FormalPara Key Points

Many drug survival studies on interleukin-17 and interleukin-23 inhibitors have emerged. This meta-analysis provides an extensive and inclusive overview of all currently available drug survival data on these biologics.

Interleukin-17 and interleukin-23 inhibitors demonstrated high drug survival rates in psoriasis treatment, with highest rates for guselkumab and risankizumab.

Data from registry/electronic health records provided more information and had less risk of bias than pharmacy/claims databases in the context of drug survival.

1 Introduction

In patients with psoriasis, interleukin (IL)-17 and IL-23 play a major role in the pathogenesis of the disease [1]. The most recently developed biologics for psoriasis target the IL-17 and IL-23 pathway. Four IL-17 inhibitors (secukinumab, brodalumab, ixekizumab, bimekizumab) and three IL-23 inhibitors (guselkumab, tildrakizumab, risankizumab) are currently approved by the European Medicines Agency and US Food and Drug Administration. These drugs showed very good results in randomised clinical trials [2,3,4]. However, this does not necessarily reflect their effectiveness in daily practice. Clinical trials are known for their strict inclusion and exclusion criteria, creating a homogeneous study population. This can impair the generalisability of trial results to the real-world population, which is often more heterogeneous [5]. In addition, differences in adherence to medication can lead to variations between outcomes of clinical trials and the real world [6]. To evaluate treatment in a real-world setting, drug survival, also known as “drug retention” or “drug persistence”, is a commonly used measure. Drug survival is defined as the time that patients remain on the prescribed drug and is visualised using Kaplan–Meier curves. The outcomes of drug survival analyses can give insights in the number of patients discontinuing their treatment, but also in the reasons for discontinuation in daily practice. Main reasons for discontinuation are ineffectiveness and side effects. In addition, various patient-related variables can affect drug survival such as sex, body mass index, the presence of psoriatic arthritis or prior experience with other biologics [7].

Previously published systematic reviews on drug survival in patients with psoriasis focused on tumor necrosis factor-α inhibitors and the IL-12/23 inhibitor ustekinumab, except for Mourad et al. [8,9,10], who included secukinumab, ixekizumab and guselkumab. Since that time, two more IL-17 inihibitors and two more IL-23 inhibitors have become available, resulting in many new publications on drug survival of IL-17 inhibitors and IL-23 inhibitors. A review and meta-analysis on the drug survival of the newer biologics (IL-17 and IL-23 inhibitors) are not yet available. The advanced methodology used in this meta-analysis summarised the total course of drug survival curves. This provides more robust and precise summary drug survival estimates that enhance the reliability of findings. For patient-tailored treatment, a comprehensive overview of the newer biologics is essential in making evidence-based choices among the newer biologics available for psoriasis.

2 Methods

A systematic review and meta-analysis of real-world evidence on drug survival of IL-17 and IL-23 inhibitors for the treatment of psoriasis was conducted. The literature search and reporting were done according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guideline checklists [11,12,13]. The study protocol was registered in PROSPERO (CRD42021297356).

2.1 Literature Search

A literature search of PubMed, EMBASE, Cochrane Library and Web of Science was developed with the help of an institutional librarian and conducted by two authors (ST and LB, last search 27 December, 2023) to identify eligible studies. The search terms included several alternatives for drug survival analysis, such as ‘Kaplan–Meier estimate’, ‘drug adherence’, ‘drug failure’, ‘drug retention’, ‘drug persistence’ and ‘drug discontinuation’, combined with synonyms for psoriasis and the available biologics. The full search strategy can be viewed in Table 1 of the Electronic Supplementary Material (ESM).

2.2 Study Selection

Two authors (ST/LB) independently screened and selected relevant studies by using the Rayyan web tool [14]. Inclusion criteria were (1) the design was a cohort study, (2) the study subjects were patients aged ≥ 18 years with plaque psoriasis and (3) drug survival of at least one of the following biologics was described: secukinumab (SEC), ixekizumab (IXE), brodalumab (BRO), risankizumab (RIS), guselkumab (GUS) or tildrakizumab (TIL). Exclusion criteria were (1) studies with a primary focus on patients with psoriatic arthritis (e.g. selected from a rheumatological cohort), (2) studies with fewer than ten study subjects and (3) studies in another language than English.

