Skip Navigation Bar
 

Activities To-Date, Mockups, and Proposed Data Entry Process

Session 3: Basic Results Database

Deborah A. Zarin, M.D.

Background

Specific Results Data Elements

Data Entry and Display Mockups

Timeline showing that the Basic Results Database is to be launched within 1 year of enactment of Public Law 110-85 (i.e., by September 27, 2008)

Activities to Date

Results Reporting

Results Reporting: Preliminary Activities

  • Trans-NIH Working Group on Clinical Trial Results Reporting in 2004
  • "Scientific Challenges in the Registration of Clinical Trials" Expert Meeting in November 2006
  • Commissioned review of existing results databases

Conclusions from Preliminary Activities

  • Two basic models: structured narrative vs. tabular
  • Data input should be formatted in a way that facilitates provision of comprehensive and valid data
  • Tabular data (vs. narrative) would be easier to curate

Activities since Sept. 2007

  • FDAAA mandated "tabular form" for "basic results"
  • Specific content of tables identified in law
  • Challenge is to determine method for data entry that would facilitate provision of structured data for
    • Heterogeneous trial types
    • Heterogeneous variable types

Basic Results Tables

  • Demographic and baseline characteristics
    • Overall and by arm
  • Participant flow
  • Values for primary and secondary outcomes
    • By arm
    • Scientifically appropriate tests of statistical significance
Timeline for development of basic results data entry and data display systems. December 2007 – begin development of basic data entry table shells. April 2008 – freeze requirements for data entry system; begin data display system. May 2008 – XML DTD (alpha) for data entry system. June 2008 – Data entry prototype; Data display mock-up. July 2008 – XML DTD (beta ) for data entry system. August 2008 – Beta data entry system; Beta data display system. September 2008 – XML DTD (final) for data entry system. September 27, 2008 – Production data entry and data display systems

Current NLM Thinking

  • Developing a model that would allow the registrant to enter information about the trial and the variables, and then enter data into tables
  • Professional Services Contracts at NEMC and OHSU
    • Evaluate templates using 40 clinical trials, starting with parallel design intervention studies
    • Process of iterative revisions
    • Process for public comment and review

Results Database Mockups

Proposed Approach

Data in Mock-ups Based on Published Results

Image of title page from journal article reporting on a trial that was used in the mock-ups of the results database:  JE Richter, et al, "Efficacy and Safety of Esomeprazole Compared With Omeprazole in GERD Patients With Erosive Esophagitis: A Randomized Controlled Trial," American Journal of Gastroenterology, 2001.

Results Record

Background

Image of Mocked-Up Results Record in ClinicalTrials.gov showing the title of the study, sponsor, information provider, NCT number, study purpose, study design, study start date, and study completion date.

Published Baseline Data

Image of Table 2 from the referenced publication, summarizing demographic and baseline characteristics of patients evaluable for efficacy, by arm of the trial (Esomeprazole, Omeprazole).

Results Record

Baseline Demographics

 EsomeprazoleOmeprazole
N-1,216N=1,209
Age    
Mean Age (SD): 46.5 (12.5) 46.8 (12.3)
Number >= 65 Years: 108 121
Number < 65 Years: 1,108 1,088
Gender    
Male: 494 449
Female: 722 760
Ethnicity    
White: 1,134 1,133
Other: 82 76

Results Record

Baseline Demographics

Image illustrating linking from the ClinicalTrials.gov results record to structured product label information in NLM’s DailyMed database for the drugs studied in the trial.

Results Record

Baseline Clinical Characteristics

  EsomeprazoleOmeprazole
N-1,216N=1,209
Positive Biopsy for H. pylori    
Number with Characteristic 90 96
Number without Characteristic 1,126 1,113
Los Angeles Classification    
Grade A: 427 386
Grade B: 470 502
Grade C: 257 240
Grade D: 60 80
Missing 2 1
History of GERD    

Results Record

Baseline Clinical Characteristics

Illustration of link from Baseline Clinical Characteristics table to background information on the Los Angeles Classification.

Published Participant Flow

A CONSORT diagram showing patient randomization and disposition: Number screened, Number randomized/Not randomized, Number allocated to arm 1 (Esomeprazole) and arm 2 (Omeprazole), Number completing treatment (by arm), Number not completed treatment (by arm) and reasons for not completing

Results Record Participant Flow

Table shows how participant flow information will be captured in the results database. Information is provided on target enrollment, actual enrollment, number completing, and the number discontinued because they dropped out or were excluded. Data are collected by arm.

Published Primary Outcome

Figure illustrates a graphical (bar chart) representation of the outcome of the clinical trial as published in a journal article. Results for patients in arm 1 and arm 2 are shown at two different points in time (week 4 and week 8).

Results Record

Primary Outcome

Tabular representation of the primary outcome showing the outcome measure (healing at 8 weeks) and the number with the event and without the event, by arm. The table also includes a note indicating the statistical significance of the result and the statistical test used (e.g., log-rank test).

