Outcomes following switching from etanercept originator to etanercept biosimilar in 1024 patients with RA: a matched-analysis of the BSRBR-RA

Abstract

Objectives

Adults with RA are being switched from etanercept originator to biosimilar in non-medical/cost-saving switching. This analysis aims to investigate outcomes in these patients, including (i) drug survival and (ii) disease activity at 6 months and 12 months, compared with those who remain on the originator.

Methods

Using BSRBR-RA, those who switched directly from etanercept originator to biosimilar were identified and matched to patients receiving the originator, based on gender, age, disease duration and originator start year. Drug survival was calculated; Cox-proportional hazard models assessed differences in drug persistence between those who switched vs remaining on originator. Change in DAS28 after 6 months and 12 months was compared between cohorts. Multiple imputation was used.

Results

A total of 1024 adults with RA switching from etanercept originator to biosimilar were included, with a matched cohort of patients remaining on the originator. Patients who switched onto a biosimilar product were no more likely to discontinue etanercept treatment vs those who remained on the originator; hazard ratio 1.06 (95%CI 0.89–1.26), with 65% of patients remaining on treatment at three years. Ninety-five (9%) patients switched back to the originator within the first year. After 6 months and 12 months, biosimilar patients were no more likely to have a worsening of DAS28 (>0.6 units) compared with those who remained on the originator.

Conclusions

This is the largest matched comparative effectiveness analysis showing patients switching from etanercept originator to biosimilar appearing to do just as well with regard to disease activity and drug persistence compared with those who remained on the originator. These data will be reassuring to clinicians and patients regarding non-medical switching.

Graphical Abstract

Rheumatology key messages.

• The largest matched comparative effectiveness analysis of 1024 RA patients switched from etanercept originator to biosimilar (non-medical switch).

• Disease activity and drug persistence was similar compared with those who remained on the etanercept originator.

• This information is reassuring to clinicians and patients regarding the impact of non-medical biological switching.

Introduction

Etanercept, a TNF inhibitor, was one of the first licenced biologic DMARDs (bDMARDs) for use in adults with RA in the early 2000s [1]. Since then, it has remained one of the most commonly used TNF inhibitors [2]. Once the patent for the original etanercept therapy (originator) expired in 2015, other pharmaceutical companies have marketed ‘biosimilars’ of the drug. While these versions of the therapy are similar, with phase III clinical trials demonstrating equivalent clinical efficacy [3–6], they can never truly be identical due to the complex manufacturing process [7]. The reduced cost of biosimilars has resulted in many patients already receiving the etanercept originator being switched onto the biosimilar for non-medical reasons, i.e. not for ineffectiveness or adverse events.

There are small studies comparing the originator vs biosimilars of the TNF-inhibitor infliximab [8–10], the first anti-TNF biosimilar authorised for RA. A Danish study found no difference in disease activity in 403 patients with RA after switching from the infliximab originator to a biosimilar compared with pre-switch disease activity [11]. A US retrospective cohort study of 849 patients who remained on the infliximab originator and 838 who switched to the biosimilar found patients who switch onto the biosimilar were more likely to discontinue treatment compared with those who remained on the originator [12]. While this study included other inflammatory arthritis and inflammatory bowel disease patients, the authors state that subgroup analyses of those with RA yielded similar results. Many of the patients who discontinued biosimilar therapy restarted the originator, potentially due to what the experts have coined the nocebo effect, i.e. perceived lack of efficacy [13]. This is supported by many open label studies also finding discontinuation rates in infliximab biosimilars to be higher compared with those on the originator [14]. Intervention studies have found that tailored communication with nurses may reduce this nocebo effect [15].

When considering patients switching from the etanercept originator to biosimilar [16–21], very few studies include a comparison of patients who do not switch and those that do tend to be small in size [5, 22, 23]. Therefore, the aims of this analysis were to investigate what happens to patients with RA following a non-medical switch from the etanercept originator to an etanercept biosimilar with respect to (i) drug survival and (ii) disease activity at 6 months and 12 months following the switch, compared with a cohort of matched patients who remain on the etanercept originator.

