Rituximab biosimilar in rheumatoid arthritis: an enhanced-evidence assessment to evaluate equivalence with the originator based on network meta-analysis

The number of patients evaluated in pre-approval trials is more limited for biosimilars than for originators. For this reason, some physicians are reluctant to employ biosimilars in clinical practice and continue to prefer originators. ¹

An original approach to enhance the clinical evidence supporting biosimilars has recently been described. ² , ³ Using this method, a network meta-analysis was carried out, that included not only the equivalence study comparing the biosimilar with the originator, but also randomized studies that compared the originator with the previous standard of care (SOC).

We retrospectively applied this approach to the approval of the rituximab biosimilar (CT-P10) in the treatment of active rheumatoid arthritis (RA) in combination with methotrexate in patients not responsive to methotrexate monotherapy. In particular, we compared rituximab biosimilar CT-P10 with MabThera/Rituxan (originator).

The clinical data on CT-P10 were extracted from the randomized trial carried out by Yoo and colleagues, ⁴ while the meta-analysis published by Hazlewood and colleagues ⁵ provided the data on both the originator MabThera/Rituxan and the SOC, i.e. methotrexate monotherapy. ⁵ The endpoint was the response at 24 weeks according to the American College of Rheumatology (ACR)50. Our network meta-analysis was based on the Bayesian method proposed by the National Clinical Guideline Centre. ⁶ Odds ratio (OR) for all pairwise comparisons was the output of the analysis together with the ranking histogram and 95% credible intervals (CrIs). Since no heterogeneity was found in the clinical material, the Bayesian statistics were performed with a fixed-effect model.

Data on ACR50 response from the three randomized trials selected by Hazlewood and colleagues are shown in Table 1. Our network meta-analysis estimated an OR of 1.30 (95% CrI: 0.636–2.719) for the comparison originator versus biosimilar, 0.21 (95% CrI: 0.089–0.485) for SOC versus biosimilar, and 0.274 (95% CrI: 0.177–0.414) for SOC versus originator. In terms of effectiveness, CT-P10 ranked: first in 76% of Bayesian simulations; second in 24%; third in 0%; the originator: first, 24%; second, 76%; third, 0%; SOC always ranked third (Figure 1). The number of evaluated patients was increased by this approach from 154 (trial of Yoo and colleagues ⁴ ) to 820 (network meta-analysis).

Table 1.

Data of American College of Rheumatology 50 response at 24 weeks reported in the four randomized trials included in our network meta-analysis.

Study	American College of Rheumatology 50 response
	Rituximab biosimilarCT-P10	Rituximab originator	Standard of care
Yoo and colleagues (2017) 4	37/100	15/48
SERENE(Emery and colleagues 2010) 7		44/170	16/172
DANCER (Emery and colleagues 2006) 8		41/122	16/122
Edwards and colleagues (2004) 9		17/40	5/40
Overall	37/100	117/380	37/334

Figure 1.

Comparative effectiveness of three treatments (MabThera plus methotrexate, rituximab biosimilar plus methotrexate, and methotrexate monotherapy) in patients with active rheumatoid arthritis with inadequate response to methotrexate monotherapy evaluated according to Bayesian network meta-analysis (four randomized studies; endpoint = American College of Rheumatology 50 response at 24 weeks). The figure shows the histogram of rankings estimated for each of the three treatments according to the Bayesian probabilistic analysis (fixed-effect model). Each panel indicates, in a series of simulations, how often the treatment concerned ranked first or second or third in effectiveness. MTX, methotrexate.

The 95% CrI estimated by the Bayesian meta-analysis for the above comparison (OR = 1.30; 95% CrI: 0.64–2.72) is close to the 95% confidence interval (CI) reported in the equivalence trial (OR = 1.29; 95% CI: 0.62–2.69): hence, the results of the network meta-analysis concerning this comparison (along with their variability) confirm those found in the equivalence randomized trial (see also Figure S1 in the supplementary material, which is available from the authors upon request).

The clinical endpoint chosen for our network meta-analysis was the rate of ACR50 response. It is well known that three different endpoints defined by the ACR can be employed for assessing the effectiveness of treatments for RA: ACR20, ACR50, and ACR70. In our network meta-analysis, we adopted the endpoint based on ACR50; this endpoint is suitable for the naïve patients with active disease that were enrolled in the various trials reported in Table 1. The appropriateness of this endpoint is confirmed by the lack of heterogeneity that we found in our analysis.

In conclusion, apart from its simplicity, the methodological approach described herein is advantageous because it relies on a clinical material that, in most cases, is already available. On the other hand, the clinical consequences generated by our network meta-analysis should be seen as a reinforcement of the clinical evidence available on the biosimilar and as a confirmation of the homogeneity of the overall clinical material included in this effectiveness assessment.