[1] 24w The Clinician's Interview-Based Impression of Change (Plus caregiver input) and goal attainment in two dementia drug trials: Clinical meaningfulness and the initial treatment response

[1] 144w INTRODUCTION:The Clinician's Interview-Based Impression of Change Plus caregiver input (CIBIC-Plus) has been widely used in dementia drug trials to evaluate cognition, behavior, and function. New trials of symptomatic drugs forecast renewed interest in this measure. METHODS:To test its clinical meaningfulness, we examined how CIBIC-Plus performed in two cholinesterase inhibitor trials compared to goal attainment scaling Scale (GAS) scores, a patient-reported outcome measure. RESULTS:Net goal attainment was seen for all but one GAS domains in subjects who improved on the CIBIC-Plus. Subjects who improved initially on CIBIC-Plus scores were likely to remain improved across all other outcomes for each trial's duration, except for Disability Assessment for Dementia scores. DISCUSSION:The initial response to treatment, as assessed by CIBIC-Plus, remained stable for most outcome measures. Even small CIBIC-Plus improvement changes are associated with clinically meaningful change as assessed by GAS. Other tests detect decline better than improvement.

[1] 166w The Clinician's Interview-Based Impression of Change Plus caregiver input (CIBIC-Plus) scale is a widely used global assessment in trials of symptomatic drugs for people living with late-life dementia. [1][2][3][4] It baseline measures disease severity, through the Clinician's Interview-Based Impression of Severity (CIBIS), and the degree of change with a 7-point judgment-based rating scale. 5 It evaluates cognition, behavior, and function, to yield written and numerical summaries from semistructured interviews. 5 In its heyday, the CIBIC-Plus was a common co-primary outcome measure in dementia drug trials. [1][2][3][4]6 With a move to standardized measures, seen as more objective, it is now used much less. 7 The CIBIC-Plus was tarnished as being merely about symptoms, in an era in which it was imagined that disease-modifying drugs could prevent their emergence. Indeed, that belief saw Alzheimer's disease (AD) itself reframed to be diagnosed independently of symptoms, signaled instead by given levels of biomarkers. 8,9 Even so, there is reason to doubt that diseasemodifying drugs can make consideration of symptoms obsolete.

[2] 98w More recently, there are positive reasons to consider allowing patients and caregivers the opportunity to help assess the efficacy of candidate anti-dementia treatments. Most important is that symptomatic treatments are again being tested, and not all are being found wanting. [10][11][12] Further, as elaborated below, carefully diagnosed clinical dementia is not strongly related to dementia neuropathology at autopsy-arguably the ultimate in biomarker-defined disease. 14,15 Finally, the question of clinical meaningfulness is enjoying renewed attention in dementia research. [16][17][18][19][20][21] For these reasons, we must still consider whether anti-dementia drug treatments lead to changes that are clinically meaningful to patients.

[3] 161w Despite its traditional use, whether improvement on the CIBIC-Plus is clinically meaningful is unclear. It notably has only variably informed contemporary consideration of whether to offer treatment to people with AD. 22,23 To assay clinical meaningfulness, we can consider that any treatment that meets the expectations of patients, their caregivers, and their physicians is likely to be clinically important. [24][25][26] Goal attainment scaling (GAS) offers a way to evaluate such expectations. 27 With GAS, patients, caregivers, and/or clinicians identify and track individualized treatment goals throughout a trial. Goals are monitored and quantified with a formula that categorizes the degree of goal attainment. Achieving personalized goals through GAS is clinically meaningful to both patients and caregivers in AD drug trials. 28,29 Further, the initial treatment response (good or bad) forecasts outcome at the trial's end. 18 To evaluate the clinical meaningfulness of the CIBIC-Plus, we first compared CIBIC-Plus scores with clinically meaningful changes in clinician-rated and patient/carer-rated GAS in people with mild-

