Adalimumab, etanercept, infliximab, rituximab and abatacept for the treatment of rheumatoid arthritis after the failure of a tumour necrosis factor inhibitor: a systematic review and economic evaluation

ACR response criteria

Modern clinical trials rely on composite end points such as the American College of Rheumatology (ACR) definition of improvement and the Disease Activity Score (DAS). The ACR response requires an improvement in the counts of the number of tender and swollen joints (using designated joints) and at least three items from the following: observer evaluation of overall disease activity; patient evaluation of overall disease activity; patient evaluation of pain; a score of physical disability [such as the Health Assessment Questionnaire (HAQ); see below]; and improvements in blood acute phase responses [e.g. erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP)].

Response is defined as ACR20, ACR50 or ACR70, where figures refer to the percentage improvement of these clinical measures. This creates a dichotomous outcome of responders and non-responders. Achieving an ACR20 response has been regarded as a low hurdle, but in clinical practice patients who achieve this hurdle often gain a worthwhile clinical response, especially in early RA. ACR response criteria are described in more detail in Appendix 1.

DAS response criteria

The DAS score is calculated using a formula that includes counts for tender and swollen joints, an evaluation by the patient of general health (on a scale of 0 to 100) and blood acute phase (usually a log of the ESR, but more recently using CRP). DAS response criteria are described in more detail in Appendix 1. Originally DAS was based on an assessment of 53 joints for tenderness and 44 joints for swelling. Disease Activity Score 28 (DAS28), based on an evaluation of 28 joints, is used widely in routine clinical practice, partly as a result of NICE guidance on use of TNF inhibitors. DAS28, like DAS, is a continuous scale with a theoretical range from 0 to 10. Thresholds have been suggested for the scale such that a score greater than 5.1 is regarded as indicating high disease activity, a score of less than 3.2 low disease activity and a score of less than 2.6 remission. 34,35 Achieving a DAS28 score of less than or equal to 3.2 and improving the score by greater than 1.2 is regarded to be a good response while achieving a score of less than or equal to 3.2 and improving by greater than 0.6 but less than 1.2 is regarded as a moderate response. Current NICE guidance for TNF inhibitors demands that patients should improve DAS28 by 1.2 in order to justify continuing treatment. It has been suggested that NICE guidance should be altered to allow patients who have attained a moderate response to continue treatment with a TNF inhibitor. 36

While DAS28 scores are a very valuable tool for assessing treatment responses in groups of patients, scores have important limitations when used for individual patient decisions. For example, DAS28 does not incorporate ankle and foot disease. Thus, a patient with disease localised here may not attain a sufficiently high score to be eligible for a TNF inhibitor. DAS28 also shows poor concordance with clinical judgement (based on a wide range of parameters). 37 In addition, the degree of measurement error in a test–retest reliability study indicates that the faith placed in DAS28 as the sole decision-making tool is misplaced. 38 For example, the smallest detectable difference which should be exceeded if a clinician is to be 95% confident that a change exceeds measurement variability was 1.32 for DAS28.

The Health Assessment Questionnaire

The HAQ is a family of questionnaires designed to assess functional capacity of patients. 39 The most widely used version of HAQ is the modified HAQ (MHAQ) score which comprises eight items such as an ability to dress, get in and out of bed, lift a cup, walk outdoors and wash. MHAQ is reported as an average score across the eight categories such that 0 indicates an ability to achieve tasks without difficulty and 3 reflects an inability to achieve tasks. Scores therefore range between 0 and 3 with an interval of 0.125. Low scores indicate better function. Care is needed in the interpretation of HAQ scores in published studies because there are several modifications to HAQ. The HAQ score is described in more detail in Appendix 1.

Radiographic measures

Radiographic outcomes are believed by many to be the most important outcome measure in RA. However, variation in joint inflammation has a more profound and immediate impact on disability compared with the slow and cumulative effect of radiographic damage on disability. 40 The most commonly used tools for assessing joint damage are the Sharp and Larsen methods and their modifications (see Appendix 1), which rely on evaluations of plain radiographs of hands and feet. Plain radiographs are rather insensitive to change but are cheap and widely available. A majority of patients show only mild or no progression on plain radiographs over periods of 1–2 years, highlighting one of their limitations in modern clinical trials. 41

Adalimumab

Adalimumab is a recombinant monoclonal antibody, made from human peptide sequences, which neutralises the biological functions of tumour necrosis factor alpha (TNFα) by binding to TNF cell-surface receptors. ADA is licensed for use in RA, juvenile idiopathic arthritis, psoriasis, psoriatic arthritis, ankylosing spondylitis and Crohn’s disease.

Etanercept

Etanercept is a combination protein consisting of the extracellular portion of two TNFα receptors (75-kDa TNF receptors) combined with a human fragment crystallisable (Fc) portion of the human immunoglobulin G1 (IgG1). ETN inhibits TNFα activity by binding soluble and cell-bound TNFα with high affinity and by competing with natural TNFα receptors. ETN is licensed for use in RA, psoriatic arthritis, psoriasis and ankylosing spondylitis.

Infliximab

Infliximab is a recombinant chimeric human–murine monoclonal antibody that binds soluble and membrane-bound TNFα thereby, inhibiting the functions of TNFα. IFX is licensed for use in RA, Crohn’s disease, ulcerative colitis, ankylosing spondylitis, psoriatic arthritis and psoriasis.

Other tumour necrosis factor inhibitors

Certolizumab pegol has been granted a marketing authorisation in the European Union (EU) for the treatment of moderate-to-severe RA. It is administered by subcutaneous injection. Certolizumab pegol was the subject of a separate NICE single technology appraisal (STA),58 with guidance published in February 2010. Golimumab is currently being assessed by the European Medicines Agency. A positive opinion has been given for the granting of marketing authorisation in RA. Golimumab has been referred to NICE, but the appraisal has been suspended because the manufacturer is not in a position to submit evidence to NICE.

Special precautions for use of tumour necrosis factor inhibitors

TNFα is a key component of host defence against Mycobacterium tuberculosis (MTB), especially by forming granulomas and preventing dissemination of mycobacteria. 59,60 Inhibition of TNFα increases the risk of MTB and other granulomatous diseases, such as those due to Listeria monocytogenes (a bacterium associated with food-borne diseases) and Histoplasma capsulatum (a fungus which, in endemic areas, causes lung disease in people with a compromised immune system). Recommendations for screening patients for tuberculosis (TB) before treatment have been published. 61 In the UK this is done most commonly by taking a medical history focusing on TB and a pre-treatment chest radiograph. Some centres also perform a tuberculin skin test,62 although interpretation of such tests is complicated by the UK’s previous vaccination programme for TB prevention and also the fact that many patients with RA respond poorly to tuberculin (possibly because of current immunosuppressive therapy but also because of the disease). 63

Routine monitoring of blood tests is not necessary for patients taking TNF inhibitors, but is needed for concomitantly used DMARDs such as MTX. TNF inhibitors can induce anti-nuclear and anti-double-stranded DNA antibodies in the blood of some patients treated with TNF inhibitors. These antibodies are associated with systemic lupus erythematosus (SLE), a potentially serious rheumatic disease. Cases of drug-induced SLE have been reported with TNF inhibitors, but are rare. 64

Rituximab

Rituximab is a chimeric monoclonal antibody which binds the CD20 cell surface marker found on B lymphocytes and depletes these cells. CD20 occurs on normal and malignant B lymphocytes (as in non-Hodgkin’s lymphomas). Normal plasma cells, an important component of host defence, and haematopoietic stem cells do not carry CD20. RTX is licensed for use in RA, non-Hodgkin’s lymphoma and chronic lymphocytic leukaemia. A small number of cases of progressive multifocal leucoencephalopathy, a rare but usually fatal demyelinating brain disease, have been reported in RA patients following RTX treatment. 65

Abatacept

Abatacept is a fusion protein consisting of CTLA-4 (cytotoxic T-lymphocyte-associated antigen-4) linked to a modified Fc portion of the human IgG1. ABT works by blocking activation of certain populations of T lymphocytes. ABT is currently licensed for use only in RA.

Tocilizumab

Tocilizumab was the subject of a separate NICE STA,66 with guidance published in August 2010. This guidance is likely to have a key impact on the treatment pathways considered in this review. TOC is a humanised monoclonal antibody that inhibits the activity of the cytokine interleukin-6 (IL-6). In the EU it is licensed for use only in moderate-to-severe RA patients who are intolerant, or have responded inadequately, to one or more DMARDs or TNF inhibitors. The drug is recommended for use in combination with MTX, but may be used alone in patients intolerant of MTX or for whom it is contraindicated. TOC is given by intravenous (i.v.) infusion over 1 hour once a month indefinitely.

Outcomes of interest

Selected outcomes of interest were specified in the review protocol, based upon the final scope issued by NICE for this technology appraisal. These were:

• treatment withdrawal (and reasons for withdrawal)

• ACR20, ACR50, ACR70

• disease activity (e.g. DAS28 or DAS)

• physical function (e.g. HAQ)

• joint damage/radiological progression (measured by a valid scoring system)

• pain

• fatigue

• extra-articular manifestations of the disease

• serious AEs (including death)

• other adverse effects potentially associated with treatment

• HRQoL.

Handling of data and presentation of results

Assessment of publication bias

All manufacturers of the interventions provided a list of all company-sponsored RCTs and other non-randomised or uncontrolled studies that are relevant for this appraisal. Requests of clarification of trial data that are potentially available but not reported in published papers were also made to the manufacturers of RTX and ABT.

The number of relevant studies for individual technology was too small to allow a formal assessment of publication bias.

Sensitivity analyses

The protocol specified that if evidence permits sensitivity analyses may be carried out taking into account the following factors:

• quality measures of studies such as blinding and randomisation

• factors associated with the characteristics of the study population

• factors associated with study design such as study duration and drug doses

• exclusion of data supplied as commercial/academic in confidence.

However, sensitivity analyses were not performed as no pooling of study results was undertaken.

Overview of evidence

Five studies in six publications92–97 met the inclusion criteria. No RCT was found. Four studies had comparator arms in which the patients were TNF inhibitor naive. 92–94,96 These arms were excluded here. One of the four studies93 also had a small comparator arm of nine patients, which did not meet the inclusion criteria of this report of greater than or equal to 20 patients for an arm to be included; thus, data from this arm were excluded.

