0.49
If you only remember one figure from this paper, it will be this one.
People with type 2 diabetes taking an SGLT2 inhibitor and a GLP-1 receptor agonist together had an odds ratio for death from any cause of 0.49 compared with people on just one of the two. The 95% confidence interval ran from 0.41 to 0.60. The p value was below 0.00001.
In plain language: the odds of dying during follow-up were roughly half as high on the combination.
The combination beat SGLT2 inhibitors alone (p = 0.0003) and beat GLP-1 drugs alone (p = 0.03). It was not a case of simply beating the weaker partner.
Where that number comes from
Now for the part that matters just as much as the number itself.
The mortality result is built on two studies. Not thirteen. Two.
They happen to be the two biggest in the review. One is the Danish national registry study by Jensen and colleagues, following 66,807 people for up to 20 years. The other is Riley and colleagues, a five year analysis of electronic health records drawn from TriNetX, a global network, covering 2.2 million patients in total. Between them, the GRADE summary table counts 418,630 people contributing to the mortality comparison.
That is a lot of people. It is also only two sources. When one of them is as enormous as a multinational records network, it can dominate the pooled estimate.
The authors also report an I² of 93% for this outcome. I² is a measure of how much the included studies disagree with each other beyond what chance would explain. Anything above 90% is classed as high heterogeneity. So the two studies both point toward benefit, but they do not agree closely on how big it is.
Turning it into something you can picture
The GRADE table offers an absolute figure: an estimated 22 deaths per 1,000 people on the combination, with a range of 19 to 27.
Absolute numbers are always more useful than relative ones when you are trying to understand your own situation. "Half the odds" sounds enormous. "About 22 in 1,000" gives you something you can actually hold in your head and discuss with a doctor.
Why the result is plausible
It would be easy to wave away a number this big as too good to be true. But the direction of the finding lines up with what we already know from large, properly randomised trials of each drug class on its own.
The paper walks through them. Empagliflozin reduced cardiovascular death and all-cause mortality in its outcome trial. Liraglutide reduced cardiovascular and all-cause death in LEADER. Subcutaneous semaglutide lowered cardiovascular deaths in SUSTAIN-6. The SELECT trial, with 17,604 people who had heart disease and obesity but not diabetes, found semaglutide cut the rate of cardiovascular death, non-fatal heart attack and non-fatal stroke from 8.0% to 6.5%.
The paper also cites an earlier meta-analysis of outcome trials that reported combination therapy linked to a 30% reduction in major cardiovascular events, 31% for cardiovascular death or heart failure hospitalisation, and 57% for all-cause mortality compared with a single drug.
So you have two drug classes that each, independently, save lives in rigorous trials. They work by different mechanisms. It would be surprising if stacking them gave no additional benefit at all. The real-world data says the extra benefit is there.
The honest reading
Here is how I would put it to a friend.
The combination is strongly associated with living longer. The association shows up in two very large real-world datasets. It fits with what the randomised trials showed for each drug separately. That is a meaningful, positive signal.
What it is not, yet, is proof that adding the second drug causes that halving. People who get prescribed both drugs may differ from people who get one. They may have better access to care, more attentive doctors, or be seen at clinics that manage everything more aggressively. Observational data cannot fully separate the drug from the kind of person who ends up on it.
The authors themselves say the mortality figures are unadjusted. That limitation gets its full airing in Part 7.
For now, hold two things at once. The number is real and it is big. The number is also resting on a narrower base than its size suggests. Being able to hold both of those at the same time is the whole skill of reading medical research.
In Part 4 we move from life and death to the measurements your doctor actually checks at every visit.
This series discusses a peer-reviewed meta-analysis of observational studies. Nothing here is medical advice. Do not start, stop or combine prescription medications without your doctor.