[1] 264w Data from both trials were combined and plotted as a function of the CIBIC-Plus score (Figure 1). Clinician-rated and patient/carer-rated GAS scores exhibited a strong relationship with CIBIC-Plus scores at the endpoint (respectively, rho = -0.70 P < .001, Figure 1A; rho = -0.56 P < .001, Figure 1B). The relationship with change in the MMSE was also monotonic but much weaker (rho = -0.25, P = .011, Figure 1C). In contrast to the GAS and MMSE scores, the ADAS-Cog, normalized daily function, and DAD scores (Figure 1D-F Subjects who improved on the CIBIC-Plus (scores <4) also improved significantly in GAS and ADAS-Cog scores. Average clinician-rated GAS scores were 60.8 ± 9.0, patient/carer-rated GAS scores were 60.7 ± 10.6, significantly different from the baseline anchor GAS score of 50 (P < .001 for both). The mean difference in ADAS-Cog scores from baseline to endpoint was -1.84 ± 4.9 (P = .004), reflecting net improvement in subjects with CIBIC-Plus improvement. In contrast, there was no relationship between improvement in the CIBIC-Plus and either normalized daily function scores (mean change = 1.38 ± 7.6, P = .275) or DAD scores (mean change = -2.36 ± 10.1, P = .253). Interestingly, subjects who were classified as "no change" in the CIBIC-Plus (scores = 4) saw net goal attainment in patient/carer-rated GAS (54.8 ± 8.5, P = .007) and significant decline in normalized daily function (-3.08 ± 7.0, P = .042). Subjects who were classified as "worse" on the CIBIC-Plus (scores >4) showed net worsening across all measures except patient/carer-rated GAS. These data are summarized in Table 3.

[2] 207w To determine whether the response to treatment varied by goal domains, the mean GAS difference score from 50 was plotted (± 95% CI) by goal domain for subjects whose CIBIC-Plus scores improved, showed no change, or worsened at the endpoint(Figure 2). When clinician-rated GAS was investigated, goals were attained across all goal domains in subjects who improved on CIBIC-Plus (behavior: mean = 62.0 ± 8.8, P < .001; cognition mean = 56.3 ± 9.6, P < .001; daily function mean = 58.1 ± 10.0, P < .001; executive function mean = 58.4 ± 9.8, P < .001; physical manifestations mean = 65.1 ± 6.9, P < .001; Figure 2A). Patient/carer-rated GAS improved significantly in all goal domains except physical manifestations (52.3 ± 12.9, P = .496; Figure 2B) when CIBIC-Plus improvement was present. When CIBIC-Plus scores declined, significantly worse goal attainment was seen in clinician-rated cognition (43.7 ± 9.1, P < .001; Figure 2A) and daily function goals (44.5 ± 10.9, P = .003; Figure 2A). By contrast, patient/carer-rated GAS did not change in any goal domain when CIBIC-Plus decline occurred. Importantly, subjects saw significant goal attainment in patient/carer-rated executive function goals (59.2 ± 10.4, P < .001; Figure 2B), even when CIBIC-Plus scores did not change.

[3] 177w We next compared initial severity as measured with the CIBIC-Plus at 2 to 3 months to the endpoint treatment responses (8-9 months) as determined by CIBIC-Plus, GAS, ADAS-Cog, normalized daily function, or DAD (Figure 3). Of subjects who initially improved on CIBIC-Plus, 63% remained improved at the endpoint (OR = 5.68, 95% CI 2.8-11.5, Outcome measure N Mean (SD) P ADAS-Cog Improved 63 -1.84 (4.9) .004 No change 29 0.21 (4.0) .781 Worse 62 3.63 (7.8) .001 Clinician GAS Improved 61 60.8 (9.0) <.001 No change 27 51.3 (6.1) .277 Worse 55 44.5 (10.0) <.001 Patient GAS Improved 61 60.7 (10.6) <.001 No change 27 54.8 (8.5) .007 Worse 55 47.8 (9.9) .099 Normalized daily function Improved 37 1.38 (7.6) .275 No change 24 -3.08 (7.0) .042 Worse 42 -8.57 (11.5) <.001 DAD Improved 25 -2.36 (10.1) .253 No change 5 -3.6 (12.1) .542 Worse 19 -7.74 (11.8) .01 Abbreviations: ADAS-Cog, Alzheimer's Disease Assessment Scale-Cognitive subscale; CIBIC-Plus, Clinician's Interview-Based Impression of Change Plus caregiver input; DAD, Disability Assessment for Dementia; GAS, goal attainment scaling; SD, standard deviation.

