Introduction

Parkinson’s disease (PD) is a progressive, neurodegenerative, and disabling disorder characterized by the loss of dopamine-producing brain cells leading to the core motor features of the disease [1, 2]. Treatment with dopamine replacement agents, such as the dopamine precursor levodopa and dopamine agonists (DA), is very effective in reducing motor symptoms, but can cause motor and non-motor complications. These side-effects include motor fluctuations and dyskinesias and non-motor symptoms such as sleep disturbances, psychosis, and impulse control disorders (ICDs), the latter particularly associated with DA [3,4,5].

There are four major ICDs in PD: pathological gambling, compulsive buying, compulsive sexual behavior, and binge/compulsive eating, and can generate considerable burden to patients, their families, and society. Most patients do not spontaneously offer information about ICDs, either because of shame or unrecognition, so family/caregiver input can be particularly helpful. A useful tool to detect ICDs is the Questionnaire for Impulsive–Compulsive Disorders in Parkinson’s disease (QUIP), a validated quick screening questionnaire that focus in the multiple aspects of impulsivity [6].

Continuous levodopa–carbidopa intestinal gel (LCIG) infusion has become a therapeutic option for advanced PD, as it effectively smooths levodopa plasma concentrations and improves motor fluctuations [7]. Levodopa infusions led to greater reductions in ‘‘Off’’ time compared with conventional pharmacotherapy and improved motor and non-motor symptoms and quality of life (QoL) [8, 9]. However, data about its effects on ICDs severity are scarce. We aimed to assess the effect of 6-month treatment with LCIG infusion in patients with advanced PD in terms of control ICD symptoms, sleep disturbances, and impact on patients’ and caregivers’ QoL.

Methods

Study design and participants

This was a multicenter, prospective, and observational study conducted at 13 Spanish outpatient neurology clinics from December 2012 to July 2014. Participants were adults diagnosed with idiopathic PD according to the UK Brain Bank criteria [10]. They received levodopa therapy and displayed advanced motor symptoms with fluctuations and/or dyskinesias despite optimized medical therapy, and were candidates for switching from oral medication to LCIG infusion. Exclusion criteria were presence of atypical parkinsonian features, cognitive impairment (Mini Mental State Evaluation [MMSE] [11] score < 24) or other relevant medical problems. Additionally, the main caregiver had to remain the same person throughout the observation period.

The study was approved by the Ethics Committee of the Hospital Clínico San Carlos (Madrid, Spain) and was conducted in accordance with the principles of the Declaration of Helsinki. All patients and caregivers provided written informed consent before any study procedure was implemented.

Procedures

Clinical observations on medical history, PD characteristics, previous PD treatment(s) and reason to start duodenal levodopa infusion were collected upon inclusion to the study. Data were recorded at baseline (BL) prior to initiation of LCIG, and at follow-up visits one (M1), three (M3), and 6 months (M6) after the definitive gastrostomy procedure. Treatment with LCIG (consisting of a water-based suspension containing micronized levodopa (20 mg/mL) and carbidopa (5 mg/mL) in methylcellulose (Duodopa™) was initiated according to standard clinical procedures in patient routine care at each center. LCIG infusion was started by a temporary nasoduodenal tube and its effectiveness was assessed for 2–3 days. Subsequently, eligible participants underwent jejunal placement of a gastrojejunostomy tube through a percutaneous endoscopic gastrostomy under sedation and local anesthesia. Duodenal levodopa infusion was administered for approximately 16 waking hours using a portable pump. Continuing use of other PD drugs as concomitant treatment to LCIG infusion was allowed at the discretion of the treating physician. Dosing increment of insomnia medication was not permitted.

