Improving the specificity and precision of PANSS factors · N=1,710 subjects | 5 studies | acute schizophrenia RCTs {lurasidone, placebo, olanzapine, quietiapine} 0123456 HOS DIS

Improving the specificity and precision of PANSS factors:

One approach to facilitate development of novel treatments in

schizophrenia

Seth C. Hopkins, PhDExecutive Director Translational MedicineSunovion Pharmaceuticals

1

• Industry-sponsored RCTs generate large clinical databases – which can be better examined to advance the field and

improve public health

Use of Clinical Trial Databases to Improve RCT Efficiency

tens of thousands of unique research subjects • acute studies

• long-term safety studies

• different compounds• range of doses

volume of clinical observations

2

• The potential for clinically-meaningful differentiation is an incentive for innovation, especially for a novel MoA

How to apply clinical trial databases to facilitate drug development, including drugs with novel mechanisms of action?

more accurate/specific descriptions of treatment effect across sub-domains of schizophrenia symptoms?

better understanding of effects in stable/specific patient types?

subsub-domains

subpopulations

sub-populations

3

Historical view of the accumulated database in schizophrenia drug development

Drug launches for schizophrenia

year of launchThompson Reuters | Integrity Database

19501960197019801990200020102020

Chlorpromazine Promazine

ThioridazineHaloperidolClopenthixol

ThiothixeneSulpirideSpiperonePimozide

ClozapineFluphenazine

Carpipramine Bromperidol

ZotepineAmisulpride

LevosulpirideEmonapride

RisperidoneOlanzapineQuetiapine

ZiprasidonePerospironeAripiprazole

SertindolePaliperidone

BlonanserinAsenapineIloperidoneLurasidoneBrexpiprazole

CariprazineCum

ulative drug launches (world-w

ide) for schizophrenia

year of launch

Dopamine D2Serotonin 5-HT2A

4

The cumulative number of research subjects is large

ClinTrials.orgindustry-sponsored | schizophrenia | phase 2 or phase 3 | randomized, placebo-controlled | primary endpoint efficacy

Dopamine D2 based mechanism

Subj

ects

enr

olle

d(c

umul

ativ

e) cumulative number of schizophrenia subjects enrolled in RCTs for industry-sponsored drug development programs (ClinTrials.org)

5

ClinTrials.org | industry-sponsored | schizophrenia | phase 2 or phase 3 |randomized, placebo-controlled

Total investment in clinical development for new schizophrenia compounds is large

all others

drug-development RCTs in schizophrenia (start dates)

Dopamine D2 based mechanism

Subj

ects

enr

olle

d(c

umul

ativ

e)

6

ClinTrials.org | industry-sponsored | schizophrenia | phase 2 or phase 3 |randomized, placebo-controlled

Total investment in compounds with novel mechanisms is also large

all others all others

Dopamine D2 based mechanism non-D2 mechanisms

Subj

ects

enr

olle

d(c

umul

ativ

e)

7

ClinTrials.org | industry-sponsored | schizophrenia | phase 2 or phase 3 |randomized, placebo-controlled

The success of compounds with non-D2 MoA’s has been poor relative to D2 compounds

all others all others

launches

Dopamine D2 based mechanism non-D2 mechanisms

Subj

ects

enr

olle

d(c

umul

ativ

e)

8

ClinTrials.org | industry-sponsored | schizophrenia | phase 2 or phase 3 |randomized, placebo-controlled

Development of compounds with non-D2 MoA’s has focused on unmet needs of Cognition and Negative Symptoms

Dopamine D2 based mechanism non-D2 mechanisms

all others all others

launches

PANSS totalCognition

Negative Symptoms

Relapse

RelapsePrimary Endpoints as portion of total enrolled subjects

PANSS total

Subj

ects

enr

olle

d(c

umul

ativ

e)

Drug development RCTs for new treatments in schizophrenia have incorporated new measurements of specific symptom domains

9

How to apply clinical trial databases to facilitate new treatments, including drugs with novel MoA’s?

