Comparative safety of istradefylline in Parkinson’s disease: A systematic review of randomized controlled trials and real-world studies

Highlights:

• •An up-to-date assessment of comparative safety of PD adjuncts was conducted by NMA.
• •Istradefylline had numerically lower odds of TEAEs/TRAEs vs other adjuncts included.
• •All estimates excluding the null value from the 95 % CrI favored istradefylline.
• •Istradefylline exhibited a favorable safety profile relative to other adjuncts.

Abstract

Introduction

Istradefylline offers a novel mechanism (adenosine A_2A receptor antagonism) to treat OFF episodes in Parkinson’s disease (PD). It may potentially offer improved tolerability versus other adjuncts, but comparative safety data are lacking.

Methods

A systematic review and Bayesian network meta-analysis (NMA) incorporating RCTs of PD adjuncts until January 10, 2024, was conducted to estimate relative safety. Inconsistency was assessed and heterogeneity evaluated by global I²-statistic and between-study heterogeneity. Incidences of safety outcomes were summarized from RWE identified according to the same criteria.

Results

100 RCTs and 55 RWE publications were identified; 76 RCTs were included in NMAs. Istradefylline demonstrated lower odds of serious AEs (odds ratio [OR] = 0.56; 95 % CrI: 0.32, 0.99), treatment-emergent AEs (0.43; 0.25, 0.73), treatment-related AEs (0.33; 0.19, 0.56), hallucinations (0.25; 0.06, 0.97), and withdrawal due to AEs (0.37; 0.19, 0.68) versus amantadine. Istradefylline showed lower odds of dyskinesia (0.63; 0.41, 0.99) and hypotension (0.19; 0.03, 0.82) versus catechol-O-methyl transferase inhibitors (COMTi), lower odds of nausea (0.58; 0.33, 0.99) versus dopamine agonists (DA), and lower odds of hypotension (0.09; 0.01, 0.52) versus monoamine oxidase-B inhibitors (MAO-Bi). Sensitivity analysis of RCTs published since 2000 found a reduction in odds of dyskinesia and hallucinations for istradefylline versus DA. RWE were heterogeneous but demonstrated lower incidence of certain AEs with istradefylline, specifically dyskinesia (versus MAO-Bi), somnolence (versus DA and COMTi), peripheral edema and hallucinations (versus amantadine), and nausea (versus all comparators).

Conclusion

Istradefylline exhibits a favorable safety profile versus other PD adjuncts, as demonstrated by RCTs and RWE.

1. Introduction

Parkinson’s disease (PD) is a neurodegenerative disease related to loss of dopaminergic neurons and characterized by motor and non-motor symptoms [1]. While levodopa in combination with dopamine decarboxylase inhibitors (carbidopa, benserazide) is well established as standard of care for these patients [2], long-term treatment is associated with diminishing therapeutic efficacy and increased safety and tolerability issues [3]. These complications may arise due to the dopaminergic mechanism of levodopa, and include motor fluctuations such as wearing-off and dyskinesia, as well as safety issues such as confusion, hypotension, somnolence, dyskinesia, impulse control disorder, hallucinations and more [3]. PD adjunctive therapies are available and include dopamine agonists (DA), monoamine oxidase type B (MAO-B) inhibitors, catechol-O-methyl transferase (COMT) inhibitors [2], istradefylline and amantadine. However, the dopaminergic-related side effects of nearly all (or most) of these agents necessitate finding a balance between effective control of symptoms and limiting associated safety and tolerability issues [1]. These PD adjuncts are associated with a plethora of side effects, ranging from mild to severe, including nausea, confusion, hypotension, somnolence, hallucinations, peripheral edema, impulse control disorder (ICD), dry mouth, and dyskinesia, among others [4], [5], [6], [7], [8], [9].

Istradefylline is a first-in-class adenosine A_2A receptor antagonist approved in the US and Japan as an adjunct to levodopa/decarboxylase inhibitor treatment in adult patients with PD experiencing OFF time/wearing off [10] that acts through a distinct, non-dopaminergic pathway [10]. Adenosine A_2A receptors are strongly involved in regulation of the central nervous system, and are found co-localized with dopaminergic D₂ receptors in the striatum [11]. This creates an environment for an antagonistic interaction between adenosine and dopamine [11], with A_2A activation opposing the effect of D₂ activation [12]. While the precise mechanism by which istradefylline exerts its therapeutic effect in PD is unknown, blocking adenosine A_2A receptors reduces abnormal overactive striatopallidal output in PD, to reduce OFF time [13]. In addition, adenosine A_2A receptors can be found in the shell of the nucleus accumbens [14]. A_2A receptor antagonists are proposed to promote wakefulness by acting on these receptors to block the sleep-promoting effects of adenosine, and therefore promote arousal, which may explain the reduction in daytime somnolence observed with istradefylline [14], [15].

