Return to site

rapidGRADE: An easy to use evidence grading system for any test or therapy

Oliver Zolman MD

Introducing rapidGRADE

GRADE (Grading quality of evidence and strength of recommendations) is the gold standard for evaluating how effective a treatment is.

However, it seems too complex for many people to use, especially non-professional scientists, and can be time intensive to do. So I invented rapidGRADE.

rapidGRADE is a faster, simplified version of doing GRADE that anyone can do with minimal training.

It also extends the original GRADE to safety (rapidGRADE SAFETY) and diagnostic test accuracy (rapidGRADE TEST ACCURACY), on top of the rapidGRADE EFFICACY

rapidGRADE can be done with minimal training such that even journalists, bloggers/vloggers, patients with no scientific background can use it, or time strapped clinicians, students or trainees can do it .

Each of rapidGRADE EFFICACY, rapidGRADE SAFETY, rapidGRADE TEST ACCURACY are scored out of 5, with 5 being the best score for efficacy, safety or accuracy.

rapidGRADE can only be applied to specific outcomes and populations 

Note, like GRADE, for different outcomes and different subpopulations you need to calculate a separate rapidGRADE score.

So for example you would need to calculate a rapidGRADE separately for each of the following

Metformin for diabetics preventing stroke

Metformin for diabetics preventing raise in HbA1c

Metformin for pre-diabetics preventing raise in HbA1c

Metformin in non-diabetic, non-prediabetics for preventing stroke

rapidGRADEs MUST be only applied to a specific population and outcome!


Effectiveness score out of 5

1/5 = animal data effectiveness

2/5 = any non randomised peer reviewed human data effectiveness

3/5 = small effectiveness in randomised human data (0.2 - 0.5 standard deviation (SD) difference)*

4/5 = as above, but medium effect size (e.g. 0.5 - 0.8 SD difference)*

5/5 = as above, but large effect size (e.g. > 0.8 SD difference)*

effect size can be measured by cohens D or hedges G, basically its just how many standard deviations the therapy changed the outcome from the control group, this is a simple equation done in excel - see this video on cohens D

* effect sizes must exceed a minimum clinically significant difference (MCSD) otherwise they are scored at 2/5... i.e. can have a large statistically significant standard deviation increase that reduces systolic blood pressure by 1 mmHg, which is clinically insignificant as the MCSD is 2 mmHg


Diagnostic accuracy score out of 5

1/5 = animal data accuracy

2/5 = any non randomised peer reviewed human data

3/5 = over 0.7 correlation against gold standard index test (e.g. ROC curve score or R correlation coefficients) accuracy in randomised human data)

4/5 = over 0.8

5/5 = over 0.9

So whoop watch would score 5/5 for nighttime deep sleep 5 minute rMSSD HRV based on their radnomised diagnostic test accuracy peer reviewed study

but it would only score 4/5 for deep sleep

but 5/5 for REM sleep

whereas Oura ring would score 0/5 for WASO (wake after sleep onset)


Safety score out of 5

For all data randomised controlled trial side effect data should be searched for systematically first, and then Phase 4 data and case reports, SIDER EMBL and pubmed Clinical Queries are good resources for RCT data; MHRA/FDA databases, (USA VPN required), patients like me, any search in pubmed are good for case reports; normal Google, and Google Scholar or EMBASE/ CINAHL/ PSYCHINFO/ Chinese Databases is good for more extensive searching

1/5 = Only animal evidence of non-life long relevant animal model safety data

2/5 = Only animal safety evidence of: Near life long / full life long mammal animal safety data AND safety data in relevant 2D or ideally 3D human cell/tissue/organ models

2/5 = good evidence of common serious organ damage side effects (>2% rate) in humans, at any followup