Methods

Study setting

The British Society for Rheumatology Biologics Register for RA (BSRBR-RA) is an observational prospective cohort study for patients with RA starting biologic or targeted-synthetic (b/ts) DMARDs, established in 2001. The aim of the register is to monitor the long-term effectiveness and safety of b/tsDMARDs nationally in routine clinical practice. At the start of b/tsDMARD therapy (registration), detailed information is collected on patient demographic information (age, gender), disease duration, disease activity using core outcome variables [swollen joint count, tender joint count, patient global assessment, erythrocyte sedimentation rate (ESR) and/or C-reactive protein (CRP)] and the 28-joint disease activity score (DAS28), and current or previous anti-rheumatic therapies. Follow-up data are collected every 6 months for 3 years, then annually, on disease activity, anti-rheumatic therapy and adverse events. The general BSRBR-RA methodology has already been published [24]. Patients must be enrolled in the register at the point of starting a new therapy, which may include the point of switching to a biosimilar, but do not have to join at the point of starting their first b/tsDMARD therapy. The register was approved by the UK Northwest Multicentre Research Ethics Committee in December 2000 (MREC 00/8/53), and all participants provide written informed consent upon registration.

Patient inclusion criteria/matching

Patients were selected for this analysis if they had RA and had started etanercept originator up to 30 November 2022 (date of data extract for analysis). All patients were eligible for the originator ‘comparator’ cohort. Patients who switched directly from etanercept originator to etanercept biosimilar at any point after 31 January 2016 were eligible for the ‘biosimilar’ cohort. Patient characteristics of all patients eligible for matching, both biosimilar and originator cohort, at the start of initial etanercept originator therapy are shown in Supplementary Table S1, available at Rheumatology online.

Each biosimilar cohort patient was matched 1:1 to an originator patient based on gender, age (±4 years), disease duration (±1 year) at etanercept originator start, and etanercept originator start year (±2 years). For the purposes of the current analyses, the date they switched to biosimilar is referred to as ‘index date’ for both the biosimilar patient and the matched originator patient (had to still be on etanercept originator therapy). Originator patients could only be matched up to three times to avoid over-representation of a small number of patients.

All patients were included in the treatment persistence analysis. For the disease activity analysis, patients were required to have disease activity data available from at least one of the following time points: index date (range: up to 6 month prior, and 7 days after), 6 months following index date (range: from 1 week after index date to 9 month after), and/or 12 months following index date (range: from 9 months to 18 months after index date). See the flow diagram (Fig. 1) for visualisation of cohort selection process.

Figure 1.

Flow diagram of patient inclusion into this analysis. ^*See Supplementary Table S2, available at Rheumatology online: Patients in the biosimilar cohort who could not be matched were those starting originator as older patients with longer disease duration, or those starting the originator more recently. Conversely, the originator patients who could not be matched to a biosimilar patient were those who started originator in the early 2000s

Statistical analysis

Drug survival

Drug survival was assessed using Kaplan–Meier estimates. Patients entered at index date (switch date for biosimilar patients or matched date for originator patients). Patients were censored at drug stop date if they stopped therapy for remission (to avoid combining negative and positive outcomes), drug stop date if they stopped therapy to switch from etanercept originator to etanercept biosimilar (switched within 32 days; originator patients only assumed non-medical switch), or at final follow-up date, death date, or cut-off date for this analysis if they remained on therapy, whichever was first. Where stop date of etanercept therapy was not recorded but the patient was known to have started a subsequent b/tsDMARD, the start date (minus one day) of the next b/tsDMARD therapy was used as the etanercept stop date. Cox-proportional hazard models were used to assess whether there was a difference in the proportion of patients who stayed on etanercept over time between those who switched onto a biosimilar vs those who stayed on the originator.