One multicentre study was conducted in 12 countries, 11 of which were European, including the UK. Other studies were conducted in the UK, Sweden and Greece. It was unclear in which country one of the studies was conducted.

Sample sizes were small, ranging from 24 to 41 patients, that are relevant to the review in four studies; in one study there were 899 patients. Patients included all had previous treatment with either one or two TNF inhibitors, most frequently IFX. Reasons for switching TNF inhibitors were lack of efficacy only in one study,93 lack of efficacy or intolerance in two studies96,97 and lack of efficacy or AEs in two studies. 92,94 Details on ADA treatment were not reported in one study; in all the other studies ADA was given 40 mg subcutaneously every other week. Study duration ranged from 12 weeks to over 1 year. Further details are outlined in Table 3.

Patient characteristics

Data on patient characteristics can be found in Table 4. Characteristics of the patients included in the five studies varied in some aspects:

• Where reported 81%–92% were female.

• The mean age of the patients ranged from 50 to 57 years.

• The mean RA duration ranged from 11.6 to 16.6 years, but was not reported in two studies.

• The percentage of RF-positive patients was reported only in two studies (63% and 72%).

• Concomitant DMARDs: where reported 37%–85% patients were on MTX other DMARDs included CyA (4%), leflunonide (3%–13%), HCQ (3%) and AZA (1%).

• The percentage of patients on concurrent steroids was reported in two studies and ranged from 77% to 100%.

• Where reported the mean number of previous DMARDs used ranged from 2 to 5.

• The mean number of previous TNF inhibitors was greater than or equal to 1 in the biggest study, and it was exactly 1 in all the other studies.

• The HAQ scores ranged from 1.29 to 2.07 in four studies, but were not reported in one study.

• The mean DAS28 scores were very similar, ranging from 5.5 to 6.3.

• The mean number of tender and swollen joints at baseline was reported in three studies and ranged from 6 to 15 and from 8 to 11, respectively.

• Baseline ESR was reported in only one study (41.7 mm/hour) and CRP in only two studies (25.1 mg/dl and 43.9 mg/dl).

Quality assessment

The studies were all uncontrolled; four of them were prospective and one was retrospective. 93 Criteria for patient inclusion were clearly stated in four studies; however, in three of these it was unclear whether consecutive patients were included. The highest percentage of patients withdrawn from a study was 26.8%. There were no withdrawals from the retrospective study. In general, the higher withdrawal rates occurred with the longer follow-up durations. Further details on the quality assessment of the studies are given in Table 5.

Results

Tables 6 and 7 show what outcomes were measured in each study. Outcomes in Table 6 are reported and discussed in the main text of this report and those in Table 7 are reported in Appendix 10 only.

Withdrawals

Withdrawal rates are presented in Figure 1. At 3 months, the percentage of patients withdrawn was very similar in the two studies that reported this outcome (9.9% and 9.8%). No patients withdrew in a retrospective study during 6 months. Withdrawal rates reported at 1 year were 12.5% and 26.8% in the two studies that reported this outcome. The percentage of patients withdrawn owing to lack of efficacy and owing to AEs at 3 months was reported only in the biggest study and was 2.9% and 5.6%, respectively. The percentage of patients withdrawn owing to lack of efficacy and owing to AEs at 12 months was measured in two studies: 8.3% and 17.1% withdrew because of lack of efficacy and 8.3% and 14.6% withdrew because of AEs.

FIGURE 1.

Adalimumab: withdrawals from studies by reason. LCI, lower confidence interval; UCI, upper confidence interval.

One study92 reported withdrawal data based on all 70 patients, including 44 patients who received a prior TNF inhibitor as well as TNF inhibitor-naive patients; the withdrawal data were not included in this report.

ACR20 response

The ACR20 response was assessed in four studies (Figure 2). Two studies assessed it at 3 months and the response was achieved by around half of the patients (46% and 60%). In the other two studies, the percentage of patients who achieved ACR20 response was 70% at 6 months and 75% at 12 months.

FIGURE 2.

Adalimumab: ACR (20, 50, 70) responses. LCI, lower confidence interval; UCI, upper confidence interval.

DAS28

One study measured DAS28 at 3 and 6 months and another study at 12 months; the mean scores were 4.5, 4.2 and 3.2, respectively. See Figure 3 for details. The mean changes from baseline to 3 months and to 6 months [note: in the Bennett et al. study92 it was measured after mean treatment duration of 8.5 (range 1–19) months], were reported in four studies including the biggest study. They all showed that treatment with ADA significantly improved DAS28 scores (mean changes ranged from –1.30 to –1.90). See Figure 4 for details.

FIGURE 3.

Adalimumab: DAS28 scores. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

FIGURE 4.

Adalimumab: mean changes in DAS28 scores. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

EULAR response

Two studies reported EULAR response at 3 months; most of the patients had a good/moderate response (76% and 78%) and 17%–23% had a good response. The Bennett et al. study92 measured EULAR response after a mean treatment duration of 8.5 months (range 1–19 months); the response rate was 65%, of whom 46% had a moderate response and 19% had a good response. See Figure 5 for details.

FIGURE 5.

Adalimumab: EULAR response. (a) Nikas et al. 94 only reported ‘EULAR response’ without providing further detail. LCI, lower confidence interval; UCI, upper confidence interval.

Health Assessment Questionnaire

Mean change in HAQ score was reported in three studies. Figure 6 shows that the mean HAQ score measured at 3 months in two studies, including the biggest study, and at mean 8.5 months (range 1–19 months) in the Bennett et al. study92 in all cases showed a significant decrease, ranging from –0.21 to –0.48, with the largest improvement observed in the biggest study.

FIGURE 6.

Adalimumab: mean change in HAQ scores. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

Summary

Five uncontrolled studies were identified for the assessment of effectiveness of ADA in comparison with standard care. Follow-up duration ranged from 3 months to over 1 year. All patients included in the studies were generally similar. The main results are summarised in Table 8.

Overview of evidence

No RCT was found. Seven uncontrolled observational studies98–104 were identified that assessed efficacy of ETN.

In the studies by Buch et al. 99 and Bingham et al. 104 lack of efficacy was the primary reason for switching to ETN. In studies by Haraoui et al. ,98 Cohen et al. 100 and Buch et al. 101 patients discontinued IFX owing to a lack of efficacy or safety. In Iannone et al. ,102 patients had to have responded to prior IFX treatment but later switched to ETN due to side effects. The patient population in this study was therefore different from the other studies. In Laas et al. ,103 patients discontinued IFX owing to a lack of efficacy, safety or non-medical reasons. The group of patients who discontinued IFX owing to non-medical reasons (46%, 23/49) had responded well to IFX, but switched to ETN for practical reasons such as convenience (e.g. no need for hospital visit to receive infusion). Two studies99,101 were carried out at the same centre (Leeds Teaching Hospitals) in the UK. These studies were described separately in this section although it is possible that patients included in Buch et al. 200599 were a subgroup of the cohort included in Buch et al. 2007. 101 The other studies were carried out in France,100 Italy,102 Finland103 and the USA. 98 One study104 was a multicentre study that enrolled patients from both the USA and Canada. The length of follow-up varied from 12 weeks to more than 9 months. Further details are provided in Table 9.

Patient characteristics

Full details of patients’ characteristics are reported in Table 10. The number of patients included in the studies varied from 24 to 201. Patient characteristics differed across the seven studies:

• Where reported, the percentage of female patients ranged from 60% to 88%.

• Where reported, the mean age ranged from 49 to 57 years.

• Where reported, the mean disease duration ranged from 8.3 to 12.2 years.

• Where reported, the percentage of RF-positive patients ranged from 44% to 75%.

• Where reported concomitant DMARDs were: 88–99% MTX, other DMARDs included HCQ (9%) and sulfasalazine (5%).

• Where reported, 40%–88% of patients were receiving corticosteroids.

• Where reported, the mean/median number of previously used conventional DMARDs varied from 4.1 to 7.0.

• All the studies included patients previously treated with IFX.

• Where reported the mean baseline HAQ ranged from 0.90 to 2.16.

• The mean baseline DAS28 score ranged from 5.6 to 6.6.

• One study102 reported baseline DAS44 (mean value was 2.7).

• Where reported, the mean number of tender and swollen joints was variable (tender: 10.0–17.8 and swollen: 8.6–14.3).

• Baseline ESR was reported only in two studies (21 mm/hour and 30 mm/hour).

• Where reported, CRP ranged from 0.6 (median) to 6.2 (mean) mg/dl.

The baseline values listed in Table 10 for Iannone et al. 102 were measured 8 weeks before patients switched from IFX to ETN (while they were still responding to IFX) and thus the values may not be comparable with those from the other studies.

Quality assessment

All the seven studies were uncontrolled studies. Five were prospective98,99,101,103,104 and two were retrospective. 100,102 Full details of the quality assessment are reported in Table 11. With the exception of Laas et al. ,103 studies stated clearly their inclusion criteria. Only Buch et al. 200599 and Buch et al. 2007101 clearly stated that consecutive patients were included in the studies; this information was unclear in Bingham et al. 104 and Haraoui et al. 100 Only one study104 reported the percentage of patients lost to follow-up (0.5%).

Results

Table 12 and Table 13 show what outcomes were measured in each study. Outcomes in Table 12 are reported and discussed in the main text and in Table 13 are reported in Appendix 10 only.

Withdrawals

Five out of seven studies reported withdrawals and the reasons for withdrawing from treatment. The percentages and reasons for withdrawing from the study after commencing ETN are shown in Figure 7. The percentage of patients who withdrew for any reason ranged from as low as 6.5% (at 3 months) to as high as 58.3% (at 12 months). The percentage of patients who withdrew because of AEs and lack of efficacy ranged from 0% to 16.3% and from 0% to 29.2%, respectively.

FIGURE 7.

Etanercept: withdrawals in the studies by reason. LCI, lower confidence interval; UCI, upper confidence interval.

ACR20 response

The ACR20 response was assessed in four studies (Figure 8). The percentage of patients treated with ETN after IFX failure that achieved ACR20 response after 3 months ranged from 37.5% to 72.0%.

FIGURE 8.

ACR20: responses in patients receiving etanercept. LCI, lower confidence interval; UCI, upper confidence interval.

ACR50 response

The ACR50 response was assessed in five studies, but results from Iannone et al. 102 are not presented here, as explained above (Figure 9). The proportion of patients achieving a ACR50 response after 3 months ranged from 18.4% to 64.0%.