The safety and tolerability of istradefylline in PD have been evaluated in several placebo-controlled trials [16]. However, comparative data on the safety of istradefylline versus other adjuncts are lacking. Previous studies have sought to conduct indirect comparisons between PD treatments, though these have limitations. Stowe et al. 2011 [17] conducted a meta-analysis to evaluate the comparative benefits and risks of these agents. Whilst this supplemented a gap arising from the lack of head-to-head trials in this space, the study presented only a limited focus on safety and included adjuncts approved up to 2010. In 2023, Sako et al. [18] used network meta-analysis (NMA) to compare the efficacy and safety of PD adjuncts, though this reported on only a limited number of safety outcomes (adverse events [AEs], discontinuation due to AEs, dyskinesia, hallucinations, and orthostatic hypotension) and did not include all relevant comparators. An updated analysis was conducted using RCTs published up to 2019, which included newly approved adjuncts such as istradefylline, opicapone, and amantadine extended release (ER) [19], though a further update is required to include studies from the last five years.

The objective of this study was to assess the relative safety of istradefylline compared to other PD adjuncts based on randomized controlled trials (RCTs) and real-world evidence.

2. Methods

A systematic literature review (SLR) and NMA was conducted according to a protocol, and the results were reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) harms extension [20].

2.1. Systematic literature review search strategy

The electronic databases Embase®, MEDLINE® (including MEDLINE® Epub Ahead of Print, MEDLINE® In-Process & Other Non-Indexed Citations, and MEDLINE® Daily) and EBM (Evidence-Based Medicine) Reviews (including the health technology assessment [HTA] database, the National Health Service Economic Evaluation Database [NHS EED], Cochrane Central Register of Controlled trials [CENTRAL], Database of Abstracts of Reviews of Effects [DARE], and the Cochrane Database of Systematic Reviews) were searched on January 10, 2024, via the Ovid platform. Full search terms are provided in Supplementary Tables 1–4.

Conference proceedings from the last three years, reference lists of included publications, HTA bodies, and clinical trial registries were hand-searched to identify relevant evidence.

2.2. Study selection

Citations were screened against defined eligibility criteria by two analysts independently, with any discrepancies resolved by consensus. Eligible studies included randomized controlled trials (RCTs) and real-world evidence (RWE) studies including patients with PD of any age who were receiving levodopa (any duration) and had developed motor complications. Interventions included COMT inhibitors, MAO-B inhibitors, DA, istradefylline, and amantadine. Studies must have reported at least one relevant safety outcome. Full eligibility criteria are provided in Table 1.

Table 1.

Eligibility criteria.

Criteria	Inclusion criteria	Exclusion criteria
Patient population	Patients with PD of any age who were receiving levodopa (any duration) and had developed motor complications	Populations not detailed in the list
Study design	•RCTs •Real-world/observational studies (no restriction on study perspective [retrospective/prospective] and to include both comparative and non-comparative studies†)	•Non-randomized studies conducted in a clinical setting (to include Phase 1/2 studies) •RCTs with a duration of ≤4 weeks (to be excluded and tagged at full publication review)
Interventions	Licensed doses of istradefylline (Nourianz®)	Interventions not included in the list
Comparators	Licensed doses of the following treatments: •COMT inhibitors (entacapone [Comtan®], tolcapone [Tasmar®], opicapone [Ongentys®]) •MAO-B inhibitors (rasagiline [Azilect®], selegiline/sublingual selegiline [Eldepryl®, Zelapar®], safinamide [Xadago®]) •DA (apomorphine [Apokyn®, Kynmobi®], bromocriptine [Cycloset®, Parlodel®], cabergoline [Dostinex®], pramipexole [Mirapex®], ropinirole [Requip®], rotigotine [Neupro®]) •Amantadine/amantadine ER (Osmolex®, Gocovri®)	Comparators not included in the list
Outcomes	Overall incidence of SAE, TRAE, TEAE, withdrawal due to AEs, and all withdrawals Individual AEs: •Dyskinesia •Hallucination •Hypotension/orthostatic hypotension •Somnolence/drowsiness/sleepiness •Falls •Fracture •Nausea •Peripheral edema •Dry mouth •Confusion •Impulse control disorder •Musculoskeletal and connective tissue disorders	Studies assessing outcomes not relevant to the review
Time-limits	Studies published 1975 and later	Studies published prior to 1975
Countries	No limits
Language	English language only§

Abbreviations: AE, adverse event; COMT, catechol-O-methyltransferase; DA, dopamine agonist; DET, data extraction table; ER, extended release; MAO-B, monoamine oxidase-B; PD, Parkinson’s disease; RCT, randomized controlled trial; SAE, serious adverse event; SLR, systematic literature review; TEAE, treatment-emergent adverse event; TRAE, treatment-related adverse event; UPDRS, Unified Parkinson’s Disease Rating Scale.