The proportion of patients who stopped etanercept were calculated. In the biosimilar cohort, stop reason was categorised as: (i) switched back to the etanercept originator; (ii) started another b/tsDMARD therapy (other than etanercept); or (iii) remained off b/tsDMARD therapy until their final (most recent) follow-up.

Disease activity

Patient characteristics and disease activity data, including DAS28 state [remission (DAS28 < 2.6), low (DAS28 ≥ 2.6 and ≤3.2), moderate (DAS28 < 3.2 and ≤5.1) and high (DAS28 > 5.1)] [25, 26] were presented at index date; the point of switching for those switching from originator to biosimilar, or matched date (switch date of their matched patient) for the originator cohort. Disease activity data were presented 6 months and 12 months after index date, including DAS28 state, change in DAS28 (continuous) and change in DAS28 categorised; improved (by >0.6 units), no change (±0.6 units), worsening slightly (by >0.6 and ≤1.2 units), or severely worsened (by >1.2 units) [25].

Logistic regression was used to identify whether patients switching to a biosimilar were more likely to have worsening of their disease activity by >0.6 units (vs no change/improving) compared with those who remained on the originator, at 6 and 12 months. A second model was run, reclassifying patients who stopped etanercept prior to the time point as ‘worsening’, regardless of DAS28 (i.e. non-responders), unless the patient switched from the originator product to the biosimilar within 32 days of stopping the originator (i.e. for a non-medical switch). This was to ensure that for patients who stopped therapy prior to DAS28 assessment and switched to an alternative b/tsDMARD which then improved DAS28, this improvement was not incorrectly attributed to etanercept. Patients who stopped treatment for remission were reclassified as ‘improving’ regardless of DAS28 (i.e. responder). These regression models were repeated twice: (i) adjusting for DAS28 at index date; and (ii) adjusting for the index date variables of age, gender, disease duration, line of therapy (of originator), time on originator, tender joint count, ESR, CRP, patient global assessment and DAS28.

Multiple imputation using chained equations (20 datasets) was used to account for missing data. Imputed variables included tender joint count, swollen joint count, ESR, CRP, pain VAS and DAS28 at index date, 6 and 12 months. Also included in the multiple imputation were complete variables; gender, age, disease duration, time on originator prior to index date, and line of therapy. DAS28 state at index date, 6 and 12 months, change in DAS28 at 6 and 12 months, and change in DAS28 state at 6 and 12 months were calculated using imputed values.

Sensitivity analysis: The analysis was repeated including only those matched patients starting etanercept as their first-line therapy.

All analyses were completed in Stata version 14 [27].

Patient and public involvement

This analysis was designed to address a priority research area set by our patient partners.

Results

There were 1024 patients who switched from etanercept originator to etanercept biosimilar who were matched 1:1 to originator cohort patients (see Fig. 1). Patient characteristics of those who were unable to be matched and therefore not included in this analysis are presented in Supplementary Table S2, available at Rheumatology online. The biosimilar patients that were excluded tended to be those with a longer disease duration before starting the originator (median 25 years), while most of the originator patients that were excluded were those who started etanercept originator prior to 2010.

For all the included patients, characteristics at the point of switching (or switch date of matched patient for the originator cohort; index date) were similar with respect to age (median 64 years), gender (76% female) and disease duration (median 19 years) (see Table 1). Most patients had started etanercept originator as their first-line b/tsDMARD therapy (71% biosimilar cohort, 69% originator cohort), with 23% and 24% starting the originator as their second-line b/tsDMARD therapy, respectively. The biosimilar cohort had a median time on the originator before switching of 9.3 years and the originator cohort similar at 9.6 years. The biosimilar cohort patients had a slightly longer follow-up in the BSRBR-RA after the switch date (median: 2.7 years vs 2.2 years).

Table 1.

