P15 Using SAS© macros to automate data reports in multi-center clinical trials

80S

Abstracts

leads to increased flexibility and responsiveness. If a parameter changes, it requires only a data dictionary value change, not a series of programming modifications.

PI4 MONITORING DATA ERROR RATES: A TOOL FOR MAINTAINING HIGH STANDARDS IN DATA QUALITY Sharon Lawlor, Manual Lombardero Gerald Swanson and Nancy Remaley University of Pittsburgh Pittsburgh~ Pennsylvania Despite stringent data collection, entry, and editing procedures designed to ensure quality data, a number of data errors, inevitably, will exist in most, if not all, databases. The identification and correction of errors occurs on a routine basis during the creation and initialization of a database. When data collection and entry procedures have stabilized, quality control procedures should be in place to concentrate on the detection and correction of existing and new data errors. In addition, monitoring the rate or percentage of errors detected in a regularly scheduled data editing cycle can reveal whether error levels increase markedly relative to previously established norms. The error rate within an editing cycle for each different data collection form can be defined as: #ERRORS IDENTIFIED #RECORDS EDITED * #EDIT CHECKS APPLIED The average error rate for each form can be estimated after a number of data editing cycles and, subsequently, used to determine the expected number of errors in future cycles. A statistical comparison between the expected and the observed number of errors can identify the existence of unusually high levels of erroneous data which will alert data management to the possibility of problematic data collection or entry practices. Prompt remedial attention directed to the problem will assure the highest quality data. The presentation will describe how this error monitoring system is implemented.

PI5 U S I N G SAS © M A C R O S T O A U T O M A T E D A T A R E P O R T S IN M U L T I - C E N T E R C L I N I C A L T R I A L S C h r i s T h o m p s o n , Georgia Saylor, Virginia A n d r e w s a n d Michael E. Miller Bowman Gray School of Medicine Winston-Salem, North Carolina In multi-center clinical trials, each of multiple clinics collect similar data on randomized participants. Sending monthly reports to clinics can insure timely data cleanup and provide clinics with feedback on progress related to recruitment and follow-up. When a trial involves numerous clinics, then generation of these reports on a monthly basis can result in an overwhelming amount of work for coordinating center personnel. The P R E V E N T trial, a placebo-controlled, multi-center investigation of the progression of atherosclerosis, involves 16 clinics, follow-up visits every 3 months, and a total of 3 years of follow-up on participants. In this trial, SAS © macros have been developed and used by the coordinating center to facilitate the production of monthly clinic reports. We describe several unique applications of SAS © macros that permit generation of automated monthly r e p o t s providing feedback to clinics on topics ranging from the percentage

Abstracts

81S

of participants completing specific follow-up visits to an identification of what participants within each center have missing forms. Application of the macro language is well suited to this type of trial where the same variables are measured at multiple points in time, and the processing of data is done in exactly the same manner for each clinic. By changing the definition of parameters passed to the macros, the macros are generally transportable to other clinical trials.

P16 AN AUTOMATED EDITING PROCESS FOR RESEARCH DATABASES Pamela S. Moke and C. Hendricks Brown

Jaeb Center for Health Research Tampa, Florida Editing of clinical trial databases is a complex and time consuming process involving the detection, evaluation, and tracking of possible errors as well as updating datasets with confirmed edits. Not only must the database be updated, but a record of all changes that have been applied to the data must be maintained as part of the trial's documentation. As part of the Longitudinal Optic Neuritis Study (LONS), we developed a UNIXbased editing process that evaluates data received, generates edit messages for the source clinics, and provides a paper trail of changes when updating the existing database. The primary component of the process is a generic C program that generates SAS code based on a given study's edit criteria. The resulting SAS program code then outputs data that are processed with nroff/troff tormats and UNIX tools to create a uniquely numbered, traceable, query message tor the source clinic. Specifications tor the program are listed in a simple ASCII text file and can include the standard editing steps (valid values, range checks, etc.) as well as conditional statements (e.g., comparison against a master database), edit message information (date, form number, variable label, variable location on the form, etc.), and specific SAS code to be executed during the editing procedures. We have now extended this editing process to other studies and will present examples of program input and output including edit specifications, query messages, and update reports. The program is adaptable for use by interested parties. P17 ISSUES IN USING A U N I V E R S I T Y DATA C O O R D I N A T I N G C E N T E R F O R AN I N D U S T R Y S P O N S O R E D C L I N I C A L T R I A L

Murray D. Barnhart

Bristol-Myers Squibb Company Prh~ceton, New Jersey. The Bristol-Myers Squibb Company has sponsored four large clinical trials where universities acted as the Data Coordinating Centers. Issues involving data management were common during the conduct and reporting of these trials, often because of the need to preserve both the scientific and regulatory integrity of the study. Two of the trials were conducted outside of the United States and added distinct challenges to the process. Data Management topics to be discussed include processing serious adverse reactions, compatibility of data with other studies conducted by the Sponsor, availability of blinded data to the Sponsor prior to the end of the trial and logistics in the preparation of registrational dossiers.