GRAPHICAL ABSTRACT

INTRODUCTION

Continuous dopaminergic delivery has become a cornerstone in the management of advanced Parkinson’s disease (aPD), with the goal of reducing motor fluctuations and improving quality of life when oral therapies become insufficient [1,2]. Among the available device-aided therapies (DATs), continuous subcutaneous foslevodopa/foscarbidopa infusion (CSFLI) has recently emerged as a promising therapeutic option [3-5].

Phase 3 clinical trials and open-label extensions have demonstrated the efficacy of CSFLI in reducing OFF time and improving motor complications over several months of treatment [3,5]. These studies also reported favorable safety and tolerability profiles, with most adverse events related to infusion-site reactions. However, the neuropsychiatric and cognitive safety of this therapy remains less well understood. Participants with significant cognitive impairment, a history of psychosis, or active psychiatric symptoms were typically excluded from these trials, limiting the generalizability of the findings to more fragile patient populations [3,5,6].

Some real-world reports have raised concerns about potential neuropsychiatric complications under CSFLI, including hallucinations, confusion, delusions, behavioral changes, and cognitive decline [6-8]. These adverse effects (AEs) were observed in up to 55% of patients in small retrospective cohorts [7] and in approximately 8 to 17% of participants in larger open-label studies, although detailed neurocognitive assessments were not systematically conducted [5,8] and these effects were the most common AEs (mostly hallucinations) leading to drug discontinuation [5]. These reports underscore the potential vulnerability of certain patients to neuropsychiatric worsening under CSFLI, particularly those with a prior cognitive or psychiatric history. However, the prevalence, severity, and risk factors for such AEs remain poorly characterized. No study to date has systematically explored clinical predictors of intolerance or worsening, nor has it provided robust guidance for clinical practice regarding risk stratification before initiating CSFLI.

To address this gap, we conducted a retrospective observational study of 36 patients treated with CSFLI in a real-life clinical setting, all of whom completed six months of therapy. Our primary objective was to assess the frequency of cognitive and psychiatric worsening under CSFLI and to identify baseline factors associated with neuropsychiatric intolerance.

MATERIALS & METHODS

Study design and participants

We conducted a retrospective observational study including consecutive patients with aPD who initiated CSFLI at our movement disorders center between November 2024 and January 2025. Patients were eligible if they fulfilled the Movement Disorder Society clinical diagnostic criteria for idiopathic Parkinson’s disease (PD) [9], had undergone at least six months of uninterrupted CSFLI, and had complete clinical and neuropsychiatric assessments available at treatment initiation (M0) and at six months (M6). Patients who discontinued CSFLI prematurely or whose follow-up evaluations were missing were excluded.

This research was approved by the Rothschild Hospital Review Board (IRB 00012801) under study number CE_20250527_3_GBE.

CSFLI protocol

CSFLI was administered via a portable subcutaneous pump system providing continuous infusion. The total daily duration of infusion (daytime-only or 24 hours) was determined collaboratively by the treating neurologist and the patient on the basis of motor and nonmotor symptom control. Oral dopaminergic medications, including levodopa, could be maintained or adjusted as needed. The mode of CSFLI initiation—either during hospitalization or in an outpatient setting—was recorded.

Clinical motor and nonmotor assessments

Baseline demographic and clinical data included sex, age at PD diagnosis, disease duration, age at CSFLI initiation, number of daily oral levodopa intakes, and baseline oral levodopa equivalent daily dose (LEDD). The use of dopamine agonists, clozapine, and catechol-O-methyltransferase inhibitor (COMT-I) was noted. A prior history of hallucinations or confusion and cognitive status (assessed by the Montreal Cognitive Assessment [MoCA]) were systematically collected.

Motor and nonmotor symptoms were assessed at M0 and M6 using the MDS-Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) Parts I to IV and the Parkinson’s Disease Questionnaire-8 (PDQ-8). In addition to total scores, specific subdomains were analyzed: the cognitive and psychiatric subitems from the MDS-UPDRS Part I, as well as the cognitive and psychological subscores derived from the PDQ-8.

Definitions of outcomes

The primary outcome was defined as clinically relevant neuropsychiatric and/or cognitive worsening between baseline and six months. Clinically meaningful neuropsychiatric/cognitive worsening was defined as an increase (≥1 point) in 1) the relevant MDS-UPDRS Part I items (1.1–1.6: cognitive impairment, hallucinations/psychosis, depressed mood, anxious mood, and apathy) or 2) the PDQ-8 cognitive/psychiatric items between baseline and the 6-month follow-up. This conservative threshold was chosen to maximize safety signal detection in this realworld cohort, as pivotal CSFLI trials excluded patients with significant neuropsychiatric symptoms. This approach is consistent with the established minimal clinically important difference thresholds for these scales [10,11]. Each scale was evaluated independently without summation, with different recall periods (1 week for MDS-UPDRS Part I and 1 month for PDQ-8) providing complementary temporal windows.

Patients were then dichotomized into two groups on the basis of the presence or absence of neuropsychiatric worsening at follow-up.

