Article Text
PDF
Abstract
Objective To compare EuroQol 5-Dimensions (EQ-5D) utility and quality-adjusted life-years (QALYs) in patients with early, methotrexate (MTX) refractory rheumatoid arthritis (RA), randomised to addition of infliximab (IFX) or sulfasalazine and hydroxychloroquine (SSZ+HCQ).
Methods RA-patients with symptoms <1 year were enrolled between 2002 and 2005 at 15 Swedish centres. After 3–4 months of MTX monotherapy, patients with a remaining DAS28>3.2 were randomised to addition of IFX or SSZ+HCQ and followed for 21 months. EQ-5D profiles were collected every 3 months. Between-group comparisons of utility change and accumulated QALYs were performed, using last observation carried forward (LOCF) following protocol breach. Missing data were imputed by linear interpolation or LOCF. Sensitivity analyses applying baseline observation carried forward (BOCF) or restricted to completers were conducted.
Results Of 487 patients initially enrolled, 128 and 130 were randomised to IFX or SSZ+HCQ, respectively. Mean utility in the IFX and SSZ+HCQ groups increased from 0.52 (SD 0.27) and 0.55 (SD 0.27) at randomisation to 0.66 (SD 0.25) and 0.63 (SD 0.27) at 21 months (adjusted mean difference favouring IFX 0.04; 95% CI −0.01, 0.09; p=0.15). Average accumulated QALYs were 1.10 (SD 0.37) and 1.07 (SD 0.42) in the IFX and SSZ+HCQ groups, respectively (adjusted mean difference favouring IFX 0.07; 95%CI −0.01, 0.14; p=0.07). BOCF analysis showed similar results, while differences were reversed, though remained statistically non-significant among completers. Dropout rates in the IFX/SSZ+HCQ groups were 30%/43% (p=0.01).
Conclusions Comparing addition of IFX or SSZ+HCQ to MTX in active early RA, no statistically significant differences in utility or QALY gain could be detected over 21 months.
Trial registration Registered in WHO database at the Karolinska University Hospital, number CT20080004.
- Rheumatoid Arthritis
- Early Rheumatoid Arthritis
- Anti-TNF
- DMARDs (synthetic)
- Economic Evaluations
Statistics from Altmetric.com
Introduction
Over the last decades, methotrexate (MTX) has become established as the first-line disease-modifying antirheumatic drug (DMARD) in rheumatoid arthritis (RA).1 For many patients, however, MTX alone is insufficient to alleviate disease activity, necessitating attempts with more potent regimens.1 In this situation, adding sulfasalazine and hydroxychloroquine (SSZ+HCQ) to MTX has been shown to be superior to MTX monotherapy,2 ,3 as has the combination of MTX with a tumour necrosis factor (TNF) antagonist.4–7 These approaches were recently compared in the SWEdish FarmacOTherapy (SWEFOT) trial, finding the addition of infliximab (IFX) to result in significantly better treatment response rates at 9, but not at 15 or 21 months, and to be radiologically superior when compared with the addition of SSZ+HCQ.8 ,9
While greatly expanding therapeutic options, the introduction of biologic DMARDs has led to considerably increased treatment costs.10 Thus, recent years have seen a growing interest in economic evaluations, including cost-utility analyses comparing the cost per quality-adjusted life-year (QALY) of alternative interventions. QALYs are calculated from utility, a metric of health-related quality of life (HRQoL), anchored at 1 (full health) and 0 (death), and from the time spent in each HRQoL-state. Utility, in turn, is obtained by generic instruments, of which the EuroQol 5-Dimensions (EQ-5D) is most often applied in cost-utility analyses of RA.11–13 Of note, the use of different utility instruments, or national EQ-5D preference sets, may render varying utilities in the same patient.14 ,15
Both conventional and biologic DMARD treatments have demonstrated beneficial HRQoL-effects in RA.16 Few randomised controlled trials, however, have compared QALYs derived from direct observations of utility at multiple time points.17 Instead, existing cost-utility analyses have generally applied utilities (and costs) indirectly calculated from other clinical measures, such as the Health Assessment Questionnaire (HAQ) score,13 ,18 the appropriateness of which is currently under debate.19–22 While the SWEFOT trial was primarily designed to detect between-group differences in treatment response rates, EQ-5D-derived utility was included among the prespecified secondary outcomes. HRQoL-data has not been presented in any of the previous publications from the SWEFOT trial.8–9 ,23 ,24
The aim of the current analysis was to compare the development of EQ-5D-derived HRQoL of patients assigned to addition of either IFX or SSZ+HCQ in the SWEFOT trial, a randomised, active-controlled trial of patients with early, MTX-refractory RA. Between-group differences in utility change, accumulated QALYs, and in the five EQ-5D-dimensions were assessed.
