DATA -> DITA -> DOCXA Real Story of Efficiency from the Pharma Industry

LavaCon 2016 | Las Vegas | Steffen Frederiksen | steffen.frederiksen@ditaexchange.com | www.ditaexchange.com

Background

SANOFI

• Sanofi is a global healthcare company focused on patient needs. Their business includes pharmaceuticals (notably prescription drugs for diabetes, rare diseases, multiple sclerosis and oncology, consumer healthcare products and generics), vaccines and animal health.

• Sales in 2015(Rx) of 34,896 million USD makes it the 5th largest pharma in the world

What Does a Pharma Company Produce?

?

4

What Does a Pharma Company Produce?

•The “old” answer was: Molecules and content

•The current answer is: Content

5

A Typical Timeline

And How Many Pages?

•Your Best Guess?

7

And How Many Pages?

•150,000 – 250,000 pages per molecule

8

The Situation

• Huge number of hours spent

• Documents finalized late

• Documents not e-compliant

• Lack of consistency

• No standard methodology or tools for data collection & aggregation

• Lack of structure

• Long review cycles

• Multiple review iterations

• Recurrent remarks (regression) at different stages of review

The SANOFI CRUISE Project: Business Drivers

ENSURE CONSISTENCY WITHIN THE ORGANIZATION AND BETWEEN PUBLIC DISCLOSURE AND REGULATED REPORTS (NDA, IND, CTA,

PSUR...) AGAINST DATABASES

QUICKLY ASSESS VALUE AND NON-VALUE ADD ACTIVITIES AND MANAGE THE SOURCING OF THOSE ACTIVITIES WITHIN THE

CAPABILITIES PORTFOLIOS BALANCING FIXED AND VARIABLE COSTS

REDUCE CYCLE TIMES, ENABLE EARLIER AND MORE EFFECTIVE DECISION-MAKING AROUND CLINICAL DEVELOPMENT PROGRAMS,

AND REDUCE TIME TO MARKET

SHARE KNOWLEDGE BY MOVING TOWARD REUSABLE COMPONENTS OF INFORMATION THAT CAN BE MANAGED AND REUSED ACROSS

PUBLICATIONS, DEPARTMENTS AND AUDIENCES.

Short Version

• Increase Quality, Consistency and Compliance

• Faster Time to Market for New Drugs

• Reduce Excessive Costs

The Idea

• Develop modular content re-use and structured authoring solution

• Leverage Microsoft platform (Word and SharePoint)

• Enable integration of other SANOFI applications

• Refine submission documents preparation

• Implement content reuse across lifecycle and processes

• Enable a more granular content management process

• Refine review process

• Improve traceability

Action: EnCORE

13

CASE STORY: PATIENT NARRATIVES

What Are ‘Patient Narratives’?

• As Per ICH E3 guidelines, a patient safety narrative should describe the following:• the nature, intensity, and outcome of the event

• the clinical course leading to the event

• an indication of timing relevant to study drug administration

• relevant laboratory measures

• action taken with the study drug (and timing) in relation to the event

• treatment or intervention

• post-mortem findings (if applicable)

• Investigator’s and Sponsor’s (if appropriate) opinion on causality

Content in Patient Narratives

• Specifically, narratives should include the following:• patient identifier

• age and sex of patient; general clinical condition of patient, if appropriate

• disease being treated (if this is the same for all patients, this information is not required) with duration (of current episode) of illness

• relevant concomitant/previous illnesses with details of occurrence/duration

• relevant concomitant/previous medication with details of dosage

• test drug administered, including dose, if this varied among patients, and length of time administered

BEFORE

Typical Process

• Preparation of first draft narrative from patient/subject data by the Medical Writer

• Scientific and editorial peer review by the CRO project lead to check the document is accurate, complete, and consistent with

requirements and across documents

• Clinical review of draft narrative: It is recommended that this be performed by the Sponsor or designate, although the CRO can

provide this service as necessary

• Medical Writer revision based on clinical review: If the writer does not agree with clinical review comments, for example, when

requested amendments conflict with the evidence or when changes would introduce inconsistencies between narratives, or when

review comments are unclear, these should be discussed with the Sponsor or designate, as appropriate, and responses retained on

file

• Quality control (QC) review based on final patient/subject data. Given the often large number of narratives required for individual

studies and small size of each document relative to the CSR, it is recommended that a single QC review be performed toward the

end of the process, rather than QC review of the first draft and final deliverable

• Medical Writer revision based on QC review findings. Note: when significant findings are identified during QC review, these should

be discussed with the Sponsor and clinical reviewer, as appropriate, and further updates should be checked for consistency and

accuracy

• Approval by the Sponsor after a final review

The Sources for Narratives

• A Medical Writer will use various sources of information when preparing patient safety narratives. These include • Council for International Organizations of Medical Sciences (CIOMS) forms

• Case Report Forms (CRFs)

• MedWatch forms

• Data Clarification Forms (DCFs)

• Clinical database listings

Structure of patient narratives (1/2)

HEADER

MEDICAL HISTORY

AE

PV

TABLES

Generally, one or more topics make up the content in each section. A single event is described per topic.

The output from a map is a publication in MS Word format

1 map = 1 patient narrative

Patient narrative maps are organized into sections

Structure of patient narratives (2/2)Header section

Medical history section

Adverse Event

section

Pharmacovigilance

summary section

21

Before Costs of Writing Narratives

• Per narrative with manual process:• 8hours/narrative * $100/hour = 800 USD

• Per study (example):• 625 narratives (average) * 800 USD = 500,000 USD

• Per year (example):• 100 studies * 0.5 million USD = 50 million USD

22

ROI for Narratives

23

Challenge #1:

“Can you design a robust,

extensible feature that

allows us to pull together

data from many different

sources and systems into

our topics?”