Statistical analysis

Descriptive statistics were used to summarize the baseline characteristics. Continuous variables were compared using the Wilcoxon rank-sum test, and categorical variables were compared using Fisher’s exact test. Comparisons between groups (with and without neuropsychiatric worsening) were performed to identify univariate differences.

A multivariable logistic regression model was constructed to identify independent baseline predictors of neuropsychiatric worsening. Candidate variables were selected on the basis of clinical relevance: age at diagnosis, disease duration, baseline oral LEDD, motor severity (MDS-UPDRS Part III), PDQ-8 total scores, use of clozapine or dopamine agonists, a prior history of hallucinations or confusion, and CSFLI initiation setting. Owing to the limited sample size and the potential for sparse data bias, Firth’s penalized maximum likelihood logistic regression was used. Adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were reported.

To further explore the heterogeneity of neuropsychiatric outcomes, we computed a composite severity index in patients who exhibited worsening cognition and/or psychiatric status. This index was based on the cumulative sum of severity levels (rated from 0 to 4) across the cognitive and psychiatric subitems of the MDS-UPDRS Part I and the cognitive and psychological subscores of the PDQ-8 at M6. On the basis of tertiles of the resulting distribution, patients were classified into three subgroups: mild, moderate, and severe worsening. Each subgroup was compared independently to the “no worsening” group using univariate statistics. Additionally, exploratory subgrouping was performed to distinguish between patients with primarily worsening psychiatric status, patients with worsening cognitive status, and patients with mixed profiles. These three groups were also compared to the “no worsening” group in terms of baseline demographic and clinical features. All p values were corrected using the Benjamini–Hochberg false discovery rate (FDR) method. Significant results were defined as FDR-corrected p-values <0.05.

All analyses were conducted using R software (version 4.3.1; R Foundation for Statistical Computing).

RESULTS

Baseline characteristics

Thirty-six consecutive patients with aPD who underwent CSFLI during the study period were included in this analysis. All patients completed the six-month follow-up. Baseline characteristics are presented in Table 1. The cohort included elderly patients (mean age at CSFLI initiation, 67.2±10.7 years) with a long disease duration (11.7±5.5 years). Cognitive impairment was common: the mean MoCA score was 21.4±4.8, with 26 pa-tients (72.2%) scoring <26 and 17 patients (47.2%) scoring <21. Prior neuropsychiatric complications were frequent: 13 patients (36.1%) had documented medication-induced confusion, and 16 patients (44.4%) had histories of hallucinations or psychosis. Sixteen patients (44.4%) were receiving clozapine at baseline. All patients were on levodopa therapy (a mean oral LEDD of 1,678±274 mg), with 16 (44.4%) also receiving COMT-I. Monoamine oxidase B (MAO-B) inhibitors and amantadine were each used in 3 patients (8%).

Comparisons between patients with and without cognitive/psychiatric worsening

Among the 36 patients, 17 (47.2%) experienced clinically meaningful neuropsychiatric and/or cognitive worsening within the first six months of CSFLI treatment, while 19 (52.8%) remained stable. Worsening encompassed isolated cognitive decline (n=1, 6%), isolated psychiatric aggravation (n=7, 41%), and mixed cognitive–psychiatric profiles (n=9, 53%).

Patients who developed worsening did not differ from those who remained stable in terms of age at diagnosis (54.4±9.4 vs. 56.4±11.3 years, p=0.349), disease duration (12.8±4.7 vs. 10.8± 6.1 years, p=0.075), or baseline cognitive function (MoCA scores of 22.2±3.9 vs. 20.6±5.6, p=0.246). Motor severity at baseline was comparable between the groups (MDS-UPDRS Part III scores of 59.1±18.9 vs. 56.4±29.4; p=0.874).

However, critical differences emerged in psychiatric management and medication profiles. A history of medication-induced confusion was significantly more common in the stable group than in the worsening group (11/19 [57.9%] vs. 2/17 [11.8%], p=0.006). Similarly, a history of hallucinations or psychosis showed a trend in the same direction (11/19 [57.9%] vs. 5/17 [29.4%], p=0.106). Baseline clozapine use was significantly more frequent in patients who remained stable (12/19 [63.2%] vs. 4/17 [23.5%], p=0.023). Conversely, COMT-I use was markedly more common in the worsening group (12/17 [70.6%] vs. 4/19 [21.1%], p=0.006). Patients in the worsening group also had more frequent daily oral medication intake before CSFLI (5.8±0.9 vs. 4.5±2.4, p=0.052), suggesting more complex medication regimens.

Notably, all 3 patients receiving MAO-B inhibitors and all 3 receiving amantadine experienced worsening neuropsychiatric worsening, although the difference was not statistically significant given the small sample size (p=0.095 for both).

Baseline MDS-UPDRS Part I scores were paradoxically higher in the stable group (30.5±10.6 vs. 23.8±5.9, p=0.027), likely reflecting documented neuropsychiatric symptoms in patients already receiving clozapine, whereas the worsening group may have had subclinical or unrecognized vulnerabilities.

