Docstoc

White Paper - SDL Multilingual Labelling Life Sciences

Document Sample
White Paper - SDL Multilingual Labelling Life Sciences Powered By Docstoc
					White Paper




  Optimizing Multilingual
  Labeling Processes




                            Written By:
                            Matthias Heyn, SDL
                            September 2012
Optimizing Multilingual Labeling Processes

Background and Scope
This paper, intended for readers with some knowledge of global labeling procedures, presents a
framework to assess overall efforts that are necessary to manage multilingual product submissions
in the Centralized Procedure (CP) in Europe. This framework provides a better understanding of the
overall multilingual labeling process burden in the CP. The following considerations are not meant
to provide an exhaustive description of all required efforts. However, they should serve as a good
starting point for further quantifications. It should be noted that most of the following quantifica-
tions are applicable with slight modifications to other European registration procedures like the
mutual recognition procedure (MRP) or the decentralized procedure (DCP) and can be projected to
global labeling needs.

The overall findings will support the case that multilingual labeling is an excellent application field
for automation and in general for measures to reduce the amount of manual processing of related
tasks.



Table Of Contents

General activities in the Centralized Procedure...................................................................................1
Translation Processes Included...........................................................................................................1
     Typical Timeline for Submissions.................................................................................................2
     Document Set and Document Sizes.............................................................................................2
  Estimation of Coordination Efforts..................................................................................................3
  A New Marketing Authorization (NCE/NBE)....................................................................................4
     Number of documents handled...................................................................................................4
     Scenarios.....................................................................................................................................4
  Type II Variation..............................................................................................................................6
Summary of Typical Coordination Efforts: Initial Marketing Authorization & Type II Variation...........7
Multilingual Work Load per Company type.......................................................................................7
  Example 1: Small Company.............................................................................................................8
  Example 2: Medium Size Company.................................................................................................8
  Example 3: Major Pharma Company Profile....................................................................................8
Communication Overhead Assessment..............................................................................................8
An Opportunity for Business Process Improvements..........................................................................9
  Translation Management Systems and Translation Memories.........................................................9
Summary.........................................................................................................................................10
Bibliography....................................................................................................................................11
General activities in the Centralized Procedure
The centralized procedure (CP), which came into operation in 1995, allows applicants to obtain a marketing
authorization that is valid throughout the European Union. It is required for several types of drugs, such as
those intended to treat AIDS, cancer and diabetes.

As part of the Centralized Procedure, the activities for the Marketing Authorization Holder (MAH) include:

•	        Authoring & Submission: The initial product information is sent in English to the European Medicines
          Agency (EMA)
•	        Submission of Revisions: The updated English is sent to the Committee for Medicinal Products for
          Human Use (CHMP)
•	        Provisional Translations: Translations of the submission are prepared in parallel into 23 languages to
          review for Opinion
•	        Revision & Submission of Translations: Submissions in 23 languages are sent to the EMA
          simultaneously
•	        Editing & Submission of the Translation Revisions: Dependent upon national agency comment(s)
          translations are updated and sent back to the EMA
•	        Distribution & Archiving of the Commission Decision
•	        Creation of Packaging Materials
•	        Translation of Supplementary Information (e.g., patient alert cards)


Translation Processes Included
Translation plays a critical role in the centralized procedure. To simplify the quantifications in this description,
we will limit the discussion to the two most important submission types in the CP that pertain to translation:

•	        New marketing authorization procedures (NCE/NBEs)
•	        Type II Variations

The above submission types represent the predominant work items when it comes to the linguistic phases in
the CP. Renewal procedures are very similar in terms of timelines and associated linguistic efforts; therefore,
they can be included in the Type II variation assessment scenario.

Excluded from quantifications are the following procedures:


                                                               30 day procedures are typically less frequent and
     •	     Urgent (30 days) procedures
                                                               require very few linguistic adaptations.

                                                               Type IA/B Variations require very little multilingual
     •	     Type IA/B Variations
                                                               efforts (if at all).