† Comparative studies include at least two active treatment arms or comparison of an active treatment with a placebo group or no treatment; non-comparative studies comprise a single treatment group.§ Non-English citations that were potentially relevant to the SLR were assessed on an individual basis.

2.3. Data extraction

Data pertaining to study design, baseline characteristics, and treatment outcomes were extracted into an Excel® based extraction table by an analyst and independently checked by a second analyst. Risk of bias assessment of RCTs was conducted using the 2011 Cochrane tool for RCTs [21]. Risk of bias assessment was based on primary outcomes in the RCTs.

2.4. Network meta-analysis

Meta-analysis feasibility assessment was conducted using a transparent, all-inclusive, and reproducible methodology. This included a review of previously published meta-analyses, generation of a”best-case scenario” evidence network, assessment of trial heterogeneity, and generation of outcome-specific evidence networks.

The outcomes were analyzed using a Bayesian NMA conducted in WinBUGS. All statistical models were fitted by adapting code written by the National Institute for Health and Care Excellence Decision Support Unit (NICE DSU) for their evidence synthesis series [22], [23], [24]. Random effects (RE), baseline risk-adjusted RE, and fixed effects (FE) models were run for each analysis. Heterogeneity was assessed by exploring the global I² statistic using the R netmeta package and between-study standard deviation (SD) (tau). Model fit was then evaluated using I², deviance information criterion (DIC), tau, regression coefficient of RE baseline risk-adjusted models, and total residual deviance. Selection of the final model for each outcome was based on the following stepwise criteria: i) Assessment of heterogeneity: the I² statistic was used. If I² > 0, a RE model was selected. If I² = 0, the DIC was evaluated. ii) Model fit: the model with the lowest DIC was chosen if the difference was more than 3 points. If DIC values were within 3 points, the RE model was preferred. iii) Sanity check: SD and total residual deviance were reviewed as a final validation step. An overview of this assessment and final model selection for each outcome is available in Supplementary Table 5.

A sensitivity analysis was conducted to compare safety outcomes restricted to DA RCTs published after year 2000, excluding older studies which may be less reflective of current clinical practice, such as using concomitant medications that are no longer part of standard practice.

3. Results

3.1. Systematic literature review

The process of study selection is documented in Fig. 1. In total, 6,967 articles were identified for title and abstract screening. Of these, 340 full text articles were reviewed and 139 were included. A manual search of conference proceedings and reference lists of included studies yielded a further eight citations. A total of eight clinical study reports (CSR) were included. This resulted in a total of 155 relevant publications (RCT, n = 100; RWE, n = 55) reporting on 122 unique studies (RCT, n = 90; RWE, n = 32) being included in the SLR.

Fig. 1.

PRISMA flow diagram. Abbreviations: CSR, clinical study report; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT, randomized controlled trial; RWE, real-world evidence; SLR, systematic literature review.

3.2. Evidence mapping

Evidence mapping generated connected”best-case scenario” evidence networks comprised of 76 RCTs included in the SLR. A summary of the 76 RCTs included is provided in Supplementary Table 6, and a summary of the 55 RWE publications identified is provided in Supplementary Table 7. The 24 publications that were not included in the best-case scenario network were excluded for one of the following reasons: the intervention arms did not connect within the network; the publication was reporting the same study as a CSR that was included in the network; the publication reported no relevant safety data; the publication was reporting the same study as another publication and reported no unique data.

The best-case scenario network is presented in Fig. 2. Outcome specific networks are provided in Supplementary Figs. 1–16.

Fig. 2.

Best-case scenario network. Abbreviations: COMT, catechol-O-methyltransferase; DA, dopamine agonist; ER, extended release; IR, immediate release; MAO-B, monoamine oxidase-B; PD, Parkinson’s disease.

3.3. Heterogeneity assessment and feasibility analysis

There was inevitable heterogeneity between RCTs with respect to study design and patient population; however, no outliers that warranted exclusion from the best-case network were identified. Based on a qualitative assessment of inter-study heterogeneity, the studies appeared sufficiently homogeneous to combine in analyses. Of note, five trials exclusively enrolled patients with levodopa-induced dyskinesia, suggesting that these patients may have more advanced PD compared with other trials. A total of 18 trials were conducted exclusively in China or Japan and enrolled Asian patients only. All other studies enrolled primarily Caucasian patients. There was notable inter-trial variability in terms of time since diagnosis, time since levodopa initiation, levodopa dose, and concomitant medications received. The impact of these factors across the trials (in addition to unknown patient characteristics) was explored via meta-regression based on the baseline risk.