FIGURE 9.

ACR50: responses in patients receiving etanercept. LCI, lower confidence interval; UCI, upper confidence interval.

ACR70 response

The ACR70 response was assessed in five studies, but results from Iannone et al. 102 are not presented here, as explained above (Figure 10). The proportion of patients achieving a ACR70 response after 3 months ranged from 4.2% to 20.0%.

FIGURE 10.

ACR70: responses in patients receiving etanercept. LCI, lower confidence interval; UCI, upper confidence interval.

DAS

Figure 11 presents the mean changes from baseline in DAS. Four studies100,101,103,104 reported using DAS28. The mean decrease in DAS28 ranged from 1.47 to 1.80 at 3 months. One study102 reported no statistically significant decrease in DAS28 score from baseline at 12 months [mean change = –0.47, 95% confidence interval (CI) –1.06 to 0.12]. One study102 reported DAS calculated based on 44 joints (DAS44). Iannone et al. 102 found no statistically significant differences in DAS44 scores when results for 16 and 24 weeks were compared with the baseline value.

FIGURE 11.

Etanercept: mean changes from baseline in DAS. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

EULAR response

Three studies reported EULAR responses. Figure 12 shows the proportion of patients treated with ETN who achieved a good and good-to-moderate EULAR response after IFX failure. The percentage of patients who achieved a good score EULAR ranged from 12.5% to 45.8% at 3 months. The percentage of patients who achieved a good-to-moderate EULAR response ranged from 58.2% to 61.1% at 3 months.

FIGURE 12.

Etanercept: EULAR. LCI, lower confidence interval; UCI, upper confidence interval.

Health Assessment Questionnaire

Three studies reported mean changes from baseline in HAQ score (Figure 13). In Haraoui et al. ,98 the change in HAQ score was –0.45. However, it was not reported whether this change was statistically significant. For Iannone et al. ,102 the value of HAQ remained largely unchanged at 16 weeks (0.90) and 24 weeks (0.75) compared with the baseline value (0.75). In Bingham et al. ,104 there was a mean decrease in HAQ score of 0.35 at 3 months; this corresponds to a 22% decrease from baseline. This change was statistically significant.

FIGURE 13.

Etanercept: mean change from baseline in HAQ. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

One study98 reported the percentage of patients who achieved minimal clinically important difference (MCID) in physical function (Figure 14). MCID was defined as a change of at least 0.22 in HAQ score. The percentage of patients who achieved MCID was 52%. Forty per cent of patients experienced an improvement in physical function of at least twice the value considered to represent MCID.

FIGURE 14.

Etanercept: MCID physical function. LCI, lower confidence interval; UCI, upper confidence interval.

Serious adverse events

Two studies reported serious AEs (Figure 15). Haraoui et al. 98 reported that no serious AEs occurred during the study. Bingham et al. 104 found that 5% of the patients experienced a serious AE during the study period.

FIGURE 15.

Etanercept: reported serious AEs. LCI, lower confidence interval; UCI, upper confidence interval.

Infection and serious infection

Three studies reported infection and serious infection (Figure 16). One study104 reported that two patients (1%) experienced serious infections. The percentage of patients treated with ETN who reported any infection ranged from 4.1% to 8.3%.

FIGURE 16.

Etanercept: reported infection or serious infection. LCI, lower confidence interval; UCI, upper confidence interval.

Summary

For the assessment of effectiveness of ETN, seven uncontrolled studies were identified. Follow-up duration ranged from 12 weeks to over 9 months. Patients included in the studies were generally similar. The main results are summarised in Table 14.

Overview of evidence

Three studies were identified that assessed IFX in comparison with standard care: one uncontrolled prospective study107 and two uncontrolled retrospective studies. 105,106 (Note: the study by Yazici et al. 107 had a control group consisting of patients who were given their first biologic drug. This control group was not relevant to this report and, therefore, the study was utilised as uncontrolled.)

All included patients had tried one TNF inhibitor (ETN) before. Reasons for discontinuation included lack of efficacy, toxicity drug shortage, patient concerns about safety and thrombocytopenia.

All studies were conducted in the USA. Duration of follow-up was unclear in all the three studies.

Further details are provided in Table 15.

Patient characteristics

All three studies were rather small, with the number of patients treated with IFX ranging from 20 to 24. They provided very little information about the baseline characteristics of included patients. However, based on the available information there might have been some baseline differences between study populations (Table 16).

• In two studies the percentage of female participants ranged from 60% to 90%; Yazici et al. 107 did not provide any information.

• In two studies the mean age was 48 years and 61 years; it was not reported in Ang et al. 105

• In two studies disease duration was 9.3 years and 13.4 years; it was not reported in Ang et al. 105

• In two studies 34%–65% of patients were RF positive; no information was provided in Yazici et al. 107

• In Ang et al. 105 62% of patients were receiving MTX and 31% LEF; in Hansen et al. 106 all patients were receiving LEF and some of them also other DMARDs (AZA, sulfasalazine, MTX and prednisone); Yazici et al. 107 did not report concomitant DMARDs.

• Only one study (Hansen et al. 106) reported that 75% of patients were receiving concomitant prednisone.

• Two studies reported the number of previous DMARDs – it ranged from 0 to over 5; it was not reported in Hansen et al. 106

• Patients had tried one previous TNF inhibitor (ETN) in all three studies.

• None of the studies reported the baseline HAQ or DAS score.

• Only one study106 reported that patients had a mean of 14 tender and 14 swollen joints at baseline.

• Only one study106 reported the baseline ESR (mean 13 mm/hour) and CRP (mean 23.8 mg/dl).

Quality assessment

Of the three identified studies, two were uncontrolled retrospective analyses. One study was uncontrolled and prospective. None of the studies reported inclusion criteria clearly. It was unclear if consecutive patients were included in Yazici et al. 107 and this item was not applicable to retrospective studies. A total of 28.6% were withdrawn from Yazici et al. 107 and this percentage was unclear in the remaining two studies. Details of the quality assessment are reported in Table 7.

Results

Table 18 indicates which of the outcomes reported in the main text of the report were assessed in individual studies and Table 19 provides similar information for outcomes described in Appendix 10 only.

Ang et al. 105 reported results in a way that made it impossible to utilise them in this report (correlations between response to IFX and ETN).

Withdrawals

Withdrawal for any reason was assessed only in Yazici et al. ,107 withdrawal because of lack of efficacy only in Hansen et al. 106 and withdrawal because of AEs was not assessed in any of the studies. Details are reported in Figure 17. Yazici et al. 107 reported that 28.6% of patients were withdrawn from the study for any reason (follow-up unclear). Ten per cent of patients were withdrawn from Hansen et al. 106 owing to lack of efficacy (follow-up unclear).

FIGURE 17.

Infliximab: withdrawals. LCI, lower confidence interval; NR, not reported; UCI, upper confidence interval.

Infections/serious infections

Details of infections are reported in Figure 18. Fifteen percent of patients in Hansen et al. 106 experienced an infection (follow-up was unclear). No other studies reported infections. Serious infections were not reported in any of the studies.

FIGURE 18.

Infliximab: infections. LCI, lower confidence interval; NR, not reported; UCI, upper confidence interval.

Infliximab in comparison with an ongoing biologic agent

One RCT [open-label, pilot protocol of patients with rheumatoid arthritis who switch to infliximab after an incomplete response to etanercept (OPPOSITE133)] was identified that compared IFX with ongoing ETN. Although the study met the inclusion criteria of the systematic review, this comparison was not considered relevant to this report and, therefore, the study was not analysed.

It was a multicentre randomised trial and included 27 patients who had active RA and had an ‘incomplete response to etanercept’. Patients were randomised either to discontinue ETN and receive IFX (13 patients) or to continue ETN treatment (14 patients). The follow-up duration was 30 weeks. Data were collected on outcomes including ACR response, HAQ, radiological progression, serum biomarker levels and safety.

Summary

Three studies compared IFX with standard care: two uncontrolled retrospectiv105,106 and one uncontrolled prospective Yazici et al. studies. 107 They included small numbers of patients ranging from 20 to 24. Follow-up was unclear in all of them. There was little information about baseline characteristics; however, it seems that there may be some, if small, differences between studies. The main results of included studies are summarised in Table 20.

Overview of evidence

This section reports on studies that evaluated the use of TNF inhibitors as a class after the failure of the first one. No RCT was found. One controlled121–123 and six uncontrolled observational studies108–113 were identified. In Finckh et al. 136,137 lack of efficacy was the primary reason for switching TNF inhibitors. In Hyrich et al. ,121–123 Gomez-Reino et al. 108 and Blom et al. 113 patients switched to another TNF inhibitor because of a lack of efficacy or AEs. In Hjardem et al. ,110 Duftner et al. 111 and Karlsson et al. 112 patients switched TNF inhibitorsowing to lack of efficacy or AEs or for other reasons. The reason for changing from one TNF inhibitor to another was unclear in Solau-Gervais et al. 109 Hyrich et al. 121–123 used data from the BSRBR. The other studies were carried out in Switzerland, Spain, France, Denmark, Austria, Sweden and the Netherlands. The length of follow-up ranged from 3 months to up to 4 years. Further details are provided in Table 21.

Patient characteristics

Full details of patients’ characteristics are reported in Table 22. The number of patients included in the studies ranged from 70 to 818. Patient characteristics were generally similar across the eight studies:

• The percentage of female patients ranged from 67% to 89%.

• Where reported, the mean age ranged from 51 years to 58 years.

• Where reported, the mean disease duration ranged from 8.0 years to 14.7 years.

• Where reported, the percentage of RF-positive patients ranged from 51.5% to 81%.

• Where reported, 61%–75% patients were on MTX; 55%–68% of patients were receiving corticosteroids.

• Where reported, the mean number of previously used conventional DMARDs varied from 4.0 to 4.7.

• Where reported, studies included patients who previously tried IFX, ETN and ADA.

• Where reported, the mean baseline HAQ ranged from 1.4 to 1.9.

• Where reported, the mean DAS28 score ranged from 4.1 to 6.5.

• The mean number of tender and swollen joints was reported only in one study (tender 9.3 and swollen 8.4).

• The mean baseline ESR was reported in one study and was 36 mm/hour.

• The baseline CRP was reported in one study and was 2.8 mg/dl.

