Columbia Suicide Severity Rating Scale

Scoring and Data Analysis Guide

************************************************************************
Contributors: Mary E. Nilsson, Shailaja Suryawanshi, Cristiana Gassmann-Mayer, Sarah
Dubrava, Paul McSorley, and Kaihong Jiang.

Introduction

The Columbia Suicide Severity Rating Scale (C-SSRS) is an assessment tool that
evaluates suicidal ideation and behavior. This guide outlines the proposed safety
outcomes and statistical analysis strategy for the C-SSRS for an individual clinical trial.

standards, or to be consistent with the needs of a particular regulatory agency/division.

As such this document serves as a general guideline and is not a proposal for mandatory
analyses involving the C-SSRS. When analyzing the C-SSRS across studies, the same
safety outcomes can be used. However, different statistical methodology would apply to
combining data across studies, and is not discussed in this document. The briefing
document for the 13 December 2006 Food and Drug Administration (FDA) Advisory
Committee Meeting contains a discussion of meta-analytical methods to consider when
analyzing suicide-related outcomes across studies
(http://www.fda.gov/ohrms/dockets/ac/06/briefing/2006-4272b1-index.htm). The C-
SSRS may also be used to assess efficacy, but specific guidance related to efficacy
assessment will be a topic for a separate document.

As noted in the August 2012 Suicidal Ideation and Behavior:

adopted the 11
categories defined in the C-SSRS (five subtypes of suicidal ideation, five subtypes of
suicidal behavior, and self-injurious behavior without suicidal intent) as their standard.
Some data collected on the C-SSRS are not included in these proposed displays for safety
analysis (e.g., suicidal behavior lethality and suicidal ideation intensity), but are used for
individual clinical management, safety monitoring, or other research purposes.

Analysis Set:
The analysis set should
analysis set. Frequently, this means including patients who have been exposed to at least
one dose of study drug. When analyzing the C-SSRS, it is recommended to include
patients having at least 1 post-baseline C-SSRS measurement, regardless of whether they
had a baseline C-SSRS measurement. For some analyses of the C-SSRS (e.g., treatment-
emergent assessments and shift tables), a pre-treatment C-SSRS measurement is also
required.

Outcomes:
There is current debate on whether suicidal ideation and suicidal behavior should be
combined and analyzed as a single outcome. Researchers agree on the need for having
analyses that keep suicidal ideation and suicidal behavior separate, but disagree on

Version 2.0 (Finalized February 2013) 1

whether there is value to having any outcome that combines them. This document
currently includes the combined outcomes measure.

The following outcomes are C-SSRS categories and have binary responses (yes/no). The
categories have been re-ordered from the actual scale to facilitate the definitions of the
composite and comparative endpoints, and to enable clarity in the presentation of the
results.

Category 1 Wish to be Dead

Category 2 Non-specific Active Suicidal Thoughts
Category 3 Active Suicidal Ideation with Any Methods (Not Plan) without Intent to Act
Category 4 Active Suicidal Ideation with Some Intent to Act, without Specific Plan
Category 5 Active Suicidal Ideation with Specific Plan and Intent
Category 6 Preparatory Acts or Behavior
Category 7 Aborted Attempt
Category 8 Interrupted Attempt
Category 9 Actual Attempt (non-fatal)
Category 10 Completed Suicide

Self-injurious behavior without suicidal intent is also a C-SSRS outcome (athough not
suicide-related) and has a binary response (yes/no).

The following outcome is a numerical score derived from the C-SSRS categories. The
score is created at each assessment for each patient and is used for determining treatment
emergence.

Suicidal Ideation Score: The maximum suicidal ideation category (1-5 on the C-
SSRS) present at the assessment. Assign a score of 0 if no ideation is present.

Endpoints:
Composite endpoints based on the above categories are defined below.

Suicidal ideation
five suicidal ideation questions (Categories 1-5) on the C-SSRS.

Suicidal behavior
five suicidal behavior questions (Categories 6-10) on the C-SSRS.

