Principles and Practice of Pharmaceutical Medicine

Manual code the ‘no-hit’ terms and update the
glossary.

Run computer edit checks (Table 26.5) and gen-
erate queries. Reconcile the discrepancies
between the queries generated by manual review
and computer edit check. Issue the queries
(Table 26.6) to the investigators.

Query resolution and database update

When the answers to the queries are returned, the
CDC updates the database and CRF images. This is
a continuous process during the course of the
clinical trials.

Create test datasets for various
analysis population

To make theadequate inferences of the efficacyand
safety of the study drug, Federal Register (1996)
should be followed: in Section 11.1, ‘Data Sets
Analyzed’, it states:

Exactly which patients were included in each efficacy

analysis should be precisely defined, e.g. all patients

receiving any test drugs/investigational products; all

patients with any efficacy observations or with a cer-

tain minimum number of observations; only patients

completing the trial; all patients with an observation

during a particular time window; or only patients with

a specified degree of compliance. It should be clear, if

not defined in the study protocol, when (relative to

unblinding of the study) and how inclusion/exclusion

criteria for the datasets analyzed were developed.

Generally, even if the applicant’s proposed primary

analysis is based on a reduced subset of the patients

withdata,there should also be, for any trial intendedto

establish the efficacy, an additional analysis using all

randomized (or otherwise entered) patients with any

on-treatment data...A diagram showing the relation-

ship between the entire sample and any other analysis

groups should be provided.

Therefore, an algorithm has to be developed to
precisely define how each analysis population of

the dataset is defined. For example, there are at

least four analysis population datasets, for example

the intent-to-treat (ITT) population, the per-proto-

col (PP) population, the safety population and the

microbiological population, as indicated in the

diagram (Figure 26.2). During the derivation of

various analysis populations, it may be necessary

to issue new queries and update the database, based

on the resolution of the queries.

Status reporting (Table 26.7)

1. From the image files and the database, a weekly
   production report is generated.


2. A cross-check of milestones and progress
   achieved should be made and the status reported
   to the department heads for review and action.


3. The department heads may adjust the resources,
   depending on the progress report.

Create audit sheet for audit

The computer-generated audit sheet (Table 26.8)

should be formated in the same sequence as the

fields in the CRFs. This will enhance the speed of

the audit task. The QA auditor will check the audit

sheet against the CRF images.

Issue interim audit document

and audit summary

When 10% of the CRFs have been scanned and

entered into the computer, the interim audit

should be conducted in order to tune up the

CDC review manual and edit-check programs.

Findings regarding the quality of the database

should be given to the head of the data manage-

ment group for possible action. The audit docu-

ment (Table 26.9) should include patient ID, the

initials of the keyer and verifier, CDC, editing

programmer, type of audit, number of errors,

description of errors. It is a tool to find out

which records tend to produce more errors and

26.4 DATA PROCESSING 351