GOfER©

Graphical Overview for Evidence Reviews

USER GUIDE

Developed by the Peninsula Technology Assessment Group

http://sites.pcmd.ac.uk/pentag/

© 2011

CONTENTS

3

3 CONTENTS

4 1 – INTRODUCTION

5 2 – INSTALLATION5 2.1 System Requirements5 2.2 Installation

6 3 – GOfER WORKFLOW6 3.1 GOfER output6 3.2 The data viewer7 3.3 Adding and removing rows and columns7 3.4 Naming modules7 3.5 Selecting module types

8 4 – GOfER MODULES8 4.1 Module 1: ‘text’9 4.2 Module 2: ‘range’

10 4.3 Module 3: ‘design’13 4.4 Module 4: ‘proportion’14 4.5 Module 5: ‘count’15 4.6 Module 6: ‘features’16 4.7 Module 7: ‘outcomes’17 4.8 Footnotes

18 5 – THE OUTPUT WINDOW

1 – INTRODUCTION

4

Systematic reviews of evidence are a vital foundation of health policy decisions. They are used to show an overview of the empirical research that has been conducted around a health intervention. However, with a growing number of clinical trials and systematic re-views published every day, the volume of data collected in some reviews can be intimidating.

It is perhaps not surprising that in such situa-tions, graphical techniques have been used to present this data. The researchers that pro-duce systematic reviews have used visual dis-plays from the standard bar and line charts, to the techniques developed specifically with evidence review in mind, such as forest plots.

However, complex calculations are required to determine the weightings used for the differ-ent trials in forest plots. The GOfER display is designed to give a visual overview of key data on each trial, such as size, design, follow-up,

participant characteristics, study quality, and outcomes used. This allows the viewer to quickly understand the reasons for particular study weightings in a forest plot, but can also be used where meta-analysis is not possi-ble - such as where a large range of outcome measures prevent studies from being pooled.

At PenTAG, we have been able to use visual design and page layout software to create some quite sophisticated and data-rich GOf-ER displays. However, we realised that com-mercial design software is not always avail-able to those conducting evidence reviews. Therefore, we have developed this software, in the hope that it will enable researchers to construct their own GOfER displays for inclu-sion in reports.

We hope you enjoy using it, but we are always looking for ways to improve it. If you have any feedback, please use the contact form here:

sites.pcmd.ac.uk/pentag/contact.php?section=form

—Will Stahl-Timmins GOfER designer

2 – INSTALLATION

5

2.1 System Requirements

xxxxxx

2.2 Installation

xxxxxx

3 – GOfER WORKFLOW

6

3.1 GOfER outputThe main purpose of GOfER is to visually present key data on all of the trials included in a review. The skill in using GOfER is in selecting exactly which data to present, to give an overview of the quality and quantity of the evidence in the review. If GOfER is to be included in a printed document, or online PDF of a particular size, it is usually best to construct your GOfER so that all the informa-tion on a trial fits on the width of one page (whether portrait or landscape). The trials are then displayed running from top to bottom of the page, with one horizontal “row” of the display for each trial. (see sample output at the top of this page) To fit on a single page, this often means that all that can be shown is the author, date, design, size, follow-up, a couple of characteristics, the quality check-list, and the outcomes measured. However, it may also be important to show the location of the study, or the number of centres in which it was conducted. The GOfER software allows you to choose what will give the best idea of the quality of the studies in the particular context of the review.

3.2 The data viewerThe main window used while working with GOfER is called the “data editor” - shown below. GOfER creates visual displays from numerical data, but this must be entered into the data editor in a particular way.

The GOfER display represents information with a series of “modules” - each of which creates a visual display from a particular set of data. The data for each module is entered in one (or more) of the columns in the data editor. The data editor is similar in many ways to most spreadsheet software.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

7

3.3 Adding and removing rows and columnsRows (representing studies in the review) can be added by clicking the right mouse button on the row numbers at the left of the window (as shown to the right). Similarly, columns (which can be used for different modules) can be added or removed by right-clicking on the grey area at the top of the window.

3.4 Naming modulesThe first row of the data editor is called “Data Title”. This is a free text input field. Anything entered here will appear on the output as a title for a particular module. This field may be left blank, in which case nothing will appear in the output.