Outcome-specific evidence networks were feasible for five overall AE outcomes and nine specific AEs. Outcomes were reported for all drug classes of interest. None of the studies reported the incidence of fractures, and analyses of ICD, and musculoskeletal and connective tissue disorders were not feasible due to zero/low event numbers.

3.4. Risk of bias assessment

Risk of bias assessment showed that randomization was adequate in 53 % of studies, while 45 % were less adequate. Allocation concealment and blinding were well-reported in 97 % of studies, and baseline characteristics were similar in 83 % of studies. Few studies showed dropout imbalances (29 %), and outcome reporting was adequate in 85 %. Intent-to-treat (ITT) analysis was used in 75 % of studies. Conflicts of interest were declared in 40 %, but unclear in 55 % of studies. Full risk of bias results are available in Supplementary Table 8.

3.5. Network meta-analysis

Broadly speaking, the results across the NMAs favored istradefylline, and where results excluded the null value from the 95 % credible intervals (CrIs), all favored istradefylline. An overview of the results is provided in Table 2.

Table 2.

Overview of NMA results for istradefylline versus comparator.

Outcome	Istradefylline versus comparator, OR (95 % CrI)
	Amantadine	COMT inhibitors	DA	MAO-B inhibitors
Dyskinesia (n = 63)	2.31 (0.63, 7.83)	0.63 (0.41, 0.99)	0.69 (0.43, 1.09)	1.16 (0.69, 1.96)
Hypotension (n = 38)	0.22 (0.03, 1.54)	0.19 (0.03, 0.82)	0.33 (0.07, 1.30)	0.09 (0.01, 0.52)
Somnolence (n = 42)	2.35 (0.50, 11.36)	0.49 (0.23, 1.06)	0.53 (0.27, 1.07)	0.65 (0.23, 1.90)
Hallucinations (n = 46)	0.25 (0.06, 0.97)	1.24 (0.48, 3.34)	0.48 (0.19, 1.20)	0.84 (0.27, 2.56)
Falls (n = 19)	1.42 (0.25, 6.59)	1.19 (0.23, 4.66)	1.78 (0.29, 7.71)	0.81 (0.13, 3.51)
Nausea (n = 62)	0.45 (0.12, 1.58)	0.61 (0.35, 1.07)	0.58 (0.33, 0.99)	0.77 (0.37, 1.57)
Peripheral edema (n = 17)†	0.10 (0.01, 1.26)	2.90 (0.31, 52.36)	0.30 (0.04, 2.17)	1.91 (0.15, 50.07)
Dry mouth (n = 21)†	0.04 (0.01, 0.23)	0.13 (0.03, 0.63)	0.12 (0.02, 0.65)	0.13 (0.02, 0.68)
Confusion (n = 16)	0.56 (0.02, 11.15)	1.22 (0.11, 16.96)	0.66 (0.04, 9.92)	0.60 (0.02, 14.00)
TEAEs (n = 61)†	0.43 (0.25, 0.73)	0.76 (0.57, 1.01)	0.82 (0.61, 1.07)	0.99 (0.72, 1.34)
TRAEs (n = 24)	0.33 (0.19, 0.56)	0.85 (0.53, 1.32)	0.77 (0.53, 1.09)	0.87 (0.59, 1.23)
SAEs (n = 43)†	0.56 (0.32, 0.99)	0.90 (0.60, 1.34)	1.11 (0.78, 1.58)	0.74 (0.52, 1.06)
Withdrawal due to AE (n = 67)†	0.37 (0.19, 0.68)	0.68 (0.44, 1.02)	0.78 (0.49, 1.16)	1.15 (0.72, 1.85)
All withdrawals (n = 61)	1.04 (0.52, 2.04)	0.87 (0.57, 1.32)	1.34 (0.90, 1.99)	1.07 (0.68, 1.69)

Note: The unadjusted random-effects model was the optimal model of choice in most instances. ‘†’ indicates outcomes where the baseline risk-adjusted model was the optimal model of choice. See Supplementary Table 5 for full results including inconsistency checks, I² values and DIC estimates. Bold represents values where the null value was excluded from the 95% CrI. Results shaded in grey favor istradefylline.Abbreviations: AE, adverse event; COMT, catechol-O-methyltransferase; CrI, credible interval; DA, dopamine agonist; DIC, deviance information criterion; FE, fixed effect; MAO-B, monoamine oxidase-B; NMA, network meta-analysis; OR, odds ratio; RE, random effect; SAE, serious adverse event; TEAE, treatment-emergent adverse event; TRAE, treatment-related adverse event.