                                                               61(3) Notifications have a very different time frame
     •	     61(3) Notifications                                and vary significantly in terms of associated transla-
                                                               tion burden.




                                                           1
The Austrian and German label are identical, the Greek and Cypriot, the English and Irish also, and Belgium
takes the French and German translations. Therefore those local variations have to be harmonized across those
languages by the involved affiliate countries. Those coordination efforts are not explicitly included in the
following.


Typical Timeline for Submissions
The following Figure 1 provides a simplified overview of the CP timelines for typical product maintenance
(Type II, 60 or 90 days) and an initial product submission (NCE) with focus on linguistic activities.




Figure 1: Simplified Schema of the Centralized Procedure Timelines for Product Maintenance (Type II) and initial product
submission (NCE). This diagram shows a schematic representation. The length or size of the box representing a process
step is not intended to be proportional to the timescale depicted. See also: EMA – The linguistic review process of product
information in the centralized procedure human.

Given the rigid timelines, coordination efforts in the centralized procedure are significant. To fully quantify the
volume of information required as part of this procedure, the following can serve as a starting point to help
determine the effort required.


Document Set and Document Sizes
The product information (PI) set submitted to the EMA consists primarily1 of the following document types:

•	   Summary of Product Characteristics (SPC)
•	   Package Leaflet (PL)
•	   Annex II
•	   Labels (Inner, Outer, Immediate Blister, Small Immediate [e.g., Vial])


1 There can be other documents in certain situations, but these are not formally defined as part of the PI set nor are they included in the QRD
templates. An example would be the graphic user instructions in the use of an inhaler.

                                                                        2
Depending on how many forms, strengths and presentations exist for a given product, hundreds and
sometimes thousands of those documents have to be created, edited and reviewed, translated, translation
reviewed - all against very short timelines. Figure 2 illustrates how quickly the number of documents can grow,
depending on product complexity in terms of how many formulations, strength, presentations and official EU
languages are required.




Figure 2: Total number of documents depending on product complexity.

As the complexity of product information rises, so does the number of redundant documents. To help deal
with the complexity of the submission process and reduce the number of individual documents, it is common
practice to combine all documents into one single master Microsoft Word document per language. This single
master document therefore consists of many nearly identical text parts2 and subsequent editing, updating and
review steps typically happen on this single master document.


Estimation of Coordination Efforts
We now need to distinguish the two main types of submissions in the CP (leaving renewals, 61(3) Notifications
and other submissions aside):

•	   A new marketing authorization of a new chemical or biological entity (NCE/NBE)
•	   A Type II Variation as an example for product maintenance work




2 Approximately 30% redundant text parts are typical for PI. This can be computed by applying the statistical routines of a translation memory
system counting identical sentences throughout a document.



                                                                       3
A New Marketing Authorization (NCE/NBE)
The average size of an initial marketing authorization procedure has circa 40,0003 words. This corresponds
to roughly 5 weeks of translation time (26 working days for an FTE, using a 1500 words per day base). Actual
products may vary significantly from the average, with some submissions consisting of only 1500 words and
complex products with more than 100,000 words.

As a best practice, when a product has an average number of words, there are typically 5 linguistic activity
points that are considered for an initial submission.




Figure 3: Major linguistic activity points during translation in the CP. HA Comments = Health Authority comments on
the translations during the linguistic review phase. Red dots denote where activity always happens, green dots represent
optional activities.

Good practice is to start a first translation round 8 weeks before Opinion (after receipt of the Assessment
Report at day 150). If there are substantial reviewer comments, then a second hand-off of the adapted baseline
information should be done at the time of CHMP discussion (around day 180, with possible Oral Explanation
at Day 181). Shortly before Opinion (Day 210) it is good to have all accumulated changes adapted (circa 1
week or a couple of days before the CHMP meeting). The objective is to be prepared and ready to adapt any
final changes from the Opinion during the 5 official days (which include a weekend) when the translations are
finalized. It is important to note that none of the 23 language versions can be late without risking a delay and
the recognized economic effects.