3.5 Selecting module typesThe second row of the data editor is called “Data Type”. This is where the module to be used can be selected, from a drop-down list. The seven different module types will be dis-cussed in the next section.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

4 – GOfER MODULES

8

4.1 Module 1: ‘text’The text module simply places text in the centre of each trial’s row on the output. Where text should be separated into different arms of a trial, insert a comma between each arm. If you would like a comma to appear on the output, place a / before it, like so:Paracetamol/, 250mg

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

data entry output

9

4.2 Module 2: ‘range’The range module shows key values from a distribution. In the example below, the module is showing information on the age of the people in the trials.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

data entry output

Different trial reports may present different key values when describing distributions of patient characteristics. The values that can be presented as a part of this GOfER module shown below.

value

minimum and maximum

mean

median

standard deviation of the mean

upper and lower quartiles

min, max, med and quartiles (box-and-whisker plot)

mean and standard deviation

outputdata format

30-55

39

[39]

(10)

34[ ]54

30-34[39]48-55

39(10)

10

4.3 Module 3: ‘design’The design module is probably the most important past of most GOfER displays. It shows the quantity of people in a trial, and how they are split into groups (arms) - in what quantities, and over what length of time. The groups can be labelled with surrounding text modules, as shown below.

Once a design module is used, almost all other modules can be split into arms using commas. The output for each arm should then align horizontally with the arms in the design module.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

FeaturesAt

tAs

sSe

l

Outcomes (auditory)1S

T

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22

1 This wasn't clear.

data entry output

The design module can represent many different kinds of trial. Some examples are shown below:

data entry output notes

N=50Control N=28

500mg N=22

Regarded as a “gold standard” in medical trials, a randomised control-led trial (RCT) can be specified by including a single number in the first design column, and two or more numbers separated by a comma in the second column. Numerical values can (and should) be included using text columns on either side.

Randomised controlled trial (RCT)

Where RCTs are not available, survey and non-randomised trials can be specified by using a single column in the data editor.

Survey and non-randomiseddesigns

N=23

N=27

N=50

11

data entry output notes

In some cases, a review may include pre/post trials, where the same participants are measured before and after an intervention. These can be specified by adding a group with 0 people. Data on the cohort before and after intervention can then be added on two separate rows, similar to RCTs.

Pre/post design

N=50

A pre-post design can also be given a control group. In this example, three participant groups can be displayed, each horizontally aligned with the pre-intervention, post-intervention and control groups.

Note the double comma in the text field labelling the patient flow. This inserts a blank label for the second group.

Pre/post with control

control N=30

intervention N=50

N=50Control N=28

500mg N=22

Follow-up of participants over time can also be displayed. Enter your unit of time (normally months or years) in the “config A” box. The length of time for each section of the trial can be specified with a number in brackets following the group sizes. This RCT has a 6 month follow-up, and the survey has 12.

Follow-up

N=50

DesignFollow-up (years)01

The participant flow module can also show how many participants were lost to follow-up (for example between stages of a pre/post trial, or after randomisation in an RCT). In this RCT example, the pre-randomisation cohort size is entered first, and the size of the groups that were meas-ured after intervention is placed in the next column. The number of people not measured in the randomised groups are shown branching out from the study.

Including dropout

N=50 N=18N=22

12

data entry output notes

Crossovers (where participants are taken out of one group and added to another) can be specified by using extra data columns. The participant flow module automatically adds crossover in the case of simple 2-arm trial designs.

Crossover

In some cases, it may be necessary to specify exactly how many people come from a particular arm of a trial into another. For example, in this case, the first trial has three arms and individual crossover rates must be specified. The second exampleis a multi-phase trial with built-in crossover.

For the multi-phase trial, the number in square brackets represents which phase 1 group the participants come from. For the crossover in the three arm RCT, the numbers in square brackets are the group numbers that participants com from, followed by an equals sign, and then the number of participants that come from that arm.

Crossover (advanced)

N=50N=18

N=32

N=60

N=50

Follow-up (months)01 22 4

Some quite complex trials can be represented, including multi-phase trials, with crossover, dropout and phases of different lengths, .