When a full-text version was not available, or in case of other crucial missing data, authors of the specific study were contacted. All studies were carefully screened for overlapping patient populations and authors were contacted in case of doubt. In case of no response, only the cohort with the longest follow-up was analysed. Complex decisions regarding whether to include specific outcomes of separate studies were deliberated within the study team (ST, MS EJ, JR).

2.3 Data Extraction

The following data were extracted by ST and LB and implemented in a pre-designed data-extraction spreadsheet: study design, author, year of publication, location, time frame, study design, setting, information source (electronic health records [EHR]/registry data; or pharmacy/claims data), patient population size, follow-up period, patient characteristics (age, sex, body mass index, age at onset of psoriasis, disease duration, baseline Psoriasis Area Severity Index score, presence of concomitant psoriatic arthritis) percentage biologic-naive patients, type of biologic treatment (IL-17 or IL-23 inhibitor), dosage, treatment regime, treatment duration. Drug survival was depicted as overall drug survival, ineffectiveness-related drug survival, adverse event-related drug survival, and drug survival for biologic-naive or experienced patients.

2.4 Methodological Quality Assessment

The quality of the included studies was assessed with the Quality in Prognostic Studies (QUIPS) tool and the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool [15, 16]. The QUIPS tool was partly adjusted in discussion with the study group to fit the study methodology of drug survival analyses (see Appendix 1 of the ESM). Two authors (ST/LB) independently evaluated each domain for all articles, resulting in an overall risk of bias (RoB) score per domain. In case of disagreement, a third author (JR) was consulted. The quality of evidence was also summarised using the Quality Rating Scheme for Studies and Other Evidence, a modification from the Oxford Centre for Evidence-Based Medicine.

2.5 Statistical Analysis

As we considered summary drug survival curves most informative to compile drug survival studies, we used a non-parametric random-effects model to retrieve a distribution-free summary drug survival curve described in detail by Combescure et al. [17] This method obtains a distribution-free summary drug survival curve by expanding the product limit estimator of drug survival for aggregated drug survival data. The extension of DerSimonian and Laird methodology for multiple outcomes was applied to account for between-study heterogeneity [17]. The I2 statistic was used to measure the between-study variability of the arcsine transformed conditional survival estimates [18]. In contrast to a meta-analysis of drug survival at a single point in time, the homogeneity assumption is that the conditional drug survival probabilities are equal in the studies for any time t.

The main advantage of this approach over meta-analyses of drug survival probabilities at a single timepoint lies in the ability to use full drug survival curves. The estimated pooled drug survival at time t includes all studies, also studies ended before t, because the conditional drug survival probabilities before t are estimated with these same studies.

Summary drug survival curves with 95% confidence intervals (CIs) [based on Greenwood’s formula] were estimated from the drug survival rates and the numbers-at-risk extracted from studies included in the meta-analysis. Drug survival probabilities at each timepoint were extracted using a semi-automated tool (Webplotdigitizer Version 4.5; https://automeris.io/WebPlotDigitizer/) at monthly intervals. The numbers of at‐risk participants during different time intervals were calculated using the method previously described by Williamson et al. [19] and Tierney et al. [20] Heterogeneity was measured using I2 values and Cochran’s Q statistic. Statistical analyses were performed using R (R Foundation for Statistical Computing, Vienna, Austria) with package ‘MetaSurv’.

Summary drug survival (pooled) curves of all separate biologics were constructed for different discontinuation reasons (events): overall drug survival, ineffectiveness-related drug survival and adverse event-related drug survival. As biologic naivety has shown to be an important variable influencing drug survival, drug survival data on biologic-naïve and biologic-experienced patients were also extracted if available [21]. In case no Kaplan–Meier curve was available, and drug survival rates were only described at specific timepoints in the text or tables, these rates were extrapolated to earlier timepoints and incorporated in sensitivity analyses but not in the primary analyses. Additionally, separate sensitivity analyses were performed excluding studies, which were classified as a high risk of bias in the QUIPS tool and as a serious risk of bias in the ROBINS-I tool.

Studies based on data from registry/EHR databases and studies using pharmacy/claims data were analysed as separate groups as the underlying information leading to drug survival was different and might influence drug survival. In registry/EHR studies, drug survival is not derived from data on insurance claims, but from the medical records (e.g. patient registry data/medical record investigations). The actual use by the patient, reason for discontinuation (including being lost to follow-up), temporary dose changes and definitive discontinuation dates are recorded in registry/EHR databases, whereas they are mostly not recorded in pharmacy/claims databases. Albeit being less precise on these issues, pharmacy/claims databases lead to information in large groups of patients. Therefore, summary drug survival curves were constructed separately for (I) registry/EHR data and (II) pharmacy/claims data.

