Abstract
Objective. To evaluate longterm drug survival (proportion of patients still receiving treatment) and discontinuation of etanercept (ETN), infliximab (IFX), adalimumab (ADA), certolizumab pegol (CZP), and golimumab (GOL) using observational data from patients with rheumatoid arthritis (RA).
Methods. Following a systematic literature review, drug survival at 12 and 12–24 months of followup was estimated by summing proportions of patients continuing treatment and dividing by number of studies. Drug survival at ≥ 36 months of followup was estimated through Metaprop.
Results. There were 170 publications included. In the first-line setting, drug survival at 12 months with ETN, IFX, or ADA was 71%, 69%, and 70%, respectively, while at 12–24 months the corresponding rates were 63%, 57%, and 59%. In the second-line setting, drug survival at 12 months with ETN, IFX, or ADA was 61%, 69%, and 55%, respectively, while at 12–24 months the corresponding rates were 53%, 39%, and 43%. Drug survival at ≥ 36 months with ETN, IFX, or ADA in the first-line setting was 59% (95% CI 46–72%), 49% (95% CI 43–54%), and 51% (95% CI 41–60%), respectively, while in the second-line setting the corresponding rates were 56% (95% CI 52–61%), 48% (95% CI 40–55%), and 41% (95% CI 36–47%). Discontinuation of ETN, IFX, and ADA at 36 months of followup was 38–48%, 42–62%, and 38–59%, respectively. Data on CZP and GOL were scarce.
Conclusion. After > 12 months of followup, more patients with RA receiving ETN remain on treatment compared with other tumor necrosis factor inhibitors.
- DRUG SURVIVAL
- SYSTEMATIC LITERATURE REVIEW
- OBSERVATIONAL STUDY
- METAANALYSIS
- RHEUMATOID ARTHRITIS
- TUMOR NECROSIS FACTOR INHIBITOR
Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by persistent inflammation of the joints that reduces a patient’s physical functioning and quality of life1,2. The direct and indirect costs associated with RA are considerable and longterm, with continued treatment being crucial for maintaining disease remission and maximizing the cost-effectiveness of therapy3. Treatment of RA is predominantly focused on controlling inflammation and pain, as well as slowing the progression of joint destruction and disability. A range of therapeutic agents have been approved for treating RA, including corticosteroids (e.g., methylprednisolone, prednisone, and prednisolone); conventional synthetic disease-modifying antirheumatic drugs [csDMARD; e.g., methotrexate (MTX), hydroxychloroquine, leflunomide, and sulfasalazine]; biologic agents that inhibit tumor necrosis factor (TNF) or other important mediators of the inflammatory pathway, such as the interleukin 6 receptor (e.g., tocilizumab and sarilumab), CD20 (e.g., rituximab), and CTLA-4 (e.g., abatacept); targeted synthetic DMARD (tsDMARD; tofacitinib, baricitinib); and biosimilar versions of some of these agents.
The introduction of TNF inhibitors (TNFi) led to a profound improvement in outcomes for patients with RA4. The first TNFi to be approved for treating RA was etanercept (ETN) in 1998, and this was followed by the approvals of infliximab (IFX), adalimumab (ADA), certolizumab pegol (CZP), and golimumab (GOL). Discontinuation of TNFi treatment is usually reported to be due to a loss or lack of efficacy, or to incident adverse events (AE)5,6. The evaluation of survival times and discontinuation rates associated with each of the specific TNFi is necessary when evaluating their real-world effectiveness and cost-effectiveness7,8. Randomized controlled trials are relatively short compared with the chronic course of RA and therefore real-world studies and clinical registries with longer patient followup can be very useful for assessing healthcare resource use in the longterm management of the disease.
Previous studies that compared the drug survival (i.e., the proportion of patients still receiving treatment at a specific timepoint) of ETN, IFX, and ADA in patients with RA reported conflicting results. Two US studies that used data from the Corrona registry and a US insurance database for 2000–2005 showed that IFX had the highest drug survival9,10. Several European studies, however, reported a shorter drug survival for IFX compared with ETN and ADA, while other studies have reported no difference among these 3 TNFi6,11–17. To the authors’ knowledge, there has been no previous analysis that compared drug survival for all 5 currently approved TNFi. Therefore, we conducted a systematic literature review and metaanalysis with the aim of (1) identifying all publications reporting the longterm drug survival of ETN, IFX, ADA, CZP, and GOL in observational studies of patients with RA; and (2) evaluating the rates of drug survival and discontinuation in patients with RA after ≥ 12 months of followup.
