Study design

The study measured effectiveness by evaluating efficacy, safety/tolerability and other patient outcomes when used under usual conditions in patients with dementia, stroke, or TBI. Adults with PBA secondary to dementia, stroke, or TBI and a CNS-LS score ≥13 (scale range, 7 [no symptoms] to 35 [maximum]) were enrolled in an open-label, 90-day study of DM/Q 20/10 mg twice daily (once daily during Week 1). Clinical assessments were performed at baseline, Day 30, and Day 90, with a telephone contact at Day 60. This study was conducted in the United States at 74 enrolling sites, from 26Feb2013 to 30Apr2015, according to Good Clinical Practice and the Declaration of Helsinki. The protocol and study materials were approved by the institutional review board at each site and registered on www.clinicaltrials.gov. (identifier: NCT01799941).

Participants

Participants were eligible for enrollment if they had a clinical diagnosis of PBA based on published criteria and a clinical diagnosis of dementia, stroke, or TBI [2]. PBA diagnostic criteria were (1) involuntary or exaggerated episodes of emotional expression (i.e., laughing or crying); (2) development of symptoms that represent a change from the person's usual emotional reactivity occurring subsequent to a specified brain disorder; (3) episodes are incongruent with or out of proportion to the individual's mood state and independent to or in excess of a provoking stimulus; and (4) symptoms are not better accounted for by another disorder, substance abuse, or medication use. Patients were also required to have a CNS-LS score ≥13 [27, 28], the same score required for entry into phase 3 trials of DM/Q for PBA [20–22].

Eligible participants had documented diagnoses of one of the following: dementia, (including Alzheimer’s, vascular, Lewy body, or frontotemporal dementia; ischemic or hemorrhagic stroke; or mild, moderate, or severe non-penetrating TBI. There were no restrictions on types of allowed concomitant medications with the exception of those contraindicated by the US Prescribing Information [16], specifically monoamine oxidase inhibitors (MAOIs), and drugs that both significantly prolong QT interval and are primarily metabolized by CYP2D6 (e.g., thioridazine). Other medications typically used by patients with the studied neurologic conditions (including those that are metabolized by CYP2D6) were allowed with the stipulation that medications for management of dementia, such as memantine or acetylcholinesterase inhibitors should be stable for at least 6 weeks and other neuropsychiatric medications such as anticonvulsants, antidepressants, antipsychotics, anxiolytics, and sedative/hypnotics should be stable for at least 2 months prior to baseline; any medication changes, if deemed necessary, were recorded. Potential participants were excluded if they had severe dementia (Mini-Mental State Examination [MMSE] score <10); stroke within 3 months of study enrollment; penetrating TBI; severe depressive disorder; history or current symptoms of schizophrenia (including psychosis), schizoaffective disorder or bipolar disorder; substance/alcohol abuse in the preceding 3 years; systemic disease, neurologic condition or brain injury that was unstable or rapidly changing within the 3 months prior to enrollment; life expectancy ≤6 months; contraindication to DM/Q use (including known QT interval prolongation); DM/Q use during the previous 6 months; or interventional clinical study participation within the preceding 30 days. Written informed consent was obtained from all participants or from legally authorized representatives.