In addition to the linguistic points above, there is work for those languages which have received linguistic
comments by the local regulators (HA comments). In practice, it is very unusual if there are no comments at
all, so typically there will be activity at the Post-Opinion national linguistic review stage.


Number of documents handled
As discussed above, the number of documents depends on the number of presentations, so if there are 2
forms and 2 strengths, each one has a SPC, PL etc. Therefore, it is between 4 and hundreds of Microsoft Word
files that will go through the above steps


Scenarios
In the following, the term “task” (process step) means that someone internally or externally has to do
something with a document. This could be translating content, reviewing text, or resolving a linguistic
comment sent by a regulator. A task can take several weeks (translation of a 40,000 word document) or a
couple of minutes (change a few words in one paragraph).

3 During an efpia project, a sample of CP products was analyzed and computed, resulting in 39,326 average English words. See below table for
the detailed word count.

                                                                                                           SPC       Leaflet     Label    Factor      Total
 39,326      Words      Average word count for English source language per Initial MA NCE                 8,796       5,185      1,749     2.5*       39,326
 * Average factor because of additional strength and presentations, added to the set of documents. EMA data from 2004 computed to an average of 2.5


                                                                              4
All tasks have to be coordinated and some activities like “Translation” and “Update Translation” may be
outsourced to an external party.

In order to gain an understanding of the necessary efforts, two typical scenarios are distinguished:

1. Less Complex Case Scenario: A submission which represents a relatively “easy” and uncontroversial
   submission from the outset
2. More Complex Case Scenario: A submission which represents a submission that requires a high degree
   of internal agreement as well as requires more effort to pass the regulator.

The reader should be aware that any level of complication and deviations from those scenarios are possible,
but the following should help to understand the actual workload situation.


Less complex Case Scenario
Let’s assume an “easy” submission where no major changes happen to the English baseline and only two
reviews (medical and e.g., safety) are required for all languages. The following Table 1 provides a break-down
of the necessary activities.

The table should be read as follows: “On (around) day 180 we have 24 target language documents distributed
for medical review which represents 24 (internal) tasks. Typically only 2 reviews are necessary with one
medical review on late change during opinion.




Table 1: Example Task distribution in case of a submission with no or very few changes to the English baseline.



More complex Case Scenario
A more complex but highly realistic case involves intensive dispute between applicant and regulator, which
in turn triggers more update translation work. For major new product introductions, the applicant will more
likely add a legal or even a marketing review. The following Table 2 summarizes corresponding tasks.




                                                             5
Table 2: Tasks in case of a submission with major changes to the English baseline. There are additional reviews tasks
necessary during opinion and in a worst case scenario, there can be many more updates pre-opinion.



Type II Variation
If we take a 60 day Type II variation scenario we typically will only have one translation round, as the textual
changes are normally minimal.




Figure 4: Multilingual Activity Points in a 60 day Type II procedure. Red dots denote where activity always happens, green
dots represent optional activities.

Here, the chances for linguistic comments are much lower as only a few new sentences will usually differ from
the current approved labeling4 , although variations can contain large amounts of new and changed text.

In a simple case scenario (see Table 3), we only have one Update Translation task and perhaps a few scattered
Health Authority comments in some languages. In a more difficult case scenario, a translation update task and
a medical and safety review with multiple Health Authority comments for all languages may be required. Table
4 sums up the tasks for the more difficult scenario.




                                                                                                                     Table 3: Tasks in case of a
                                                                                                                      simple Type II variation.




4 In theory, the approved labeling should not receive reviewer comments as the translations were reviewed and approved by the Health Author-
ity during the previous regulatory activity. Only the new or changed text, with the resulting new translations, would be reviewed. However,
in practice, due to the subjective nature of translation, and the issues of base text inconsistency discussed in this paper, it is possible to receive
significant comments on text which is actually unchanged from the approved product information already on the market.
                                                                          6
The following Table 4 captures the more complex scenario:




Table 4: Tasks in case of a complex Type II Variation

In case of a 90 day procedure for major changes to a drug’s labeling, big pharma companies will certainly
schedule additional legal and marketing reviews.