An overview of which study was included in each outcome can be found in Table 10 of the ESM. Additionally, in all figure legends, the references of the included studies for that specific outcome were stated.

2.6 Direct Comparison Summary Drug Survival Estimates

Summary drug survival estimates from the meta-analyses were directly compared at 1, 2 and 3 years between the different biologics for the overall drug survival and ineffectiveness-related drug survival using the methodology described by Klein et al. [22], and presented as differences in drug survival estimates with 95% CIs.

3 Results

3.1 Study Characteristics

The literature search resulted in 2299 records, after screening for duplicates 1615 unique records remained. Of these, 127 full-text articles were assessed for eligibility, resulting in 69 articles included in this review (Fig. 1, Appendices 2 and 3 of the ESM).

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. *Some studies were classified with more than one exclusion reason. IL interleukin

Full size image

3.2 Quality Assessment

An overview of the quality assessment per domain using the QUIPS and ROBINS-I tool is provided in Tables 3 and 4 of the ESM. All studies that were assessed as high risk of bias in the QUIPS tool [23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74] and as serious risk of bias in the ROBINS-I tool [28, 31, 36, 45, 47, 52, 54,55,56,57, 60, 63, 66, 67, 69, 73, 75, 76] were excluded in separate sensitivity analyses. Results of the separate sensitivity analyses were in line with results of the main analyses and shown in Tables 7 and 8 of the ESM. Excluding studies marked as serious risk of bias using the ROBINS-I tool, summary survival estimates of registry/EHR studies were very similar. Pharmacy/claims database studies all had to be excluded because of their serious risk of bias assessment according to the ROBINS-I tool. When using the QUIPS tool to assess the risk of bias, many studies had to be excluded and summary survival estimates slightly changed in both directions. However, in general, results were still in line with the main analyses.

Using the Quality Rating Scheme for Studies and Other Evidence, most studies were rated with a 3: ‘case-control studies; retrospective cohort study’ (Table 5 of the ESM).

3.3 Systematic Review and Meta-Analysis

Forty-seven articles reported on SEC (23,960 patients), 31 on IXE (12,446 patients), 13 on BRO (2353 patients), 24 on GUS (8174 patients), 7 on RIS (1427 patients) and 4 on TIL (304 patients). In total, 48,704 patients were included in this literature review. The characteristics from the included studies are given in Table 2 of the ESM.

As stated, studies based on drug survival from registry/EHR data and studies using pharmacy/claims data were analysed separately. (I) For registry/EHR data, several drug survival outcomes were analysed based on the available studies per agent. Our literature search yielded studies on overall drug survival for SEC, IXE, BRO, GUS, RIS and TIL, for drug survival split for biologic naivety for SEC, IXE, BRO, GUS and RIS, for ineffectiveness-related drug survival for SEC, IXE, BRO and GUS, and for adverse event-related drug survival for SEC, IXE and GUS. Regarding the separate biologics, SEC had the most available registry/EHR studies (34), followed by IXE (21), GUS (20), BRO (10), RIS (6) and TIL (4).

(II) In pharmacy/claims databases, discontinuation reasons were missing hence only overall drug survival studies were identified. A meta-analysis of pharmacy/claims database studies could be performed for SEC, IXE, BRO and GUS as these drugs had multiple studies available. Results split for biologic naivety of patients on SEC, IXE and GUS could also be included to construct summary drug survival curves.