MATERIALS AND METHODS
Literature search
Relevant publications were identified using an approach that followed the Cochrane dual-reviewer methodology and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocols (PRISMA-P) guidelines18. The search was conducted in November 2017 using Embase, MEDLINE, the Cochrane Central Trials Register and Database of Systematic Reviews, other Cochrane Library databases, and PubMed, with no restrictions on language or the year of publication. Abstracts presented at the annual meetings of the American College of Rheumatology and European League Against Rheumatism in 2016 and 2017 were also searched. The search terms included but were not restricted to the following: rheumatoid arthritis, cohort, longitudinal, etanercept, infliximab, adalimumab, certolizumab pegol, golimumab, Flixabi, Renflexis, Inflectra, Remsima, Amjevita, Cyltezo, Imraldi, Benepali, and Erelzi. The reference lists of relevant review articles were also searched.
Articles and conference abstracts published in English and reporting data from observational studies in humans were included, while reports on studies with ≤ 10 participants or with a followup duration < 12 months were excluded. Publications reporting data from preclinical studies or randomized or nonrandomized clinical trials were excluded, as were reviews, guidelines, expert opinions, and case reports. Studies that reported only discontinuations due to treatment switching were also excluded. The publications used a variety of alternative terms to describe drug survival (i.e., the proportion of patients still receiving treatment at a specific timepoint), including retention, persistence, and continuation, and these were considered equivalent for the present analysis. The term drug adherence was not considered equivalent and data on adherence were not included.
Screening and quality assessment of published studies
Publications identified in the literature search were screened by 2 reviewers in a 2-step process: titles and abstracts were first screened against the prespecified inclusion and exclusion criteria to identify eligible publications, and then the same reviewers used the same prespecified criteria to screen the full text of each of the eligible publications. The opinion of a third reviewer was obtained in the event of any disagreement between the 2 reviewers.
The included publications all described nonrandomized, observational studies and therefore their quality was assessed using the Newcastle–Ottawa instrument19. The quality of conference abstracts was assessed using the modified Downs and Black instrument20.
Data extraction and synthesis
Data were extracted from the publications by 1 reviewer and validated by a second reviewer. Drug survival was defined as the average proportion of patients still receiving treatment at 12 and 24 months of followup, and was estimated for each TNFi by summing the percentages reported by each publication and dividing this by the number of studies reporting data. Drug survival at ≥ 36 months of followup was estimated through metaanalysis using Metaprop (Stata/IC 15.1 for Windows, StataCorp). Briefly, the Metaprop procedure pools values that are proportions and then presents weighted subgroup and overall pooled estimates with inverse-variance weights obtained from a random-effects model.
An exploratory analysis of the survival rates of TNFi monotherapy versus TNFi with concomitant DMARD was also conducted. Data from studies reporting that TNFi treatment was concomitant with MTX were pooled with those from studies in which the concomitant DMARD was not reported. Data from any treatment arm that had > 85% concomitant DMARD use were included if they were reported for a timepoint of interest. The estimates at each timepoint were made regardless of TNFi treatment line to include the maximum number of studies possible.
RESULTS
Identification of publications
The initial search yielded a total of 4412 publications (Figure 1). Subsequent screening resulted in 170 publications being included in the analysis: 133 full articles and 37 conference abstracts (see Supplementary Table 1, available with the online version of this article, for a complete reference list). The key characteristics of the studies described in the 170 publications are summarized in Table 1. Of the 133 full articles identified, the quality of all except 4 was rated as either “excellent” or “good,” with the remaining 4 rated as “fair.” Among the 37 conference abstracts, the quality of the majority (26) was rated as “fair,” with 6 rated as “good” and 5 rated as “poor.”
Identification of relevant published articles through systematic literature review. RA: rheumatoid arthritis; TNFi: tumor necrosis factor inhibitor.
Key characteristics of published studies included after the screening process.