Proposed Data Entry Process

Data Entry Mechanisms

  • Electronic Data Interchange – upload data in XML format
  • Manual Data Entry – enter data using a Web-based Results Registration System (RRS)
    • analogous to the Protocol Registration System (PRS) for the registry

Two-Step Data Entry Process

  1. Specify "parameters" of results tables
    • Baseline Demographic Categories
    • Baseline Clinical Characteristics
    • Detailed Outcome Measures
  2. Enter data

Measurement Types

Representative table from a journal article showing the different types of variables needed to collect baseline demographics and clinical characteristics of randomized patients: Dichotomous variables (e.g., male, female), Continuous variables (e.g., years of education), Multiple categories (e.g., married, previously married, never married)

Measurement Types (continued)

Graph illustrating another type of variable to be collected: time-to-event variables (or survival curves).

Basic Results Data Entry System

  • Prepopulates certain data elements from registry, if provided (e.g., Arm Description)
  • Lists some prespecified baseline demographics (e.g., gender, ethnicity)
  • Allows users to enter and edit tables
    • Column labels: e.g., Arms
    • Row labels: e.g., demographics, outcome measures
    • Cells: e.g., measurement values

Specifying Parameters

Arms & Baseline Demographics

Screenshot of data entry system for information on arms and baseline demographics. Some information is initially prepopulated from the registry (e.g., the names of the two arms of the study). Other prespecified demographic categories will also be indicated (e.g., Age, Gender, and Ethnicity). Users will be able to edit these categories and add others to reflect the categories that are most important in their study.

Entering Data

Baseline Demographics

Screenshot illustrating the entry of baseline demographic information into the results table. Data is provided for each arm of the study, as well as in total. For age as a continuous variable, the user may enter the mean and the standard deviation.

Specifying Parameters

Baseline Clinical Characteristics

Screenshot of page for entering Baseline Clinical Characteristics.  Users specify the characteristics of interest to the study. Characteristics can be added and edited.

Specifying Parameters

Baseline Clinical Characteristics

Screenshot showing pop-up window for entering a Baseline Clinical Caracteristic. The user is prompted to provide the name of the characteristic (in this case “Positive Biopsy for H. pylori”) and to select a measure type (e.g., continuous, categorical, or time to event) and unit of measurement.

Specifying Parameters

Baseline Clinical Characteristics

Screenshot showing the addition of a second Baseline Clinical Characteristic:  Los Angeles Classification.

Specifying Parameters

Outcome Measures

Screenshot of page used to create the tables of Primary Outcome Measures and Secondary Outcome Measures. Names of the measures are initially pre-populated from the registry. The user may update or modify outcome measures using the Edit and Add functions.

Entering Data

Outcome Measures

Screenshot of page used to enter results data. The example shown is for "Healed Esophagitis at Week 8."  User is prompted to input the number of subjects analyzed, the number with event, and the number without event for each arm of the study and in total. The user is prompted to provide information on the statistical test used for the analysis and the results of the statistical analysis (value, 95% confidence interval and p-value in the example).

Data Display Mechanisms

  • Summary result records displayed in a Results Database at ClinicalTrials.gov
  • Results records linked to registry records via NCT number
  • Similar search engine and "look and feel"
  • Details to be determined

Results Record Sections

  • Background: e.g.,Title, NCT, Sponsor
  • Baseline Data (Table 1)
    • Baseline Demographics by Arms
    • Baseline Clinical Characteristics by Arms
    • Participant Flow, including dropout and exclusion
  • Outcomes (Table 2), including statistical tests
    • Primary Outcomes by Arms
    • Secondary Outcomes by Arms
  • Administrative: e.g., Point of Contact, Certain Agreements

Results Record

Background

Image of Mocked-Up Results Record in ClinicalTrials.gov showing the title of the study, sponsor, information provider, NCT number, study purpose, study design, study start date, and study completion date.

Results Record

Baseline Demographics

Image illustrating linking from the ClinicalTrials.gov results record to structured product label information in NLM’s DailyMed database for the drugs studied in the trial.

Results Record

Baseline Clinical Characteristics

Image illustrating linking from the ClinicalTrials.gov results record to background information in NLM’s consumer-oriented MedlinePlus system, in this case for information on GERD (gastroesophageal reflux disease).

Results Record

Participant Flow

 TotalEsomeprazoleOmeprazole
Target Enrollment: 2,080 1,040 1,040
Actual Enrollment: 2,425 1,216 1,209
Completed: 2,316 1,161 1,155
Discontinued 109    
Dropped Out: 88 45 43
Excluded: 21 10 11

Results Record

Primary Outcome

How to Facilitate Rational Use of Data

  • Provide context
    • Study design
    • Background information
  • Consider linking to systematic reviews
  • Link to other resources

Some Other Issues to Ponder

  • What constitutes a “minimally acceptable” results record?
  • How to handle disputes over validity of information?
  • Should post-hoc analyses be allowed on database?
  • What external sources should we link to?