Summary of Typical Coordination Efforts: Initial Marketing Authorization
& Type II Variation

The following Table 5 summarizes the above results in terms of multilingual tasks.


 Number of tasks (process steps) per Submission Type              Less complex Case       More complex Case        Mean

 New Marketing Authorization (NCE/NBE)                                     144                     216               180

 Type II Variation / Renewals / Annual Re-assessments                      29                       96              62.5



Table 5: Summary on number of linguistic tasks for initial submissions and Type II Variations in the two scenarios. The mean
value assumes that there are an equal number of friendly and more complex submissions to be handled.



Multilingual Work Load per Company type

Companies differ largely in terms of pipeline, size and activity of their portfolio. The following tables predict
approximate multilingual efforts for a small company, a medium size pharmaceutical or a major big pharma
organization. Similar calculations can be easily adapted.

The following limitations need to be kept in mind when looking at the following scenarios:
3. This only captures CP and does not include NP, DCP, MRP or any global labeling activities
4. This only captures Type II Variations (Renewals and annual Re-Assessments) as well as initial marketing
   authorizations
5. Is based on assumptions on two simplified submission scenarios. Each is represented here with an equal
   distribution (of a simple and a more difficult use case).

                                                             7
Example 1: Small Company
The following table represents a sample case of a small pharmaceutical company:




Table 6: Example Company with a small but recent portfolio of products in the CP.



Example 2: Medium Size Company




Table 7: Example Company with medium size and balanced active Portfolio



Example 3: Major Pharma Company Profile




Table 8: Example large pharmaceutical Company with large Portfolio.



Communication Overhead Assessment

In addition to the time required to complete each task, there is overhead time required in order to complete
each of them. For example, each task for the 3 examples above typically requires the writing of an email
notification, which may include further instructions to translation service providers or to reviewers. Each
recipient of that email will have to process the information before acting on it. After completion of the work,
each recipient will return the work back to the coordinator and provide further clarifications via email, which
have to be processed by the original sender. Typically, both sides this process are logged into a database or
other means to track and trace the overall process.

                                                            8
If we assume a total processing time of 6 minutes per step above (creation and logging by sender, processing
and logging by recipient, returning by recipient and logging and processing by original sender, plus final
logging) overall communication overhead time computes to approximately 24 minutes per task.

The following table computes the overall work requirements for the different size organizations.

Table 9: Communication assessment per exemplary company type for the CP.




This is a simplification without taking 61(3) Notifications, 30 days urgencies, Type I, DCP, MRP, NPs or global
labeling related work into account, or the fact that some tasks are bundled together for all languages during
outsourcing activities.



An Opportunity for Business Process Improvements

It quickly becomes clear how most pharma companies can be burdened by multilingual coordination efforts,
especially efforts resulting from product maintenance and the need for internal and regulatory reviews. A large
number of small changes have to be managed over a large group of geographically distributed stakeholders.
But, not only big pharma is confronted with thousands of multilingual tasks surrounding their marketing
authorization activities: the case of a medium-sized company like Genzyme (now part of Sanofi) reconfirms
that even though the translation of text can be outsourced, the burden in terms of internal tasks remains very
high5.


Translation Management Systems and Translation Memories
Most of the communication related tasks are prime candidates for automation. There is a class of software
application systems which support translation coordination tasks systematically. This type of software
application commonly has been named “translation management systems” (TMS). A TMS manages translation
projects and the associated translation related tasks, groups all activities into workflows and automates task
distribution, email notifications, alerts, task tracking, and audit trails as well as associated financial transactions
with external translation vendors. In contrast to document management systems or general workflow
systems, a TMS has multilingual task distribution - grouping by language - and most importantly, the notion of
translation reuse mechanisms built into its core.