Clearly overlapping patient data derived from 6 data sources were reported in 14 of the 170 publications. It was clear from 3 of these 6 sources that identical patient groups were described in the associated publications and therefore only 1 of the publications based on each of these 3 sources was included in the analysis of drug survival and discontinuations. For the remaining 3 sources, all associated publications were included because the results from the overlapping patients could not be separated from the other patients in those studies. There were 64 publications originating from 25 data sources for which the extent of overlap between patient groups within the reported results was unclear. These 64 publications were treated as separate datasets for the analysis of drug survival and discontinuations.
There were only 2 of the 170 identified publications that reported data on biosimilar forms of TNFi. These 2 publications were both abstracts from 2017, with one describing data obtained following switching of patients from originator to biosimilar forms of ETN and IFX21 and the other reporting data on a biosimilar form of ETN evaluated in the DANBIO study22.
Baseline characteristics of the patient populations
The size of the patient populations described in the 170 publications identified in the literature search ranged from 18 to 17,405 patients, with a mean or median followup time ranging from 12 months to 12 years. The majority of the studies included a greater proportion of women than men (typically 70–90% women), and the majority reported a baseline mean or median age in the range of 50–60 years. Data on patient race and/or ethnicity were reported in 9 publications. The mean or median duration of disease at baseline ranged from 2.4 to 18.5 years. The majority of publications (96/170) did not report previous DMARD use. There was 1 publication that reported all patients (100%) were DMARD-naive, while 4 publications reported prior DMARD use in 1–50% of participants, 19 publications reported prior DMARD use in 51–90% of participants, and 50 publications reported that 90–100% of the patients involved were DMARD-experienced. About half of the publications that reported prior biologic DMARD (bDMARD) experience (48/104) stated that 100% of the patients were naive to treatment with bDMARD.
Data on baseline 28-joint Disease Activity Score (DAS28) were inconsistent in representing mean or median values, as well as in presenting overall scores for all treatments versus scores separated according to specific TNFi. The minimum reported mean (SD) DAS28 was 2.5 (0.7) and the maximum mean (SD) DAS28 was 7.3 (1.1). At baseline, the minimum median (interquartile range) Health Assessment Questionnaire (HAQ) score was 0.9 (0.4–1.4) and the maximum mean (SD) HAQ score was 2.1 (0.6). The Supplementary Results (available with the online version of this article) contain additional details on the baseline characteristics of the patient populations and the number of publications that reported them.
Drug survival of TNFi
Eight of the 170 publications did not report sufficient data to be included in the analysis of drug survival and discontinuations, and therefore those endpoints were evaluated from a total of 162 publications. Among these 162 publications, 51 did not report drug survival data, 1 reported the drug survival rate but not the length of followup, and 1 reported mortality only. Therefore, there were 109 publications from which drug survival rates could be estimated: 28 that reported followup data at 12 months only; 20 that reported a followup > 12 months but < 36 months; and 61 that reported a followup ≥ 36 months (1 study reported a followup of 35 mos and was included in the ≥ 36-mos group).
In the first-line treatment setting, the survival rates of the 5 TNFi ranged from 62% to 71% at 12 months and 57% to 72% at 12–24 months of followup. In the second-line treatment setting, the survival rates ranged from 55% to 69% (no rate was reported for CZP) at 12 months and 38% to 57% at 12–24 months of followup (Table 2).
Unweighted average TNFi survival rates in patients with rheumatoid arthritis.
After excluding publications that did not separate data into specific treatment lines, data from 36 publications were pooled for an exploratory metaanalysis of TNFi survival at ≥ 36 months. Estimates of drug survival for CZP (no publications) and GOL (1 publication) were not calculated owing to a lack of data. The highest rate of drug survival in the first-line treatment setting was in the ETN group (59%, 95% CI 46–72%), followed by ADA (51%, 95% CI 41–60%) and then IFX (49%, 95% CI 43–54%; Figure 2A). The ETN group also displayed the highest rate of drug survival at ≥ 36 months in the second-line setting (56%, 95% CI 52–61%), followed by IFX (48%, 95% CI 40–55%) and then ADA (41%, 95% CI 36–47%; Figure 2B).