3.3.1 Registry/Electronic Health Record Data

The registry/EHR data extracted were provided by medical records (42 studies; 11,365 patients) and patient registries (13 studies; 10,154 patients) from 29 different countries, mainly located in Europe. In Table 1, summary (drug) survival estimates (SSE) with 95% CIs per biologic regarding overall drug survival, ineffectiveness-related drug survival, adverse event-related drug survival and drug survival split for biologic naivety at 1, 2, 3 and 5 years of treatment are provided. Summary survival estimates for overall, ineffectiveness-related and adverse event-related drug survival were high for all included biologics (for instance, SSE all ≥ 0.8 at year 1) (see Table 1, Figs. 2 and 3). All estimates for biologic-naive patients were higher than the estimates of the same biologic for the experienced patients. For example, for IXE naive versus experienced at year 1, SSE were 0.83 (95% CI 0.77–0.89) and 0.72 (95% CI 0.65–0.80), respectively. Risankizumab showed the highest SSE for overall drug survival at years 1, 2 and 3 (all SSE > 0.86). Overall drug survival contained data on all biologics, whereas in the differentiated analyses (such as ineffectiveness, adverse events and biologic naivety) not all biologics were represented, especially RIS and TIL. These differentiated analyses showed that GUS consistently had the highest SSE on almost all drug survival outcomes; for example, the GUS SSE was 0.87 (95% CI 0.84–0.91) for 5-year ineffectiveness-related drug survival (Table 1). The only exception was SEC drug survival, which was highest in biologic-naive patients at year 1 with an SSE of 0.86 (95% CI 0.82–0.89). Summary drug survival estimates of the IL-17 inhibitors SEC, IXE and BRO were similar to each other for the 1-year and 2-year overall drug survival and ineffectiveness-related drug survival. One-year and 2-year adverse event-related drug survival of IL-17-inhibitors (SEC, IXE) was similar to that of GUS (SSE GUS 1 and 2 year 0.95 (95% CI 0.91–0.98) and 0.90 (95% CI 0.84–0.96)). The summary drug survival curves of the meta-analysis for adverse event-related drug survival and for drug survival split for biologic naivety are displayed in Figs. 1, 2 and 3 of the ESM. Heterogeneity between studies was low (I2 < 29%), see Table 1. Results of the sensitivity analyses, in which also studies where no Kaplan–Meier curves were provided [38, 47, 55, 77] were included, were very similar to the primary analyses, see Table 6 of the ESM.

Table 1 Summary drug survival estimates (95% confidence intervals) per drug survival outcome
Full size table
Fig. 2
figure 2

Non-parametric random-effects summary drug survival curves with 95% confidence intervals for overall survival. Summary drug survival curves in blue, separate studies in grey. A Secukinumab [24,25,26, 33, 34, 39,40,41, 44, 46, 48, 50, 51, 53, 62, 64, 68, 70, 72, 74, 80,81,82,83,84,85], B ixekizumab [23, 32, 33, 39, 40, 42,43,44, 46, 49, 62, 70, 81, 84, 85], C brodalumab [39, 40, 58, 62, 84,85,86,87], D guselkumab [29, 30, 33, 35, 37, 43, 44, 50, 59, 70, 84, 85, 88, 89], E risankizumab [27, 50, 84, 85, 90], F tildrakizumab [27, 50, 84, 91]. The numbers of at-risk patients in the whole cohort at the beginning of each year are reported

Full size image
Fig. 3
figure 3

Non-parametric random-effects summary drug survival curves with 95% confidence intervals for ineffectiveness-related drug survival. Summary drug survival curves in blue, separate studies in grey. A Secukinumab [24, 34, 65, 71, 77, 83], B ixekizumab [23, 71, 77], C brodalumab [71] (drug survival estimate instead of summary drug survival estimate), D guselkumab, E risankizumab [71] (drug survival estimate instead of summary drug survival estimate). The numbers of at-risk patients in the whole cohort at the beginning of each year are reported

Full size image

3.3.2 Pharmacy/Claims Data

Table 1 also shows drug survival data (SSE from the meta-analysis or separate Kaplan–Meier drug survival estimates) for SEC, IXE, BRO and GUS from pharmacy/claims databases. Fourteen pharmacy/claims database studies (27,521 patients) could be included from nine different countries, most of which were conducted in North America). In Figures 4, 5 and 6 of the ESM, a visualisation of summary drug survival curves is provided (SEC, IXE, BRO, GUS, RIS). Summary survival estimates of pharmacy/claims data for 1-year and 2-year overall drug survival were low compared to SSE from registry/EHR data (e.g. 1-year and 2-year overall SSE for SEC pharmacy/claims data of 0.67 (95% CI 0.61–0.75) and 0.49 (95% CI 0.41–0.59) versus 0.81 (95% CI 0.77–0.85) and 0.66 (95% CI 0.61–0.72) in registry/EHR data, respectively). Heterogeneity between studies varied greatly (range I2, 0–87%), see Table 1. The drug survival percentages of the sensitivity analysis using extrapolation of point estimates [31, 57] for pharmacy/claims databases are also reported in Table 6 of the ESM.