Dopamine dysregulation syndrome (DDS) features (MDSUPDRS Item 1.6) tended to increase in the worsening group (mean score 0.35±0.79 vs. 0.22±0.55; p=0.619), although this difference was not statistically significant.

Full baseline characteristics and between-group comparisons are detailed in Table 2.

Clinical profiles of patients with only cognitive, only psychiatric, and mixed worsening

To further explore the heterogeneity of clinical worsening, the 17 patients with cognitive/psychiatric decline were subdivided into three subgroups: cognitive-only (n=1), psychiatriconly (n=7), and mixed cognitive-psychiatric worsening (n=9). Each subgroup was independently compared to the “no worsening” group (n=19) in terms of baseline characteristics. While descriptive trends suggested more severe baseline profiles in the mixed subgroup, none of the between-group differences reached statistical significance (Supplementary Table 1).

Clinical profile at 6 months

At the six-month follow-up, motor function remained stable in both groups. The median MDS-UPDRS Part I score was 29.5 in the worsening group and 28 in the nonworsening group (p=0.68). The PDQ-8 total scores were 21.5 and 23 (p=0.74), respectively, and the MDS-UPDRS Part III scores were 49.0 and 61.0 (p=0.87), respectively. MDS-UPDRS Part IV scores tended to increase in the worsening group (11.0 vs. 9.0), although this difference was not statistically significant (p=0.07). The LEDD at six months was 1,856 mg in the worsening group versus 1,496 mg in the stable group (p=0.11). With respect to dose trajectories, patients with worsening disease experienced a mean increase in LEDD of 534±451 mg from baseline, whereas stable patients had a mean increase of 274±222 mg (p=0.18).

The infusion rates were similar across groups: the median daytime rate was 0.50 vs. 0.46 mL/h (p=0.22), and the nighttime rate was 0.20 mL/h in both groups (p=0.88). Additionally, infusion protocols were similarly distributed between groups. Among worsening patients, 8/17 (47.1%) received 24-hour continuous infusions, and 9/17 (52.9%) received daytime-only infusions, whereas 8/19 (42.1%) and 11/19 (57.9%) of stable patients received these infusions (p>0.99). The use of dopamine agonists (14.3% vs. 13.3%; p>0.99) and clozapine (42.9% vs. 6.7%; p=0.09) at 6 months also did not significantly differ between the groups.

A detailed overview of these comparisons is provided in Supplementary Figures 1 and 2.

Predictors of neuropsychiatric and cognitive deterioration (Figure 1)

Multivariate logistic regression revealed that key clinical and treatment-related variables, namely, clozapine use (OR=0.18; 95% CI: 0.04–0.77; p=0.023) and higher MDS-UPDRS Part I scores at baseline, were paradoxically protective against neuropsychiatric worsening (OR=0.91 per point; 95% CI: 0.78–0.98; p=0.026). Most patients on baseline clozapine (14/16, 87.5%) continued treatment through M6, while 2 patients initiated clozapine during follow-up. Conversely, COMT-I use was significantly associated with an increased risk of worsening (OR=9.00; 95% CI: 1.97–41.08; p=0.006). Notably, COMT-I was systematically discontinued at CSFLI initiation according to the institutional protocol.

Additional univariate analyses revealed no associations between specific variables and outcomes. Daily oral intake showed a trend toward significance (p=0.052). The total MoCA score was not significantly associated with worsening (p=0.246). With respect to the dopaminergic load, univariable logistic regression revealed a trend but did not identify the magnitude of the ΔLEDD increase from baseline (p=0.087) or the total LEDD at M6 (p=0.10) as a statistically significant predictor of worsening.

Finally, no significant associations were found for age at diagnosis (OR=1.16, 95% CI: 0.88–1.52, p=0.349), history of hallucinations/psychosis (OR=0.30, 95% CI: 0.08–1.21, p=0.107), use of a dopamine agonist (OR=3.54, 95% CI: 0.59–21.40, p=0.219), or MDS-UPDRS Part I Item 1.6 scores (DDS) (OR=0.97, 95% CI: 0.45–2.10, p=0.957).

Domain-specific analyses revealed that the protective effect of clozapine varied across neuropsychiatric domains. With respect to worsening hallucinations/psychosis, compared with 15% (3/20) of nonusers, none of the 16 clozapine users experienced worsening (p=0.238). With respect to worsening cognition, none of the clozapine users experienced worsening versus 40% (8/20) of nonusers (p=0.005). With respect to worsening depression, 6% (1/16) versus 20% (4/20) of the participants reported worsening depression (p=0.355). With respect to worsening anxiety, 6% (1/16) versus 25% (5/20) of the participants reported worsening anxiety (p=0.196). With respect to worsening apathy, 25% (4/16) versus 35% (7/20) of the participants reported worsening apathy (p=0.718). Overall, 25% (4/16) of clozapine users experienced neuropsychiatric worsening versus 65% (13/20) of nonusers (p=0.023).

