|
|
|
International Clinical Trials
|
The advent of electronic submissions to the FDA has facilitated the
streamlining of submissions to various agencies around the world, as
well as harmonising data formats across the multiplicity of documents
received by regulators from different companies. Since 2008, the FDA has
been actively encouraging companies to submit new drug applications,
biologic licence applications, investigational new drug applications and
so on electronically, using the ICHs electronic common technical
document. Most submissions are now sent to the FDA in this way.
Formatting the Variables
How
data are formatted is important. Back in 2004, the FDA implemented the
study data tabulation method (SDTM) as the preferred standard for
clinical submissions. Its binding guidance document splits relevant
observations and information made in a clinical trial into four types of
variable:
- Identifier variables pinpoint data like the name
of the study, the subject observed, the domain and the sequence number
of each record
- Topic variables include descriptions of the observation, such as the identity of a specific lab test
- Timing variables cover anything involved in the timing of the observation
- Qualifier variables can consist of any further important information, whether descriptive text, results or units used
Qualifier variables are broken down into a further five categories:
- Grouping qualifiers aggregate observations within the same domain
- Result qualifiers are used to describe specific results associated with the topic variable in question
- A synonym qualifier can be employed to note an alternative name for any variable in an observation
- Record qualifiers define additional attributes of the overall observation record
- Variable qualifiers add further description to individual variables within an observation
While,
in theory, the SDTM implementation guide (SDTMIG) can be used to submit
non-clinical toxicology data, the format can be less than ideal for
handling the fields required for data generated in animal tests. Its
format focuses on safety domains, making the recording of information
such as death diagnosis, organ measurements, palpable masses and
macroscopic findings difficult.
New Approach
As a
result, a modified standard standard for exchange of non-clinical
data (SEND) has been developed, which includes specific domains, and
others, for the information above. The FDA has now issued binding
guidance, which means this standardised electronic format will have to
be used for non-clinical data included in submissions from December 2016
onwards.
The latest version of the SEND implementation guide
(SENDIG) can be found on the Clinical Data Interchange Standards
Consortium website. The aim of the document is to lead users through the
organisation, structure and formatting of non-clinical tabulation
datasets that are consistent and suitable for sharing between all
partners, including the trial sponsor, a CRO or the regulators.
SEND
is, essentially, a more specific implementation of the same SDTM model.
Data from animal trials and in vitro testing are captured and
transformed into the tabulated form as prescribed in the guidance. Some
domains are clearly going to be exactly the same as those in SDTMIG,
such as vital signs findings, or laboratory or electrocardiographic test
results. However, animal tests also include other factors that generate
additional data fields like species, strain and sub-strain information
in the demographics domain, as opposed to SDTMIG-specific fields such as
race and ethnicity information. The trial sets (TX) domain allows the
sponsor to define the planned sets of subjects, which result in a
combination of the experimental factors of interest on a study.
Experience
gained from presenting data in SDTM format will be extremely helpful in
implementing SEND. Furthermore, some of the software tools developed
and built for SDTMIG will be transferrable, either directly or with
modifications. SEND implementation, therefore, is unlikely to involve
having to start from scratch. Creating a new metadata library for SENDIG
is entirely analogous to the creation process for a new library for
either new or upgraded versions of SDTMIG, or a sponsor-specific
implementation of SDTM.
Process Stages
There are
many process considerations that must be taken into account when
implementing SEND. The conversion process is important, and will include
mapping, merging and converting source data, such as that garnered from
electronic data capture, third-party data and raw data files. The
complete conversion process must be scheduled and followed up.
Next,
the datasets must be transferred whether these are draft, interim or
final datasets and then formatted in the appropriate way, for example
with XPT, SAS or Dataset-XML. The definition metadata must be created,
validation checks will have to be made, and trial design tables
generated.
As SDTMIG and SENDIG are based on the same model, the
implementation of the two systems ought to be comparable. Domains will
have to be created for SEND that will be populated with the appropriate
data and these will, in terms of organisation and structure, for the
most part closely resemble those that exist for SDTMIG. Many of the
automated processes that are already employed for SDTMIG should be
translatable to SENDIG.
Key Differences
Although
the two standards have their roots in the same place, there are many
high-level differences between the implementation guides for SEND and
SDTM. For example, when looking at trial design, there are no trial
visits, trial inclusion/exclusion criteria or trial disease assessment
domains but there is a TX domain for trial sets, in which there is the
ability to subdivide arms, or group multiple arms. Figure 1 (see full
PDF) shows the intersection between the domains included in the SDTMIG
and SENDIG standards.