Outcome measures

Secondary measures

Secondary measures included PBA episode count for the 7 days preceding each study visit as well as the Clinical Global Impression of Change (CGI-C) and Patient (or Caregiver) Global Impression of Change (PGI-C; 7-point scales ranging from 1 [very much improved] to 7 [very much worse]) based on overall change in the patient’s condition with respect to PBA. In addition, a quality-of-life visual analog scale (QOL-VAS) assessed the degree to which PBA episodes affected the patient’s overall quality of life (11-point scale ranging from 0 [“not at all”] to 10 [“significantly”]) during the past week and a 5-point Likert-type scale rated patient satisfaction with treatment from 1 [very dissatisfied] to 5 [very satisfied]. The Folstein Mini-Mental State Examination (MMSE; 11-item assessment of orientation, memory, attention, and language scored from 0 to 30) was included as a cognitive assessment [29]. As an additional measure, the Patient Health Questionnaire (PHQ-9; a 9-item assessment of depression symptoms with each item rated 0 [not at all] to 3 [nearly every day] based on frequency of occurrence over the past 2 weeks, for a total possible score of 27) was included [30]. Disease-specific functional measures, namely the Neurobehavioral Functioning Inventory and Stroke Impact Scale were assessed in the TBI and stroke cohorts, respectively, and will be reported separately. Caregivers were allowed to complete all assessments except the MMSE for any patients who were unable to do so. The timing of assessments is illustrated in Fig. 1.

Statistical analysis

The safety analysis set included all participants who received at least 1 dose of DM/Q. The effectiveness analysis set comprised all participants in the safety set who also met all study eligibility criteria, and completed at least 1 post-baseline CNS-LS assessment. The Medical Dictionary for Regulatory Activities (version 15.1) was used for coding, categorizing, and reporting AEs.

All data were analyzed descriptively. Changes from baseline in ratings at Day 30 and 90 were also analyzed inferentially using one-sample t-tests for rating scale measures (CNS-LS, QOL-VAS, MMSE, and PHQ-9) and a mixed-effects Poisson regression model using number of PBA episodes in the past 7 days as the dependent variable to estimate change in PBA episode counts. The mixed-effects Poisson regression model incorporated age, gender, and time (Day 30 and Day 90) as fixed effects while allowing for individual differences in baseline rate (random-subject effects). The percentage change episode count from baseline to a given visit is 1 minus the appropriate time parameter (λ).

There was no imputation method for missing data; however, if the patient had a Final Visit, that visit was included as the Day 90 Visit. If there was no Final Visit, the Day 30 visit was not carried forward as the subject’s Final Visit.

The primary analysis tested the null hypothesis that the mean change in CNS-LS score from baseline to the Day 90/Final Visit was equal to zero; the 95 % confidence interval (CI) was also reported to enable a descriptive comparison with the CNS-LS change seen in the pivotal phase 3 registration trial (STAR trial) that led to the US approval of DM/Q for PBA [20]. Pearson correlation coefficients were calculated to assess the correlation of CNS-LS scores with other endpoints at baseline, Day 30, and Day 90/Final Visit, as well as correlation of change from baseline in CNS-LS with changes in other measures.

Tests of significance were 2-tailed and carried out at the α = .05 level of significance; all analyses were completed using either SAS v9.2 (SAS Institute Inc., Cary, NC) or Stata v12 (StataCorp, College Station, TX). All patients with data for the given comparison were included.

A power calculation was based on mean (SD) CNS-LS change observed from the pivotal phase 3 trial for the same dose of DM/Q (20/10 mg twice daily): −8.2 (6.1) points vs. placebo: −5.7 (5.3) points. Based on these results, it was determined that a sample size of 100 patients per disease group would provide 80 % power to detect a CNS-LS mean change of −7.45 points (increase of 1.75 points over assumed true placebo mean change of −5.7), or 90 % power to detect a CNS-LS mean change of −7.7 points (increase of 2.0 points over assumed true placebo mean change of −5.7). An interim analysis, conducted after the first 100 patients (regardless of cohort) completed the study, supported these assumptions of magnitude of the effect and indicated that a sample size of 100 per disease group would provide sufficient (≥80 %) power to meet the protocol-specified endpoints. In addition to the results for the entire study population described here, pre-specified analyses for the three distinct diagnosis groups (Dementia, Stroke, and TBI cohorts) were conducted separately and have either been reported in full (Dementia [24]) or have manuscripts in preparation (Stroke [26] and TBI [25]).

