POPLHLTH 304 tutorial 8: Meta-analysis
Simon Thornley
5 May, 2025
A three minute primer on meta-analysis from the UK.
Malcolm Gladwell, the well-known Canadian science writer, describes some of the evidence which has turned around the science related to the influence of saturated fat intake.
Getting started
This session, we will focus on conducting meta-analyses with R.
We will learn how to:
- Summarise data for meta-analysis
- Analyse meta-analysis data in R
- Think about whether a fixed effect or random effect model is more appropriate
- Interpret the various plots associated with meta-analyses, including funnel and forest and consider the effects of publication bias on summary measures.
We will be leaning on the meta and dmetar The dmetar package
has an excellent book associated with it: here.
Rstudio cloud link
The link to this project can be found here.
Before we start
There is much debate about the role of saturated fat or animal fat in the diet. Early on in my medical training it was routinely believed that saturated fat was a direct cause of heart attacks or strokes. Eating butter was believed to increase bad cholesterol (LDL) and this led to clogging of arteries, in much the same way as fat left on dishes after cooking can clog sink drains.
However, this has undergone a reassessment recently due to the work of people such as Professor Christopher Ramsden.
Let’s have a look at the evidence for ourselves. We’ll be looking at the data from the Cochrane meta-analysis on the subject.
This will give us an idea of what the average effect is likely to be across many different trials.
There is some really nice work done in R to do meta-analysis. There is a free online textbook here. We will start by looking at the influence of reducing saturated fat on combined cardiovascular disease events.
I will then let you loose to have a look at the effect of this diet on all-cause mortality.