Heterogeneity (measured by I²) was above zero for all outcomes except somnolence, dry mouth, confusion, and SAEs. The RE or RE baseline risk-adjusted model, selected following the stepwise criteria, was chosen as the preferred model for all outcomes, with RE for nine outcomes and the RE baseline risk-adjusted model for five outcomes.

3.6. TEAE, TRAE, SAE, and withdrawals

The NMA found lower odds of treatment-emergent adverse events (TEAEs) and treatment-related adverse events (TRAEs) with istradefylline versus amantadine (TEAEs odds ratio [OR]: 0.43, 95 % CrI: 0.25, 0.73; TRAEs OR: 0.33, 95 % CrI: 0.19, 0.56), where the null value was excluded from the 95 % CrI. For all other comparators, the NMA estimates for odds of TEAEs and TRAEs were numerically favorable for istradefylline, but all estimates included the null value within the 95 % CrI (TEAEs vs COMT inhibitors, OR: 0.76, 95 % CrI: 0.57, 1.01; vs DA, OR: 0.82, 95 % CrI: 0.61, 1.07; vs MAO-B inhibitors, OR: 0.99, 95 % CrI: 0.72, 1.34; TRAEs vs COMT inhibitors, OR: 0.85, 95 % CrI: 0.53, 1.32; vs DA, OR: 0.77, 95 % CrI: 0.53, 1.09; vs MAO-B inhibitors, OR: 0.87, 95 % CrI: 0.59, 1.23).

Similarly for serious adverse events (SAEs), lower odds were found with istradefylline versus amantadine (OR: 0.56, 95 % CrI: 0.32, 0.99). The NMA estimates for SAEs were also numerically favorable for istradefylline versus COMT inhibitors (OR: 0.90, 95 % CrI: 0.60, 1.34) and MAO-B inhibitors (OR: 0.74, 95 % CrI: 0.52, 1.06), but the estimates included the null value within the 95 % CrIs. The estimates for SAEs numerically favored DA over istradefylline (OR: 1.11, 95 % CrI: 0.78, 1.58); however the estimates again included the null value within the 95 % CrIs.

The NMA found lower odds of withdrawals due to AEs with istradefylline versus amantadine (OR: 0.37, 95 % CrI: 0.19, 0.68). The NMA estimates for odds of withdrawals due to AEs were also numerically favorable for istradefylline versus DA (OR: 0.78, 95 % CrI: 0.49, 1.16) and COMT inhibitors (OR: 0.68, 95 % CrI: 0.44, 1.02), but not MAO-B inhibitors (OR: 1.15, 95 % CrI: 0.72, 1.85), however estimates included the null value within the 95 % CrI.

For all withdrawals (due to AEs or other reasons), the NMA estimates were numerically favorable for istradefylline versus COMT inhibitors only (OR: 0.87, 95 % CrI: 0.57, 1.32). The NMA estimates showed numerically higher odds of all withdrawals with istradefylline versus amantadine (OR: 1.04, 95 % CrI: 0.52, 2.04), DA (OR: 1.34, 95 % CrI: 0.90, 1.99), and MAO-B inhibitors (OR: 1.07, 95 % CrI: 0.68, 1.69). All estimates versus comparators included the null value within the 95 % CrI.

3.7. Dyskinesia

The NMA showed lower odds of dyskinesia with istradefylline versus COMT inhibitors (OR: 0.63, 95 % CrI: 0.41, 0.99) and numerically lower odds of dyskinesia with istradefylline relative to DA (OR: 0.69, 95 % CrI: 0.43, 1.09). Sensitivity analysis involving studies after year 2000 showed a lower risk of dyskinesia (OR: 0.55, 95 % CrI: 0.38, 0.82) for istradefylline compared with DA. There were numerically higher odds of dyskinesia with istradefylline versus amantadine (OR: 2.31, 95 % CrI: 0.63, 7.83) and MAO-B inhibitors (OR: 1.16, 95 % CrI: 0.69, 1.96); however, the estimates included the null value within the 95 % CrI.

3.8. Hypotension

Istradefylline was associated with lower odds of hypotension versus COMT inhibitors and MAO-B inhibitors (OR: 0.19, 95 % CrI: 0.03, 0.82; OR: 0.09, 95 % CrI: 0.01, 0.52, respectively). The NMA estimates were also numerically favorable for istradefylline versus amantadine (OR: 0.22, 95 % CrI: 0.03, 1.54) and DA (OR: 0.33, 95 % CrI: 0.07, 1.30), but the null value fell within the 95 % CrI.