[4] 143w were also more likely to show improvement on ADAS-Cog scores (3.0, 1.6-5.8, P = .001), clinician-rated GAS (4.8, 2.4-9.7, P < .001), patient/carer-rated GAS (2.5, 1.3-4.9, P = .007), and normalized daily function scores (7.7, 3.0-19.6, P < .001). On the other hand, those who initially improved on CIBIC-Plus did not show improvement in their DAD scores at endpoint (0.9, 0.3-2.9, P = .858). Mean (± SD) endpoint scores for subjects who initially improved on CIBIC-Plus were then compared to scores from those who showed no initial improvement or declined. Mean values were 3.4 ± 1.2 versus 4.3 ± 1.0 for CIBIC-Plus (P < .001; Figure 3A), -0.7 ± 7.2 versus 2.0 ± 5.9 for ADAS-Cog (P = .016; Figure 3B), 57.7 ± 11.0 versus 48.5 ± 10.2 for clinician-rated GAS (P < .001; Figure 3C), 58.3 ± 12.5 versus 51.4 ± 9.5

[5] 26w for patient/carer-rated GAS (P < .001: Figure 3D), and 0.0 ± 9.2 versus -6.3 ± 10.1 for normalized daily function scores (P = .001; Figure 3E).

[6] 22w By contrast, DAD scores showed no difference between the groups (-6.1 ± 12.9 vs -4.3 ± 10.8, P = .597; Figure 3F).

[1] 148w We assessed the clinical meaningfulness of CIBIC-Plus scores in dementia drug trials in people with mild-moderate AD, by understanding both the extent to which CIBIC+ changes reflected patient-specific goal attainment, and how they related to the standardized measures used in two clinical trials. Changes in GAS scores rated by clinicians or by patient/carers paralleled changes in CIBIC-Plus scores at the endpoint. In those rated as improved on the CIBIC-Plus, goal attainment was achieved in all GAS goal domains, save patientrated goals regarding disease-related physical manifestations. For other outcomes, CIBIC-Plus scores were best aligned when measuring decline. In addition, subjects initially rated as improved on the CIBIC-Plus had improved across all outcomes, except DAD scores, by trial endpoint. Even so, CIBIC-Plus scores were aligned with both patient and clinician daily function goals, suggesting that the CIBIC-Plus was better able to detect change in daily function compared to the DAD.

[2] 90w Together, these observations indicate that global change in dementia measured with the CIBIC-Plus scale corresponds closely with changes in GAS scores assessed by both physicians and patient/carers. As GAS is a patient-reported outcome measure that involves individualized goal setting, our results suggest that, especially when undertaken in cases in which GAS is being used, changes measured by CIBIC-Plus scores in symptomatic trials for dementia are likely to be inherently clinically meaningful if they demonstrate significant differences between the intervention and the control groups, with assessors blind to treatment group membership.

[3] 360w We compared GAS scores with global change as assessed by the CIBIC-Plus to evaluate its clinical meaningfulness. Clear patterns emerged between CIBIC-Plus scores and goal attainment. The mean clinician-rated GAS score for each point of the CIBIC-Plus was distinctly represented. For example, patients whose CIBIC-Plus scores showed minimal worsening also showed minimal net decline in goal attainment. This linear trend was skewed upward in patient-rated GAS scores. Patient-rated GAS goals often showed minimal improvement when CIBIC-Plus scores showed no change. Although this trend resulted in a larger effect size compared to clinician-rated GAS, 29 clinician-rated GAS scores could be a seen when using the ADAS-Cog; 18 however, we saw greater discrimination between subjects when using the CIBIC-Plus. This phenomenon, of the intensity of the initial treatment response, corresponds to the Matthew Effect described as the rich getting richer; it has been seen with a variety of complex systems. 32 Our data should be interpreted with caution. The two studies in which data were aggregated were of different design. Here, control arm data from VISTA were excluded from analysis and follow-up data were combined by difference from baseline. For example, the last follow-up included data from the 8-month visit in VISTA and 9-month visit in ACADIE. The two studies used different outcomes to measure function. In VISTA, the 100-point DAD evaluated the instrumental and non-instrumental activities of daily living. In ACADIE, this was achieved using the FAQ, IADL, and PSMS. Although we normalized measures of function in ACADIE to a similar 100-point scale, the distribution of scores in these two measures was markedly different and therefore could not be aggregated. Here, however, we are concerned less with efficacy and more with the relationships between outcome measures. In addition, we have proposed that attaining individualized goals is evidence of clinical meaningfulness. That is likely, but as outlined elsewhere, 33 it is not the full story. Further, although GAS is useful, it is not a perfect measure. Further work needs to be done on analysis, 34 and on the approach to validation-especially the need to test in double-blinded placebo-controlled trials, 35 and to ensure that goals be set before randomization and unalterable thereafter.