Methods
Patients
Fifteen Swedish rheumatology units participated, enrolling patients between October 2002 and December 2005. Inclusion criteria were fulfilment of the 1987 revised American College of Rheumatology criteria for RA,25 age ≥18 years, symptom duration <1 year, a disease-activity score based on the 28-joint count (DAS28) of >3.2, no previous DMARD experience, and no or stably dosed oral glucocorticoid therapy for at least 4 weeks of, at most, 10 mg daily prednisone (or equivalent). Contraindications to any of the trial drugs was an exclusion criterion.
The study was approved by the regional ethics committees of all participating centres. Written informed consent was given by all patients prior to inclusion.
Study design
The investigator-initiated SWEFOT study was designed as a multicentre, randomised, active-controlled, open-label trial, comparing the addition to MTX of IFX or SSZ+HCQ in patients with active early RA.8 Patients were followed for 2 years, including a 3-month run-in period prior to randomisation. The main study time points are henceforth referred to as inclusion (start of run-in; –3 months), randomisation (0 months) and end of follow-up (21 months).
At inclusion, all patients started oral MTX (n=487). The initial dose of 10 mg was increased by 5 mg every 2 weeks to 20 mg weekly. Further MTX dose adjustments were allowed throughout the trial according to clinical practise.
After a 3-month run-in phase on MTX monotherapy, patients with low disease-activity (DAS28≤3.2) left the trial, while those with a remaining DAS28 >3.2 were randomly assigned to addition of either IFX (3 mg/kg bodyweight, rounded up to the nearest 100 mg increment, administered intravenously at weeks 0, 2 and 6, and every 8 weeks thereafter) or SSZ (1000 mg twice daily; orally) and HCQ (400 mg daily; orally). The randomisation process was centralised and done according to a computer-generated random allocation sequence. Neither patients, nor health-care providers were blinded to the treatment allocation owing to the different administration techniques. Additional DMARDs, or changes to the initial oral glucocorticoid dose, were not allowed, while intra-articular glucocorticoid injections were permitted except during 4 weeks before study visits.
Several treatment adjustments were allowed within the trial protocol. In case of inadequate response, the SSZ dose could be elevated to 1500 mg twice daily, while intervals between IFX administrations could be decreased to every 6 weeks. Following adverse events, SSZ or HCQ could be dose-reduced or withdrawn (continuing treatment with the other). Alternatively, SSZ+HCQ could be discontinued and replaced by cyclosporin A. Likewise, IFX could be switched to etanercept (50 mg subcutaneously once weekly) following adverse reactions.
Patients could leave the study at any time due to lack of effectiveness, adverse events or own choice. If possible, the scheduled follow-up continued despite protocol breach, while treatment was thereafter decided by the responsible rheumatologist according to clinical practise.
Outcome measures
The SWEFOT study was designed to detect a between-group difference in European League Against Rheumatism treatment response rates at 1 year (9 months after randomisation).26 As part of the prespecified secondary outcomes, EQ-5D assessments were done every 3 months, allowing for the calculation of utilities and accumulated QALYs (the area under the curve of utility plotted against time).
The primary outcomes of the present work were between-group differences in utility change and accumulated QALYs from randomisation to 21 months. Analyses were also conducted including the run-in phase of MTX monotherapy. Furthermore, the distribution of answers to the five EQ-5D questions were assessed at inclusion, randomisation and end of follow-up.
The UK EQ-5D preference set was used to calculate utilities.27 For comparison, results using the US national EQ-5D tariff are provided in the online supplement.28
Statistics
In the main analysis, utility change and accumulated QALYs were assessed in all randomised patients, using last observation carried forward (LOCF) from protocol breach (referred to as ‘LOCF analysis’). Missing utility data at inclusion (n=35) or randomisation (n=34) were imputed using a linear regression model with age, sex, DAS28, HAQ and visual analogue scale (VAS) for pain scores as covariates. In case of missing HAQ (n(inclusion/randomisation)=5/6), or VAS pain scores (n(inclusion/randomisation)=2/5), these were imputed using age, sex and DAS28. Missing utility data at later time points were imputed by linear interpolation, or by LOCF when future measurements were lacking.