Quality assessment

One study was controlled. 121–123 Two studies108,109 were uncontrolled and prospective. Four studies were uncontrolled and retrospective. 110–113 Finckh et al. 136,138 was a non-randomised comparative study (TNF inhibitors vs RTX). This section presents data only for TNF inhibitors. Full details of the quality assessment are reported in Table 23. Most studies stated clearly their inclusion criteria. The inclusion criteria were unclear in two studies. 108,110 It was unclear in most studies whether consecutive patients were included in the study. Nearly one-third (140/477) of patients who met the study inclusion criteria were excluded from Karlsson et al. 112 because of dropouts/missing response data at 3 months. The exclusion of these patients may partly account for the higher rates of EULAR responses observed in this study compared with other studies (see Figure 25). The percentage of patients withdrawn was clearly reported in two studies.

Results

Tables 24 and 25 state which outcomes were measured in each study and whether they are reported in the main text or Appendix 10 of this report.

Withdrawals

Two studies reported withdrawals together with reasons for withdrawing treatment (Figure 19). The percentage of patients who withdrew for any reason ranged from 7.6% (at 3 months) to 38.6% (at 12 months). The percentage of patients who withdrew because of AEs ranged from 6.1% (at 3 months) to 10.2% (at 6 months). At 12 months, the percentage of patients who withdrew because of AEs ranged from 6.0% to 14.7%. The percentage of patients who withdrew because of lack of efficacy ranged from 1.5% (at 3 months) to 22.6% (at 12 months).

FIGURE 19.

Tumour necrosis factor inhibitor as a class: withdrawals from the studies by reason. LCI, lower confidence interval; UCI, upper confidence interval.

One study reported 1-year drug survival108 (probability of staying on treatment at 12 months) of 0.79 (95% CI 0.74 to 0.83). Two studies reported median drug survival. 110,111 Hjardem et al. 110 and Duftner et al. 111 reported that the median drug survival was 37 weeks and 8.0 months (range 0–43.7 months), respectively.

ACR20 response

The ACR20 response was assessed in one study (Figure 20). Karlsson et al. 112 reported that at 3 months ACR20 response rate was 49.0% (95% CI 43.5% to 54.4%).

FIGURE 20.

TNF inhibitors as a class: ACR20 response. LCI, lower confidence interval; UCI, upper confidence interval.

ACR50 response

The ACR50 response was assessed in one study (Figure 21). Karlsson et al. 112 reported that at 3 months ACR50 response rate was 25.8% (95% CI 21.2% to 30.8%).

FIGURE 21.

TNF inhibitors as a class: ACR 50 response. LCI, lower confidence interval; UCI, upper confidence interval.

ACR70 response

The ACR70 response was assessed in one study (Figure 22). Karlsson et al. 112 reported that at 3 months ACR70 response rate was 7.1% (95% CI 4.6% to 10.4%).

FIGURE 22.

TNF inhibitor as a class: ACR70 response. LCI, lower confidence interval; UCI, upper confidence interval.

DAS28

Three studies reported mean changes from baseline in the DAS28 score (Figure 23). The mean decrease in DAS28 ranged from 0.86 to 1.00 at 3 months and from 0.88 to 0.92 at 6 months. Two studies112,113 reported low disease activity (DAS28 less than 3.2) (Figure 24). At 3 months the percentage of patients with low disease activity ranged from 14.2% to 29.1%. One study reported DAS28 remission (DAS28 less than 2.6) (Figure 24). Karlsson et al. 112 reported that 15.4% (95% CI 11.7% to 19.7%) of patients were in remission.

FIGURE 23.

TNF inhibitors as a class: mean changes from baseline in DAS28. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval.

FIGURE 24.

TNF inhibitors as a class: low disease activity (DAS28 < 3.2) and remission (DAS28 < 2.6). LCI, lower confidence interval; UCI, upper confidence interval.

EULAR response

Three studies110,112,113 reported the percentage of patients who achieved good and good-to-moderate EULAR responses (Figure 25). The percentage of patients who achieved good EULAR response ranged from 8.6% to 22.8% at 3 months and was 9.1% at 6 months. The percentage of patients who achieved good-to-moderate EULAR response ranged from 31.5% to 64.7% at 3 months. Only one study reported good-to-moderate EULAR response at 6 months (32.5%).

FIGURE 25.

TNF inhibitors as a class: EULAR response rates. LCI, lower confidence interval; UCI, upper confidence interval.

Health Assessment Questionnaire

Only one study reported mean changes from baseline in HAQ score (Figure 26). Hyrich et al. 121–123 compared patients who discontinued TNF inhibitor within the first 12 months and did not start a subsequent TNF inhibitor or other biologic drug during the next 12 months (‘stoppers’) with patients who stopped their first TNF inhibitor within the first 12 months of therapy because of the lack of efficacy, but started a second TNF inhibitor during the subsequent 12 months (‘switchers’). The mean change in HAQ score was adjusted for differences in age, gender, disease duration, HAQ score at first failure, DAS28 at start of first TNF inhibitor and DAS28 score at first failure. ‘Switchers’ (adjusted mean change = –0.11, 95% CI –0.18 to –0.04) had significantly greater improvement in HAQ score than ‘stoppers’ (Figure 26).

FIGURE 26.

TNF inhibitors as a class: adjusted mean change from baseline in HAQ score. Adjusted for age, gender, disease duration, HAQ score at first failure, DAS28 at start of first TNF inhibitor and DAS28 score at first failure. LCI, lower confidence interval; UCI, upper confidence interval.

Serious adverse events

Only one study reported serious AEs (Figure 27). Hjardem et al. 110 reported that 6.0% (95% CI 3.3% to 9.8%) of the patients experienced a serious AE during the study period.

FIGURE 27.

TNF inhibitors as a class: serious adverse events. LCI, lower confidence interval; UCI, upper confidence interval.

Infection and serious infection

Two studies reported infection and serious infection (Figure 28). At 3 months the percentage of patients who experienced infection ranged from 27.2% to 28.1%. One study111 reported that 13.9% (95% CI 9.1% to 19.9%) of the patients experienced serious infections at 3 months.

FIGURE 28.

TNF inhibitors as a class: infections and serious infections. LCI, lower confidence interval; UCI, upper confidence interval.

Summary

For the assessment of effectiveness of TNF inhibitors as a class after failure of the first TNF inhibitor, one non-randomised comparative and seven uncontrolled studies were identified. Follow-up duration ranged from 3 months to 4 years. Patients included in the studies were generally similar. The main results are summarised in Table 26.

Overview of evidence

Seven studies were identified that assessed RTX: one RCT [randomised evaluation of long-term efficacy of rituximab in rheumatoid arthritis (REFLEX)124–126] and six uncontrolled studies. 114–118,137,139 One of these (Finckh et al. 136,137) contained a comparative arm with an alternative TNF inhibitor; the comparative data are described in the section Evidence from comparative studies. One study116 included data from patients of whom nearly half were previously TNF inhibitor naive. Only data reported separately for those who had a previous TNF inhibitor were included in this report. In another study,118 at 6 months, 17 patients (including five who were TNF inhibitor naive at original baseline) started a second course of TNF inhibitor; data for this group of patients were excluded from the report.

Data from one cohort analysis of the REFLEX RCT extension139 and one pooled analysis of all RTX development studies from the MS are also described. The REFLEX extension139 was a long-term follow-up analysis of repeated treatment data of the original RCT: it included patients who had responded to an initial course of RTX during the RCT and received open-label treatment with the same RTX regimen for up to three repeat treatment courses. (Note: responding patients in the initial REFLEX RCT124–126 after reaching the primary end point at week 24 requiring further courses of RTX treatment entered the open-extension study.) Patients from the placebo arm of the RCT were also included and received their first course of RTX within the extension study. A total of 480 patients from the RCT (308 from the RTX arm and 172 from the placebo arm) entered the extension phase.

The manufacturer’s pooled analysis combined data from patients of the REFLEX RCT,124–126 together with data from its open-label extension study, and from other studies in manufacturer’s RTX development programme. (Note: data were pooled for patients who only received the expected licensed dose of RTX two × 1,000 mg plus MTX regimen for first and subsequent courses and who received prior TNF inhibitor therapy.) It is unclear how many patients from the REFLEX trial124–126 were included in the pooled analysis.

The Keystone et al. uncontrolled study116 also reported data for up to two treatment courses; these data are presented with those from the REFLEX extension139 and the RTX pooled analysis.

The REFLEX trial was a multicentre RCT conducted in 114 counties in the USA, Europe, Canada and Israel. Of the six uncontrolled studies, one was conducted in Switzerland, one in the UK, one in Sweden, one in the Netherlands and one in France. For the studies included in the Keystone et al. analysis,116 and for those included in the manufacturer’s pooled analysis, except the REFLEX trial,139 it is unclear in which country the studies were conducted.

Further details are provided in Table 27.

Patient characteristics

Data on patient baseline characteristics can be found in Table 28. Patient characteristics were not reported for the manufacturer’s pooled analysis and were not reported separately for the patients who had previously received a TNF inhibitor in the Keystone et al. analysis. 116

The number of patients included in the REFLEX RCT124–126 was 517 and ranged from 20 to 155 in the six uncontrolled studies. Where reported, characteristics of the patients included in the studies varied in some aspects, but were generally similar:

• The percentage of female patients ranged from 77% to 86%.

• The mean age ranged from 52 to 58 years in four studies and the median age in two studies was 54–55 years.

• The mean disease duration ranged from 10 to 15 years in four studies and the median age in two studies was 12–16 years.

• The percentage of RF-positive patients ranged from 79% to 90% and was lowest in the REFELX study; one study and both analyses from the MS did not report this.

• Concomitant DMARDs were reported in five studies: 30%–100% patients were on MTX; all the patients in the REFLEX RCT124–126 were on concomitant DMARDs.

• The proportion of patients who were receiving concurrent steroids ranged from 55% to 100%; one study did not report this.

• The mean number of previously used conventional DMARDs reported in three studies ranged from 2.5 to 4.2 and median reported in the other two ranged from 3 to 4.

• Where reported, the mean number of previous TNF inhibitors was 1 or greater than 1 and the median number reported in two uncontrolled studies was 2.

• The mean baseline HAQ was reported only in the REFELX study was 1.9 and the median baseline HAQ reported in two uncontrolled studies ranged from 1.6 to 2.6.

• Where reported, the mean DAS28 score ranged from 5.0 to 6.9 and it was the highest in the REFELX study.