The following outcomes were assessed (see Table 1): Unified Parkinson’s Disease Rating Scale [12] parts II (on- and off-state) and III (on-state); Clinical Impression of Severity Index for Parkinson’s Disease (CISI-PD) [13]. Neuropsychiatric symptoms were assessed using the Scale for Evaluation of Neuropsychiatric disorders in Parkinson´s Disease (SEND-PD) [14] and the Neuropsychiatric Inventory-Questionnaire (NPI-Q) [15], the latest completed by the main caregiver. ICDs were recorded using the questionnaire for ICD in PD (QUIP-RS) [6]. This is a self-report and self-completed screening tool developed specifically to detect the presence of clinically significant impulse control (compulsive gambling, buying, sexual behavior, and eating) and related behavior (punding, hobbyism, and walkabout) symptoms reported to occur in PD. Cognitive evaluations were performed using the MMSE [11] and the lexical fluency performance [16]. To evaluate sleep disorders we used the Scales for Outcomes in Parkinson’s disease—sleep (SCOPA-S) and the Parkinson’s Disease Sleep Scale (PDSS) [17]. Psychosocial function was assessed by means of the Scales for Outcomes in Parkinson’s disease—psychosocial questionnaire (SCOPA-PS) [18]. The frequency and intensity of pain were measured by visual analog scales (VAS). Finally, in order to quantify the impact of PD on caregiver’s QoL we asked the main caregiver to fill out the Zarit Burden Inventory (ZBI), a 22-item scale that rates the impact on the carer’s physical, emotional, and socio‐economic status [19].

Table 1 Study assessments by participant
Full size table

Statistical analysis

To assess clinical outcomes, we used a within-patient control mechanism, whereby every patient acted as their own control. Due to the limited target population and the observational design of the study, all eligible patients from participant sites were enrolled into the study during 18-month period of enrollment. For descriptive analysis, categorical variables are summarized in numbers and percentages, continuous variables are presented in means, standard deviations (SD), medians, interquartile range (IQR), and range. The non-parametric Wilcoxon’s test was used for comparisons of continuous variables, with the Bonferroni correction for all pairwise comparisons. Apart from the relative changes, the effect sizes were calculated using the differences between the pre- and post-treatment means divided by the corresponding SD at the baseline. To interpret these effect sizes, the established criteria that consider effect sizes of 0.20 to < 0.50 as small, ≥ 0.50 to < 0.80 as moderate, and ≥ 0.80 as large were adopted [20]. In addition, the number needed to treat (NNT) for obtaining 1 patient improving the threshold or more for each outcome variable was calculated. The selected threshold was ½ baseline SD, which has been proposed as a threshold for a clinically significant change in health status and patient-reported outcomes [20, 21]. Finally, we used generalized linear models to examine possible associations with the change in QUIP-RS after 6 months of LCIG treatment. All statistical tests were two-sided and were considered significant when p < 0.05. Statistical analyses were performed using the SAS version 9.4 software.

Results

A summary of the demographic and clinical characteristics of the 62 participating patients is presented in Table 2. Most of the patients had at least primary school education and were retired. Mean MMSE score was 27.6 (SD 2.1) and mean lexical fluency was 10.3 (SD 5.8). Median Hoehn and Yahr stage was 3 (range 1–4). The basal mean oral levodopa dose was 1117.4 (SD 766.1) mg, and 43 patients (74.2%) were receiving DAs: 17 pramipexole, 13 ropinirole, and 13 rotigotine. Mean age of the main caregivers was 63.7 (SD 12.5) years, 74.1% were female, and all except two were patients’ relatives. Mean time spent taking care of the patient was 19.6 (SD 6.7) h/day.

Table 2 Baseline characteristics of participants
Full size table

There were 16 premature terminations, representing 25.8% of the population, due to adverse drug reactions (n = 1), procedure and device related events (n = 3), lack of efficacy (n = 5), and consent withdrawal (n = 1). During the follow-up period one patient died (sepsis, not related to treatment) and five patients were lost to follow-up.

Efficacy of LCIG in PD-related symptoms

The “off” time was reduced by 5.8 h/day (79%) whereas duration of dyskinesias was not affected, although it was numerically higher at the end of the study. Benefits of LCIG were also evident by significant improvements in the activities of daily living and motor subscales of the UPDRS (part II and part III), the effect sizes ranging between 0.30 and 0.78. The mean changes in the overall severity of the disease as evaluated with the CISI-PD were significant and considerable, with a 25.0% reduction in the total score (Table 3). At M6 patients were receiving a mean levodopa dose of 1378.2 (SD 437.1) mg and five patients (10.9%) received DAs: 1 pramipexole, 2 ropinirole, and 2 rotigotine.