[4] 155w This work fits the broad framework of understanding clinical meaningfulness, for which criteria have been proposed. 33 The CIBIC-Plus is a refined scale to assess the Clinical Global Impression of Change 34 that has been used in many valid studies for anti-dementia drugs. [1][2][3][4] In VISTA, 29 the CIBIC-Plus detected a larger effect size than the ADAS-Cog (standardized response mean = -0.40 vs -0.36) when compared to placebo. Here, with the presence of a dose-response effect, the CIBIC-Plus was able to distinguish clinically detectable differences. In addition, CIBIC-Plus scores were correlated with changes in other outcome measures at endpoint. By its nature, the CIBIC-Plus is a judgmentbased measure. The close alignment between CIBIC-Plus and GAS scores shown here builds evidence for the clinical significance of the observed effect sizes. For these reasons, even a minimal impression of global change as assessed by the CIBIC-Plus can be clinically meaningful for people with AD or their caregivers.

[5] 448w Generals are often accused of preparing for the last war, not one that is more likely to be fought. In the case of biomarker-defined dementia the dilemma instead appears to be an imagined war that might not occur. Whatever the hope that inhibiting biomarkers in people with normal cognition and no dementia symptoms might be for those people, the notion that this would serve as proof of concept for late-life dementia is not without its problems. Unless a new disease-modifying treatment is curative, not everyone with biomarker-defined dementing disease could be expected to remain asymptomatic. In that case, detecting the emergence of symptoms that constitute clinical dementia might even serve as a useful endpoint. This could especially be the case if treatments were to be effective not in eliminating cognitive impairment, but in attenuating it-in ways that might not simply recapitulate untreated disease, in a back-to-baseline fashion, but instead give rise to detectable novel symptom patterns. In short, there is no reason to expect that a disease-modifying treatment will result in no cognitive impairment, and especially in people who are worried enough about dementia to enroll in a trial while cognitively intact. Moreover, dementia in late life, even when arguably due in part to AD, typically occurs in the face of other neuropathological lesions, 37,38 and overall health, 39 that may be unaffected in an approach which focuses on a single protein abnormality. Moreover, people with phenotypically characteristic AD, who nevertheless lack pre-clinical evidence of amyloid pathology, still have symptoms. If their symptoms are to be evaluated, including in potentially paradigm-shifting treatments such as senolytic therapy, 40 we should not miss the opportunity to better understand disease expression and treatment effects from the standpoint of patients and care partners. In this, the CIBIC-Plus, implemented to capture clinical observations and assay their importance, allows us to capture the wisdom of the group. Further, clinical judgments can be made systematically, and while informed by the patient/caregiver experience, need not determined by it. Judgment that adjudicates the lived experience of people treated for dementia can help us understand what successful prevention and treatment might look like, even short of cure. 41 It is important that we learn from the cholinesterase inhibitor trials. This is once again controversial. A recent review 23 set aside most key studies used for registration of the cholinesterase inhibitors due to a high risk of bias-where specified, this reflected follow-up being too short. Here we see that people who have an early and favorable symptomatic response to cholinesterase inhibitors are likely to do better than those who did not have that response. This ability to understand the early treatment response should not be lost in new trials.

[1] 67w The Atlantic Canada Alzheimer's Disease Investigation of Expectations (ACADIE) study was a 12-month open-label trial of donepezil. 28 The Video Imaging Synthesis of Treating Alzheimer's disease (VISTA) galantamine trial was a 4-month double-blind controlled trial with a 4-month open-label extension, in which all subjects received galantamine. 29 Here, to compare responses to treatment over several months, the placebo group was excluded from analysis. For these exploratory analy-

[2] 254w ses, data were aggregated at four time points: baseline, initial response (2 months in VISTA and 3 months in ACADIE), 6 months, and endpoint (8 months in VISTA and 9 months in ACADIE; ACADIE's 12-month endpoint is excluded, to facilitate comparison). Subjects in both trials (64 from VISTA, 108 from ACADIE) had mild-moderate AD. Briefly, the open-label ACADIE study (conducted between February 1998 and November 1999) showed net goal attainment as rated by patients and caregivers for 9 months and by clinicians for 6 months. The CIBIC-Plus showed significant improvement to 3 months, and no RESEARCH IN CONTEXT 1. Systematic review: We used PubMed to search for relevant articles. Despite a modest increase in using individualized and other clinically meaningful outcomes for dementia, standardized measurements of global clinical impressions are much more common. Few studies have related the degree of change in the Clinician's Interview-Based Impression of Change Plus caregiver input (CIBIC-Plus) to the standardized measures of cognition and function typically used in dementia drug trials. 2. Interpretation: Changes in the CIBIC-Plus scale coincided with clinically meaningful changes as assessed both by patient/carer-rated goal-attainment scaling (GAS), a patient-reported outcome measure, and by clinicians for whom the CIBIC-Plus scores were masked. Further, the standardized outcome measures (with CIBIC-Plus raters masked to results) showed lesser degrees of change when patients were improving clinically compared to when they declined. 3. Future directions: This relatively small sample of patients naïve to cholinesterase inhibitors suggests that individualized measures may be able to detect meaningful treatment effects not otherwise captured.