Patient characteristics of those who switched from originator to biosimilar, and the matched originator cohort

	Etanercept Biosimilar Cohort		Matched Etanercept Originator Cohort
Total, n	1024		1024 (unique patients: 794)
	Start of originator	Start of biosimilar (index date)	Start of originator	At matched switch date (index date)
Female, n (%)	781 (76%)	—	781 (76%)	—
Age, years	56 (47, 64)	64 (56, 71)	56 (48, 64)	64 (56, 71)
Disease duration, years	9 (4, 16)	19 (10, 26)	9 (4, 16)	19 (10, 26)
Line of therapy (originator), n (%)
1	732 (71%)	—	711 (69%)	—
2	240 (23%)	—	248 (24%)	—
3	35 (3%)	—	49 (5%)	—
4+	17 (2%)	—	16 (2%)	—
Time on originator before switch date, years	—	9.3 (2.6, 12.8)	—	9.6 (2.5, 12.8)
Follow-up in BSRBR-RA following switch date, years	—	2.7 (1.9, 4.5)	—	2.2 (1.3, 4.2)
Year of originator start, n (%)
<2010	548 (54%)	—	550 (54%)	—
2010 onwards	476 (46%)	—	474 (46%)	—
Year switched to biosimilar, n (%)
2016	—	315 (31%)	—	—
2017	—	507 (50%)	—	—
2018	—	174 (17%)	—	—
2019	—	24 (2%)	—	—
2020 onwards	—	4 (<1%)	—	—

All continuous variables are reported as median (inter-quartile range).

Drug survival

Kaplan–Meier survival estimates were used to estimate etanercept treatment survival from index date; 81% of biosimilar patient and 84% of originator patients remained on treatment after one year, 72% and 74% remained on treatment at two years, and 65% and 65% remained on treatment at three years, respectively (Fig. 2). Patients who switched onto a biosimilar product were no more likely to discontinue etanercept treatment vs those who stayed on the originator; hazard ratio 1.06 (95% CI 0.89, 1.26).

Figure 2.

Etanercept treatment persistence from index date (date of switch for biosimilar patients, matched date for originator patients) stratified by cohort

During the follow-up of the BSRBR-RA, median patient follow-up time in the etanercept originator cohort was 2.3 years (IQR 1.4, 4.3). During this time, 326 patients (32% of the whole originator cohort) switched to an etanercept biosimilar in a non-medical switch, and 205 (20%) stopped etanercept treatment. Within the biosimilar cohort there was a median follow-up time of 2.7 years (IQR 1.9, 4.6), with 346 (34%) stopping therapy during this time; 127 patients (37% of those who stopped) switched back to the originator, 95 (9% of all biosimilar patients) within the first year. Most biosimilar patients stopped for ineffectiveness (n = 161), and of these most switched back to the originator (n = 68; 42%) or changed to another b/tsDMARD (n = 85; 55%) (Table 2). Patients who stopped for an adverse event (n = 96) tended to remain off b/tsDMARD therapy at their last follow-up (n = 46; 48%). Median follow-up time from biosimilar stop date to last follow-up for the 84 patients who remained off b/tsDMARD therapy at their last follow-up was 1.3 years (IQR 0.7, 2.9).

Table 2.