Changes in the neuropsychiatric composite score stratified by clozapine and COMT-I use

To directly assess the effects of medication on neuropsychiatric trajectories, we compared changes in a composite neuropsychiatric score (sum of MDS-UPDRS Part I items: cognitive impairment, hallucinations/psychosis, depression, anxiety, and apathy) between baseline and the 6-month follow-up. Clozapine users (n=16) had higher baseline scores (13.2±3.5 vs. 9.2± 3.6) but experienced score improvement (median change -2.0, IQR: -2.0 to -0.8), while nonusers remained stable (median +0.5, IQR: -2.0 to +2.0; p=0.080). Conversely, COMT-I users (n=16) had lower baseline scores (9.1±2.8 vs. 12.4±4.3) but experienced worsening (median change +1.0, IQR: -1.0 to +1.2), whereas nonusers improved (median -2.0, IQR: -2.0 to 0.0; p=0.017).

Stratification by severity of neuropsychiatric and cognitive worsening

To better characterize the heterogeneity within the worsening group, we developed a composite severity score integrating all relevant cognitive and psychiatric subitems from the MDSUPDRS Part I and PDQ-8 at M6. Each item was weighted according to its severity (0 to 4), and patients who experienced worsening were stratified into tertiles: mild (n=6), moderate (n=6), and severe neuropsychiatric deterioration (n=5). Each severity subgroup was then independently compared to the “no worsening” group (n=19) in terms of baseline clinical and treatment variables. While certain patterns were observed—such as higher oral intake frequency or more frequent use of COMT-I in more severe cases—no between-group difference reached statistical significance after Benjamini–Hochberg correction for multiple comparisons (Supplementary Table 2).

DISCUSSION

Neuropsychiatric safety remains a central concern when CSFLI is implemented in aPD patients. In our real-world series, 47% of patients experienced a clinically meaningful worsening in cognitive and/or psychiatric domains over six months, despite stable or improved motor control. These events included isolated cognitive decline (6%), isolated psychiatric aggravation (41%), and mixed cognitive–psychiatric profiles (53%). This finding contrasts sharply with pivotal trial data reporting neu-ropsychiatric adverse events in only 7%–17% of highly selected patients [3,5,6,8] but aligns with observational reports documenting up to 55% in less restrictive cohorts [7,12]. Our findings suggest that in vulnerable, real-world populations, cognitive and psychiatric trajectories require the same vigilance as motor outcomes do.

Several recent multicenter studies have reported neuropsychiatric outcomes under CSFLI, but most have considered them as secondary endpoints, often without structured, combined cognitive–psychiatric measures [6,8]. By targeting these domains specifically and integrating both clinical rating scales and patient-reported outcomes, we revealed a higher-than-anticipated burden, reinforcing the need for systematic monitoring. Notably, the initiation mode (inpatient vs. outpatient) did not influence worsening occurrence, in contrast to the trends suggested by other observational data [8]. From a pathophysiological perspective, events under CSFLI parallel mechanisms proposed for levodopa–carbidopa intestinal gel, both of which provide continuous dopaminergic stimulation that stabilizes motor fluctuations but may modulate mesocorticolimbic circuits, influencing mood, behavior, and cognition [13]. The role of continuous 24-hour levodopa delivery in hallucination development remains controversial, with studies reporting both improvement and worsening [14-17]. Expert recommendations advise lowering nocturnal infusion rates during initial adjustment to limit psychosis risk [17]. In our cohort, the mean nocturnal rate was >50% lower than the daytime rate (0.20 vs. 0.50 mL/h), yet neuropsychiatric worsening still occurred in 47% of patients. This suggests that even substantial nocturnal dose reductions may be insufficient in populations enriched with older, frailer individuals with preexisting vulnerabilities, highlighting that infusion rate adjustments alone cannot prevent such events and that comprehensive baseline screening and close monitoring remain essential. Similarly, we investigated whether the quantitative increase in dopaminergic stimulation drove these adverse events. While the worsening group had a greater mean increase in LEDD (534 mg vs. 274 mg), this difference did not reach statistical significance in logistic regression modeling. This suggests that in this high-risk population, qualitative susceptibility factors (such as COMT inhibition or lack of clozapine protection) may outweigh simple quantitative dopaminergic load in precipitating neuropsychiatric decline.

The predominance of psychiatric-only worsening in our series likely reflects excessive dopaminergic drive to limbic pathways in predisposed individuals. Sustained levodopa exposure may affect nondopaminergic systems—including serotonergic and cholinergic neurotransmission—implicated in hallucinations, confusion, and cognitive fluctuations [18-20]. Pharmacokinetic differences between subcutaneous and intestinal delivery routes could further modify plasma-to-brain levodopa dynamics, potentially exacerbating neuropsychiatric vulnerability. Recent work has demonstrated that atypical psychosis under CSFLI—characterized by auditory/somatic hallucinations and delusions—correlates with DDS and is associated with mesolimbic circuit dysfunction [12]. Clozapine’s selective D4 and 5-hydroxytryptamine receptor 2A (5-HT2A) antagonism normalizes this dysregulated circuitry without compromising the dorsal striatal D2 function required for motor control, explaining its unique ability to treat psychosis without worsening motor activity [21-23]. Our inclusion of apathy and depression alongside hallucinations/psychosis reflects the bidirectional dopaminergic modulation of mesolimbic circuits, where suboptimal levels manifest as apathy while excessive stimulation triggers psychosis [24-26].