Patients

Of the 394 participants screened, 367 participants with PBA (134 with dementia, 113 with stroke, and 120 TBI) were enrolled and received at least 1 dose of DM/Q; 271 (73.8 %) completed the study through Day 90. Participant disposition is shown in Fig. 2; early discontinuation from the study occurred most commonly because of AEs (9.3 %) and withdrawal of consent (5.7 %). Sixty-nine patients (18.8 %) were excluded from the effectiveness analysis set because of one or more of the following: lack of post-baseline CNS-LS score (n = 41), failure to meet all study eligibility criteria (n = 23), and/or site noncompliance (n = 11). A total of 297 patients were included for the Day 30 efficacy analysis, while 261 had evaluable data at Day 90. The baseline demographic and clinical characteristics of the enrolled cohort (safety analysis set) are described in Table 1. The mean (SD) patient age was 59.4 (16.5) years; 17 % lived in a skilled nursing or assisted living facility, and 45 % had a caregiver. In addition, the mean (SD) CNS-LS score at baseline was 20.5 (4.4) and median PBA episode count for the 7 days prior to baseline was 12 (range: 0, 240). Study participants were receiving a median of 7 [0–27] concomitant medications at baseline; the majority (70.8 %) were using least one psychiatric medication, most commonly antidepressants (48.5 %; Table 1). Nearly three quarters (73.0 %) had concomitant cardiovascular disease at baseline, most commonly hypertension (57.8 %) or hyperlipidemia (44.4 %), and the majority (88.6 %) had another central nervous system disorder, most commonly depression (57.5 %) and anxiety disorder (42.2 %).

Characteristic	(N = 367)
Age, mean (SD), y	59.4 (16.5)
Age category, n (%)
≥ 65 years	152 (41.4)
≥ 75 years	75 (20.4)
Gender, n (%)
Male	165 (45.0)
Female	202 (55.0)
Race, n (%)
White/Caucasian	304 (82.8)
Black/African American	50 (13.6)
Asian	3 (0.8)
Othera	4 (1.1)
Unknown	6 (1.6)
Ethnicity, n (%)
Hispanic/Latino	71 (19.4)
Presence of a caregiver, n (%)	166 (45.2)
Place of Residence, n (%)
Home	303 (82.6)
Assisted living	35 (9.5)
Skilled nursing facility	29 (7.9)
Primary diagnosis, n (%)
Dementia	134 (36.5)
Stroke	113 (30.8)
TBI	120 (32.7)
Concomitant medications at baseline (total no. of medications)
Mean	7.7
Median (min, max)	7.0 (0, 27)
Psychopharmacologic medication use,b n (%)	260 (70.8)
Any antidepressant	178 (48.5)
SSRIs	103 (28.1)
Other	83 (22.6)
Non-selective (tricyclic)	14 (3.8)
Sedative/hypnotics/anxiolytics	124 (33.8)
Any benzodiazepinec	109 (29.7)
Antipsychoticsd	66 (18.0)
Anticonvulsants	92 (25.1)
CNS-LS scoree,f
Mean (SD)	20.5 (4.4)
Median (min, max)	20 (13, 34)
PBA episode count (over 7 days prior to baseline)f
Mean (SD)	21.4 (25.0)
Median (min, max)	12 (0, 240)
MMSE score, mean (SD)f	23.9 (5.9)
QOL-VAS, mean (SD)	5.9 (2.6)
PHQ-9,g mean (SD)	13.5 (5.9)

CNS-LS Center for neurologic study-lability scale, MMSE mini-mental state examination, PBA pseudobulbar affect, PHQ-9 patient health questionnaire-9, QOL-VAS quality-of-life visual analog scale, SD standard deviation, SSRI selective serotonin reuptake inhibitor

^aOther includes American Indian, Alaskan Native, Native Hawaiian, or other Pacific Islander

^bPsychopharmacologic medications included anticonvulsants, antipsychotics, antidepressants, sedatives/hypnotics or anxiolytics, and benzodiazepine

^cIncludes benzodiazepines as sedatives/hypnotics plus clonazepam as an anticonvulsant