Frequency and reasons for all patients discontinuing etanercept biosimilar

	All patients discontinuing etanercept biosimilar	Patients who switched back to the originator	Patients who started a b/tsDMARD other than etanercept	Patients who discontinued all b/tsDMARD therapy according to the latest follow-up
Total	346/1024 (34%)	127	135	84
		12% of whole cohort	13% of whole cohort	8% of whole cohort
		37% of those stopped	39% of those stopped	24% of those stopped
Time on biosimilar therapy, years	0.9 (0.4, 1.7)	0.6 (0.3, 0.9)	1.2 (0.5, 2.5)	1.2 (0.7, 2.0)
Time from end of biosimilar therapy to end of recorded BSRBR-RA follow-up, years	1.9 (1.1, 3.4)	2.1 (1.3, 3.6)	2.1 (1.0, 3.3)	1.3 (0.7, 2.9)
Stop reason (according to the clinician)
Ineffectiveness	161 (47%)	68 (54%)	88 (65%)	5 (6%)
Adverse event	96 (28%)	25 (20%)	25 (19%)	46 (55%)
Death	6 (2%)	—	—	6 (7%)
Other	53 (15%)	26 (20%)	8 (6%)	19 (23%)
Missing	30 (9%)	8 (6%)	14 (10%)	8 (10%)
Time to restart, days	—	0 (0, 7)	44 (10, 123)	—

All continuous variables are reported as median (inter-quartile range).

Disease activity

For the patients with disease activity data available (505 patients per cohort), disease activity on the index date (date of switch/matched date) was low and similar between the cohorts; median tender joint count 1, median CRP 5 mg/l, and median DAS28 2.9 (see Table 3). Similar proportions of patients were in DAS28 remission (DAS28 <2.6); 38% in the biosimilar cohort and 39% in the originator cohort.

Table 3.

Disease activity at the index date, after 6 months and 12 months of treatment; stratified by cohort