The high worsening rate observed likely reflects our cohort’s composition: patients were older, more polymedicated, and had frequent hallucination or confusion histories—factors systematically underrepresented in pivotal studies [3-6]. This finding highlights a recurrent theme in DAT research: safety profiles established under controlled conditions may not translate to heterogeneous, comorbid populations encountered in clinical practice. Our findings support the use of structured neuropsychological screening before CSFLI initiation and close, multidimensional follow-up including regular cognitive testing, systematic caregiver input, and proactive medication adjustment when early signs of deterioration emerge. Critically, 58% of the neuropsychiatrically stable patients had prior confusion or hallucination histories versus only 12%–29% in the worsening group, while 63% of the stable patients received baseline clozapine versus 24% who experienced worsening. These findings suggest that psychiatric history per se does not confer risk; rather, unmanaged psychiatric vulnerability represents the true risk factor. Appropriate clozapine prophylaxis appears to enable access to safe advanced therapy in populations traditionally considered unsuitable candidates. Importantly, neuropsychiatric worsening occurred without consistent motor decline, supporting the hypothesis of partially independent underlying mechanisms.

COMT-I use emerged as the strongest baseline predictor of neuropsychiatric worsening—a potentially modifiable risk factor. COMT-I increases central dopamine availability by reducing peripheral levodopa metabolism, potentially amplifying mesolimbic overstimulation in predisposed individuals [27]. Additionally, COMT inhibition affects noradrenergic and serotonergic neurotransmission, systems implicated in hallucinations and anxiety [27]. The combination with continuous high-dose CSFLI may create synergistic neuropsychiatric vulnerability. These findings suggest that COMT-I discontinuation or dose reduction could be considered when CSFLI is initiated in patients with neuropsychiatric vulnerabilities, although prospective validation is needed.

MAO-B inhibitors and amantadine suggested safety signals despite their limited use (n=3 each, 8%). Compared with 42% of the patients who did not receive these medications, all patients who received these medications experienced worsening, although this difference was not statistically significant. MAOB inhibitors are known to cause confusion, hallucinations, and orthostatic hypotension when combined with high-dose dopaminergic therapy [28]. Amantadine is associated with documented neuropsychiatric risks, including hallucinations, confusion, and delirium [29]. These preliminary findings suggest caution when combining these agents with CSFLI, particularly in neuropsychiatrically vulnerable patients.

The protective role of clozapine warrants particular attention. Our finding that 63% of the stable patients were on baseline clozapine versus 24% of the worsening patients (p=0.023) challenges traditional assumptions. Patients with documented confusion (58%) or hallucinations (58%) in the stable group tolerated CSFLI well when managed with clozapine, inverting the conventional risk model and suggesting that a psychiatric history should prompt enhanced prophylaxis rather than therapeutic exclusion. Domain-specific analyses revealed that the protective effects of clozapine extend across multiple neuropsychiatric domains beyond psychosis. Clozapine significantly prevented cognitive worsening (0% vs. 40%) and showed protective effects against depression (6% vs. 20%), anxiety (6% vs. 25%), and hallucinations/psychosis (0% vs. 15%), although the effects on apathy were modest (25% vs. 35%). Unlike quetiapine, which impairs cognition in PD patients, clozapine preserves cognitive function [30], possibly through neuroprotective mechanisms beyond D4/5-HT2A antagonism, including anti-inflammatory and neurotrophic effects [12]. These exploratory findings suggest that proactive clozapine therapy before CSFLI initiation in patients with a history of neuropsychiatric symptoms warrants investigation in prospective trials.

Our study has several limitations that warrant acknowledgment. First, the small sample size (n=36) limits statistical power, particularly for multivariable modeling. We addressed this by presenting univariable analyses as primary findings and multivariable models as exploratory only. Second, the retrospective design relies on clinical documentation quality and may be subject to ascertainment bias. However, our use of standardized scales (MDS-UPDRS, PDQ-8) and verification through medical record review mitigated this concern. Third, our outcome measure combines cognitive and psychiatric domains of varying severity, creating a heterogeneous endpoint. We addressed this through subgroup analyses stratifying by phenotype (cognitive-only, psychiatric-only, and mixed) and severity (mild, moderate, and severe), although small subgroup sizes limited interpretability. Fourth, the neuropsychiatric assessment was limited to routinely collected clinical scales (MDS-UPDRS Part I, PDQ-8). Specialized instruments such as the MDS-Non-Motor Rating Scale, Neuropsychiatric Inventory, or validated psychosis and impulse control disorder screening tools (e.g., QUIPCS) are not used in standard clinical practice at our center, despite their demonstrated value in predicting atypical psychosis risk [12]. Fifth, our single-center design may limit generalizability to centers with different practice patterns. Finally, a six-month follow-up may not capture late-onset complications, although most neuropsychiatric events in DATs occur within the first 3–6 months. Additionally, the absence of a control group receiving conventional oral therapy prevents the definitive attribution of neuropsychiatric worsening to CSFLI versus natural disease progression over the 6-month period. However, several arguments support a predominantly treatment-related contribution: 1) our 44.4% worsening rate substantially exceeds typical 6-month progression rates reported in longitudinal advanced PD cohorts, 2) the strong associations with baseline clozapine and COMT-I use implicate dopaminergic mechanisms, and 3) the relatively short observation period makes substantial natural progression less probable.