Table 3 Efficacy of LCIG on PD severity and related symptoms (per protocol population; n = 46)
Full size table

Effects of LCIG on patient- and caregiver-reported outcomes

Treatment with LCIG infusion was associated with progressive and significant improvements in ICD symptoms over the study period (Fig. 1a). The total QUIP-RS score was reduced by a mean of 5.1 points (53.9% reduction) by month 3 and by 7.0 points (64.4% reduction) by month 6. Mean reductions at month 6 were also significant in the eating, hobbyism, punding, and medication subscores. No effect was seen in the gambling, sex, and buying domains. The relative changes in the QUIP-RS scores and subscores were of high magnitude in terms of percentages of change, although the effect sizes were small to moderate (Table 4). Figure 1b shows QUIP-RS total scores for patients that were receiving and not receiving DAs in basal visit, without evidence of different patterns of response among DAs (Supplementary Table 1). The intended comparative analysis could not be conducted due to the low sample size of the subgroups.

Fig. 1
figure 1

Changes in QUIP-RS total score in the whole population (a) and stratifying by DA treatment in basal visit (b). QUIP-RS Questionnaire for Impulsive–Compulsive disorders in Parkinson’s disease—rating scale, DA dopamine agonists, BL baseline, M1-6 month 1-6. *p < 0.001 vs BL

Full size image
Table 4 Effects of LCIG as measured with neurocognitive and psychopathological scales (per protocol population, n = 46)
Full size table

The reductions in the psychotic and ICD domains of SEND-PD were significant, with percentage reductions of 54.5 and 46.8, respectively, which corresponded to effect sizes of 0.43 and 0.42 (Table 4). Other neuropsychiatric symptoms were also reduced according to both domains of the NPI-Q rated by the caregivers. A positive effect on sleep disturbances was evident as evaluated with the SCOPA-S and the PDSS, with moderate-to-large effect sizes that ranged from 0.73 for daytime symptoms to 0.99 for the total SCOPA-S score (Table 4). Patient psychosocial function improved, but caregiver’s burden remained unchanged. Frequency and intensity of pain were similar at baseline and final visits.

Regression models for ICDs

In order to understand the relationship of different variables to the change observed in ICDs after treatment with LCIG, we performed separate stepwise logistic regression analyses taking final minus basal QUIP-RS total score as the dependent. Variables included in the regression model were age, sex, disease duration from diagnosis, DA treatment, and levodopa dose > 1000 mg. The results showed that only two variables were significantly and independently related to the QUIP-RS score change in our sample: disease duration from diagnosis (p = 0.043) was positively related, while age (p = 0.001) was negatively related.

Discussion

During the last years, there has been increasing awareness that PD patients may be at increased risk of developing ICDs or related phenomena, which can lead to shattering consequences, such as financial ruin, divorce, loss of employment, and diverse health risks. ICDs are also associated with increased functional impairment [22], decreased QoL [23], and increased caregiver burden [24]. The results of this observational, prospective study suggest that LCIG is effective for the improvement of ICDs and related behaviors associated with advanced PD and/or its treatment. LCIG also ameliorated motor and psychotic symptoms and produced clinically relevant improvements in sleep impairment, QoL, and disability. LCIG was well tolerated in our group of patients, with only one participant discontinuing due to side-effects related to LCIG.

ICDs are relatively frequent in PD patients and most often appear or worsen after initiation or dose escalation of dopamine replacement therapy [25]. However, the relationship is far from causal, because the vast majority of patients do not develop ICDs even when taking very high doses of the medication for long periods. Genetic differences in dopamine receptors and metabolism and different personality traits most likely play a role in determining which patients are at greatest risk [22, 26]. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria the prevalence of ICDs is about 14% of PD patients receiving DA [5]; when the QUIP [6] is used, prevalence elevates to around 40% [27]. The QUIP-RS, as the questionnaire itself, focusses attention to those behaviors that the patient identifies as problematic, and uses frequency as the main indicator of severity. This approach could miss instances in which a patient minimizes the impact of an ICD or underestimates a behavior that is relatively infrequent but intense. Furthermore, an important issue to consider when assessing ICD symptoms is that there may be discordance between patient and informant reporting [28, 29]. In our study, we minimized this potential bias as the main caregivers, always the same, assisted for the completion of the neuropsychiatric assessments.