Exploratory metaanalysis of drug survival at ≥ 36 months for first-line TNFi treatment in patients with RA. The vertical broken line represents the mean of the pooled treatment effect from the random-effect analysis, across all studies and regardless of treatment. Diamonds represent the upper and lower 95% CI around the pooled treatment effect. RA: rheumatoid arthritis; TNFi: tumor necrosis factor inhibitor.
Exploratory metaanalysis of drug survival at ≥ 36 months for second-line TNFi treatment in patients with RA. The vertical broken line represents the mean of the pooled treatment effect from the random-effect analysis, across all studies and regardless of treatment. Diamonds represent the upper and lower 95% CI around the pooled treatment effect. RA: rheumatoid arthritis; TNFi: tumor necrosis factor inhibitor.
There were 25 publications that reported drug survival rates at 48 months, with the rates reported showing a wide range. Drug survival with ETN in the first-line setting ranged from 23% to 69% in 15 of the 17 publications reporting data on this agent, with the remaining 2 publications reporting rates of 3% and 95%. First-line drug survival with ADA ranged from 27% to 54% in 13 of the 15 publications reporting data on this agent, with the remaining 2 publications reporting rates of 4% and 100% (the same 2 publications that also reported the much lower and much higher rates for ETN). Drug survival with IFX in the first-line setting ranged from 18% to 79% in the 20 publications reporting data on this agent. One publication reported a first-line survival rate of 86% for GOL. There were 4 studies that reported survival rates at 48 months of followup in the second-line setting, with rates of 42–55% for ETN, 25–41% for ADA, and 37–39% for IFX.
Exploratory analysis of TNFi survival rates in patients receiving ADA, ETN, or IFX monotherapy showed that ETN had the highest survival rate at each timepoint evaluated (Table 3). Drug survival rates in patients receiving TNFi with concomitant DMARD were also highest in those who received ETN for each timepoint evaluated, except at 12 months when a slightly higher rate was observed with IFX and the highest rate was observed in the much smaller number of patients who received GOL. There was a trend toward improved drug survival in patients who received a TNFi with concomitant DMARD compared with those who received TNFi monotherapy at each timepoint evaluated, except in the case of ADA at 48 months. Drug survival in the 2 groups (TNFi monotherapy and TNFi with concomitant DMARD) at 36 months was investigated by metaanalysis, with the data showing that survival rates, both with or without concomitant DMARD, were higher with ETN than with ADA or IFX (Figure 3). There was a trend toward similar or lower drug survival rates in patients receiving ETN, IFX, or ADA monotherapy compared with those who received 1 of these TNFi with concomitant DMARD. There was significant heterogeneity among the studies included in the metaanalysis, with the inclusion of data on any TNFi treatment line at each timepoint being one potentially important contributing factor.
Exploratory metaanalysis of drug survival at 36 months in patients with RA and receiving either TNFi monotherapy or TNFi with concomitant DMARD. The vertical broken line represents the mean of the pooled treatment effect from the random-effect analysis, across all studies and regardless of treatment. Diamonds represent the upper and lower 95% CI around the pooled treatment effect. DMARD: disease-modifying antirheumatic drug; RA: rheumatoid arthritis; TNFi: tumor necrosis factor inhibitor.
Unweighted average TNFi survival rates according to concomitant DMARD use (all treatment lines included).
Discontinuation of TNFi treatment
Among the 162 publications included in the analysis of outcomes, there were 134 that reported data on a measure of treatment discontinuation. Of these 134 publications, 37 did not report data suitable for the analysis because they presented either a combined group of all TNFi, treatment switching only, deaths only, discontinuations according to treatment courses only and not according to individual patients, or data for which the time frame of the discontinuations was unclear.
At 36 months of followup, IFX treatment was associated with the highest rate of total discontinuations (42–62%), followed by ADA (38–59%), and then by ETN (38–48%) and GOL (35%, derived from a single study involving 109 patients). Regarding discontinuations due to AE at 36 months of followup, patients receiving ETN displayed a lower rate of discontinuation (8–16%; 3 publications) than those who received ADA (8–26%; 7 publications) or IFX (15–27%; 7 publications). Six publications (1 for ETN and 5 for IFX) reported the proportion of patients experiencing an infection that resulted in TNFi discontinuation, across any treatment line, for followup times ≥ 36 months. Followup times varied considerably among these 6 publications (36–120 months), thus limiting formal comparison. There were 10 publications that reported discontinuations due to inefficacy at 36 months of followup, and these ranged from 12% to 15% for ETN, 5–24% for IFX, and 11–33% for ADA.