More complex trials

N=150

Follow-up (months)06 12

Measurement points can be inserted using a separate sub-column in the design section - select “time points” from the drop-down list. While these points do not indicate what was measured, they can still give an overview of how frequently outcomes were meas-ured in a trial.

Measurement points

N=50N=28

N=22NOT IMPLEMENTED

IN THIS VERSION

13

4.4 Module 4: ‘proportion’The proportion module is used for showing the percentage of people with a particular characteristic. This could be the presence of a particular gene, or as in the example below, whether they are male or female.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

data entry output

As with most modules, the characteristics of people in multiple arms can be shown by sepa-rating the data with commas. Labels for the left and right hand groups can be included in the Config A field, separated with / marks.

14

4.5 Module 5: ‘count’The fourth module, count, simply shows the number of something used. In the sample below, it is the number of centres at which the trial was conducted, but it could also be used to show the number of countries in which these centres were located, for example

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

data entry output

A ? symbol can be used where information is not reported clearly.

15

4.6 Module 6: ‘features’The features module is intended to display study quality checklists commonly used in systemat-ic reviews. However, if used in a public health context, they might also be used to show whether certain features of an intervention were present or not

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

?

data entry

outputFor this module, the ‘Config A’ row is used to categorise the features. It is recommended that category names should be shortened to just a few characters, so that the subtitle row of the output does not become too large. In the example above, Selection bias, Assessment bias and Attrition bias are shortened to Sel, Ass and Att. If a ‘Config A’ cell is left blank, that column will be part of the previous category, appearing below the last column on the output.

The ‘Config B’ row is used purely to remind the reviewer which column is which, and the text here does not appear on the output.

There are five possible values that can be used in each feature cell (shown to the right). When used for quality checklists in a systematic review, the intention is that the more black in the grid, the higher the quality of the trial.

value

positive

negative

partial

unreported/unclear

not applicable

outputdata format

+

0

-

0

(leave blank)NOT IMPLEMENTED IN THIS VERSION

16

4.7 Module 7: ‘outcomes’The outcomes module allows an overview of all the different outcomes measured by the trials in a review.

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

1

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

?

data entry

outputThe ‘Config A’ row is used to show which outcomes are represented by each column. To categorise these outcomes, separate outcomes modules should be used, so that each one has a different title, as shown above and left.

Outcomes which have different significance in different arms can be shown by separating values with commas.

There are four possible values that can be used in each outcome cell (shown below).

value

statistically significant outcome

non-statistically significant outcome

negative significant outomce (favours placebo)

not measured

outputdata format

1

0

-1

(leave blank)

17

4.8 FootnotesA footnote can be included in any cell of the GOfER data entry window, using an asterisk (*) symbol.

The footnotes will appear at the bottom of the page, automatically numbered.

?

Title

Able Baker et al. 2010

Dogg's Hamlet 2009

Echo Bazaar 2006

Gender

M F

M F

M F

M F

Ages

20 30 40 50 60 70

Sites

?

Features

Att

Ass

Sel

Outcomes (auditory)

1ST

2ST

CAP

MM

S

CR

ISP

IPSy

n

Outcomes (other)

APH

AB

AQoL

N=50

N=150

N=25

Design

Control N=28

500mg N=22foo

bar

cube

1 This wasn't clear.

1

data entry output

18

5 – THE OUTPUT WINDOWSelecting “Generate Output...” from the “File” menu, or clicking the right-hand facing arrow icon in the GOfER window brings up the “Generate Diagram” window:

This window allows you to specify a paper size for the output, and a number of rows (trials) to be included on each page. Finally, this window allows you to choose a location to save this out-put. To change from the default location, click “Choose file”. Clicking the “OK” button will save your GOfER diagram to the specified location.

The GOfER software outputs to a PDF file type, which can be scaled to any size without los-ing print quality. To import this into the Microsoft Word program, using the 2007 or 2010 edi-tion choose Insert > Object (in the “text” group). Click “Create from File, click “Browse”, then browse to your GOfER output file, and click “insert”. Click “OK” to embed the PDF file as an image, which you can move or resize as needed.