The aspect of systematic translation reuse allows for further significant improvements in a labeling context.

5 See: Raun Kupiec, Translation & the Critical Path to Centralized Product Labeling in the European Union. Session Presentation, 45th Annual
Meeting of the DIA 2009, San Diego




                                                                       9
PI is one of the most highly repetitive and redundant text types across industries: As the complexity of a
pharmaceutical product rises (i.e., the number of distinct formulations, strengths and pack sizes), so does the
number of highly redundant text parts. Therefore, PI is a prime candidate for the application of a software
system which automatically reduces - if not eliminates - the manual handling of redundant or altered text parts.
Such a mechanism is well known as a “translation memory” (TM). Such a TM performs (among other things)

•	   Storing regulator approved translations for future reuse
•	   Automatically detecting changed text parts in variations and protect unchanged text parts
•	   Systematically ensuring that all redundant text parts are translated consistently throughout the submission
•	   Enforcing consistent terminology use


Until very recently, it was very difficult - if not impossible - to apply translation memories on PI in the CP, DCP
or MRP. Commercial systems failed to support the necessary mechanisms of handling tracked changes and
comments in the underlying Microsoft Word documents6 as required by the European regulators.

With this barrier removed, today a TMS system offers an opportunity to significantly reduce the multilingual
burden from handling and communicating the thousands of multilingual tasks during marketing authorization
procedures. Affiliates as well as central labeling staff can be liberated via a TMS allowing such expert staff
members to spend more time on higher value activities.



Summary

This paper has made an attempt to provide a framework for quantifying the overhead stemming from the
multilingual labeling process burden in the CP. Each individual organization needs to adapt such calculation
considering their specific environment. Nevertheless, especially for medium to large size pharmaceuticals, it
becomes obvious that multilingual labeling is a prime candidate for future business process improvements and
the application of a TMS system is self-evident.




6 See: Matthias Heyn: Applying Language Technology to Multilingual Labeling. SDL Whitepaper 2012.


                                                                   10
                    Bibliography
                    1. EMA - European Medicines Agency - Doc.Ref: EMEA/5542/02/Rev 4.1. (2011, February 03). The
                       linguistic review process of product information in the centralised procedure - human. London,
                       United Kingdom.
                    2. Heyn, M. (2006). Moving to efficient creation of multilingual product information. Regulatory
                       Rapporteur (TOPRA) - Vol 3, No. 6, 2-8.
                    3. Heyn, M. (2007). Best Practices for Addressing pharmaceutical Multilingual XML Requirements
                       (Session #119). 43rd Annual Meeting of the DIA. Atlanta.
                    4. Heyn, M. (2007). GIM and PIM. Next Generation Pharmaceutical Europe NGP, Issue 4, online
                       magazine (http://www.ngpharma.eu.com/focus-area/Issue-4/Information-Technology/), 1-2.
                    5. Heyn, M. (2009). The Multilingual Dimension: Best Practices for Addressing Global Content and
                       Global XML. 4th Annual Meeting DIA. San Diego.
                    6. Heyn, M. (2012). Applying Language Technology to Multilingual Labeling. An SDL Whitepaper.
                       Maidenhead, United Kingdom.
                    7.       Kupiec, R. (2009). Raun Kupiec, Translation & the Critical Path to Centralized Product Labeling
                             in the European Union. Session Presentation, 45th Annual Meeting of the DIA 2009, San Diego.
                             45th Annual Meeting of the DIA 2009. San Diego: DIA.




SDL_wp_OptimizingMultilingual_EN_LTR           Copyright © 2012 SDL plc. All Rights Reserved. All company product or service names referenced herein are properties of their respective owners.

				
DOCUMENT INFO
Description: For most Life Sciences organizations, the multilingual labeling process can be painful, expensive, and error-prone. In this White Paper, SDL examines how SDL Global Information Management (GIM) solutions—combining best practice translation management, translation reuse, and translation review interfaces—provide the essential business strategy and tools for managing the escalating requirements for local language delivery.