^dTypical antipsychotic use in 1.9 %, and atypical antipsychotic use in 16.6 %; categories were not mutually exclusive

^eThe CNS-LS scale ranges from 7 to 35, with higher scores indicating increased frequency and severity of PBA episodes

^fEffectiveness analysis set (n = 298)

^gPHQ-9 scores range from 0 to 27, with higher scores indicating increased severity of depression

Primary efficacy endpoint

Figure 3 depicts the mean (SD) CNS-LS scores at each assessment. Compared with the baseline CNS-LS score, mean (SD) CNS-LS scores at Day 30 (n = 297) and Day 90/Final visit (n = 261) were 15.0 (5.0) and 12.8 (5.0) and represent a statistically significant improvement from baseline (mean [SD] change at Day 30 of −5.4 [5.5]; 95 % CI: −6.1, −4.8; P < .001 and Day 90/Final visit of −7.7 [6.1; 95 % CI: −8.4, −6.9]; P < .001). Improvement at Day 90/Final Visit was consistent with that reported for the same dose of DM/Q in the 12-week phase 3 pivotal trial that enrolled patients with PBA secondary to ALS or MS (mean [95 % CI] change from baseline, −8.2 [−9.4, −7.0]) and represents an improvement over the placebo group from that study (−5.7 [−6.8, −4.7]).

Secondary analyses

PBA episode counts over the 7 days prior to study visit decreased from a median of 12 at baseline to 4 at Day 30 and 2 at Day 90. PBA episode frequency by study visit is shown in Fig. 4, using the mixed-effects Poisson model estimates. PBA episodes were reduced overall by an estimated 57.5 % at Day 30 and 72.3 % at Day 90 compared with baseline (P < .001 for both). Remission of PBA episodes (defined as no reported episodes in the week before assessment) was reported by 20.3 % of the sample at Day 30 and 35.4 % at Day 90.

Global ratings by clinicians and participants/caregivers indicated the majority believed that during study treatment there had been substantial overall improvement with respect to PBA (Fig. 5). At Day 90/Final Visit, 76.6 % of patients were rated by their clinician as “very much improved” or “much improved” on CGI-C and 72.4 % of participants (or caregiver proxies) rated themselves as “very much improved” or “much improved” with respect to PBA compared with baseline. In contrast, 2 patients reported being “minimally worse” or “much worse” on PGIC (0.8 %) and 3 were reported to be “minimally” or “much worse” on CGIC (1.1 %). No patient was reported as “very much worse” on either measure. When asked about satisfaction with the treatment, 47.5 % of participants (or caregivers) indicated they were very satisfied, 28.0 % somewhat satisfied, 11.5 % neutral, 5.4 % somewhat dissatisfied, and 7.7 % very dissatisfied.

The mean QOL-VAS rating of PBA episode impact on quality of life, the PHQ-9 depressive symptom measure, and the MMSE cognitive measure all improved during the study. QOL-VAS scores improved from a mean (SD) of 5.9 (2.6) at baseline to 2.7 (2.6) at Day 90/Final Visit (mean [SD] change from baseline, −3.1 [3.2]; P < .001). PHQ-9 scores improved from mean (SD) 13.5 (5.9) at baseline (moderate depression) to 7.5 (5.5) at Day 90/Final Visit (mild depression), representing a mean [SD] change from baseline of −5.6 [6.2] points (P < .001)); mean (SD) MMSE scores increased from 23.9 (5.9) at baseline to 24.3 (6.0) at Day 90/Final Visit (mean change from baseline, 0.60 (3.0); P < .01). For the correlation analysis, CNS-LS score reduction from baseline to Day 90 was significantly correlated with improvements in all secondary outcomes (weekly PBA episode rate, QOL-VAS, PGI-C, CGI-C, PHQ-9, and treatment satisfaction; all P < .001) except change in MMSE (Table 2).