It has been proposed that, compared with other advanced therapies for PD, LCIG may be suitable for all patients with long duration of disease [30]. The enteral infusion of levodopa/carbidopa gel formulation provides more stable plasma levels when compared with oral levodopa administrations and eliminates pulsatile dopaminergic stimulation [7, 31]. Diverse studies have suggested that ICDs could be related to the acute release of dopamine in the ventral striatum associated with a pulsed dopaminergic therapy and that continuous infusion of the drug may reduce this concern [32]. The improvement observed in our results is in line with those described in previous works. In a recent observational study of 19 patients treated with LCIG, ICDs resolved in 30% of cases and no new cases developed during 3-year follow-up, although two patients had persistent compulsive gambling and punding [33]. In a small retrospective study of 8 PD subjects (6 had ICDs and 3 had DDS), dopamine infusions were administered 15 h a day for each patient; DDS and ICDs improved in all subjects, and concomitant punding was also noted to improve [34]. However, ICD behaviors persisted in some patients who had pre-procedure discontinuation of dopamine agonist, although continued use of high doses of levodopa and amantadine use may have contributed to this. Another report of 15 patients that received LCIG for advanced disease described the occurrence of ICDs in 27% of cases [35], all of them with a previous psychiatric diagnosis and a higher daily levodopa intake at six months of drug infusion [35].

An additional risk factor associated with ICDs and related behaviors in PD is young age at disease onset, which is associated with slower progression of motor symptoms and longer disease course with preserved cognition, but also with earlier appearance of motor fluctuations, dyskinesias, and psychiatric symptoms [36]. In our sample, PD duration was positively related to changes in the QUIP-RS score indicating that those who probably were more affected showed greater improvement.

Sleep disturbances are among the most common non-motor complications of PD, and increase in frequency with advancing disease. They are associated with a poor QoL for the patient and the caregiver [37]. Established causes of insomnia include nocturnal motor symptoms, pain, psychiatric complications such as depression and hallucinations, and also medication effects, which may disrupt the sleep architecture and lead to motor fluctuations [38]. As could be expected, the decreased in motor and neuropsychiatric symptoms associated with LCIG treatment was accompanied by an improvement in sleep quality in our population, regardless of pain frequency or intensity that remained unchanged. All these improvements may have conducted to the better psychosocial function experienced by the patients at the end of the study. By contrast, caregiver burden was not significantly modified by levodopa infusion. It has been proposed that neuropsychiatric symptoms are important determinants of carer burden, particularly mood/apathy and psychosis, while ICDs could play a lesser role, although disruptive in some cases [39]. In fact, cognitive behavioral therapy for patients with ICDs, although effective for patients, did not attenuate caregiver burden [40].

An important strength of the present study is that participants represented a typical routine-care population of advanced patients with PD suitable for switching from conventional medication to LCIG infusion, so selection bias was unlikely. However, the criterion for inclusion of patient without cognitive impairment could have introduced an unintended selection bias, as it has been shown that neuropsychiatric disorders are more prevalent and burdensome in patients with dementia [39].

The major limitations of this study were its open design and the lack of a control group. However, reduction/discontinuation of DAs as the sole strategy for ICDs improvement does not seem the most ethical approach in patients with advanced PD because of significant motor worsening. The utilization of self-reported measures from patients and caregivers carries the inherent variability associated with their subjective status. For this reason, we selected from the available tools questionnaires and scales with consistent psychometric properties. Finally, we could not conduct a comparative analysis between patients receiving and not receiving DAs due to the small sample of the subgroups. Big samples with this kind of treatment are very scarce and a low number of investigations were specifically focused on the ICDs and sleep disorders. Although the present results do not allow assuring that LCIG is itself the cause of ICD symptoms resolution, they clearly show that the patient improves from the motor point of view, and that by the action of continuous dopaminergic stimulation, withdrawal of DAs, or both, ICDs improve.

In summary, management of ICDs in advanced PD is challenging. As many unwanted side-effects of medications, ICDs often start when dopaminergic medications are increased or when a second dopaminergic drug is added, so it is important for the physician to continue to ask for a patient and/or caregiver report about ICDs at follow-up visits, especially after dose changes, and to institute suitable treatments. Currently, there is no consensus regarding treatment of ICDs because of paucity of randomized clinical trials, phenotypic heterogeneity of impulsive behaviors, limited comparison tools across studies, and intrinsic differences among PD patients. This study provides evidence regarding the benefits of LCIG as a specific therapy for ICDs in patients with advanced PD, together with improvements in motor symptoms and sleep quality, without a significant increase in levodopa dose. However, this treatment may present complications related to the infusion system, which may require frequent revision. Additional studies with a controlled sample of patients are needed.