• The mean number of tender joints was 34 and swollen joints was 23 in the REFLEX trial; 124–126 the median number was 26 and 13 respectively in Jois et al. ;115 other studies did not report the baseline number of tender and swollen joints.

• The baseline mean ESR was 48 mm/hour in REFLEX124–126 and the median value 37 mm/hour and 56 mm/hour in other two studies.

• The mean CRP was 3.7 mg/dl in the REFELEX trial and 3.2 mg/dl in another study; median CRP was 1.9 and 2.9 in the other two studies.

Quality assessment

Results

Tables 31 and 32 present what outcomes were measured in the studies. Outcomes in Table 31 are reported and described in the main text of this report and those in Table 32 are reported in Appendix 10 only. Outcome data from the RTX arm in the RCT are also included in the section on uncontrolled studies for comparison purposes. As data from the REFLEX extension cohort139 and the RTX pooled analyses were analysed according to RTX treatment courses, the results of these analyses are described separately from the results of the uncontrolled studies.

Withdrawals

Randomised controlled trial

Withdrawal rates are presented in Figure 29. At week 24, there were significantly fewer withdrawals for any reason in the RTX arm than in the placebo arm of the REFLEX RCT124–126 (RR = 0.39, 95% CI 0.29 to 0.51). Risk of withdrawal because of AEs tended to be higher in the RTX than in the placebo group; however, the difference was not statistically significant (RR = 2.71, 95% CI 0.58 to 12.65).

FIGURE 29.

Rituximab: withdrawals in the REFLEX RCT124–126 at 24 weeks by reason.

Non-randomised controlled trials

Withdrawal rate for any reason at 6 months was reported in only one uncontrolled study115 and it was 10%. For comparison, 17.9% of patients in the RTX arm in the REFLEX RCT124–126 withdrew at 6 months for any reason and 2.6% withdrew because of AEs (Figure 30). In one study114 the total number of patients withdrawn by reason was not reported, but it was stated that one patient discontinued RTX treatment after a second infusion (week 4) because of severe headache and stomach pain. Two patients who had a medical history of chronic myocardial ischaemia died of myocardial infarction, one within the first month and the other at 13 months.

FIGURE 30.

Rituximab: withdrawals in uncontrolled studies by reason. LCI, lower confidence interval; UCI, upper confidence interval.

ACR20

Randomised controlled trial

In the REFLEX trial,124–126 the percentage of patients who achieved ACR20 response at week 24 in the RTX group was nearly three times that in the placebo group and the difference was statistically significant (RR = 2.85, 95% CI 2.08 to 3.91). At week 48, the response rate based on observed data (of a smaller number of patients) favoured the RTX group, but the difference was not significant (RR = 1.53, 9% CI 0.84 to 2.76); when analysed based on non-responder imputation data, the response rate in the RTX group was nearly five times of that in the placebo group and the difference was significant (RR = 4.92, 95% CI 2.40 to 10.09). Details can be found in Figure 31.

FIGURE 31.

Rituximab: ACR20 response in the REFLEX study124–126 (observed data and non-responder imputation data are from MS).

Non-randomised controlled trials

In the Keystone et al. 116 pooled analysis, 24 weeks after the first course of RTX, 65.2% patients had an ACR20 response, while in the RTX arm of the REFLEX trial124–126 the figure was 51% (Figure 32). None of the other uncontrolled studies reported ACR20 responses.

FIGURE 32.

Rituximab: ACR20 response in cohorts 24 weeks after first course of RTX. LCI, lower confidence interval; UCI, upper confidence interval.

ACR50

Randomised controlled trial

At week 24 in the REFLEX trial,124–126 the percentage of ACR50 responders in the RTX group was nearly five and a half times that of the placebo group and the difference was statistically significant (RR = 5.40, 95% CI 2.87 to 10.16). The effect persisted at week 48, analysed based on either observed data (RR = 4.11, 95% CI 1.06 to 15.85) or non-responder imputation data, and based on non-responder imputation data the response rate in the RTX group was over 13 times that of the placebo group (RR = 13.23, 95% CI 3.23 to 54.20). Details are presented in Figure 33.

FIGURE 33.

Rituximab: ACR50 response in the REFLEX study124–126 (the observed data and non-responder imputation data were from MS).

Non-randomised controlled trials

In the Keystone et al. 116 pooled analysis, 24 weeks after the first course of RTX, ACR50 response was observed in 32.9% patients, while in the RTX arm of the REFLEX trial124–126 it was 26.8% (Figure 34). None of the other uncontrolled studies reported ACR50 response.

FIGURE 34.

Rituximab: ACR50 response in uncontrolled studies 24 weeks after the first course of RTX. LCI, lower confidence interval; UCI, upper confidence interval.

ACR70

Randomised controlled trial

At week 24 the percentage of patients achieving ACR70 response in the RTX group in the REFLEX trial124–126 was over 12 times of that of the placebo group and the difference was statistically significant (RR = 12.14, 95% CI 2.96 to 49.86). At week 48 the beneficial effect of RTX was not significant based on observed data for a much smaller patient group (RR = 3.37, 95% CI 0.47 to 24.2), but was significant based on non-responder imputation data (RR = 10.86, 95% CI 1.45 to 81.24). See Figure 35 for details.

FIGURE 35.

Rituximab: ACR70 response in the REFLEX study124–126 (those based on observed data and non-responder imputation data were from manufacturer’s submission).

Non-randomised controlled trials

In the Keystone et al. pooled analysis the percentage of ACR70 responders 24 weeks after the first course of RTX was 12.3%; it was similar to that reported in the RTX arm of the REFLEX trial124–126 (12.1%) (Figure 36). No other uncontrolled study reported ACR70 responses.

FIGURE 36.

Rituximab: ACR70 response in uncontrolled studies 24 weeks after the first course of RTX.

EULAR response

Randomised controlled trial

EULAR responses are presented in Figures 37 and 38. In the REFLEX trial,124–126 at week 12 the percentage of patients achieving good or moderate response in the RTX group was over twice that of the placebo group, as was the percentage achieving a good response; the effects were statistically significant (RR = 2.02, 95% CI 1.64 to 2.49 and RR = 2.23, 95% CI 1.12 to 4.41, respectively). At week 24 the percentage of patients achieving a EULAR good or moderate response in the RTX group was nearly three times that of the placebo group and the effect was significant (RR = 2.96, 95% CI 2.25 to 3.89); the rate of achieving a good response was also higher in the RTX group and the difference was statistically significant (RR = 7.59, 95% CI 2.77 to 20.77).

FIGURE 37.

Rituximab: EULAR response at week 12 in the REFLEX study124–126 (data from the manufacturer’s submission).

FIGURE 38.

Rituximab: EULAR response at week 24 in the REFLEX study. 124–126

Non-randomised controlled trials

None of the uncontrolled studies reported EULAR response at 3 months, whereas three reported it at 6 months. At 3 months in the RTX arm of the REFLEX RCT,124–126 68.5% of patients had moderate or good response, with 11.1% having achieved a good response; at 6 months the rates remained similar (64.8% and 15.1%, respectively). At 6 months the percentage of good or moderate EULAR responders in four uncontrolled studies including the RTX arm of the REFLEX trial124–126 ranged from 64.8% to 82%, and the good response rate ranged from 15.1% to 36%. The REFLEX trial124–126 had the lowest percentage of responders in both categories. One study also reported EULAR low disease activity and remission at 6 months (13.3% and 5.7%, respectively). See Figure 39 for details.

FIGURE 39.

Rituximab: EULAR response in uncontrolled studies (3-month data for the REFLEX trial124–126 from MS). LCI, lower confidence interval; UCI, upper confidence interval.

DAS28

Randomised controlled trial

In the REFLEX trial,124–126 at week 24, the RTX arm had a significantly smaller mean DAS28 score and significantly greater reduction in the mean DAS28 score from baseline than the placebo arm (–1.40, 95% CI –1.67 to –1.13, and – 1.50, 95% CI –1.74 to –1.26, respectively). At week 24, the proportion of patients with DAS28 improvement in the RTX group was over five times that in the placebo group and the difference was statistically significant. See Figures 40–42 for details.

FIGURE 40.

Rituximab: DAS28 score at week 24 in the REFLEX trial124–126 (last observation carried forward, data from MS). SD, standard deviation.

FIGURE 41.

Rituximab: DAS28 score change from baseline at week 24 in the REFLEX trial124–126 (last observation carried forward, data from MS). SD, standard deviation.

FIGURE 42.

Rituximab: percentage of patients with DAS28 improvement from baseline at week 24 in the REFLEX trial124–126 (last observation carried forward, data from the MS).

Non-randomised controlled trials

DAS28 score at 3 months was available in only one uncontrolled study (median DAS28 = 5.60). DAS28 score at 6 months was measured in three studies. The mean score was 5.0 in one study and it was the same as that of the RTX arm of the REFLEX trial. 124–126 Two studies provided a median score and it was 5.50 and 3.97. See Figure 43 for details. (Note: for the Jois et al. 115 and Assous et al. 117 studies scores were reported as medians.)

FIGURE 43.

Rituximab: mean DAS28 in uncontrolled studies. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

In Finckh et al. 136 the change in mean DAS28 score from baseline at 6 months was reported only for a subgroup of 50 patients. It was similar to that reported for the RTX arm of the REFLEX trial124–126 and both showed significant improvement (–1.90, 95% CI –2.08 to –1.72, and –1.61, 95% CI –1.98 to –1.24, respectively). See Figure 44 for details.

FIGURE 44.

Rituximab: DAS28 scores change from baseline in uncontrolled studies (data for REFLEX124–126 from MS). LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Health Assessment Questionnaire

Randomised controlled trial

In the REFLEX trial,124–126 the RTX group had significantly more reduction in mean HAQ score from baseline at week 24 compared with the placebo group (mean difference = –0.30, 95% CI –0.40 to –0.20; Figure 45).

FIGURE 45.

Rituximab: mean change in HAQ scores from baseline at week 24 in REFLEX trial. 124–126 SD, standard deviation.

The percentage of patients who showed HAQ improvement, defined as a decrease in score from baseline of greater than 0.25, in the RTX group of the REFLEX trial124–126 was nearly twice that of the placebo group at week 12, and over two and a half times as high at week 24; both effects were statistically significant (RR = 1.63, 95% CI 1.29 to 2.07, and RR = 2.55, 95% CI 1.89 to 3.43, respectively). See Figure 46 for details.