In a real-world aPD population enriched with older individuals with cognitive impairment, and patients with prior psychiatric complications, neuropsychiatric worsening occurred in nearly half of the patients who underwent CSFLI over six months. However, our analysis revealed that such complications are not inevitable even in high-risk patients. Proactive clozapine therapy enabled safe CSFLI use in patients with documented psychiatric histories, challenging traditional exclusion criteria. Conversely, COMT-I emerged as a modifiable risk factor warranting caution. Motor outcomes remained stable regardless of neuropsychiatric status, confirming the independence of motor and neuropsychiatric trajectories under continuous dopaminergic stimulation. These exploratory findings suggest a potential shift from risk-based exclusion to risk-stratified prophylaxis in CSFLI candidate selection, although prospective validation is essential. The patient’s psychiatric history may warrant enhanced monitoring and consideration of prophylactic clozapine rather than therapeutic exclusion. With such strategies, CSFLI can be safely extended to populations previously deemed unsuitable, thereby democratizing access to transformative advanced therapies. Prospective studies with larger samples and standardized psychiatric screening (e.g., QUIP-CS) are needed to validate these findings and refine clinical protocols.

Supplementary Materials

Notes

Conflicts of Interest

The authors have no financial conflicts of interest.

Funding Statement

None

Acknowledgments

None

Author Contributions

Conceptualization: Guillaume Baille. Data curation: Guillaume Baille. Formal analysis: Quentin Salardaine, Clément Desjardins, Guillaume Baille. Investigation: all authors. Project administration: Jean-Philippe Brandel. Supervision: Jean-Philippe Brandel. Validation: Guillaume Baille, Jean-Philippe Brandel. Visualization: Quentin Salardaine, Clément Desjardins. Writing—original draft: Quentin Salardaine, Clément Desjardins. Writing—review & editing: all authors.

Figure 1.

Baseline predictors of neuropsychiatric and/or cognitive worsening under CSFLI. Forest plot showing odds ratios (OR, logarithmic scale) and 95% confidence intervals from univariable logistic regression models for the association between baseline characteristics and neuropsychiatric/cognitive worsening at 6 months. Blue points indicate significant protective factors (OR<1, p<0.05), red points indicate significant risk factors (OR>1, p<0.05), and gray points indicate non-significant associations (p≥0.05). The vertical dashed line represents the null value (OR=1). Significant predictors included clozapine use (protective, OR=0.18, p=0.023), higher MDS-UPDRS Part I scores (protective, p=0.027), and COMT-I use (risk factor, OR=9.00, p=0.006). CSFLI, continuous subcutaneous foslevodopa/foscarbidopa infusion; LEDD, levodopa equivalent daily dose; MDS-UPDRS, Movement Disorder Society–Unified Parkinson’s Disease Rating Scale; MoCA, Montreal Cognitive Assessment; PDQ-8, Parkinson’s Disease Questionnaire-8; COMT-I, catechol-O-methyltransferase inhibitor.