Dopamine D2 based mechanism non-D2 mechanismsPANSS total

Cognition

Negative Symptoms

Relapse

Relapse

PANSS total

IMPROVEMEASUREMENTS

ACADEMIAPHARMA

NIMH/NIH 

RDOC MATRICS

SCALES

Use the existing PANSS scale, to measure, with enhanced specificity, treatment effects across key symptom domains of schizophrenia

Drug development RCTs for new treatments in schizophrenia have incorporated new measurements of specified symptom domains

10

uncorrelatedscores UPSM factor scores

N=1,710 Sunovion dataset

acute schizophrenia, 6-weeks{lurasidone, placebo, olanzapine, quietiapine}

factor analysis of change-from item scoresretain score matrix coefficients

• (ORTHOGONAL) represent symptom change specific to each dimension

• (FACE VALID) corresponds to recognizable/established dimensions

• (APPLICABLE) UPSM can transform external PANSS data sets

UPSM

PANSS factors are correlatedCan we transform them to be more specific?

Uncorrelated PANSS Score Matrix (UPSM)

dom

ain

A

domain B

11

Pseudospecificity: A major confound in the assessment of efficacy in schizophrenia

SUNOVION DATA (N=1,710 patients from 5 placebo-controlled schizophrenia trials)Marder PANSS factor pos dis neg hos/exc anx/dep totPositive Symptoms 1Disorganized Thought 0.74 1Negative Symptoms 0.57 0.62 1Hostility/Excitement 0.64 0.59 0.43 1Anxiety/Depression 0.52 0.45 0.40 0.46 1PANSS Total 0.90 0.86 0.77 0.77 0.66 1

OTSUKA DATA (N=1,368 patients from 5 placebo-controlled schizophrenia trials)Marder PANSS factor pos dis neg hos/exc anx/dep totPositive Symptoms 1Disorganized Thought 0.73 1Negative Symptoms 0.59 0.67 1Hostility/Excitement 0.64 0.58 0.42 1Anxiety/Depression 0.58 0.51 0.49 0.48 1PANSS Total 0.89 0.87 0.81 0.75 0.71 1

TAKEDA DATA (N = 240 patients from 1 placebo-controlled schizophrenia trials 20 mg TK-063)Marder PANSS factor pos dis neg hos/exc anx/dep totPositive Symptoms 1Disorganized Thought 0.65 1Negative Symptoms 0.46 0.50 1Hostility/Excitement 0.46 0.41 0.23 1Anxiety/Depression 0.55 0.50 0.45 0.47 1PANSS Total 0.85 0.80 0.71 0.66 0.76 1

Correlations Among Marder PANSS Factor Scores (Week 6 Change from Baseline)

12

TAKEDA DATA (N = 240 patients from 1 placebo-controlled schizophrenia trials 20 mg TK-063)Transformed PANSS factors POS DIS NAA NDE HOS ANX DEPPOSITIVE 1DISORGANIZED 0.07 1NEG APATHY/AVOLITION 0.18 -0.07 1NEG DEFICIT OF EXPRESSION -0.02 -0.07 0.15 1HOSTILITY -0.01 -0.13 -0.15 -0.02 1ANXIETY 0.28 -0.01 0.24 -0.18 0.44 1DEPRESSION 0.29 -0.16 0.29 -0.06 0.09 0.36 1PANSS TOTAL SCORE 0.59 0.14 0.50 -0.23 0.38 0.61 0.65

UPSM transform reduces between‐factor correlations in endpoint PANSS factor change scores

SUNOVION DATA (N=1,710 patients from 5 placebo-controlled schizophrenia trials)Transformed PANSS factors POS DIS NAA NDE HOS ANX DEPPOSITIVE 1DISORGANIZED 0.20 1NEG APATHY/AVOLITION 0.10 0.08 1NEG DEFICIT OF EXPRESSION 0.04 0.12 0.22 1HOSTILITY 0.21 0.12 0.07 -0.02 1ANXIETY 0.09 0.04 -0.01 -0.08 0.13 1DEPRESSION 0.10 0.00 0.12 0.13 0.04 0.27 1PANSS TOTAL SCORE 0.60 0.45 0.45 0.36 0.53 0.45 0.46

OTSUKA DATA (N=1,368 patients from 5 placebo-controlled schizophrenia trials)Transformed PANSS factors POS DIS NAA NDE HOS ANX DEPPOSITIVE 1DISORGANIZED 0.17 1NEG APATHY/AVOLITION 0.12 0.12 1NEG DEFICIT OF EXPRESSION 0.03 0.22 0.37 1HOSTILITY 0.08 0.18 0.12 -0.04 1ANXIETY 0.06 0.12 0.05 -0.01 0.16 1DEPRESSION 0.21 0.08 0.26 0.28 0.07 0.27 1PANSS TOTAL SCORE 0.47 0.52 0.57 0.48 0.49 0.46 0.57