FIGURE 46.

Rituximab: percentage of patients with a decrease in HAQ score > 0.25 from baseline in the REFLEX study124–126 (data from MS).

At week 24, the observed percentage of patients with minimal clinically meaningful improvement in HAQ, defined as a decrease in HAQ score of 0.22, in the RTX group of the REFLEX trial,124–126 was over 1.6 times that of the placebo groups and the difference was significant; whereas observed at week 48 there was no significant difference (Figure 47).

FIGURE 47.

Rituximab: percentage of patients with a clinically meaningful improvement in HAQ, 24 and 48 weeks after the first course of RTX (observed data from MS).

When analysed based on non-responder imputation data, the percentage of patients with minimal clinically meaningful improvement in HAQ at week 24 and week 48 in the RTX group was over two and a half and over three and a half times that of the placebo group (58% vs 23% and 23% vs 6%, respectively) and both differences were statistically significant (Figure 48).

FIGURE 48.

Rituximab: percentage of patients with a clinically meaningful improvement in HAQ, 24 and 48 weeks after the first course of RTX (non-responder imputation data from MS).

Non-randomised controlled trials

Two uncontrolled studies reported HAQ score. The median HAQ score in one study115 was 2.13 (range 0.63–2.88) at 3 months and decreased to 1.86 (range 1–3) at 6 months; however, in both cases, the reduction compared with baseline was not significant. In the Keystone et al. study,116 the percentage of patients with a decrease in the mean HAQ score of greater than or equal to 0.22 from baseline at week 24 (after one course of RTX treatment) was 71.8%, which is very similar to the observed rate reported in the RTX arm of the REFLEX trial124–126 (70.5%) (Figure 49).

FIGURE 49.

Rituximab: percentage of patients with clinically meaningful improvement in HAQ score from baseline at week 24. LCI, lower confidence interval; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Quality of life

Randomised controlled trial

Mean Short Form questionnaire-36 items (SF-36) mental and physical health scores measured at week 24 in the REFLEX trial124–126 were both significantly higher in the RTX group than in the placebo group (Figure 50). The RTX group increased mean SF-36 physical health score by 5.16 and mean SF-36 mental health score by 3.07 higher than in the placebo group, and the differences were statistically significant (Figure 51).

FIGURE 50.

Rituximab: mean SF-36 items scores at week 24 in the REFLEX trial124–126 (last observation carried forward, data from MS). SD, standard deviation.

FIGURE 51.

Rituximab: change in SF-36 items scores from baseline to week 24 in REFLEX trial. 124–126 SD, standard deviation.

Non-randomised controlled trials

None of the uncontrolled studies reported QoL.

Serious adverse events

Randomised controlled trial

In the REFLEX trial,124–126 the percentage of patients with serious AEs was lower in the RTX group than in the placebo group; the difference was not statistically significant (RR = 0.74, 95% CI 0.42 to 1.31). See Figure 52 for details.

FIGURE 52.

Rituximab: serious AEs at week 24 in the REFLEX trial. 124–126

Non-randomised controlled trials

In one 12-month study,114 one patient (2%) had severe headache and stomach pain 1 day after RTX infusion and this led to a discontinuation of treatment. A 6-month study,115 stated that no major side effects were found during the study. During a 6-month period the Thurlings et al. 118 study reported five serious AEs (16.7%): two severe infusion reactions, one arterial embolism, one pulmonary embolism and one toxic hepatitis. The other studies did not report information on serious AEs.

Any infection/serious infection

Randomised controlled trial

In the REFLEX trial124–126 both the percentage of patients with any infections and the percentage of patients with serious infections were greater in the RTX group than in the placebo group; however, none of the differences was statistically significant (RR = 1.08, 95% CI 0.87 to 1.35 and RR = 1.58, 95% CI 0.41 to 6.05, respectively). See Figure 53 for details.

FIGURE 53.

Rituximab: any infection and serious infection at week 24 in the REFLEX trial. 124–126

Non-randomised controlled trials

In the Bokarewa et al. study114 3 months after the treatment with RTX, pneumonia requiring hospitalisation was reported in one patient (2.0%). In Thurlings et al. 118 the incidence of infection per patient-year was 0.9: 48 infections requiring antibiotic, antimycotic, or antiviral treatment and one serious infection requiring i.v. antibiotics occurred among 30 patients over 2 years of follow-up. One serious infection requiring i.v. antibiotics was observed in this study.

Injection site reaction/infusion reaction

Randomised controlled trial

In the REFLEX trial,124–126 the percentage of patients with acute infusion reactions did not differ significantly between groups (RR = 1.24, 95% CI 0.90 to 1.83, for the first course and RR = 0.74, 95% CI 0.43 to 1.24, for the second course). See Figure 54 for details.

FIGURE 54.

Rituximab: percentage of patients who had acute infusion reactions after the first and second infusion.

Non-randomised controlled trials

One study (Finckh et al. ,137 subgroup of 50 patients) reported three mild-to-moderate infusion reactions. Another study118 reported two severe infusion reactions. The other studies did not report information on infusion site reactions.

Data reported by treatment course

Pooled analysis (data from Keystone et al.)

In the Keystone et al. study,116 based on evaluable data, the percentage of patients achieving ACR responses increased from course 1 to course 2 of RTX measured 24 weeks after each course (Figure 55). A similar pattern was seen for the percentage of patients with EULAR response 24 weeks after course 1 and course 2 (Figure 56).

FIGURE 55.

Percentage of patients achieving ACR responses 24 weeks after course 1 and course 2 – based on evaluable patients who had prior TNF inhibitor. LCI, lower confidence interval; UCI, upper confidence interval.

FIGURE 56.

Percentage of patients with EULAR responses 24 weeks after course 1 and course 2 – based on evaluable patients who had prior TNF inhibitor. LCI, lower confidence interval; UCI, upper confidence interval.

The percentage of patients who achieved meaningful improvement in HAQ, i.e. had a decrease of HAQ scores at least 0.22 from baseline, were similar 24 weeks after course 1 and course 2 of RTX treatment (Figure 57).

FIGURE 57.

Percentage of patients with a decrease in HAQ score of ≥ 0.22 at week 24 after course 1 and course 2 – based on evaluable patients who had prior TNF inhibitor. LCI, lower confidence interval; UCI, upper confidence interval.

Summary

For the assessment of effectiveness of RTX in comparison with standard care, one RCT and six uncontrolled studies were identified. Follow-up duration ranged from 3 months to 24 months. Patients included in the studies were generally similar. The main results of the seven studies are summarised in Table 33.

Overview of evidence

Three studies were identified that assessed ABT: one RCT [abatacept trial in treatment of anti-TNF inadequate responders (ATTAIN127–132)], an extension of this RCT (ATTAIN LTE119) and an uncontrolled study [abatacept researched in rheumatoid arthritis patients with an inadequate anti-TNF response to validate effectiveness (ARRIVE)120].

Patients were included in the ATTAIN LTE119 after completing 6 months of the RCT. It was reported that in total 74.4% of the placebo group and 86.4% of the ABT group were included in the extension.

Patients in the studies were non-responders to at least one TNF inhibitor. In the ATTAIN RCT127–132 and LTE119 lack of efficacy was the primary reason for switching biologic agents. In ARRIVE120 patients discontinued the previous TNF inhibitor because of lack of efficacy, safety concerns or intolerability.

All studies were carried out in North America and Europe. ARRIVE120 additionally included Mexican patients. No information was provided if these studies included UK patients. Follow-up was 6 months for the ATTAIN RCT127–132 and ARRIVE study. 120 In the ATTAIN LTE119 patients were followed up for up to 5 years; however, there was no published data beyond 2 years. Further details are provided in Table 34.

Patient characteristics

Full details of patient characteristics are reported in Table 35.

The number of patients included in the studies was 391 in the ATTAIN RCT,127–132 317 in its LTE119 and 1,046 in the ARRIVE study. 120 Patient characteristics were generally similar across studies and study arms:

• The percentage of female patients ranged from 78% to 81%.

• The mean age ranged from 53.0 to 54.4 years.

• The mean disease duration ranged from 11.6 to 11.9 years.

• In two studies the percentage of RF-positive patients ranged from 61.3% to 73.2%; it was not reported in the ATTAIN LTE. 119

• Concomitant DMARDs were reported in detail in ATTAIN127–132 and ARRIVE:120 69.8%–77.7% patients were on MTX; other DMARDs included HCQ (8.9%–15.0%), LEF (8.7%–2.8%) and sulfasalazine (8.0%–8.8%). In the ARRIVE study,120 AZA (4.1%) and gold (0.5%) were also used.

• In two studies 58.4%–68.3% of patients were receiving corticosteroids; this information was not reported in detail in the ATTAIN LTE. 120

• The number of previously used conventional DMARDs was not reported in any of the studies.

• The number of previous TNF inhibitors ranged from one to two in the ATTAIN127–132 and ATTAIN LTE119 studies and from one to three in the ARRIVE study. 120

• The mean baseline HAQ ranged from 1.7 to 1.8.

• The mean DAS28 score ranged from 6.2 to 6.5.

• The mean number of tender and swollen joints ranged from 17.8 to 31.8 and from 13.6 to 22.3, respectively.

• Baseline ESR was not reported in any of the studies.

• CRP ranged from 2.1 mg/dl to 4.4 mg/dl.

Quality assessment

Results

The RCT and non-randomised studies were analysed separately. Data from the ABT arm of the ATTAIN RCT are included in all figures referring to uncontrolled studies for comparison.

Table 38 indicates which of the outcomes reported in the main text of the report were assessed in individual studies and Table 39 provides similar information for outcomes described in Appendix 10 only.

Withdrawals

Randomised controlled trial

There were significantly fewer withdrawals for any reason in the ABT arm than in the placebo arm of the ATTAIN RCT127–132 (RR = 0.53, 95% CI 0.35 to 0.81). There were also significantly fewer withdrawals in the ABT group because of lack of efficacy (RR = 0.27, 95% CI 0.15 to 0.49). The risk of withdrawal because of AEs was similar in both groups (RR = 0.93, 95% CI 0.32 to 2.71). Details of the analysis are presented in Figure 58.

FIGURE 58.

Abatacept: withdrawals by reason in the ATTAIN RCT127–132 at 6 months.