Suicidal ideation or behavior o

any one of the ten suicidal ideation and behavior questions (Categories 1-10) on
the C-SSRS.

Comparative endpoints of interest are defined below.

is used for
outcomes that include events that first emerge.

Version 2.0 (Finalized February 2013) 2

 Treatment-emergent suicidal ideation compared to recent history: An increase in
the maximum suicidal ideation score during treatment from the maximum suicidal
ideation category during a specified pre-treatment period (C-SSRS scales taken
during the specified pre-
Baseline C-SSRS scale or Baseline/Screening C-SSRS scale).

Treatment-emergent serious suicidal ideation compared to recent history: An

increase in the maximum suicidal ideation score to 4 or 5 on the C-SSRS during
treatment from not having serious suicidal ideation (scores of 0-3) during a
specified pre-treatment period (C-SSRS scales taken during the specified pre-
treatment period; e -SSRS scale or
Baseline/Screening C-SSRS scale).

Emergence of serious suicidal ideation compared to recent history: An increase

in the maximum suicidal ideation score to 4 or 5 on the C-SSRS during treatment
from no suicidal ideation (scores of 0) during a specified pre-treatment period (C-
SSRS scales taken during the specified pre-
scores from the Baseline C-SSRS scale or Baseline/Screening C-SSRS scale).

Improvement in suicidal ideation at a time point of interest compared to baseline:

An improvement in this endpoint can be considered as a decrease in suicidal
ideation score at the time point of interest (e.g., the last measurement during
treatment) from the baseline measurement (e.g., the measurement taken just prior
to treatment. This analysis should only be performed for studies in which a
baseline C-SSRS can be defined (i.e., having improvement from the worse event
over a lifetime is not clinically meaningful).

Emergence of suicidal behavior compared to all prior history: The occurrence of

suicidal behavior (Categories 6-10) during treatment from not having suicidal
behavior (Categories 6-10)
m the Baseline C-SSRS scale, Screening C-SSRS scale, or
Baseline/Screening C-
Visit C-SSRS scales taken prior to treatment).

Additional comparative endpoints for consideration are defined below. These are not
necessarily recommended for all treatment programs, but if used, should follow the
nomenclature.

Treatment-emergent suicidal ideation compared to all prior history: An increase

in the maximum suicidal ideation score during treatment from the maximum

scores from the Baseline C-SSRS scale, Screening C-SSRS scale, or

Baseline/Screening C-
Visit C-SSRS scales taken prior to treatment).

Version 2.0 (Finalized February 2013) 3

 Emergence of serious suicidal ideation compared to all prior history: An increase
in the maximum suicidal ideation score to 4 or 5 during treatment from no

-SSRS scale, Screening C-SSRS scale, or

Baseline/Screening C-
Visit C-SSRS scales taken prior to treatment).

Treatment-emergent serious suicidal ideation compared to all prior history: An

increase in the maximum suicidal ideation score to 4 or 5 on the C-SSRS during
treatment from not having serious suicidal ideation (scores of 0-3) prior to
om the Baseline C-
SSRS scale, Screening C-SSRS scale, or Baseline/Screening C-SSRS scale, and
-SSRS scales taken prior to
treatment).

Outcomes that can be used for clinical management and safety monitoring or potentially
for research purposes are described below.

Suicidal behavior lethality rating taken directly from the C-SSRS

Suicidal ideation score

Any score greater than 0 is important and may indicate the need for mental health
intervention. The protocol procedures related to clinical care of patients with
treatment emergent suicidal ideation and behaviors will then be implemented to
ensure pro
example, a score of 4 (active suicidal ideation with some intent to act) or 5 (active
suicidal ideation with specific plan and intent) on suicidal ideation can be used to
indicate serious suicidal ideation and can be used to trigger further evaluation and
ental health practitioner (for non-psychiatry
trials this may be used to trigger a prompt referral to a mental health professional)
and/or possibly the emergency room.

Suicidal ideation intensity rating

Add the five intensity item scores to create a total score (range 0 to 25) to
represent the intensity rating. If the patient did not endorse any suicidal ideation
set the intensity rating to 0.