REFERENCES

• 1. LeWitt PA. Rethinking the role of continuous dopaminergic stimulation in Parkinson disease therapy. Parkinsonism Relat Disord 2025;139(Suppl 1):107354.ArticlePubMed
• 2. van Wamelen DJ, Grigoriou S, Chaudhuri KR, Odin P. Continuous drug delivery aiming continuous dopaminergic stimulation in Parkinson’s disease. J Parkinsons Dis 2018;8(S1):S65–S72.ArticlePubMedPMCPDF
• 3. Soileau MJ, Aldred J, Budur K, Fisseha N, Fung VS, Jeong A, et al. Safety and efficacy of continuous subcutaneous foslevodopa-foscarbidopa in patients with advanced Parkinson’s disease: a randomised, double-blind, active-controlled, phase 3 trial. Lancet Neurol 2022;21:1099–1109.ArticlePubMed
• 4. Fung VSC, Aldred J, Arroyo MP, Bergquist F, Boon AJW, Bouchard M, et al. Continuous subcutaneous foslevodopa/foscarbidopa infusion for the treatment of motor fluctuations in Parkinson’s disease: considerations for initiation and maintenance. Clin Park Relat Disord 2024;10:100239.ArticlePubMedPMC
• 5. Aldred J, Freire-Alvarez E, Amelin AV, Antonini A, Bergmans B, Bergquist F, et al. Continuous subcutaneous foslevodopa/foscarbidopa in Parkinson’s disease: safety and efficacy results from a 12-month, single-arm, openlabel, phase 3 study. Neurol Ther 2023;12:1937–1958.ArticlePubMedPMCPDF
• 6. Aldred J, Bouchard M, Martínez-Castrillo JC, Soileau MJ, Spiegel AM, Bergmann L, et al. Efficacy and safety of foslevodopa/foscarbidopa monotherapy in patients with Parkinson’s disease. Mov Disord Clin Pract 2026;13:181–190.ArticlePubMedPDF
• 7. Brohée S, Roze E, Grabli D, Letrillart H, Mantisi L, Foucard C, et al. Cognitive and psychiatric adverse effects of foslevodopa/foscarbidopa in patients with Parkinson’s disease. Mov Disord Clin Pract 2025;12:1028–1030.ArticlePubMedPMC
• 8. Rukavina K, Ebersbach G, Gruber D. Foslevodopa/foscarbidopa continuous subcutaneous infusion in Parkinson’s disease: real-world short-term data on tolerability, infusion rate adjustments and concomitant medication. Mov Disord Clin Pract 2025;12:2317–2323.ArticlePubMedPMCPDF
• 9. Postuma RB, Berg D, Stern M, Poewe W, Olanow CW, Oertel W, et al. MDS clinical diagnostic criteria for Parkinson’s disease. Mov Disord 2015;30:1591–1601.ArticlePubMed
• 10. Horváth K, Aschermann Z, Kovács M, Makkos A, Harmat M, Janszky J, et al. Minimal clinically important differences for the experiences of daily living parts of Movement Disorder Society-Sponsored Unified Parkinson’s Disease Rating Scale. Mov Disord 2017;32:789–793.ArticlePubMedPDF
• 11. Horváth K, Aschermann Z, Kovács M, Makkos A, Harmat M, Janszky J, et al. Changes in quality of life in Parkinson’s disease: how large must they be to be relevant? Neuroepidemiology 2017;48:1–8.ArticlePubMedPDF
• 12. Ge L, Kimura Y, Kakuda K, Ogawa K, Kajiyama Y, Asai K, et al. Atypical psychosis in Parkinson disease: a retrospective study on 24-hour continuous subcutaneous infusion of foslevodopa/foscarbidopa. Neurol Clin Pract 2025;15:e200534. ArticlePubMedPMC
• 13. Antonini A, D’Onofrio V, Guerra A. Current and novel infusion therapies for patients with Parkinson’s disease. J Neural Transm (Vienna) 2023;130:1349–1358.ArticlePubMedPMCPDF
• 14. Ricciardi L, Bove F, Espay KJ, Lena F, Modugno N, Poon YY, et al. 24-hour infusion of levodopa/carbidopa intestinal gel for nocturnal akinesia in advanced Parkinson’s disease. Mov Disord 2016;31:597–598.ArticlePubMed
• 15. Nyholm D, Jansson R, Willows T, Remahl IN. Long-term 24-hour duodenal infusion of levodopa: outcome and dose requirements. Neurology 2005;65:1506–1507.ArticlePubMed
• 16. Cruse B, Morales-Briceño H, Chang FCF, Mahant N, Ha AD, Kim SD, et al. 24-hour levodopa-carbidopa intestinal gel may reduce troublesome dyskinesia in advanced Parkinson’s disease. NPJ Parkinsons Dis 2018;4:34.ArticlePubMedPMCPDF
• 17. Thakkar S, Fung VSC, Merola A, Rollins M, Soileau MJ, Kovács N. 24-hour levodopa-carbidopa intestinal gel: clinical experience and practical recommendations. CNS Drugs 2021;35:137–149.ArticlePubMedPMCPDF
• 18. Zahodne LB, Fernandez HH. Pathophysiology and treatment of psychosis in Parkinson’s disease: a review. Drugs Aging 2008;25:665–682.ArticlePubMedPMC
• 19. Fénelon G, Mahieux F, Huon R, Ziégler M. Hallucinations in Parkinson’s disease: prevalence, phenomenology and risk factors. Brain 2000;123(Pt 4):733–745.ArticlePubMed
• 20. Diederich NJ, Fénelon G, Stebbins G, Goetz CG. Hallucinations in Parkinson disease. Nat Rev Neurol 2009;5:331–342.ArticlePubMedPDF
• 21. Factor SA, Brown D. Clozapine prevents recurrence of psychosis in Parkinson’s disease. Mov Disord 1992;7:125–131.ArticlePubMed
• 22. Parkinson Study Group. Low-dose clozapine for the treatment of druginduced psychosis in Parkinson’s disease. N Engl J Med 1999;340:757–763.ArticlePubMed
• 23. Pollak P, Tison F, Rascol O, Destée A, Péré JJ, Senard JM, et al. Clozapine in drug induced psychosis in Parkinson’s disease: a randomised, placebo controlled study with open follow up. J Neurol Neurosurg Psychiatry 2004;75:689–695.ArticlePubMedPMC
• 24. Hirano S. Clinical implications for dopaminergic and functional neuroimage research in cognitive symptoms of Parkinson’s disease. Mol Med 2021;27:40.ArticlePubMedPMCPDF
• 25. Wetmore JB, Arbelo JM, Catalán MJ, Valldeoriola F, Rodriguez-Blazquez C, Martinez-Martin P. Psychometric properties of the apathy scale in advanced Parkinson’s disease. Parkinsons Dis 2019;2019:1965394.ArticlePubMedPMCPDF
• 26. Valldeoriola F, Catalán MJ, Escamilla-Sevilla F, Freire E, Olivares J, Cubo E, et al. Patient and caregiver outcomes with levodopa-carbidopa intestinal gel in advanced Parkinson’s disease. NPJ Parkinsons Dis 2021;7:108.ArticlePubMedPMCPDF
• 27. Samudra N, Patel N, Womack KB, Khemani P, Chitnis S. Psychosis in Parkinson disease: a review of etiology, phenomenology, and management. Drugs Aging 2016;33:855–863.ArticlePubMedPMCPDF
• 28. Jankovic J, Stacy M. Medical management of levodopa-associated motor complications in patients with Parkinson’s disease. CNS Drugs 2007;21:677–692.ArticlePubMed
• 29. Perez-Lloret S, Rascol O. Efficacy and safety of amantadine for the treatment of L-DOPA-induced dyskinesia. J Neural Transm (Vienna) 2018;125:1237–1250.ArticlePubMedPDF
• 30. Yunusa I, Rashid N, Seyedin R, Paratane D, Rajagopalan K. Comparative efficacy, safety, and acceptability of pimavanserin and other atypical antipsychotics for Parkinson’s disease psychosis: systematic review and network meta-analysis. J Geriatr Psychiatry Neurol 2023;36:417–432.ArticlePubMedPDF