Accumulated QALYs were estimated by the trapezoid method, calculating the area under the utility curve. Between-group differences in utility change and accumulated QALYs were assessed by analysis of covariance (ANCOVA), including treatment group (IFX vs SSZ+HCQ), age, sex, EQ-5D, DAS28 and HAQ scores at randomisation (or inclusion for analyses encompassing the run-in period).29 ,30
To examine the robustness of results and improve comparability with other studies, sensitivity analyses, applying baseline observation carried forward (BOCF; using utility at randomisation; referred to as ‘BOCF analysis’) or restricted to patients completing the trial per protocol (referred to as ‘completer analysis’) were also conducted. Only non-imputed data were utilised in the completer analysis, and available EQ-5D data at randomisation, and at least every 7.5 months thereafter, were required for completer status regarding the QALY assessment. Furthermore, to reflect clinical practise, an additional analysis was conducted including all available EQ-5D data, regardless of previous protocol breach (referred to as ‘pure intention to treat’). Due to the variability, therapies used after study cessations were not further specified. All ANCOVA analyses included the covariates detailed above.
Assessment of the five EQ-5D-dimensions were performed by analysis of all randomised patients using LOCF. Missing data at inclusion (n=35) and randomisation (n=34) were imputed using the mode at the respective time point. Distributions at the end of follow-up were compared using the Mann–Whitney U-test.
Between-group difference in time on allocated treatment was assessed by Kaplan–Meier curves and Cox proportional hazard regression (adjusting for age, sex, EQ-5D, DAS28 and HAQ scores at randomisation), while paired t tests were used for paired comparisons of utility change. All analyses were performed using SPSS (V.17.0) or STATA (V.11.0).
Results
Between October 2002 and December 2005, 493 patients were screened for inclusion, of whom 487 were enrolled and initiated on MTX monotherapy. After the 3-month run-in phase, 258 patients with a remaining DAS28 >3.2 were randomised to addition of IFX (n=128) or SSZ+HCQ (n=130). Recruitment and flow through the trial is displayed in online supplementary figure S1.
Patient characteristics are described table 1, showing the treatment arms to be well balanced at both inclusion and randomisation. From randomisation to the end of follow-up, dropout was significantly higher in the SSZ+HCQ group (IFX/SSZ+HCQ=30%/43%; figure 1 and online supplementary figure S1; HR 1.7 (95% CI 1.1 to 2.6; p=0.01)). Included among the dropouts, a few patients in each group (n(IFX/SSZ+HCQ)=8/5, categorised as ‘other’ in online supplementary figure S1) never received the allocated treatment. Reasons for this varied, but included subjects expressing preference for the alternative regimen or belatedly deciding to deny additional therapy.
Characteristics of the randomised population*
Kaplan–Meier survival curves from randomisation showing adherence to the randomly allocated therapies. Number of patients still on study treatments according to protocol shown below the graph (with the number of dropouts in each interval, within parentheses).
As allowed by the trial protocol, the frequency of IFX infusions was reduced to once every 6 weeks in 13 patients,9 while the SSZ dose was increased to 1500 mg twice daily in one case. In response to adverse events, 10 patients switched treatment from IFX to etanercept, and six subjects from SSZ+HCQ to cyclosporin A.
Utility change
Mean utility development of the two treatment arms are displayed in figure 2, while figure 3 (panels A and B) shows individual utility changes. Utility data were missing in 18% (IFX/SSZ+HCQ=19%/17%) of scheduled measurements in patients still receiving the allocated treatment, rising to 39% (IFX/SSZ+HCQ=38%/40%) following protocol breach.
Mean (95% CI) utility development of the treatment arms (UK EQ-5D preference set). Descriptive data at inclusion, randomisation and end of follow-up detailed in the lower section table. (A) Including only observed data from patients still on allocated treatment according to protocol. Number of observations detailed below the graph. (B) Analysis of all randomised patients using last observation carried forward. *Not to be confused with the completer analysis, restricted to patients completing the 21-month follow-up while on study medication. EQ-5D, EuroQol 5-Dimensions; HCQ, hydroxychloroquine; IFX, infliximab; IQR (25th–75th percentile); SSZ, sulfasalazine.
Individual utility changes and accumulated quality-adjusted life-years (QALYs) (UK EQ-5D preference set). Upper section: analysis of all randomised patients using last observation carried forward. Lower section: completer analysis. (A) Utility changes from study inclusion to end of follow-up. (B) Utility changes from randomisation to end of follow-up. (C) Frequency distribution of accumulated QALYs from randomisation to end of follow-up. HCQ, hydroxychloroquine; IFX, infliximab; IQR (25th–75th percentile); SSZ, sulfasalazine.