Subject pooling is another key difference,
which does not appear in SDTMIG. In non-clinical studies, it is common
that a single finding may be captured for multiple subjects not just
to a single unique subject, but a pool of them. In SENDIG, pools of
subjects are defined in the POOLDEF domain. POOLID values are unique for
a given set of subjects. Clinical signs for group-housed subjects may
contain cage level findings, for which a particular individual cannot,
or has not, been identified.
For example, the technician may
notice liquid stool in the cage, but did not see which animal produced
the stool. In small animal studies, there may be clinical chemistry
tests scheduled where a single subject may not be able to provide the
volume of blood needed for testing. Therefore, blood from multiple
subjects may be drawn to get the appropriate volume. In some SENDIG
domains (LB, FW, PC and CL), the POOLID may be used instead of the
USUBJID. Those two fields are mutually exclusive, so either the POOLID
or USUBJID is populated in those domains.
A number of domains
are very specific to the nature of toxicology studies. The tumor.xpt
file must be created according to regulatory guidelines for the
submission of carcinogenicity data. Numerous domains are required to
produce this tumor.xpt file, including demographics, disposition,
exposure, microscopic findings, tumour findings and TX.
Other
domains, where there are obvious differences between the two, include
the demographics domain; if the exact age of an animal is unknown, it is
possible to give an age range, creating variability in a domain that is
not available in SDTMIGs demographics domain.
Data Capture
The
other main difference between clinical and non-clinical studies is the
way in which data are captured. In a clinical trial, information is
typically collected using some form of electronic data capture (EDC)
system. All patient data are entered into this system increasingly via
automated processes which makes generating and activating SDTM
datasets relatively straightforward. This is not the case in the
non-clinical world. There is rarely a single EDC system, plus data will
be arriving from different laboratories across multiple sites rather
than being collected at a single, central point. This already adds to
the difficulty of data being presented in multiple different formats,
posing the challenge of getting it all into a single standard form.
The
tools used for formatting transferred datasets in SDTM should also be
applicable to SEND. This will allow datasets suitable for internal
exchange to be created, as well as those in XPT or Dataset-XML for
submission to the regulators. Another compatible tool will allow define
metadata to be generated, and it should be possible to extend validation
checks from SDTM to SEND.
As an example, the process for
mapping, merging and converting source data should be interchangeable
between SDTMIG and SENDIG. These data could include EDC outputs, data
from third parties and raw data files, and it also permits the
conversion process to be scheduled in advance or followed up. Similarly,
if the sponsor requires interim datasets, the transfer process
facilitates their creation, whether from the whole dataset or a subset.
For the Benefit of All
Although
the FDA is already accepting data in SEND format and has been since
2011 there is currently no enforced standardised format for
non-clinical data, and companies are free to submit electronic documents
in a number of specified formats. However, from December 2016, SEND
will become mandatory for FDA submissions making the approval process
more streamlined, with less time being wasted on the process of studying
the documents. Other regulatory agencies such as the EMA and the
Pharmaceuticals and Medical Devices Agency in Japan have expressed
interest in recommending the use of SEND.
Although it is
creating more effort and work for companies submitting the data now, in
the long run, it will save time and money for everyone by streamlining
the data flow between sponsors, CROs and regulatory authorities, while
standardising terminology across the industry and creating opportunities
for harmonised and standardised software tools.
|
Read full article from PDF >>
|
|
 |
|
| Rate this article |
You must be a member of the site to make a vote. |
|
Average rating: |
0 |
| | | | | |
|
|
 |
News and Press Releases |
 |
Tempus Announces Agreement With Janssen R&D to Leverage AI/ML and Real-World Evidence to Enhance the Discovery and Development of Oncology Therapies
Tempus, a leader in artificial intelligence and precision medicine,
today announced the expansion of its multi-year agreement with Janssen
Research & Development, LLC (Janssen).
More info >> |
|
 |
White Papers |
 |
|
Spotlight on Quality in Study Startup
Oracle Health Sciences
This white paper addresses the growing interest in quality in clinical trial execution and how workflows play an essential role by building in the steps needed to comply with TMF guidelines, reducing downstream problems. This proactive strategy limits issues caused by siloes, yielding process improvements measurable by performance metrics.
More info >> |
|
|