3.9. Hallucinations

The NMA found lower odds of hallucinations with istradefylline versus amantadine (OR: 0.25, 95 % CrI: 0.06, 0.97). The NMA estimates for hallucinations were also numerically favorable for istradefylline versus DA (OR: 0.48, 95 % CrI: 0.19, 1.20) and MAO-B inhibitors (OR: 0.84, 95 % CrI: 0.27, 2.56), but not COMT inhibitors (OR: 1.24, 95 % CrI: 0.48, 3.34); however, estimates included the null value within the 95 % CrI. Sensitivity analysis involving studies after year 2000 showed a reduced risk of hallucinations (OR: 0.41, 95 % CrI: 0.16, 0.99) for istradefylline compared with DA.

3.10. Nausea

Istradefylline was found to be associated with lower odds of nausea versus DA (OR: 0.58, 95 % CrI: 0.33, 0.99). The NMA estimates were also numerically favorable for istradefylline versus all other comparators (istradefylline vs amantadine, OR: 0.45, 95 % CrI: 0.12, 1.58; vs COMT inhibitors, OR: 0.61, 95 % CrI: 0.35, 1.07; vs MAO-B inhibitors, OR: 0.77, 95 % CrI: 0.37, 1.57), but the null value was within the 95 % CrI.

3.11. Somnolence

With respect to somnolence, the NMA estimates were numerically favorable for istradefylline versus COMT inhibitors (OR: 0.49, 95 % CrI: 0.23, 1.06), DA (OR: 0.53, 95 % CrI, 0.27, 1.07), and MAO-B inhibitors (OR: 0.65, 95 % CrI: 0.23, 1.90), but not amantadine (OR: 2.35, 95 % CrI: 0.50, 11.36), however the null value was within the 95 % CrI.

3.12. Falls

Istradefylline was associated with numerically lower odds of falls versus MAO-B inhibitors (OR: 0.81, 95 % CrI: 0.13, 3.51), and higher odds versus amantadine (OR: 1.42, 95 % CrI: 0.25, 6.59), COMT inhibitors (OR: 1.19, 95 % CrI: 0.23, 4.66), and DA (OR: 1.78, 95 % CrI: 0.29, 7.71), but the null value was within the 95 % CrIs.

3.13. Peripheral edema

The NMA estimates were numerically favorable for istradefylline versus amantadine (OR: 0.10, 95 % CrI: 0.01, 1.26) and DA (OR: 0.30, 95 % CrI: 0.04, 2.17), but not COMT inhibitors (OR: 2.90, 95 % CrI: 0.31, 52.36) or MAO-B inhibitors (OR: 1.91, 95 % CrI: 0.15, 50.07). All estimates included the null value within the 95 % CrI.

3.14. Dry mouth

The NMA found istradefylline had lower odds of dry mouth compared with amantadine (OR: 0.04, 95 % CrI: 0.01, 0.23), COMT inhibitors (OR: 0.13, 95 % CrI: 0.03, 0.63), DA (OR: 0.12, 95 % CrI: 0.02, 0.65) and MAO-B inhibitors (OR: 0.13, 95 % CrI: 0.02, 0.68), with the null value excluded from the 95 % CrI for all comparisons.

3.15. Confusion

The NMA estimates were numerically favorable for istradefylline versus amantadine (OR: 0.56, 95 % CrI: 0.02, 11.15), DA (OR: 0.66, 95 % CrI: 0.04, 9.92), and MAO-B inhibitors (OR: 0.60, 95 % CrI: 0.02, 14.00), but not COMT inhibitors (OR: 1.22, 95 % CrI: 0.11, 19.96), however all estimates included the null value within the 95 % CrI.

3.16. Real world evidence

In total, 55 RWE publications were identified, reporting on 32 unique studies. Relevant outcomes were extracted to supplement the findings of the NMA. Publications were identified that provided RWE for relevant outcomes related to istradefylline (n = 4), MAO-B inhibitors (n = 14), COMT inhibitors (n = 28), DA (n = 7), and amantadine (n = 2). The publications covered Europe (n = 37), Asia (n = 10), North America (n = 7), and South America (n = 1). Of the studies, six were conducted retrospectively and so may be limited by incomplete data. No RWE studies were identified reporting on fractures or confusion. An overview of the RWE studies is provided in Supplementary Table 7, and the incidences of individual AEs by study are provided in Supplementary Figs. 17–31.

The incidence of TEAEs was reported by 32 publications (17 unique studies). While incidences varied between studies of the same drug, they ranged from 11.1 % with safinamide (MAO-B inhibitor) to 97.5 % with apomorphine (DA). The incidence of TRAEs was reported by 32 publications (13 unique studies) and ranged from 11.5 % with tolcapone to 65.2 % with opicapone (both COMT inhibitors). The incidence of TRAEs with istradefylline was 20.8 %.