3.4 Direct Comparison Summary Drug Survival Estimates

Risankizumab had statistically significantly higher SSE for the overall drug survival at years 1, 2 and 3, compared with SEC and IXE, and higher rates at year 2 and 3 compared with BRO [estimated differences and 95% CIs SEC-RIS at years 1, 2 and 3; − 0.11 (95% CI − 0.17, − 0.04), − 0.22 (95% CI − 0.32, − 0.12), and −0.33 (95% CI − 0.49, − 0.17), respectively, IXE-RIS at years 1, 2, and 3; − 0.12 (95% CI − 0.21, − 0.03), −0.21 (95% CI − 0.33, − 0.08), and − 0.24 (95% CI − 0.39, − 0.10), respectively, BRO-RIS at years 2 and 3; −0.17 (95% CI − 0.34, − 0.00), and − 0.23 (95% CI − 0.44, − 0.02), respectively]. Guselkumab also had statistically significantly higher SES for the overall drug survival at years 2 and 3, compared with SEC and IXE [estimated difference and 95% CI SEC-GUS at years 2 and 3; − 0.15 (95% CI − 0.25, − 0.04) and − 0.24 (95% CI − 0.40, − 0.08), respectively, IXE-GUS at years 2 and 3; − 0.13 (95% CI − 0.27, − 0.00) and − 0.16 (95% CI − 0.31, − 0.00), respectively], and higher rates at years 1, 2 and 3 for ineffectiveness-related drug survival compared with SEC [estimated difference and 95% CI SEC-GUS at years 1, 2 and 3; − 0.06 (95% CI − 0.11, − 0.01), − 0.16 (− 95% CI 0.23, − 0.09) and − 0.25 (95% CI − 0.36, − 0.15), respectively]. At 3 years, the ineffectiveness-related drug survival of IXE was significantly higher than that of SEC [estimated difference and 95% CI − 0.18 (− 0.32, − 0.03)]. An overview of all pairwise comparisons is displayed in Table 9 of the ESM.

4 Discussion and Conclusions

This systematic review and meta-analysis was performed to investigate the drug survival of IL-17 and IL-23 inhibitors in patients with psoriasis. A total of 69 studies including 48,704 patients were systematically reviewed to assess the drug survival of IL-17 and IL-23 inhibitors. Detailed summary drug survival curves were constructed to provide insight into the drug survival curves per drug over time, analysed separately for different discontinuations reasons (ineffectiveness and adverse events) and biologic naivety. Summary drug survival estimates, also for ineffectiveness-related drug survival, were similar for SEC, IXE and BRO, but ineffectiveness-related drug survival of IXE was significantly higher than drug survival of SEC at 3 years, indicating that patients on IXE are less likely to discontinue their drug because of ineffectiveness than patients on SEC. Risankizumab had the highest SSE for overall drug survival at 1, 2 and 3 years. Guselkumab had the highest SSE at 1, 2 and 3 years for almost all differentiated (e.g. ineffectiveness-related and adverse event-related drug survival) outcomes compared with the other biologics. Note that in the differentiated outcomes, such as ineffectiveness-related drug survival, not all biologics were consistently present. In line with previous findings, drug survival for biologic-naive patients was superior to that of biologic-experienced patients. Estimates of drug survival based on pharmacy/claims databases were substantially lower, indicating a worse performance of these drugs compared to the analyses based on registry/EHR data. By utilising the method by Combescure et al. in our meta-analysis, we were able to implement drug survival probabilities from each month of the full reported follow-up duration, constructing precise drug survival estimates.

In previous systematic reviews and meta-analyses on the efficacy of biologics for the treatment of psoriasis, favourable outcomes have been reported. [8,9,10] However, analyses were performed at specific timepoints (e.g. at 1 and 2 years), which results in an under-representation of studies reporting drug survival at other timepoints. As stated, to overcome this limitation the Combescure method was used, which permits inclusion of the full drug survival curves. Mourad and Gniadecki performed a comparative meta-analysis of hazard ratios specifically for the drug survival of biological treatment. The 2-year and 5-year drug survival data at that time were insufficient to compare drug survival for the IL-17 and IL-23 inhibitors yet. Their 1-year pairwise comparisons showed a superiority of SEC over IXE [10]. Prior extensive reviews on the drug survival of tumor necrosis factor-α inhibitors and ustekinumab reported lower pooled annual drug survival rates for the tumor necrosis factor-α inhibitors etanercept, adalimumab and infliximab after 1 (all < 0.74) and 3 years (all <0.54) compared to our IL-17 and IL-23 rates (all point estimates ≥ 0.8). For ustekinumab, similar drug survival rates to our results for SEC, IXE and BRO were reported. Guselkumab and RIS overall drug survival rates in our study were substantially higher than previously reported etanercept, adalimumab, infliximab and ustekinumab rates up to 5 years [8, 9]. Guselkumab and RIS drug survival rates were also statistically significantly higher than SEC, IXE and BRO rates in this meta-analysis, which is possibly related to the upstream effect of IL-23 inhibitors in the IL-23/IL-17 cytokine pathway [78].