	Etanercept Biosimilar Cohort	Matched Etanercept Originator Cohort
Total, n	505	505 (unique patients: 488)
Female, n (%)	381 (75%)	381 (75%)
Age, years
At start of originator	57 (48, 64)	57 (48, 64)
At start of biosimilar/matched date	64 (55, 70)	64 (55, 71)
Disease duration, years
At start of originator	8 (4, 15)	8 (4, 15)
At start of biosimilar/matched date	17 (8, 24)	17 (8, 24)
Line of therapy (originator)
1	364 (72%)	372 (74%)
2	106 (21%)	99 (20%)
3	24 (5%)	25 (5%)
4+	11 (2%)	9 (2%)
Time on originator before switch date, years	5.1 (2.0, 12.5)	4.6 (1.7, 12.6)
Follow-up in BSRBR-RA following switch date, years	2.6 (1.8, 4.6)	2.2 (1.4, 4.1)
Disease activity (imputed data)
28 Tender joint count
Index date	1 (0, 4)	1 (0, 3)
6 months after index date	1 (0, 4)	1 (0, 4)
12 months after index date	1 (0, 4)	1 (0, 4)
Erythrocyte sedimentation rate (ESR), mm/h
Index date	16 (8, 27)	15 (8, 28)
6 months after index date	14 (7, 27)	13 (6, 27)
12 months after index date	12 (6, 27)	15 (7, 29)
C-reactive protein (CRP), mg/L
Index date	5 (2, 9)	5 (2, 10)
6 months after index date	5 (2, 9)	5 (2, 10)
12 months after index date	5 (2, 10)	5 (2, 14)
Patient’s global assessment (PGA), (0–100) mm
Index date	35 (18, 54)	30 (12, 50)
6 months after index date	40 (20, 61)	37 (17, 60)
12 months after index date	34 (16, 54)	42 (18, 64)
28-joint Disease Activity Score (DAS28)
Index date	2.9 (2.2, 4.1)	2.9 (2.2, 3.9)
6 months after index date	3.1 (2.3, 4.4)	2.8 (2.0, 4.3)
Change (index date to 6 months)	0.1 (–0.7, 1.0)	0.0 (–0.9, 1.0)
12 months after index date	2.8 (2.1, 3.8)	3.0 (2.2, 4.1)
Change (index date to 12 months)	–0.1 (–1.0, 0.8)	0.1 (–0.9, 1.1)
DAS28 state, %
Index date
Remission (DAS28 < 2.6)	38%	39%
Low (≥2.6 and ≤3.2)	21%	23%
Moderate (<3.2 and ≤5.1)	30%	27%
High (>5.1)	11%	11%
After 6 months
Remission (DAS28 < 2.6)	34%	42%
Low (≥2.6 and ≤3.2)	19%	15%
Moderate (<3.2 and ≤5.1)	32%	29%
High (>5.1)	15%	13%
After 12 months
Remission (DAS28 < 2.6)	42%	38%
Low (≥2.6 and ≤3.2)	20%	18%
Moderate (<3.2 and ≤5.1)	28%	32%
High (>5.1)	10%	12%
Change in DAS, %
After 6 months
Improved (by >0.6 units)	26%	32%
No change (±0.6 units)	38%	34%
Worsened slightly (by >0.6 units)	14%	13%
Severely worsened (by >1.2 units)	22%	21%
After 12 months
Improved (by >0.6 units)	36%	31%
No change (±0.6 units)	34%	32%
Worsened slightly (by >0.6 units)	12%	13%
Severely worsened (by >1.2 units)	18%	24%
Model 1: Odds of DAS28 worsening (>0.6 units)
6 months
Proportion worsening, %	36%	34%
Crude logistic regression; odds ratio (95% CI)	1.05 (0.72, 1.54)	[reference]
Adjusted for DAS28 at index date	1.10 (0.77, 1.57)	[reference]
Adjusted for age, gender, disease duration, line of therapy, time on Enbrel, TJC, ESR, CRP, PGA and DAS28 at index date	1.10 (0.76, 1.61)	[reference]
12 months
Proportion worsening, %	30%	37%
Crude logistic regression; odds ratio (95% CI)	0.73 (0.47, 1.15)	[reference]
Adjusted for DAS28 at index date	0.74 (0.47, 1.16)	[reference]
Adjusted for age, gender, disease duration, line of therapy, time on Enbrel, TJC, ESR, CRP, PGA and DAS28 at index date	0.71 (0.44, 1.15)	[reference]
Model 2a: Odds of DAS28 worsening (>0.6 units)
6 months
Proportion worsening, %	41%	39%
Crude logistic regression; odds ratio (95% CI)	1.06 (0.75, 1.51)	[reference]
Adjusted for DAS28 at index date	1.09 (0.78, 1.52)	[reference]
Adjusted for age, gender, disease duration, line of therapy, time on Enbrel, TJC, ESR, CRP, PGA and DAS28 at index date	1.11 (0.77, 1.59)	[reference]
12 months
Proportion worsening, %	41%	46%
Crude logistic regression; odds ratio (95% CI)	0.80 (0.54, 1.19)	[reference]
Adjusted for DAS28 at index date	0.81 (0.55, 1.19)	[reference]
Adjusted for age, gender, disease duration, line of therapy, time on Enbrel, TJC, ESR, CRP, PGA and DAS28 at index date	0.81 (0.54, 1.21)	[reference]

All continuous variables are reported as median (inter-quartile range), unless stated otherwise. Index date: date of switch for biosimilar patients, or matched date for originator patients.

^aPatients who stop treatment were reclassified as ‘worsening’ (i.e. non-responders) unless stopped originator for a non-medical switch to an etanercept biosimilar (within 32 days). Patients who stopped for remission were reclassified as ‘improving’ (i.e. responders).

After six months, median disease activity remained similar between the cohorts; median tender joint remained 1, median CRP remained 5 mg/l, and median DAS28 was 3.1 (biosimilar) and 2.8 (originator), with a median change in DAS28 of 0 units. The proportion of patients in DAS28 remission was 34% in the biosimilar cohort and 42% in the originator cohort, while 26% of the biosimilar patients and 32% of the originator cohort had an improvement in DAS28 by >0.6 units, respectively. Patients who switched to biosimilar were no more likely to have a worsening of DAS28 (by >0.6 units) by 6 months in either model compared with those who remained on originator therapy.