The incidence of SAEs was reported by 27 publications (16 unique studies), and ranged from 0 % in one study of tolcapone and one study of pramipexole (DA) to 27.2 % with selegiline (MAO-B inhibitor). The incidence of SAEs with istradefylline was 9 %.

The incidence of withdrawal due to AEs was reported by 31 publications (28 unique studies) and ranged from 1.8 % with pramipexole to 28.8 % with opicapone. With istradefylline, the incidence of withdrawal due to AEs was reported to be from 9.7 % to 13.7 %. In terms of total withdrawals, the incidence was reported by 25 publications (22 unique studies) and ranged from 3 % with entacapone to 65 % with selegiline, while withdrawals from istradefylline treatment were between 22.6 % and 32.6 %.

The incidence of dyskinesia was reported by 27 publications (16 unique studies) and ranged from 1.3 % in one study with rasagiline to 29.3 % with apomorphine. The incidence of dyskinesia with istradefylline was 4.9 % to 6.5 %, however these were adverse drug reactions rather than AEs, so the incidence of dyskinesia AEs may be higher. Though direct comparisons are challenging given heterogeneity between the RWE studies, a higher incidence of dyskinesia was observed with MAO-B inhibitors compared with istradefylline.

The incidence of hallucinations was reported by 17 publications (13 unique studies) and ranged from 0.9 % with rasagiline (MAO-B inhibitor) to 24.2 % with amantadine. The reported incidence of hallucinations with istradefylline ranged from 2.4 % to 6.5 %. Compared with istradefylline, a higher incidence of hallucinations was observed with amantadine.

The incidence of nausea was reported by 20 publications (18 unique studies) and ranged from 0.5 % in a study of males receiving safinamide to 25 % in patients receiving ropinirole (COMT inhibitor). The reported incidence of nausea with istradefylline was 0.6 %. Overall, a higher incidence of nausea was observed with DA, COMT inhibitors, and amantadine, when compared with istradefylline.

The incidence of somnolence was reported by eight publications (eight unique studies) and ranged from 0.9 % with rasagiline to 20 % in one study of safinamide. Compared with istradefylline (incidence of 1.1 %), a higher incidence of somnolence was observed with DA and COMT inhibitors.

The incidence of ICD was reported by two publications (two unique studies). The two publications assessed the incidence of ICD with safinamide and rasagiline, both of which are MAO-B inhibitors. The incidence was 0.2 % with rasagiline and 0.6 % with safinamide; however, both publications reported ICD adverse drug reactions, and it is possible that the incidence of ICD AEs would be higher than these reported values.

The incidence of hypotension was only reported by one publication, and the incidence of orthostatic hypotension was reported by four publications (four unique studies). The incidence of hypotension and orthostatic hypotension ranged from 0 % with bromocriptine to 8.2 % with apomorphine, both DA. No RWE studies were identified reporting on the incidence of hypotension in patients treated with istradefylline.

The incidence of falls was reported by seven publications (seven unique studies), and ranged from 0.2 % in one study with rasagiline to 32.7 % in one study of patients treated with amantadine. No RWE studies were identified reporting on the incidence of falls in patients treated with istradefylline.

The incidence of peripheral edema was reported by four publications (four unique studies) and ranged from 0.8 % with istradefylline to 16.1 % with amantadine.

The incidence of dry mouth was reported by nine publications (five unique studies) and ranged from 0.9 % with rasagiline to 12.2 % with opicapone. No RWE studies were identified reporting on the incidence of dry mouth in patients treated with istradefylline.

The incidence of musculoskeletal and connective tissue disorders was reported by three publications (two unique studies) and ranged from 0.2 % with rasagiline to 8.5 % with safinamide. No RWE studies were identified reporting on the incidence of musculoskeletal and connective tissue disorders in patients treated with istradefylline.

4. Discussion

The results across the NMAs generally favored istradefylline, and all findings where the null value was excluded from the 95 % CrI favored istradefylline’s safety profile over other PD adjuncts included in the study. Notably, the results provided no evidence to suggest that any of the comparators were favorable versus istradefylline for any of the safety outcomes analyzed. Istradefylline was associated with numerically lower odds of TEAEs and TRAEs compared with all other adjuncts included in this analysis. Istradefylline was also associated with lower odds of SAEs compared with all other adjuncts except DA.

The findings from this NMA are in line with previous reports that adjuvant therapy with DA, COMT inhibitors, MAO-B inhibitors and amantadine is often associated with higher incidences of AEs. In comparison to previous analyses that included the relative safety of PD adjuncts (Stowe 2011 and Sako 2023 [17], [18]), this manuscript reports an in-depth analysis of a wider range of safety outcomes of relevance to patients with PD and is inclusive of the most recent PD adjuncts.

