Abstracts 115s

P39 A QUALITY CONTROL METHOD FOR SINGLE-DATA ENTRY

USING REMOTE DATA ENTRY SYSTEMS IN LARGE COMPASSIONATE CLINICAL TRIALS

Inder J. Sharma and Ronald W. Helms Cato Research

Durham, North Carolina

Traditionally, in clinical trials, double-data entry procedures are used to ensure that transcription of data from case report forms or source records has been captured electronically, error-free or with a miniium acceptable error rate. For pivotal trials, it is highly justified to use thii approach to ensure hiih quality of data. However, for large compassionate, non-pivotal trials using remote data entry, cost and time devoted to data entry can be minimized without significantly compromising the quality of data.

A quality control approach using acceptance sampling methodology is proposed to control the error rate for single data entry utilizing remote data entry systems in large compassionate, non-pivotal clinical trials. The objective of this quality control procedure is to help ensure that the data entry and error detection and correction procedures are functioning in such a manner as to produce data with an error rate below a specified “maximum allowable error rate”, say at approximately 0.001 errors per field or one error per thousand fields.

Quality control batch sampling methods are used for comparison of single-entry of data with source records. A decision criteria to accept completely or conditionally or reject a batch will be presented utiliiing confidence limits approach on error rate.

P40 THE DESIGN OF A WINDOWS-BASED CAT1 SOFTWARE SYSTEM

Robert B. Goldstein, Cheryl Caswell, John Lee, Azii Mboya, Brian Schuth, Maureen O’Keefe and Susan Staples-Smith

New England Research lnstites Watertown, Massachusetts

Computer Aided Telephone Interviewing (CATI) is an essential tool for cost effective recruitment screening of patients for clinical trials and for the conduct of epidemiological studies. We examined existing CATI systems such as CASES, Sawtooth, Blaise and QPL, identifying the pros and cons of each. In order to identify the unique requirements of a CAT1 environment we conducted focus groups composed of Interviewers, Field Supervisors, Research Assistants, Statisticians and Programmers. Comprehensive lists of features were given to the participants and the features were ranked in order of importance.

Using the results of the focus groups as a guide, we designed a Windows-based CAT1 system. The system components include Programmer, Interviewer and Manager interfaces. The Interviewer interface consists of question screens with controls to display respondent summary information, interview status, interview statistics and context- sensitive help. The Manager interface consists of call scheduling, contact records and report specXwations. The Programmer interface consists of data structure definition screens, Pascal-like language extensions and a methodology for specifying questionnaire flow control.

We designed a sample interview containing representative selections of different structural types of questions. We developed a functional prototype of the Interviewer

interface that implemented this sample questionnaire. A computer running the prototype will be available for attendees to use at the poster session.

P41 WEEDING OUT BAD APPLES: MONITORING DATA QUALITY IN RCTs

M.A. FIann, A. Tinazzi and V. Torri lstito di Ricerche Famacologiche “‘Mario Negri”

Milano, Italy

A great deal of organizational efforts are required in the conduct of multicenter RCTs. The involvement of a large number of centres implies the need for a coordinating headquarter, able to collect the relevant patient information and provide timely feedback to each participating institution on the development of the study. The monitoring of the quality of data provided by the participating centres represents a key aspect in the coordination of mRCTs, with staff dedicated to routinely checking the consistency as well as the completeness of the information related to patient characteristics, vital status, etc. Even though in general, within the framework of large cooperative trials, the organizational commitment is to focus only on relevant items of information, such an effort is resource and time consuming. Thus, it would be highly desirable to make it more efficient by targeting the monitoring system on centres more likely to be not compliant with the expected quality standard. That in turn would require the ability to identify some key centres’ characteristics (hospital size, volume of patients enrolled into the study, research or university affiliation etc.) related to the quality of their participation, expressed both in terms of completeness of information and in degree of compliance with the study protocol.

Using data from three ongoing national multicentre RCTs on lung, colon and ovarian cancer treatment, we explored if, and to what extent, the quality of centres participation (operationally defined as proportion of relevant information items actually provided over the overall amount requested) is related to the above mentioned hospital characteristics. In addition, the predictivity of individual centres performance on some specific items of information (rate of patient forms returned to the coordinating centre, rate of missing data on prognostic factors, rate of compliance to the protocol) has been also analysed.

P42 LOGISTICS OF INTERIM MORTALITY ASCERTAINMENT

Sharon W. Crow, Nancy W. Fishman, AIIan D. Rosen and Regina M. Hard&on University of Pittsburgh

Pittsburgh, Pennsylvania

Mortality is the primary outcome in the Bypass AngiopIasty Revascularization Investigation (BARI) Study, a multicenter clinical trial comparing safety and efficacy of coronary angioplasty and bypass surgery. Eighteen clinical centers randomized 1829 patients. For an interim report of five year mortality, the clinical sites obtained vital status on every BAR1 participant as of June 5, 1995. This was necessary to avoid bias caused by a longer time Iag to report deaths among angioplasty patients.

A supplemental protocol and form were designed for vital status ascertainment (VSA). The Coordinating Center assisted the sites in several ways: a checklist of patient IDS was provided, comphance was monitored weekly, and follow-up calls to low comphance sites were made. Equifax, a private database search company, performed a deathonly search