Map= Patient

Narrative

Topic SAE: Date: 1-Feb 2012

Topic AE leading to treat. Disc.:

Date: 10-March 2012

Topic Overdose: Date: 1-Feb 2012

Topic Cancer: Date: 5-Feb 2012

Topic PV: Date: 1-Feb 2012

24

Case Study - Patient Narrative CreationA single narrative will be provided for a given patient: The system associates all events(SAS and PV summary) related to same Patient (following business rules) in order to create the complete Patient Narrative. Patient Narratives will be grouped by the system following a configured priority: Deaths>SAEs>AE leading to treat. Disc.>AESIs>Other – files used in CSR

SAS data from system are retrieved

in Topics for each event

PV data from database

are retrieved in Topics

Generation of maps

Map= Patient Narrative 1

Topic SAE:

Topic Ae leading to treat. Disc.

Map= Patient Narrative 3 (id: 0003)

Topic Severe hypoglycemia

Topic pregnancy

Topic Cancer

Map= Patient Narrative 9 (id: 0009)

Topic overdose

Topic overdose

Map= Patient Narrative 2

Topic DeathFiles in repositoryFile Death:

Patient Narrative 2

File SAE:

Patient Narrative 1

File AE leading to treatment

Discontinuation:

Patient Narrative …..

File AESI:

Patient Narrative 3

Patient Narrative 9

File Other:

Patient Narrative …..

Improved Quality: Data, unstructured and structured information aggregated automatically for reuse

Improved productivity: automated generation of maps, style sheets to apply formats quickly

Response To Challenge #1

DITA’s content referencing (conref) mechanism

+

Microsoft SharePoint Business Connectivity Services(SharePoint BCS)

+

Our extension to conref resolving

(aka DxLiveData)

=

A general, robust and manageable mechanism for bringing in data from any LoB system into your topics

25

SharePoint BCS Overview

XXXXX Systems

“The nuts and bolts…”

OXML

DITAVAL

DITA XML

SharePoint

DitaExchange

Clinicaltrials.govFDA: 1572 FDA: Patient Narrative

Product wwwCountry wwwLanguage www

SubmissionsReview

Apps:ReviewApprove

IOSAndroidW8…

SubmissionsArchive

LoB

LoB

LoB

LoB

AFTER

After Costs of Writing Narratives

• Per narrative with automation:• 0.95 hours/narrative * $100/hour = 95 USD (before 800 USD)

• Per study (example):• 625 narratives (average) * 95 USD = 59,375 USD (before 500,000 USD)

• Per year (example):• 100 studies * 0.06 million USD = 6 million USD (before 50 million USD)

29

ROI for Narratives

Just One Small Detail…Say Hello To Reality!

30

Just One Small Detail…Say Hello To Reality!

31

To Hell with

DITA XML!

32

Challenge #2:

“OK, so we would like to

keep all that is good with

DITA XML, just give it to us

with normal Word

documents – and please

do not say DITA XML to

any of our writers in the

future…Can you do that,

please?”

Response To Challenge #2

DITA’s content architecture

+

ISO/IEC 29500 (aka Office Open XML, aka DOCX, aka Word)

+

Our authoring app for Word

(DxAuthor for Word)

+

Dx4 Publishing etc.

=

All that is good from DITA and much, much less of the bad…

33

DITA: The 80% That Are Really Good?

• Modular, topic-based architecture

• Flexible metadata and data mining

• DITA maps for organizing modules

• Content reuse (topics, maps, conrefs, DxLiveData)

• Variables

• Semantic tagging of content:

• Folksonomy (for yourself)

• Taxonomy – using SharePoint Term Store to manage taxonomies

• Schema-based tagging (IDMP, Redaction, … and many more)

• Conditional publishing and filtering, including DxTimeLinePublishing

• XML-based format, allows automation (and XSLT is till an option)

• Publish to many layouts/formats/media, using Word templates for formatting

DITA: The 20% Bad And Ugly

• Unfamiliar XML tools : FIXED

• Writers, reviewers and approvers have to deal with XML: FIXED

• Suffocating rules: “Invalid element…”: FIXED

• Expensive and very complex XSLT/ publishing templates FIXED

• Expensive and complex DTD-specializations: FIXED

36

I can work with this!

Results Delivered

• Narratives @ Sanofi

• Delivered narratives for 16 studies

• Time reduction in QC process

• Use of industry standard data model as input for Narratives process

• Additional time benefits for on-demand narratives

• Partnering with FDA to develop standardized output (HL7)

• The Big Win

• Faster time to market

38

Challenge #3:

“OK then. Our reviewers

and approvers will ONLY

work on complete

documents, not on topics.

But we need their

approved feedback

decomposed back to the

topics. Please?”

In Development…

Map & Topics Publish for Review Reviewers Use

Word

DxReview collates

feedback

Edits/Comments

back into Topics

40

Challenge #4:

“OK then. We would like to

eliminate the writing

phase for patient

narratives all together –

maybe by using AI/NLP or

content rules - leaving us

with only review and

approval. Please?”

DxLiveData Conreffing Content Automation Example

Data import layer

Data preparation layer

Calculation layer

1st Language

Generation layer

2nd Language

Generation layer

Dynamically Integrating ML/AI Text into Topics

43

The Vision of DitaExchange

"Bringing the Power of Structured Content Management to Everybody"

Data -> DITA -> Docx -> Thank You!A Real Story of Efficiency

Steffen Frederiksen | steffen.frederiksen@ditaexchange.com | www.ditaexchange.com