Existing literature highlights that adjunctive treatment with a DA often precipitates a wide range of side effects in patients with PD, ranging from mild to debilitating [4]. Reports highlight nausea, orthostatic hypotension, somnolence, hallucinations, peripheral edema, and ICD as common side effects, among others [4], [5]. In this analysis, relative to DA, istradefylline was associated with numerically lower odds of most AE outcomes analyzed, including hypotension, somnolence, hallucinations, and peripheral edema, as well as overall TEAEs and TRAEs. In particular, there were lower odds of nausea and dry mouth with istradefylline versus DA. Istradefylline was not, however, associated with lower odds of SAEs relative to DA. Avoiding or minimizing the odds of such side effects may increase the likelihood of patients remaining on treatment and thus promote better overall outcomes for patients.

Though perceived to have a better tolerability profile than DA [9], MAO-B inhibitors are similarly hampered by side effects including, but not limited to, dry mouth, nausea, orthostatic hypotension, hallucinations, and dyskinesia [8]. Istradefylline was found to be associated with lower odds of two of these key side effects: hypotension and dry mouth. Numerically lower odds of somnolence, hallucinations, falls, nausea, and confusion, as well as overall TEAEs, TRAEs, and SAEs, were also found with istradefylline versus MAO-B inhibitors.

Some of the main side effects of COMT inhibitors relate to their indirect dopaminergic effects, including worsening of levodopa-induced AEs such as dyskinesia, nausea, vomiting, orthostatic hypotension, somnolence, and hallucinations [7]. Notably, the NMA found lower odds of dyskinesia, hypotension, and dry mouth with istradefylline relative to COMT inhibitors, alongside numerically lower odds of somnolence, nausea, overall TEAEs, TRAEs, SAEs, withdrawals due to AEs, and all withdrawals. Lower odds of dyskinesia in patients with PD is a particularly notable finding, as levodopa-induced dyskinesia is one of the most commonly reported challenges in patients with PD, in many cases significantly disabling enough to negate the beneficial effects of treatment [25]. Treatment with a COMT inhibitor has previously been found as a risk factor for developing dyskinesia in patients with PD [26].

As with other PD adjuncts, amantadine is associated with a number of side effects; of note, in clinical trials, amantadine was associated with visual and auditory hallucinations in over 20 % of patients [6]. The NMA found lower odds of overall TEAE, TRAEs, SAEs, and withdrawals due to AEs, hallucinations, and dry mouth with istradefylline versus amantadine.

Apart from initial tolerability and adherence to treatment, the lower odds of AEs with istradefylline versus other adjuncts may extend the therapeutic window in which treatment does not need to be withdrawn. For all comparators except MAO-B inhibitors, istradefylline was associated with numerically lower odds of withdrawals due to AEs. Such an outcome supports use of istradefylline as an early adjunctive to levodopa and carbidopa/benserazide.

With regards to RWE, when compared with comparators of interest, istradefylline was associated with favorable safety outcomes, in line with the findings from RCTs. However, these findings were impacted by the paucity of data for some safety outcomes of interest.

4.1. Limitations

Consideration must be given to certain limitations when interpreting the results of this NMA. As noted, the analysis is based on a heterogeneous dataset, with varying trial lengths, levodopa dose, and duration and severity of PD in participants, which may limit the validity of the indirect comparisons between studies. Furthermore, some outcomes (dry mouth, confusion, peripheral edema, and hypotension) were associated with low event numbers, resulting in estimates of relevant treatment effects with large levels of uncertainty. Another notable limitation is the potential issue of lack of statistical power to detect rare AE outcomes, which often results in wide confidence intervals and limits the precision and robustness of treatment effect estimates [27], meaning that some of the results should be interpreted with caution. Future research should explore alternative methods better suited to handle rare binary outcomes in NMAs [28]. Finally, there may be challenges with the grouping of drugs by class in this analysis, due to potential differences in safety and tolerability profiles within a class.

5. Conclusions

Overall, this study provides an up-to-date assessment of the comparative safety of PD adjuncts. Istradefylline is associated with a generally favorable safety profile relative to other adjunct medications in this study.

CRediT authorship contribution statement

Sagari Betté: Writing – review & editing, Validation. Joyce Qian: Writing – review & editing, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Hannah Cummings: Writing – review & editing, Validation, Supervision. Hiroo Shimoda: Writing – review & editing, Validation, Formal analysis. Katsumi Shinoda: Writing – review & editing, Validation, Formal analysis. Ashley Thai: Writing – review & editing, Validation, Formal analysis. Sarah Batson: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Gabrielle Redhead: Writing – review & editing, Writing – original draft, Methodology, Investigation, Data curation, Conceptualization. Alexander Hodkinson: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Daniel Truong: Writing – review & editing, Validation.

Funding

This work was funded by Kyowa Kirin Inc.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: