (ESR)) in order to be labelled as a relapse; others required only a rise in inflammatory markers or symptoms suggestive of relapse. Remission was defined in all studies as the patient being free of symptoms after withdrawal of GCs. Baseline variation in the characteristics of the population being studied and the lack of individual participant level data are further limitations.

The differences in definition for both relapse and infection could introduce bias, the main form of which is ascertainment bias. For example, studies that only include a rise in inflammatory markers without correlation to clinical symptoms may conclude a worse relapse rate to those where such features were recorded. In addition, studies that only included serious infections (i.e. those that required hospitalisation) may appear superior to those that included all infections. These problems are dealt with by performing a meta-analysis since the control and intervention arms of the study are compared as relative risks or risk ratios. Therefore, as long as the definitions for infection or relapse are the same for both groups, this does not alter the results. The problem arises when crude numbers for the trials are compared to another trial which uses different definitions. A forest plot limits this particular problem and is why meta-analyses are useful. Another problem can be publication bias. This is usually assessed with a funnel plot, but due to the low numbers of studies, this was not practicable.

Other possible reasons for the lack of efficacy of the various alternative treatment regimens seen in this analysis may be due to the different underlying biological mechanism of action. For example, intervention targeting IL-6 inhibition using tocilizumab may prove to be more successful, and there are several promising case reports [3133]. A clinical trial using tocilizumab is underway. Recent data also supports a role for IL-21 [34]. IL-21 modulates the effects of TH-17 which has been well recognised in a number of inflammatory conditions including GCA [3537].

GCA is the most common vasculitis in older people. The lack of agreed outcomes inhibits drawing robust conclusions from the existing data. As has been recognised by the OMERACT committee [38], the development of validated outcomes in GCA, in particular the need to develop standard definitions of remission, is urgently needed if treatment regimens are to be developed based on sound evidence.

Disclosures

None.

Key messages

The available data from the clinical trials are of varying quality.

In remission induction, the data from the clinical trials show no additional benefit for combination therapy. Development of validated outcomes in GCA is urgently required.

Copyright information

© Clinical Rheumatology 2013