How to create a summary of findings table with GRADEPro

GRADE provides a structure for clinicians and healthcare analysts to understand the quality of evidence for an outcome. An outcome can be associated with a number of studies, and quality appraisal in terms of whether additional evidence must be needed before a definitive conclusion about an intervention-outcome can be arrived at is done on the basis of conducting GRADE tables. GRADEpro as a software enables construction of grade tables, and has two parts. In part one of this series, we showed how to construct certainty of evidence tables. In this part, we discuss how to create summary of findings tables with GRADEpro for dichotomous outcome variables.


Introduction
In the previous article in this series, we presented a step by step process of entering data to GRADEpro to appraise a body of evidence. There, we have shown that you can enter data to GRADEpro and conduct an appraisal of the certainty. In evidence based health and in GRADE process (GRADE is a shorthand form of Grading Recommendations for Appraisals and Development of Evaluations), the focus is on using outcomes as a first principles of critically appraising evidence [1] .
In appraising the quality of evidence, a study can include a number of outcomes; likewise an outcome can be studied by a number of studies. Hence, the focus is on specific outcomes that we would like to study.
In that process, here, we are going to focus on how to use GRADEpro to correctly write a summary of findings table. The summary of findings table is the second part of the evidence portfolio; the first part of evidence portfolio is the uncertainty assessment where for each outcome and outcome measurement, we assessed what was the study design, what biases were identified, whether the studies were consistent in what they found, and other issues. In the end, we came to a conclusion whether we would need more information about the association between the intervention and outcomes so that more studies would be necessary or whether the current body of studies would be sufficient in themselves to arrive at a conclusion whether an intervention I works to achieve an outcome O.
The summary of findings here help to quantify the association we observe. As before, let's start with an article and work step by step.
Here we have selected an article by Philip Tonneson et.al. (2012) where they evaulated the efficacy of nicotine mouth spray for smoking cessation in a randomised double blind trial [2] . Obtain the full text of this article and follow along.
They measured the outcome as "Continuous abstinence" How was that outcome measured?
From the article, we get to see that Continuous abstinence was defined as self-reported ''no smoking'', verified by a CO value ,10 ppm, from week 2 up to and including the given visit. Any subject who missed the visit(s) at week(s) 8, 16 and/or 20, or for some other reason had missing CO value(s) at one or more of these visits, was not regarded a treatment failure if the subject was verified continuously abstinent at a later visit.
(See pp 550, in the Statistics section) This should tell you that "continuous absence" is a binary variable.

And why so?
Think of yourself as the study investigator. If you asked a participant whether she maintained "no-smoking" status and the participant replied, "yes", and then you verified that the participant was indeed "non-smoking" status using the CO value, then you would either (1) label that person as a non-smoker or (2) if the participant lied and failed the CO test, you'd label that person as "smoker".
There are TWO choices, hence this is a dichotomous variable ("binary" means two).
The intervention was "Nicotine mouth spray".
How many people were in the intervention arm?
How many people were in the placebo arm?
See Table 2, row 1 (the row beginning with "Subjects n"), column 3, N = 161 We will input these numbers in GRADEpro table. You will find all data in Table 4  In this article, we have discussed how to code it if it's binary and presents data from a randomized controlled trial. In subsequent papers we will expand the scope to what happens when we have other study designs including metaanalyses If it's binary, as it was in this case, then pay attention to how many people were in each of treatment and control arm (based on initial enrolment), and how many people experienced the outcome at the LAST or FINAL assessment.
Identify the appropriate numbers and plug in those numbers to the GRADEpro summary of findings tables.