[3] 158w change from baseline to month 9. 28 The double-blind VISTA trial (October 2001 to March 2005) showed a significant difference in favor of galantamine at months 2, 4, and 6 with the CIBIC-Plus, at months 4 and 6 with clinician-rated GAS, and only at month 6 (2 months after the end of the double-blind period) with the patient/caregiver-rated GAS. Even so, we can combine across studies on several grounds. First, each trial had similar training, and was led by the same team. Second, we used the same approach to setting and rating GAS, and in each case had this done independently by patients and carers, and by the treating physicians. Third, the classification of goal types was done working from extensive qualitative coding dictionaries in which reliability was assayed, 30,31 a process in which the initial five goal categories (cognition, function, behavior, leisure, and social interactions) were amended as behavior, cognition, daily function, executive function, and physical manifestations.

[4] 22w ACADIE and VISTA were both multicenter trials that used CIBIC-Plus, ADAS-Cog, and GAS at all time points. In VISTA, the 100-point Disabil-

[5] 230w ity Assessment for Dementia (DAD) was used to measure function. In ACADIE, we normalized scores from the Functional Activities Questionnaire (FAQ); Lawton-Brody Instrumental Activities of Daily Living (IADL); and Lawton-Brody Physical Self-Maintenance Scale (PSMS), TA B L E 1 Examples of goals set by clinicians, patients, and carers Goals set, N (%) Goal domain Clinician GAS Patient/carer GAS Example goal titles Behavior 85 (15) 59 (5) Anger/agitation, delusions, paranoia, repetitive behaviors, wandering Cognition 170 (31) 357 (32) Communication, disorientation to time/place, misplacing objects, short term memory, verbal repetition Daily function 187 (34) 442 (40) Decline in hobbies, household chores, meal preparation/cooking, personal care/hygiene, shopping Executive function 84 (15) 206 (18) Impaired judgment, independence, personality changes, issues at church/religion, social interaction Physical manifestations 25 (5) 51 (5) Exercise, incontinence, problems with eating or appetite, pain/physical complaints, sleep disturbances Abbreviation: GAS, goal attainment scaling. consisting of 91 items as: (1 - This personalization means a different set of goals for each individual. To compare between individuals, goals were categorized as above (Table 1). Baseline GAS was anchored at 50 in both trials. Follow-up GAS scores ≥50 indicate net goal attainment whereas GAS scores ≤50 indicate that goals were not met. In both ACADIE and VISTA, the individual who performed the CIBIC-Plus was blinded to all standard scores, in ACADIE to the patient/caregiver GAS scores, and in VISTA to the clinician-rated GAS. 28,29

[6] 155w Baseline subject characteristics were reported as the mean (standard deviation [SD]) for continuous variables and the number and proportion of subjects (%) for categorical variables. Characteristics were reported for the ACADIE and VISTA trials individually and for the total, combined analysis population. Change scores were expressed as the difference from baseline, and tested with paired t-tests. Correlations were evaluated with Spearman's rank-order correlation (rho). Differences between means were assessed with Student's t-test. Where appropriate, data were fit with either an exponential or linear regression. To assess the stability of the initial response to treatment, subjects were categorized as initially improved (CIBIC-Plus <4) or no change/worse (CIBIC-Plus ≥4) at 2 to 3 months. The initial response was considered to have persisted if the initially improved group outperformed the no change/worse group at the endpoint. This association was summarized using odds ratios (ORs) with 95% confidence intervals (CIs). Resulting P values < .05 were considered statistically significant.

[7] 11w R statistical software (version 3.3.3) was used to perform all analyses.

[8] 10w Written informed consent was provided by all subjects and/or their

[1] 47w These exploratory analyses were undertaken by DGI Clinical. Each author is associated with that company: JS and TD as full-time employees, SEH as a part-time contract employee, and KR as founder and chief science officer. The decision to undertake the work and publish was that of KR.

[2] 45w Of the 238 subjects in the two trials, 172 met the criterion of having received active treatment (donepezil or galantamine, respectively) and were included (ie, no placebo-arm data and no imputed scores). Most study subjects were older women with established cognitive and functional impairment (