Analytical and Statistical Methods:

nsistently.
Suicidal ideation and suicidal behavior are relatively rare events in individual, double-
blind, controlled clinical trials, except possibly for a population particularly susceptible to
this safety risk. In general, no formal statistical hypothesis testing is recommended for
individual studies as only few events are typically observed and descriptive summaries
will then suffice [9].

Version 2.0 (Finalized February 2013) 4

See the mock tables, the mock listing, and mock figure for suggested summaries and
presentation of the data. Note that these mock tables can be easily extended to include
multiple treatment groups/comparisons as needed. Tables should be modified to fit the
needs of the specific analyses performed.

For very small studies (e.g., Phase 1 studies), or indications where suicide-related events
are expected to be rare, it may be sufficient to plan on providing a listing only (see
Listing 1). Otherwise, Tables 1 and 2 as shown may be provided.

The specific statistical method (or lack of an inferential method) should be consistent
When inferential analyses are not performed the p-value
column should be eliminated. Tables 3-4 (options for a shift table) will likely be used
when there are findings to understand in more detail and/or when the expected number of
suicide-related events is large either due to having a large study or having a population at
high risk. A graphical presentation, for example a stacked bar chart, may also be used
when appropriate (see Figure 1).

A summary of pre-treatment data is generally desirable and can be added to existing

planned tables for patient characteristics (e.g., percent of subjects with lifetime suicidal
ideation, baseline suicidal ideation, lifetime suicidal behavior, baseline suicidal behavior,
lifetime self-injurious behavior without suicidal intent, and baseline self-injurious
behavior without suicidal intent). Alternatively, if all categories of suicidal ideation and
suicidal behavior are of interest for pre-treatment summaries, Table 1 can be repeated for
providing a summary of lifetime outcomes and again for baseline outcoments. The
specific timeframe used for defining baseline will need to be defined.

When inferential statistics are used, methods for consideration include: the Miettinen and
Nurminen method [1], an unconditional, asymptotic method; the Fisher's exact test; 95%
confidence intervals (CIs) based on Wilson's score method; and forest plots to present
effect across treatment arms or subgroups of interest.

A further consideration for inferential analysis may be the requirement that at least 4
patients in any treatment group exhibit the event [2]. Because many 95% CIs may be
provided without adjustment for multiplicity, the CIs should be regarded as a helpful
descriptive measure to be used in review, not a formal method for assessing the statistical
significance of the between-group differences in adverse experiences. Appropriate
cautionary statements on the interpretation of inferential statistics for purposes of
summarizing information or signaling trends should be made.

Methods for assessing dose-response relationships should be consistent with the sponsor's
standards. Calculating and presenting incidence rates (e.g., n/patient-years) should be
considered, especially for studies with long patient exposures and expected differential
drop-out rates between treatment groups.

In long-term trials where the suicidality scale is administered frequently, a time-to-event

analysis where the event is any suicidal event (either ideation or behavior whichever
occurs first) can be performed using methods such as Kaplan-Meier method [3], log-rank

Version 2.0 (Finalized February 2013) 5

test [4], generalized Wilcoxon test [5] and Cox proportional hazards model [6] adjusting
for a limited number of covariates. It is noted that event type-specific analysis (i.e., time-
to-first-ideation and time-to-first-behavior analyses) may not be interpretable due to the
presence of-competing risks [7] and informative censoring [8]. In situations that warrant
an assessment by age, gender and/or other risk factors summary statistics of scores on
suicidality endpoints of interest can be presented by visit. For studies with longer
duration, scores can be summarized by visit or time period (treatment phase, post-
treatment phase etc.).

Based on the individual program strategy, active treatment groups may be pooled for
comparisons to placebo/active comparator, or for performing all pair-wise dose
comparisons. Analysis methods for assessing effects over several trials, subgroups or
indications are presented in [9].

Note that missing data should not be imputed.

Version 2.0 (Finalized February 2013) 6