13

-0.9 -0.6 -0.3 0.0 -0.9 -0.6 -0.3 0.0

drug vs. placebo effect size ± 95%CI

worseningimprovement 

POSITIVE

HOSTILITY

NEGATIVE

DISORGANIZED

DEPRESSION/ANXIETY

PANSS TOTAL TOTAL FACTOR SCORE

-0.38

-0.44

-0.32

-0.36

-0.33

-0.46

-0.35

-0.19

-0.22

-0.04

-0.27

-0.18

-0.14

-0.45

DEFICIT OF EXPRESSION

ANXIETY

DEPRESSION

APATHY/AVOLITION 

worseningimprovement 

MARDER PANSS FACTORS UPSM TRANSFORMED PANSS FACTORSoverlapping drug effects

Towards specificity in antipsychotic drug treatment effectsN=5 acute schizophrenia trials PANSS change from baseline, at Week 6 endpoint, lurasidone active doses vs. placebo

Hopkins et al. 2017 Schizophrenia Bulletin

subsub‐domains

more specificity

Attributions of specific treatment effects confounded by correlations among improvements in Marder PANSS Factor Scores

greater heterogeneity in effect size estimates across the sub-domains

14

Discussion Point: Treatment effects are separable among the dimensions of schizophrenia

If symptom dimensions can be separately identified and assessed even in the acute schizophrenia clinical trial setting...

...then therapeutics with novel efficacy profiles across the sub-dimensions of schizophrenia psychopathology can be evaluated

UPSM-transformation identified 2 PANSS-negative symptom sub-factors showing differential treatment effects, with larger effect sizes observed for the apathy/avolition sub-factor compared to the deficit of expression sub-factor

DRUG A DRUG B

subsub‐domains

15

Understanding of specificity within PANSS to advance the fieldand facilitate development of new treatments

more specific breakdown of total symptoms improvements according to the domains of most-prominent effects

Forest Forest Plots with specificity

• specificity among symptom sub-domains

subpopulations

sub-populations

Does an understanding of specificity of symptom change over time help to identify more-meaningful patient types?

Address issues of:• differentiation • dose-response• benefit/risk

Use during:• drug development

decisions• meta-analyses

Objective of this analysis direction:• examine failed/negative trials• test for differential treatment responses among types• develop “a priori” categories for future analysis plans of

novel MoA’s

16

are patient-types stable?do they persist post-baseline?

track response of patient-types post-baseline

PATIENT TYPES baselinePANSSbaselinePANSS

cluster(K-MEANS)

transform(UPSM)

Identified patient types at baseline who are prominent along specific UPSM factor scores

POS

HOS

DIS

ANX

DEP

NAA

NDE

subpopulations

sub-populations

PROMINENTLYPOSITIVE

POS

HOSTILE

DISORGANIZED

NEG

BASELINE

mean UPSMfactor scores

ANX DEP

train classifier(SVM)

post-baselinePANSS

post-baselinePANSS

post-baseline

PATIENT TYPES

apply classifier(SVM)

transform(UPSM)

17

Patient-types persist over time, even in the context of overall symptom (PANSS total) improvements

PROMINENTLYPOSITIVE

HOS

DIS

ANX

POS

HOS

DIS

ANX

DEP

NAA

NDE

NEG

DEP

PATIENT-TYPE PROMINENTLY POSITIVE

BASELINE SEQUENTIAL WEEKLY VISITS ENDPOINTN=1,710 subjects | 5 studies | acute schizophrenia RCTs

{lurasidone, placebo, olanzapine, quietiapine}

0 1 2 3 4 5 6

HOS

DIS

ANX

NEG

DEP

POS

DISCONTINUATION

Patient types persist over time and over treatment

POS

mean UPSMfactor scores

N=400

number of subjects

N=200

duration

Toward better definitions of patient-types within PANSS to better understand and characterize treatment response

Objectives of this analysis• examine failed/negative trials• test for differential treatment responses among types• develop “a priori” categories for future analysis plans of novel MoA’s

18

• An understanding of the specificity of symptom change in PANSS ... can facilitate the evaluation of therapeutics with novel efficacy profiles across the sub-dimensions of schizophrenia psychopathology

Future directions in the use of clinical trial databases for improving RCT efficiency

subsub-domains

subpopulations

sub-populations

Towards collaborative analyses and shared research questions for our existing clinical trial databases

19