Table 2 displays adjusted between-group differences in utility change. Irrespective of analysis method, no significant difference was found between the treatment arms.
EQ-5D derived* utility change and accumulated QALYs according to treatment allocation
Despite not reaching a DAS28 ≤3.2, a mean utility gain of 0.18 (SD 0.32) was seen during the run-in period of MTX monotherapy. Indeed, when assessing all randomised patients, mean utility improved significantly more during the run-in phase than from randomisation to 21 months (mean 0.18 (SD 0.32) vs 0.11 (SD 0.26); mean difference 0.07, 95% CI 0.01 to 0.12, p=0.02 for the LOCF analysis; non-significant when analysing completers only).
Accumulated QALYs
Mean accumulated QALYs and adjusted between-group differences in accumulated QALYs are also presented in table 2. Again, no significant difference was detected between the study groups, though point estimates of the LOCF and BOCF analyses, encompassing randomisation to the end of follow-up, favoured the IFX arm (approximately 26 and 22 quality-adjusted life-days more in the respective analysis). Results of the completer assessment were reversed as compared with the LOCF/BOCF estimates, though remained far from statistical significance. The frequency distribution of accumulated QALYs is displayed in figure 3 (panel C).
As compared with the rest of the study population, patients ceasing the trial prematurely, due to inefficacy or adverse events, accumulated significantly fewer mean QALYs, while a similar trend was seen in subjects stopping treatment due to other reasons (see online supplementary table S1).
EQ-5D-dimensions
The distribution of answers (1=no problems; 2=moderate problems; 3=extreme problems) to the EQ-5D questions at inclusion, randomisation and end of follow-up are displayed in table 3, showing improvements in all dimensions of both treatment arms from randomisation to the end of follow-up. Overall, the worst scores were seen for pain/disability, while self-care was least affected at all time points. No significant between-group difference was observed at 21 months for any dimension.
Distribution of answers to the five EQ-5D-dimensions
Discussion
Main findings
In this randomised, active-controlled, open-label trial, comparing the addition of IFX or SSZ+HCQ to patients with active (DAS28 >3.2) early RA despite 3 months of MTX monotherapy, no statistically significant differences in utility or QALY gain could be detected over 21 months. Mean EQ-5D utility improved in both treatment arms following randomisation, though interestingly, an even larger mean utility gain occurred already during the 3-month run-in phase of MTX monotherapy, signalling at least a partial response to MTX on the group level.
By contrast with the completer assessment, point estimates of QALY accumulation in the LOCF and BOCF analyses, though not statistically significant, favoured the IFX arm. This between-group difference, thus, mainly emanates from the patients ceasing the trial prematurely. Protocol breach was significantly more common in the SSZ+HCQ group, mostly due to a higher frequency of treatment inefficacy. Furthermore, as previously reported, radiological progression was significantly greater in the SSZ+HCQ group at 21 months.9 Whether these differences would also impact HRQoL-estimates, and would make it possible to detect a significant between-group difference in a larger study or over an extended timeframe, can only be speculated. Irrespective of this, our results demonstrate that in a majority of patients with early MTX-refractory RA—and, in particular, in subjects adhering to therapy, indirectly signalling treatment effect—similar HRQoL-gains appear to be achieved by both the studied treatment strategies over 21 months.
Previous research
The use of different generic utility instruments, or national EQ-5D preference sets, may generate disparate utilities in the same patient.14 ,15 Results of studies using different instruments or national tariffs are, thus, not easily compared.
Studies of RA reporting EQ-5D development in relation to start of antirheumatic therapy are summarised in online supplementary table S2. Few previous randomised controlled trials have reported EQ-5D data, none comparing the treatment strategies assessed in the SWEFOT study. In early RA, 2-year utility (using the Dutch national EQ-5D tariff) and QALY (Dutch and UK tariff) data are available from the BeSt trial, assessing four DMARD sequences using a tight control scheme,17 ,31 while in established RA, utility development has been published from the ADd enbrel Or REplace methotrexate (ADORE) trial, comparing etanercept monotherapy to etanercept+MTX over 16 weeks.32
Compared with the present results (from randomisation), the slightly larger utility gains seen after anti-TNF initiation in the ADORE trial and observational studies of established RA, are mainly a consequence of patients starting from a lower mean utility level.32–36 Apart from differences in patient characteristics, this may also be explained by the different study designs, with a substantial utility improvement occurring during the SWEFOT run-in period. Data on EQ-5D development in relation to start of conventional DMARD therapy remain sparse.