Variable	Pearson Correlation	P value
Weekly PBA episode rate change	0.21	<.001
QOL-VAS change	0.40	<.001
PGI-C	0.47	<.001
CGI-C	0.48	<.001
PHQ-9 change	0.32	<.001
Patient satisfaction score	−0.22	<.001
MMSE change	0.01	.877

CGI-C clinical global impression of changes, CNS-LS Center for neurologic study—lability scale, MMSE mini-mental state exam, PBA pseudobulbar affect, PGI-C patient/caregiver global impression of change, PHQ-9 9-item patient health questionnaire, QOL-VAS quality-of-life visual analog scale

Safety

Clinical implications

Although the present study was open label, results appear to be clinically meaningful, consistent across study cohorts, and consistent with those seen in well-controlled, phase 3 trials of DM/Q. The clinical relevance of CNS-LS score reductions from baseline to Day 90/Final Visit is reflected in corresponding reductions in PBA episodes and improvements on other secondary measures, including clinician and patient global impressions with respect to PBA and satisfaction with treatment. CNS-LS reduction was moderately but significantly correlated with improvements in all secondary outcomes (weekly PBA episode rate, QOL-VAS, PGI-C, CGI-C, PHQ-9, and treatment satisfaction; all P < .001) with the exception of the MMSE. While the measured outcomes are directionally related, the lack of strong correlation (all Pearson’s coefficients were < .5) suggests that changes in CNS-LS and other outcomes are not solely reflective of changes in PBA episode number and may reflect other aspects of PBA episodes, such as episode intensity, subjective perception, functioning related to these episodes, or other factors not readily apparent from these data.

A pre-specified analysis for this trial assessed the 95 % CI for mean change from baseline in CNS-LS scores in order to enable descriptive comparisons with the earlier placebo-controlled Phase 3 trial in patients with PBA secondary to ALS or MS [20]. Although drug performance cannot be directly compared across different clinical trial populations and settings, PBA symptom improvement with DM/Q in the current open-label trial is consistent with what was observed with DM/Q in the pivotal Phase 3 trial published by Pioro et al. [20], with a magnitude of improvement larger than observed for the group randomized to placebo in that study. These findings are also consistent with two other earlier Phase 3 randomized, controlled trials examining the effect of DM/Q for PBA in patients with MS or ALS (range, 7.4–7.7; Fig. 6). CNS-LS reductions were similar across all 3 disease cohorts included in this PRISM II trial (dementia, stroke, and TBI; range, 7.2–8.5; Fig. 6) [24–26]. Taken together, data from PRISM II expand on prior observations of the utility of DM/Q for PBA in MS and ALS and provide consistent and clinically meaningful evidence of DM/Q effectiveness for PBA in patients with 3 additional common, and disparate neurological etiologies, namely dementia, TBI and stroke.

DM/Q was associated with low incidence of overall AEs (36 %), treatment-related AEs (15 %), and AEs leading to withdrawal (9.8 %). Commonly encountered AEs were consistent with the safety data from controlled studies with DM/Q and information included in the approved product label; no new safety signals emerged in this clinical trial, and none of the reported serious AEs (6.3 %) were deemed by investigators to be treatment related.

Limitations of this study are primarily related to its open-label design without an active or placebo comparator group. In addition, use of self-reported measures may be susceptible to subjective bias based on observations by patient, caregiver, or investigator. Lastly, the criteria for the effectiveness analysis population that required patients to meet all exclusion and inclusion criteria may have introduced bias.

The CNS-LS is validated as a predictive screening assessment for PBA episode frequency and severity based on studies in patients with ALS and MS; [21, 22] however, this scale has not been specifically validated for use in screening in patient populations with dementia, stroke, and TBI. Even so, changes in CNS-LS in this study were consistent with improvements in several other measures of clinical relevance for PBA (e.g. frequency of PBA episodes), and prior epidemiological studies have shown that CNS-LS scores increase with perceived PBA episode burden [5].