DISCUSSION
Our present analysis of 133 full publications and 37 conference abstracts included a total of 270,000 patients with RA and shows that treatment with ETN is associated with the highest rate of drug survival among the 5 TNFi agents currently approved for treating the disease. ETN was also associated with the lowest rate of treatment discontinuation. The results from the present analysis are consistent with data published in previous systematic reviews by Arora, et al and Souto, et al23,24, although the present analysis goes further than these studies by reporting data from both first- and second-line treatment settings and by including data on CZP and GOL, albeit to a limited extent because of a lack of published data on these 2 agents. Evidence-based insight into the drug survival and discontinuation rates associated with each of the specific TNFi agents is key when evaluating the real-world effectiveness and cost-effectiveness of these agents when used in longterm dosing regimens for chronic conditions such as RA7,8. Within an environment of increasing pressure on healthcare budgets, data on drug survival can help inform decision making aimed at achieving optimal disease outcomes while limiting costs, and this is especially important in countries that limit the number of bDMARD reimbursed.
Discontinuation of TNFi treatment in patients with RA is most commonly reported to be due to a loss or lack of treatment efficacy and/or due to incident AE. It can also occur when patients achieve remission and wish to reduce their drug exposure; or it can be part of the preparation for surgery or a live vaccine. The higher rates of drug survival with ETN demonstrated in the present analysis may be explained, at least in part, by the lower reported incidence of antidrug antibody development when compared with IFX and ADA25, and/or the potentially lower incidence of serious infections in patients treated with ETN26, although data on rates of infection with the various TNFi are complex to interpret and compare27. It is possible that the underlying cause of the variability in the safety profiles of the currently approved TNFi agents is due to fundamental differences in their biochemistry, which may result in each of them modifying TNF-mediated signaling in a subtly different way as well as conferring on each a slightly different pharmacokinetic and pharmacodynamic profile.
The key strengths of our present analysis are the large number of publications that were reviewed and the systematic and broad approach that was adopted to obtain the data, which were derived entirely from longterm observational studies of TNFi in patients with RA. As well as being a strength of the analysis, the observational characteristics of the data were also a limitation in that a comparison of the clinical outcomes achieved with the 5 different TNFi was not possible. Another limitation is the possibility that some of the patients may have been double counted when identical or similar data were reported within multiple publications included in the analysis. A final limitation is that there were very few data available for either CZP or GOL for some of the variables evaluated despite both these agents having received regulatory approval from the US Food and Drug Administration 9 years ago.
Patients with RA who receive ETN are more likely to continue treatment and less likely to discontinue treatment after > 12 months of followup when compared with those receiving other TNFi.
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
Acknowledgment
The authors thank all the patients, physicians, and researchers who participated in the studies described in this analysis. The authors also thank Abigail Paine of Zedediah Consulting for expert statistical support. The literature search was conducted by Catherine Rolland, PhD, and Eva Scholtus, MS, of Envision Pharma Group and was funded by Pfizer. Medical writing support was provided by David Wateridge, PhD, of Engage Scientific Solutions and was funded by Pfizer.
Footnotes
The systematic literature review to support this manuscript was sponsored by Pfizer. PE declares consultancy fees from AbbVie, Bristol-Myers Squibb, Eli Lilly & Co., and Novartis. BV, PS, MT, HEJ, JW, PVSE, and LM are employees of Pfizer, or were employees of Pfizer when this study was conducted, and hold stock or stock options with Pfizer. AG holds stock in AbbVie, Amgen, Celgene, GSK, Johnson & Johnson, Pfizer, and Regeneron, and declares consultancy fees from AbbVie, Celgene, Merck & Co., and Pfizer, and has served on a speakers’ bureau for AbbVie, Celgene, Flexion, Merck & Co., Novartis, and Pfizer. CR is an employee of Envision Pharma Group and was a paid consultant to Pfizer in connection with the development of the systematic literature review. GC declares consultancy fees from Bristol-Myers Squibb and Pfizer.
- Accepted for publication May 17, 2019.