In the BeSt trial, a treatment strategy starting with MTX+IFX resulted in 1.41 mean accumulated QALYs over 2 years, significantly more than the 1.32 QALYs produced by the second-best approach commencing with MTX+SSZ+prednisone.17 Of interest though, QALY accumulation did not differ significantly between those groups during the first study year. None of the assessed strategies started with MTX+SSZ+HCQ. As compared with SWEFOT results, the higher QALY estimates of the BeSt trial are likely explained by the initial use of combination therapy as well as the frequent treatment adjustments allowed by the tight control protocol. Furthermore, the randomised SWEFOT population had all responded inadequately to 3 months of MTX monotherapy, signalling a more treatment-refractory disease.
Cost-utility analyses assessing treatments of RA are summarised in two recent reviews.13 ,18 Apart from that based on the BeSt trial,17 most use utilities indirectly calculated from other clinical measures, and model trial results over longer time horizons, rendering QALY estimates difficult to compare with our directly observed data.
Strengths and limitations
The SWEFOT trial addresses a central clinical question of early RA—whether the addition of a costly anti-TNF agent is superior to addition of further conventional DMARDs in MTX-refractory disease. Recurrent EQ-5D measurements allowed for a direct comparison of accumulated QALYs, something not previously reported for the treatment combinations under study, and, indeed, rarely available from any randomised controlled trial of RA.
The rationale behind the study design, including a run-in period of MTX monotherapy, has been discussed elsewhere, as has the limited use of low-dose glucocorticoids.8 ,9 The open-label design is an important limitation of the SWEFOT trial. This may have led some patients to abandon the SSZ+HCQ treatment arm, or physicians to assess with a bias. The use of blinded assessors (‘single-blinding’) was considered, but was rejected due to limited personnel resources at the smaller participating units. Other limitations, entailing a risk of type 2 error, include the relatively short time horizon, and that the trial was designed and powered mainly to detect a difference in treatment response rates, rather than in HRQoL-effects. At the 5% significance level, the SWEFOT sample size gives a power of 80% to detect a between-group difference of 0.14 QALYs (51 quality-adjusted life-days) over 21 months. The status, as a secondary outcome measure, may also explain the 18% missing utility data, which, to a part, have been countered by the inclusion of multiple sensitivity analyses to improve interpretability of results. Finally, substantial improvement in EQ-5D utility has been demonstrated already 2 weeks after anti-TNF initiation,33 while it remains unknown whether the same is true of MTX+SSZ+HCQ. Thus, since the first postrandomisaton measurement occurred after 3 months, potential between-group differences during this early period may have been missed.
Conclusions
Comparing addition of IFX or SSZ+HCQ to MTX in active early RA, the SWEFOT trial demonstrated HRQoL improvements in both treatment arms, but did not detect any significant between-group differences in utility or QALY gain during 21 months. Superior radiological outcome and adherence to treatment, but not better clinical response, was seen in the IFX group. While our results raise doubts about the short-term cost-effectiveness of adding IFX to MTX in this situation, further studies are needed to assess the long-term perspective.
Acknowledgments
We are indebted to all patients, colleagues and staff who made the SWEFOT trial possible.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Footnotes
-
Handling editor Tore K Kvien
-
Contributors All authors contributed to data interpretation and writing of the report. JAK contributed to quality control of data, data analysis, and had main responsibility for the manuscript drafting. MN contributed to quality control of data, data analysis, and had main responsibility for the manuscript drafting. JÅN contributed to quality control of data and data analysis. IFP, JB, RFvV and SE contributed to trial design, recruitment of patients, inclusion and follow-up visits, and were members of the trial's steering committee. PG contributed to trial design, recruitment of patients, inclusion and follow-up visits, quality control of data, data-analysis, and was a member of the trial's steering committee.
-
Funding The study was supported in part by a grant from the Swedish Rheumatism Association, and in part by an annual unrestricted grant from Schering–Plough Sweden (now MSD). The latter was used to support the study coordinator, a medical monitor and a research assistant, to arrange investigator meetings and for logistics associated with the radiographic analyses. The sponsors did not have any role in the study design, collection, analysis or interpretation of data, writing of the report, or in the decision to submit the article for publication.
-
Competing interests RFvV has received research support and honoraria from Abbott, GSK/HGS, MSD, Pfizer, Roche and UCB Pharma, IFP honoraria from Abbott, Pfizer and UCB Pharma, and MN honoraria from Pfizer. The rest of the authors declare no conflicts of interest.
-
Ethics approval The study was approved by the regional ethics committees of all participating centres.
-
Provenance and peer review Not commissioned; externally peer reviewed.