Outcome based GRADE appraisal

class: center, middle, inverse, title-slide

# Outcome based GRADE appraisal
## What is it?
### Arindam Basu
### (updated: 2021-09-12)

---

class: center, middle
# Discuss Evidence, outcomes, and quality appraisal

---

# What we have covered so far
.pull-left[
<div id="htmlwidget-58b7907114dd2e79bac2" style="width:504px;height:504px;" class="DiagrammeR html-widget"></div>
<script type="application/json" data-for="htmlwidget-58b7907114dd2e79bac2">{"x":{"diagram":"\n        graph TD\n          P[PICO question]-->Search\n          Search-->Identify\n          Identify-->Appraise\n        "},"evals":[],"jsHooks":[]}</script>
]

.pull-right[
- Frame a PICO question
- Search Literature
- Identify a study
- Appraise

]

---

# Which one of the following is a PROBLEM with appraising ONE study

- One study may not be comprehensive of every study possible
- A study can be small in size 
- A study can have serious biases
- All of the above

---

# Answer to the above question

**All of the ideas**

- One study is never comprehensive
- Studies vary in size
- If a study has biases, we have a wrong answer
- One study does not support causal inference
- We need high quality evidence

---

# What would you consider?

> Suppose you are suffering from high blood pressure and your doctor has prescribed a new Drug (let's say the name of the drug is "Xarista"). He wants to discuss with you. Which of the following is MOST important for you to discuss with your doctor?

- (A) What is the effectiveness of the Xarista?
- (B) What is the estimate of benefit from Xarista?
- (C) How confident is the doctor in these estimates?

---

# Answer to the previous question

- All three, A, B, and C are important
- C, how confident is the doctor, is VERY important!

---

# Quality of Evidence is best discussed in terms of outcomes

.pull-left[

<div id="htmlwidget-4a50fb1593a228b325f4" style="width:504px;height:504px;" class="DiagrammeR html-widget"></div>
<script type="application/json" data-for="htmlwidget-4a50fb1593a228b325f4">{"x":{"diagram":"\n        graph TD\n          S1-->O1\n          S1-->O2\n          S2-->O1\n          S2-->O3\n          O1-->S2\n          O2-->S3\n          S3-->O2\n          O1-->S3\n          \n        \n        "},"evals":[],"jsHooks":[]}</script>
]

.pull-right[
* One study can have many outcomes

* One Outcome can be linked to many studies

* Hence, it is best to pair outcomes with many studies

* For Quality of Evidence Appraisal

]

---

# Problems of NOT paying attention to Quality

- Relying on low quality of evidence

- Hormone replacement therapy in women
    - Use of Flecainide for treatment of sudden death
    - Caused by disorders of heart rhythm
    - Both of these were later revised
    
- **NOT** relying on high quality evidence
    - Example: Blood thinning therapy for heart disease
    - Got delayed for years

---

# How best can we convey information on quality?

.pull-left[
<div id="htmlwidget-46356a0d78cae9c36091" style="width:504px;height:504px;" class="DiagrammeR html-widget"></div>
<script type="application/json" data-for="htmlwidget-46356a0d78cae9c36091">{"x":{"diagram":"\n        graph TD\n        v[Very low]-->l[Low]\n        l[Low]-->m[Medium]\n        m[Medium]-->h[High]\n        "},"evals":[],"jsHooks":[]}</script>
]
.pull-right[
- Quality scores are CONTINUOUS

- Making them DISCRETE can lose meaning

- OTOH, making them discrete makes it 
    - easy to interpret
    - Vivid
    - Useful
]

---

# Strength of Recommendations

.pull-left[
<div id="htmlwidget-1b7e7319be03bc8fc702" style="width:504px;height:504px;" class="DiagrammeR html-widget"></div>
<script type="application/json" data-for="htmlwidget-1b7e7319be03bc8fc702">{"x":{"diagram":"\ngraph LR\n        r[RCT]-->o[Observational Study]\n        strong-->weak\n        Desirable-->Undesirable\n        "},"evals":[],"jsHooks":[]}</script>
]

.pull-left[
- Quality of evidence is high

- Desirable effects OUTWEIGH undesirable effects

- Balanced Uncertainty

- Need a simple yet balanced system

]

---

# Enter GRADE

- Grading of (G)

- Recommendations (R)

- Assessment (A)

- Development and (D)

- Evaluation (E)

---

# What does GRADE do?

- Classifies Quality of evidence in FOUR groups
- Starts with RCT as HIGHEST quality of evidence, then
- Downgrades based on five features

- Study Limitations
    - Inconsistency of results
    - Indirectness of Evidence
    - Imprecision
    - Reporting Bias
    
---

# Quality appraisal criteria

![](data:image/png;base64,#qualbox.png)
---

# Summary

- Evidence base depends on asking, searching, appraising

- One study is not enough

- We base our quality appraisal on outcomes and many studies

- GRADE is a consolidated system that balances various features