Figure & Data

References

Citations

Citations to this article as recorded by  

• Sex Differences in the Treatment of People with Parkinson’s Disease with a Device-Aided Therapy: A Prospective Real-World Study
  Diego Santos García, Ángela Solleiro Vidal, Marta Blázquez Estrada, Pablo Mir, Nuria López Ariztegui, Déborah Alonso Modino, Inés Legarda, Alejandro Peral, Rocío García-Ramos, Iria Cabo, Pilar Sánchez Alonso, Jorge Hernández-Vara, Javier Ruíz Martínez, Ma
  Medical Sciences.2026; 14(2): 217.     CrossRef
• Real-world outcomes and early discontinuation of foslevodopa/foscarbidopa in Parkinson’s disease
  Takashi Tsuboi, Katsuo Kimura, Kensuke Ikenaka, Toru Baba, Keita Matsuura, Yuta Kajiyama, Daiki Kamiyama, Takashi Uematsu, Katsuya Abe, Keita Kakuda, Hiroyuki Takenaka, Koichi Miyashita, Akiko Saito, Masahito Mihara, Akihiro Shindo, Takafumi Hasegawa, Yas
  Journal of Neurology.2026;[Epub]     CrossRef
• Successful use of continuous subcutaneous Foslevodopa/Foscarbidopa in Parkinson's disease with bipolar disorder without psychiatric destabilization
  Joanna Siuda, Justyna Gawryluk, Amelia Krupczak, Mateusz Toś
  Clinical Parkinsonism & Related Disorders.2026; 15: 100475.     CrossRef
• Neuropsychiatric Adverse Events Associated With Foslevodopa/Foscarbidopa Continuous Subcutaneous Infusion in Clinical Practice: A Multicenter Study
  David Campo‐Caballero, Jon Rodriguez‐Antiguedad, Arnau Puig‐Davi, Javier Ruiz‐Martinez, Ana Vinagre‐Aragón, Elisabet Mondragón, Lara Pardina Vilella, Nuria López‐Ariztegui, Nuria Caballol, Guillermo González‐Ortega, Iria Cabo, Rocío García Ramos, Inés Leg
  European Journal of Neurology.2026;[Epub]     CrossRef
• Real-world implementation of continuous subcutaneous foslevodopa/foscarbidopa in advanced Parkinson’s disease: first clinical experience from the Middle East
  Shivam Om Mittal, Kiran Waqar, Emma De Freitas, Yasmin Abdelmajid, Arshiya Anjum, Leanne Brown, Moncy Thomas
  Therapeutic Advances in Neurological Disorders.2026;[Epub]     CrossRef
• Psychosis After Weight Loss During Continuous Subcutaneous Foslevodopa/Foscarbidopa Therapy
  Laura Hartmann, Klaus Seppi, Lukas Gattermeyer-Kell, Petra Katschnig-Winter, Daniela Kern, Mariella Kögl, Elisabeth Sokol, Christian Enzinger, Petra Schwingenschuh
  Journal of Movement Disorders.2026; 19(3): 349.     CrossRef
• Real-world experience with foslevodopa/foscarbidopa in Germany: therapeutic insights and lessons learned
  David Weise, Tobias Warnecke, Carsten Buhmann, Georg Ebersbach, Carsten Eggers, Björn Falkenburger, Günter Höglinger, Brit Mollenhauer, David Pedrosa, Martin Südmeyer, Daniel Weiss, Jan Kassubek, Wolfgang H. Jost
  Neurological Research and Practice.2026;[Epub]     CrossRef

Comments on this article