Remarkably, summary drug survival estimates of pharmacy/claims databases were noticeably lower than these estimates from registry/EHR databases (e.g. 2-year overall summary drug survival estimate for SEC pharmacy/claims data of 0.49 (95% CI 0.41–0.59) vs 0.66 (95% CI 0.61–0.72) in registry/EHR data), and the I2 statistic for heterogeneity was higher in pharmacy/claims database studies compared with registry/EHR studies. As the total number of patients in the many registry/EHR studies is comparable to the total number of patients in the pharmacy/claims studies, the difference in drug survival outcomes is not likely explained by a difference in precision of the estimate. In pharmacy/claims database studies, the administrative claims for medication are used, and the actual medication use of the patient is not verified at a physician and patient level. Missing information on the cause of discontinuation and the exact date of discontinuation might pose possible hazards in the interpretation of data from pharmacy/claims databases. Events that are not related to the drug performance, for example, insurance issues, relocation or factors such as family planning, cannot be distinguished from drug-related issues. As for detailed drug survival analyses, the nature and timing of discontinuation events are utterly important and may outweigh the advantage of including large populations from claims databases. Moreover, especially for claims databases, results are generalisable to the insured population, but not necessarily to uninsured patients, or patients with other insurance types.

The study’s literature search was constrained to English-language publications, potentially introducing language bias. Drug survival studies reporting on bimekizumab were not yet available. Tildrakizumab was included in our study, however at the time of our search, only short-term follow-up drug survival studies were available. RoB assessments using the QUIPS and ROBINS-I tool led to a subset of studies with high/serious RoB, which were excluded in separate sensitivity analyses. There were no studies which could be classified as low RoB using both tools, this should be taken into consideration when interpreting results. Furthermore, most studies included in this review reported on overall drug survival. It should be kept in mind that several discontinuation reasons underlie this outcome which may not all be related to the performance of the drug itself, like wish for pregnancy or financial reasons [77, 79]. The coverage and reimbursement policies of health insurance plans and formularies can influence drug survival rates. Restrictions imposed by insurance companies, such as prior authorisation requirements or limited formulary options, may create barriers to accessing certain medications. Patients who face difficulties in obtaining insurance coverage for a prescribed medication may be more likely to discontinue treatment. This further underscores the importance of drug survival analyses with a focus on specific discontinuation reasons (ineffectiveness, adverse events). It is crucial to register financial reasons separately in order to prevent them from influencing drug survival rates with regard to ineffectiveness and adverse events. In future studies, we would strongly encourage reporting drug survival separately for different discontinuation reasons (instead of combining all reasons in an overall drug survival) and effect modifiers.

Furthermore, we want to highlight that a given drug is always both prescribed and discontinued in a landscape of competing drugs. The quantity of available alternatives likely affects the decisions made by doctors. When the selection of alternative options is restricted, doctors are likely more inclined to maintain their patients on a specific drug. In contrast, when many treatment options are available, treating physicians as well as patients could decide easier for switching to a consecutive biological treatment. In addition to doctors adjusting their prescription practices, patients' perspectives can also evolve, as they might strive for higher therapeutic objectives, potentially leading to earlier consideration of switching. In the current study timeframe however, there were consistently multiple 'older' biologic alternatives (such as ustekinumab, adalimumab, etanercept, infliximab) as well as the small-molecule apremilast available alongside the newest biologics included, which also entered the market rapidly during the studied period. The number of patients who continued with their IL-17/IL-23 inhibitor because there were no alternatives available was considered very low, thereby minimising the potential impact of drug availability on our findings.

This meta-analysis showed that investigated IL-17 and IL-23 inhibitors had high drug survival rates, with highest rates for GUS and RIS drug survival. Attention for effect modifiers (biologic naivety), and source of data (registry/EHR data vs pharmacy/claims databases) is relevant when interpreting drug survival studies.