After one year, median disease activity remained similar between the cohorts; median tender joint count remained 1, median CRP remained 5 mg/l, and median DAS28 was 2.8 (biosimilar cohort) and 3.0 (originator cohort), with a median change in DAS28 of 0 units. The proportion of patients in DAS28 remission was 42% in the biosimilar cohort and 38% in the originator cohort, while 36% of the biosimilar patients and 31% of the originator cohort had an improvement in DAS28 by >0.6 units, respectively. Patients who switched to biosimilar were no more likely to have a worsening of DAS28 (by >0.6 units) by 12 months in either model compared with those who remained on originator therapy.

Similar results were seen in the sensitivity analysis, where patients were restricted to only those who started etanercept originator as their first b/tsDMARD therapy (Supplementary Tables S3 and S4, available at Rheumatology online).

Discussion

This is the largest matched comparative effectiveness analysis of outcomes following 1024 patients switching to a biosimilar for non-medical reasons compared with a matched cohort of patients remaining on the originator, set within a national cohort study. The results suggest that these patients do just as well as those who remained on their etanercept originator over the same period with regard to treatment persistence and disease activity. Patients who switched onto a biosimilar product were no more likely to discontinue etanercept treatment vs those who stayed on the originator, with over 72% of the biosimilar patients and 74% of the originator patients remaining on their therapy at two years. However, one-in-ten (9%) of the biosimilar cohort switched back to the originator product within the first year (12% over the course of follow-up). Overall, patients switching from etanercept originator to biosimilar were no more likely to have a worsening of DAS28 (by >0.6 units) by 6 or 12 months compared with those patients who remained on the originator therapy.

There have been a number of smaller studies in patients with RA switching from etanercept originator to biosimilar [16–21] although very few included a comparison with patients who did not switch. In a study of 358 RA patients switching from originator to biosimilar (no comparison), most (81%) of whom were in remission (DAS28 ≤ 2.6) at the time of switching showed no change in mean DAS28 after 3 and 6 months [22]. A cohort of 933 RA patients switched from etanercept originator to a biosimilar were compared with 286 RA patients who did not switch (reason for not switching unknown) in the Danish observational cohort study DANBIO [23]. This study saw no difference in the proportion of patients flaring 3 months pre- or post-switch. It highlighted that 78% of patients who switched to the biosimilar remained on therapy at one year, similar to the 81% reported in the current analysis. In addition, DANBIO found no difference in retention rates of those RA patients who switched to a biosimilar compared with those who did not switch (adjusted-HR 0.8 [95% CI 0.6, 1.1]). In this current analysis, results show that median change in DAS28 between cohorts is comparable overall, with patients switching from etanercept originator to biosimilar no more likely to have a worsening of DAS28 (by >0.6 units) by 6 or 12 months compared with those patients who remained on originator therapy.

In the UK, the transition for all patients on the etanercept originator to an etanercept biosimilar occurred gradually over a few years throughout the entire NHS. While most patients in this analysis were switched by 2017, 20% were not switched until 2018 or later. Due to the pandemic and a transition to digital data collection within the study, it is possible that a further proportion of patients on the etanercept originator have since switched to biosimilar, but the data have not yet been collected. Data were censored at the date of completion of the patient’s most recent study follow-up form to avoid any bias. As patients within the originator cohort had the potential to start a biosimilar during their follow-up, patients were censored if they switched to a biosimilar within 32 days, rather than considered ‘failing’ treatment. The same was not applied to patients in the biosimilar cohort switching back to the originator as this switch would not be considered a non-medical switch, but rather the patient would be switching back due to an issue with their biosimilar product.