Non-randomised control trials

At 6 months 17.8% patients withdrew from the ARRIVE study. 120 This percentage was slightly higher than in the ABT-treated arm of the RCT. At 2 years, 30% of patients had withdrawn from the ATTAIN LTE. 119 In both studies, more patients withdrew because of lack of efficacy than because of AEs. A similar relationship was observed in the ABT arm of the RCT. Full details are presented in Figure 59.

FIGURE 59.

Abatacept: withdrawals in uncontrolled studies by reason. LCI, lower confidence interval; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

ACR20 response

Randomised control trial

ATTAIN127–132 reported ACR20 response at 3 and 6 months. At both follow-up times the risk of an ACR20 response was over two and a half times higher in the ABT group than in the placebo group and the difference was statistically significant (for 3 months, RR = 2.53, 95% CI 1.72 to 3.73; for 6 months, RR = 2.56, 95% CI 1.77 to 3.69). Details can be found in Figure 60.

FIGURE 60.

Abatacept: ACR20 response in the ATTAIN RCT127–132 at 3 and 6 months.

Non-randomised control trials

Of the uncontrolled studies, only the ATTAIN LTE119 reported ACR20 response. Results are reported by subgroup based on whether patients were originally randomised to ABT or placebo in the randomised phase (see Figure 61). After 6 months of ABT treatment, 57.3% patients in the group initially randomised to ABT and 63.6% in the group initially randomised to placebo achieved an ACR20 response. This was slightly more than in the ABT arm of the RCT (50.0%). After 6 months, there was a further increase in the percentage of ACR20 responders at 12 months in those initially randomised to ABT followed by a decrease up to 5 years (30.3%). Among those initially randomised to placebo, there was a decrease in the percentage of responders from 12 months onwards, and at 54 months 30.3% of patients were ACR20 responders.

FIGURE 61.

Abatacept: ACR20 response in non-RCTs. LCI, lower confidence interval; n/a, not available; PL, placebo; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

If only patients for whom data were available at different time points were analysed, the increase in percentage of ACR20 responders continued to 3 years (82.1%) and then decreased to 65.6% at 5 years for patients initially randomised to ABT. In the same analysis, among patients initially randomised to placebo there was an increase in the percentage of ACR20 responders up to 42 months (82.0%), and at 54 months 78.9% were ACR20 responders.

ACR50 response

Randomised controlled trial

At 6 months the percentage of ACR50 responders was over five times higher in the ABT group than in the placebo group of the ATTAIN trial127–132 and the difference was statistically significant (RR = 5.36, 95% CI 2.19 to 13.10). Details are presented in Figure 62.

FIGURE 62.

Abatacept: ACR50 response in the ATTAIN RCT127–132 at 6 months.

Non-randomised contolled trials

Of the uncontrolled studies, only the ATTAIN LTE119 reported ACR50 response. Results are reported by subgroup based on whether patients were originally randomised to ABT or placebo in the randomised phase (see Figure 63). This outcome was achieved at 6 months by 22.9% patients in the arm initially randomised to ABT and 37.4% in the arm initially randomised to placebo. For comparison, this outcome was achieved by 20.2% of patients in the ABT arm of the RCT. In the arm initially randomised to ABT, the percentage of ACR50 responders increased up to 18 months (33.9%) and then decreased to 20.6% at 5 years. In the arm initially randomised to placebo, there was a decrease after 6 months to 21.2% achieving ACR50 response at 48 months.

FIGURE 63.

Abatacept: ACR50 response in non-RCTs. LCI, lower confidence interval; n/a, not available; PL, placebo; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

In the analysis based on the observed data, only the percentage of ACR50 responders among those initially randomised to ABT increased up to 3 years (51.1%) and then it was 46.1% at 4 years and 51.1% at 5 years. Among those initially randomised to placebo there was an almost constant increase up to 48 months (53.8%).

ACR 70 response

Randomised controlled trial

In the ATTAIN RCT,127–132 the percentage of patients achieving ACR70 response at 6 months was almost seven times higher in the ABT group than in the placebo group (RR = 6.70, 95% CI 1.62 to 27.8). This difference was statistically significant; however, it needs to be highlighted that the CIs were very wide (see Figure 64).

FIGURE 64.

Abatacept: ACR70 response in the ATTAIN RCT127–132 at 6 months.

Non-randomised controlled trials

Of the uncontrolled studies, only the ATTAIN LTE119 reported ACR70 response. After 6 months of treatment the percentage of ACR70 responders was 11.5% among patients initially treated with ABT and 13.1% among patients initially treated with placebo. For comparison, it was 10.1% in the ATTAIN RCT. 127–132 In the arm initially randomised to ABT, there was a further increase to 17.0% at 12 months followed by a decrease to 9.6% at 5 years. In the arm initially randomised to placebo, there was an increase up to 15.2% at 30 months followed by a decrease to 7.1% at 54 months. Analysis based on observed data only provided more favourable results, with the highest percentage of ACR70 responders being 23.4% at 36 months in the arm initially randomised to ABT and 25.9% at 30 months in the arm initially randomised to placebo. See Figure 65 for details.

FIGURE 65.

Abatacept: ACR70 response in non-RCTs. LCI, lower confidence interval; n/a, not available; PL, placebo; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

DAS28

Randomised controlled trial

The mean change from baseline in DAS28 was –1.98 in the ABT group and –0.71 in the placebo group. The difference between these values was –1.27 (95% CI –1.62 to –0.93, p < 0.001). These data were provided in the industry submission only. No further information was provided and therefore analyses could not be undertaken.

As indicated in Figure 66, over twice as many patients achieved a clinically meaningful DAS28 improvement (defined as greater than or equal to 1.2) in the ABT arm as in the control arm (RR = 2.15, 95% CI 1.54 to 2.99).

FIGURE 66.

Abatacept: patients with clinically meaningful (≥ 1.2) DAS28 improvement in the ATTAIN RCT127–132 at 6 months.

The ATTAIN study127–132 also reported percentages of patients who, based on DAS28, achieved a low score (DAS28 less than or equal to 3.2) or remission (DAS28 less than 2.6). At 6 months, patients in the ABT arm were over five times more likely to have a DAS28 less than or equal to 3.2 than those in the placebo arm and the difference was statistically significant (RR = 5.67, 95% CI 2.08 to 15.44). They were also over 13 times more likely to have a DAS28 less than 2.6 than the placebo group and the difference was statistically significant (RR = 13.40, 95% CI 1.84 to 97.69); however, the CIs were wide. See Figure 67 for details.

FIGURE 67.

Abatacept: patients with final DAS28 of ≤ 3.2 and of < 2.6 in the ATTAIN RCT127–132 at 6 months.

Non-randomised controlled trials

Change in the DAS28 score was assessed in both uncontrolled studies. Details are presented in Figure 68. After 6 months of treatment, there was a mean change of –1.99 in the arm initially randomised to ABT and of –2.14 in the arm initially randomised to placebo in the ATTAIN LTE,119 and of –2.00 in the ARRIVE study. 120 This was similar in the RCT. 127–132 In the ATTAIN LTE,119 DAS28 further decreased with time and the mean change was –2.90 at 5 years in the arm initially randomised to ABT and –2.96 at 54 months in the arm initially randomised to placebo.

FIGURE 68.

Abatacept: DAS28 change from baseline in uncontrolled studies. LCI, lower confidence interval; n/a, not available; PL, placebo; SD, standard deviation; UCI, upper confidence interval.

ARRIVE120 measured clinically meaningful DAS28 improvement. It was defined as a decrease of greater than or equal to 1.2 or a score of less than or equal to 3.2. At 6 months, 56.1% of patients in ARRIVE120 achieved this outcome. This was slightly more than in the ABT group of the RCT127–132 (although in ATTAIN127–132 this was defined as a decrease of greater than or equal to 1.2 only). See Figure 69 for details.

FIGURE 69.

Abatacept: clinically meaningful DAS28 improvement in non-randomised studies at 6 months.

Both uncontrolled studies reported percentages of patients who, based on DAS28, achieved a low score (DAS28 less than or equal to 3.2) or remission (DAS28 less than 2.6). Full details are reported in Figure 70.

FIGURE 70.

Abatacept: patients with final DAS28 values ≤ 3.2 and < 2.6 in uncontrolled studies. LCI, lower confidence interval; n/a, not available; PL, placebo; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

At 6 months a DAS28 score of less than or equal to 3.2 was achieved by 10.6% of patients initially randomised to ABT in the ATTAIN LTE,119 by 22.2% of patients initially randomised to placebo in the ATTAIN LTE119 and by 22.4% of patients in ARRIVE. 120 For comparison, this was 17.1% of patients in the ABT arm of ATTAIN. 127–132 The percentage of patients initially randomised to ABT in ATTAIN LTE119 who achieved a DAS28 of less than or equal to 3.2 increased up to 18 months (28%) and then decreased up to 5 years (15.1%). In the arm initially randomised to placebo, the percentage of patients with low DAS28 decreased up to 54 months (7.1%).

A DAS28 of less than 2.6 was achieved at 6 months by 10.6% and 17.2% in the ATTAIN LTE119 (initial ABT and placebo, respectively) and by 13.0% in ARRIVE. 120 For comparison, 10.1% of the ABT arm of the RCT achieved this outcome. In the ATTAIN LTE119 arm initially randomised to ABT, the highest percentage of patients with DAS28 less than 2.6 was recorded at 18 months (17.0%), following which it decreased to 9.6% at 5 years. In the arm initially randomised to placebo, the highest percentage of patients with DAS28 less than 2.6 was recorded after 6 months of treatment, and at 54 months it was 6.1%.

Health Assessment Questionnaire

Randomised controlled trial

At 6 months, the HAQ change from baseline in the ATTAIN RCT127–132 was –0.45 in the ABT group and –0.11 in the placebo group and the difference between the two groups was reported to be statistically significant (p < 0.001). No data on uncertainty of individual assessments were provided in the study and therefore further analyses could not be undertaken.

This study also assessed clinically meaningful HAQ improvement, defined as a decrease in HAQ score of at least 0.3 (details are reported in Figure 71). Clinically meaningful HAQ improvement was over two times more frequent in the ABT group than in the placebo group and the difference was statistically significant (RR = 2.01, 95% CI 1.44 to 2.81).

FIGURE 71.

Abatacept: clinically meaningful improvement (≥ 0.3) in HAQ score.