While we can never truly know what might have happened to an individual patient had they not switched therapies, by matching patients who switched onto a biosimilar with similar patients on the originator we are presenting the best possible alternative to establish ‘what could have happened’. The sample size in this study was large, including 1024 matched biosimilar patients. Of the total 1092 biosimilar patients available to match, there were 56 who were unable to be matched either because there was no possible match for them in the originator cohort (n = 48), or there was a match but the originator patients were better matched to a different biosimilar patient (n = 8). Ten patients had no further follow-up available following the switch in the register. Those biosimilar patients who could not be matched were those who had a long disease duration before starting the originator (median 25 years), or those who had started the originator more recently compared with the vast cohort of available originator patients who were more likely to initiate treatment in the early 2000s. Propensity score matching had been explored but the propensity score summarised the patient’s characteristics, which lost the granularity at the individual patient level, with patients being matched with similar propensity scores but markedly different individual characteristics. As with all observational cohort studies, there were missing data. Where stop date of etanercept therapy was not recorded, likely because the clinician had immediately prescribed the next b/tsDMARD, the start date (minus one day) of the next b/tsDMARD therapy was used. While there were missing disease activity data either at the point of switching therapy (or matched date in originator patients), or after 6 or 12 months, this could be accounted for using multiple imputation, taking into account patient characteristics.

In conclusion, this large analysis of 1024 adults with RA who switched from etanercept originator to etanercept biosimilar in a non-medical switch is the first to compare outcomes with a matched cohort of patients who did not switch. Patients who switched to a biosimilar were no more likely to discontinue etanercept treatment, with less than one-in-ten switching back to the originator within the first year, and no more likely to have a worsening of disease activity after 6 or 12 months compared with those patients who remained on originator therapy. These data will be reassuring to clinicians and patients regarding any non-medical switch required of them. More data are needed to investigate impact of non-medical switch not just with regard to arthritis disease activity, but also to patient well-being and quality of life, in terms of both short-term outcomes as well as over the long term.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The data that support the findings of this study are available from the British Society for Rheumatology. Restrictions apply to the availability of these data (see: https://www.rheumatology.org.uk/practice-quality/registers/requesting-registers-data).

Contribution statement

All the authors have provided substantial contributions to the conception or design of the work, the acquisition of the data and the interpretation of data. L.K-F. and A.R. performed the statistical analysis, and M-F.T, S.S.Z., M.L., K.D.W, and K.L.H. all contributed to the analysis interpretation. L.K-F. wrote the first draft. All the other authors participated in the final drafting of the work or revising it critically for important intellectual content. All authors contributed to the final approval of the version published.

Funding

This work was supported by the British Society for Rheumatology (BSR). The BSR commissioned the BSR Biologics Register in Rheumatoid Arthritis (BSRBR-RA) as a UK-wide national project to investigate the safety of biologic and other targeted therapies in routine medical practice. K.L.H. is the chief investigator. BSR currently receives restricted income from UK pharmaceutical companies, including Abbvie, Amgen, Celltrion Healthcare, Eli Lilly, Galapagos, Pfizer, Samsung Bioepis and Sanofi and in the past Hospira, MSD, Roche, Sandoz, SOBI and UCB. This income finances a wholly separate contract between the BSR and the University of Manchester. The chief investigator and the BSRBR-RA team at the University of Manchester have full academic freedom and are able to work independently of pharmaceutical industry influence. The funding source did not play any role in study design, collection, analysis and interpretation of data, or writing of the manuscript. Members of the BSRBR-RA University of Manchester team, BSR trustees, committee members and staff complete an annual declaration in relation to conflicts of interest. All relevant information regarding serious adverse events outlined in the manuscript have been reported to the appropriate pharmaceutical company as per the contractual agreements/standard operating procedures. K.L.H. is supported by the NIHR Manchester Biomedical Research Centre (NIHR203308). S.S.Z. is supported by a National Institute for Health Research Clinical Lectureship.

Disclosure statement: K.L.H. reports honoraria from Abbvie and grants from Pfizer and BMS outside the submitted work and is supported by the NIHR Manchester Biomedical Research Centre (NIHR203308). S.S.Z. reports travel support and consultancy fees from UCB outside the submitted work. All other authors report no conflicts of interest.