Non-randomised control trials

Change in HAQ score was assessed in both uncontrolled studies (however, in the case of for ARRIVE120 only data for a subgroup of 43 US patients receiving monotherapy were reported; ABT monotherapy is licensed in the USA but not in Europe). Figure 72 presents the mean changes from baseline in HAQ score. The mean change from baseline at 6 months was –0.51 in the arm of ATTAIN127–132 initially randomised to ABT, –0.40 in the arm of ATTAIN127–132 initially randomised to placebo and –0.38 in the monotherapy subgroup of ARRIVE. 120 The results for the ABT arm of the RCT were similar. In the arm initially randomised to ABT in the ATTAIN LTE,119 the change decreased up to 3 years (–0.65) and then started slowly increasing (to –0.58 at 4 years and to –0.56 at 5 years). In the group initially randomised to placebo, there was a decrease up to 54 months of treatment (–0.71).

FIGURE 72.

Abatacept: mean changes from baseline in HAQ score. LCI, lower confidence interval; PL, placebo; SD, standard deviation; UCI, upper confidence interval.

Both uncontrolled studies reported the number of patients who achieved a clinically meaningful improvement in HAQ (details are provided in Figure 73). The ATTAIN LTE119 defined this outcome as an improvement of at least 0.3 in the HAQ score, while in ARRIVE120 it was an improvement of at least 0.22. After 6 months of treatment with ABT, the percentage of patients who achieved this outcome was 52.8% in the ATTAIN LTE119 arm that comprised patients initially randomised to ABT, 49.5% in the ATTAIN LTE119 arm comprising patients initially randomised to placebo and 46.7% in the ARRIVE study. 120 For comparison, it was 46.9% in the ABT arm of the RCT. Analysis of the data from the ATTAIN LTE119 using a non-responder imputation showed a decrease in the percentage of patients who achieved a clinically meaningful HAQ over time, with 24.8% of patients initially randomised to ABT achieving clinically meaningful HAQ improvement at 5 years and 27.3% of patients initially randomised to placebo achieving clinically meaningful HAQ improvement at 54 months. When the analysis in both groups included only patients in whom HAQ improvement was measured at different time points, there was a slight increase in the percentage over time, with a decrease in the last outcome measurement.

FIGURE 73.

Abatacept: clinically meaningful improvement in HAQ score (≥ 0.3 in ATTAIN studies119,127–132 and ≥ 0.22 in ARRIVE120) LCI, lower confidence interval; PL, placebo; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Quality of life

Randomised controlled trial

The ATTAIN RCT127–132 assessed patients’ QoL using the SF-36 scale. Patients in the ABT arm improved significantly more in both the physical component (mean difference = 5.50, 95% CI 3.74 to 7.26) and the mental component (mean difference = 3.70, 95% CI 1.45 to 5.95). Details are presented in Figure 74.

FIGURE 74.

Abatacept: SF-36 items changes from baseline in components in the ATTAIN RCT127–132 at 6 months. SD, standard deviation.

For all individual SF-36 items there was a significantly higher improvement in the ABT arm than in the placebo arm. Details for each item are presented in Figure 75.

FIGURE 75.

Abatacept: SF-36 items changes from baseline in items at 6 months in the ATTAIN RCT. 127–132 SD, standard deviation.

Non-randomised controlled trials

Of the uncontrolled studies, change in SF-36 was assessed only in the ARRIVE study120 (however, it was reported only for a subgroup of 43 patients receiving monotherapy; ABT monotherapy is licensed in the USA but not in Europe). For the physical component of the SF-36 scale, there was improvement of 4.80 for the monotherapy subgroup of ARRIVE. 120 For the mental component, the improvement was 7.34. For comparison, in the ABT arm of ATTAIN127–132 it was 6.50 and 5.40, respectively. Further details are provided in Figure 76. Data for individual items were not reported in ARRIVE. 120

FIGURE 76.

Abatacept: SF-36 items changes from baseline in components. LCI, lower confidence interval; SD, standard deviation; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Serious adverse events

Randomised controlled trial

In ATTAIN,127–132 there was no significant difference at 6 months between ABT and placebo in the risk of experiencing a serious AE (RR = 0.93, 95% CI 0.51 to 1.68). Details are presented in Figure 77.

FIGURE 77.

Abatacept: serious AEs in the ATTAIN RCT127–132 at 6 months.

Non-randomised controlled trials

Serious AEs were assessed in both uncontrolled studies. At 6 months the percentage of patients who had experienced a serious AE was 10.4% in ARRIVE. 120 It was similar in the ABT arm of the ATTAIN RCT127–132 (10.5%). At 2 years, 32.5% of patients in the ATTAIN LTE119 had experienced a serious AE. Full details are presented in Figure 78.

FIGURE 78.

Abatacept: serious adverse events in non-randomised studies. LCI, lower confidence interval; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Infections/serious infections

Randomised controlled trial

At 6 months there was no statistically significant difference between ABT and placebo in the risk of infection (RR = 1.16, 95% CI 0.87 to 1.56) or serious infection (RR = 1.03, 95% CI 0.26 to 4.06). Details are presented in Figure 79.

FIGURE 79.

Abatacept: infections in the ATTAIN RCT127–132 at 6 months.

Non-randomised controlled trials

Both uncontrolled studies reported infections. The percentages of patients who experienced any infection were similar at 6 months in the ABT arm of ATTAIN127–132 and in the ARRIVE study120 (37.6% and 38.9%, respectively). Of these 2.3% and 2.4% were serious. At 2 years 73.8% of patients in the ATTAIN LTE119 experienced an infection of any kind and 7.9% a serious infection. Details are reported in Figure 80.

FIGURE 80.

Abatacept: infections in non-randomised studies.

Injection/infusion reaction

Randomised controlled trial

At 6 months there was no statistically significant difference between ABT and placebo in the risk of infusion reaction (RR = 1.68, 95% CI 0.56 to 5.04). Details are reported in Figure 81.

FIGURE 81.

Abatacept: infusion reactions. LCI, lower confidence interval; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Non-randomised controlled trials

Of the uncontrolled studies, infusion reactions were reported only in ARRIVE. 120 At 6 months, 5.4% patients had experienced infusion reactions. For comparison, this figure was 5.0% in the ABT arm of ATTAIN. 127–132 Details are provided in Figure 82.

FIGURE 82.

Abatacept: infusion reactions. LCI, lower confidence interval; UCI, upper confidence interval. Bold type indicates that the study was an RCT.

Abatacept in combination with other biologic drugs

Two RCTs [Weinblatt et al. 134 and abatacept study of safety in use with other rheumatoid arthritis therapies (ASSURE)135] were identified that assessed ABT in combination with previously tried biologic drugs. Although both studies met the inclusion criteria of the systematic review, combination therapy was not considered relevant to this report and, therefore, they were not analysed.

The study by Weinblatt et al. 134 was a multicentre placebo-controlled randomised trial and included 121 patients who had active RA despite treatment with ETN. Patients were randomised to receive ETN and ABT or ETN and placebo and were followed up for 1 year. Afterwards they could enter a LTE (data provided for 2 years of the extension study). Data were collected on outcomes including ACR response, HAQ, SF-36 and safety.

ASSURE135 was a multicentre placebo-controlled randomised trial and included 167 patients who had active RA in spite of receiving therapy with biologic agents (ETN, IFX, ADA and anakinra), ‘warranting additional therapy at the discretion of the investigator’. (Note: it also included 1,274 patients who received background DMARDs and were probably biologic naive.) Patients continued their treatment and in addition to that were randomised to receive ABT or placebo. They were followed up for 1 year. The study assessed outcomes including HAQ Disability Index, pain, patient and physician global assessment and safety.

Summary

Three studies assessed ABT in comparison with standard care: one RCT (ATTAIN127–132) and two uncontrolled studies (ATTAIN LTE119 and ARRIVE120). Follow-up ranged from 6 months to 5 years. All studies included patients with similar baseline characteristics. The main results of the included studies are summarised in Table 40.

Overview of evidence

One prospective cohort study was identified to compare RTX with TNF inhibitors as a class. 136,137

Included patients had tried at least one TNF inhibitor (ADA, ETN or IFX) before and discontinued treatment owing to inadequate response. The study was conducted in Switzerland and the median duration of follow-up was 11 months. Full details of this study are provided in Table 41.

Patient characteristics

Full details baseline characteristics are reported in Table 42. The study included 318 patients and:

• The proportion of women was 77.5%.

• The mean age was 55 years.

• The mean disease duration was 11.3 years.

• The proportion of RF-positive patients was 82.4%.

• Concomitant DMARDs used were MTX (63.9%), LEF (18%) and other (4.5%).

• The proportion of patients receiving steroids was 56.5%.

• The number of previous DMARDs was not reported.

• The number of previous TNF inhibitors ranged from one to over two.

• The mean baseline HAQ score was 1.5.

• The mean baseline DAS28 score was 4.5.

• No information was provided on CRP and ESR.

Quality assessment

Full details of quality assessment are reported in Table 43. The study was a prospective cohort. It had clearly defined inclusion criteria. It was; however, unclear if consecutive patients were included in the study and what percentage of patients were withdrawn.

Results

Table 44 indicates which of the outcomes reported in the main text of the report were assessed in the Finckh et al. study. 136,137 No outcomes apart from the ones reported in Table 44 were assessed.

DAS28

There was a trend favouring TNF inhibitors over RTX for change from baseline in DAS28; however, this difference was not statistically significant (mean difference = –0.35, 95% CI –0.71 to 0.01). The follow-up for this outcome was unclear. See Figure 83 for details.

FIGURE 83.

Tumour necrosis factor inhibitors versus rituximab: DAS28 change from baseline. SD, standard deviation.

Injection/infusion reaction

Data for injection/infusion reactions were reported only for a subgroup of 116 patients. 136 Dermatological complications (mainly injection site reactions) occurred in one RTX patient and nine TNF inhibitor patients. Infusion reactions were reported in three RTX and none of the TNF inhibitor patients. Data from both categories were analysed together to compare AEs associated with drug administration (Figure 84). There was no statistically significant difference between groups (RR = 1.70, 95% CI 0.56 to 5.22).

FIGURE 84.

Tumour necrosis factor inhibitors versus rituximab: injection/infusion site reactions.

Summary

One prospective cohort study136,137 compared TNF inhibitors as a class with RTX. The median follow-up was 11 months; however, it was not clearly stated when outcomes were assessed. The main results of the study are summarised in Table 45.

Lack of response (primary failure) versus loss of response (secondary failure)