They are taken together, routinely, and the largest pooled comparison available puts the pair ahead of either drug alone on every cardiorenal outcome it measured [1]. The catch sits in how that ranking was built: nobody was randomized to the combination, and the authors call their own finding hypothesis-generating in the last line of the abstract. The other catch is arithmetic rather than statistical, because a second prescription is a second chance for the chain to break at the pharmacy counter [3].
This desk prices GLP-1 drugs and nothing else, so the combination is not a purchase anyone can make here. What follows is the evidence a reader can take into a consultation, in the order that decides how much weight it carries. It starts with the randomized kidney evidence for GLP-1 drugs on their own, which is the firmest ground in this whole area.
How the comparison was built
Sixteen randomized trials with more than a year of follow-up, each reporting outcomes broken down by whether participants happened to be taking an SGLT2 inhibitor or a GLP-1 in the background. Those subgroups were then assembled into a network, which allows comparisons between strategies that no single trial tested directly.
Randomization inside each trial governs who got the study drug, and it does not govern who was already on the other one. That was decided by each participant’s own clinician, before the trial, for reasons the analysis cannot see. So the combination is an observational category sitting inside randomized studies.
What the numbers show
Against placebo, every active strategy reduced major cardiovascular events and heart failure hospitalization, all at p<0.05. On composite renal events, the combination reached RR 0.63, 95% CI 0.41 to 0.97, and SGLT2 inhibitors alone RR 0.70, 95% CI 0.61 to 0.79.
Look at those two intervals rather than the two point estimates. The combination’s runs from 0.41 to 0.97, nearly touching no effect, while the SGLT2 inhibitor’s runs from 0.61 to 0.79 and is far tighter. The lower number is the less certain one, which is what you expect when it comes from subgroups rather than from randomized arms.
In pooled head-to-head comparisons the combination beat SGLT2 inhibitors alone on major events, RR 0.83, 95% CI 0.72 to 0.96, and on heart failure hospitalization, RR 0.73, 95% CI 0.56 to 0.94. Against a GLP-1 alone it showed a better kidney filtration slope, a mean difference of 2.29 mL/min/1.73 m² per year, 95% CI 0.14 to 4.44.
Which of the two is better on its own
Neither, as far as anyone can currently show. A separate network meta-analysis restricted to people with type 2 diabetes and a previous heart attack put SGLT2 inhibitors at a hazard ratio of 0.87 against placebo and GLP-1 drugs at 0.83 [2]. Set against each other indirectly, major cardiovascular events came out at 1.05, 95% CI 0.92 to 1.20.
That interval includes differences a person would care about in both directions, so the honest reading is that the evidence cannot separate the classes rather than that the classes are equivalent. The heart-attack comparison in that paper rests on four studies, exactly one of which is an SGLT2 inhibitor trial, and the trial that would settle it has never been run.
What a SUCRA of 1.000 does and does not mean
The combination ranked first across outcomes with SUCRA values from 0.928 to 1.000. SUCRA summarizes how often a treatment comes out on top across simulated orderings, so a value near 1 means it almost always ranked best.
It says nothing about by how much. A strategy can rank first consistently while the margin over second place is small, uncertain, or clinically irrelevant, and ranking statistics are the part of a network meta-analysis most often quoted without their intervals. A subgroup restricted to established cardiovascular disease attenuated the combination’s ranking on major events, which is the population where the benefit would matter most.
Two prescriptions, and who actually ends up on both
The PRECIDENTD feasibility phase randomized 173 insured adults to an SGLT2 inhibitor, a GLP-1, or both, and then made them fill the prescriptions through their own insurance [3]. At four months, 84% of those assigned one drug had filled it, against 53% of those assigned two. Over a ten-month median the figures were 87% and 68%.
Among the people who did fill, 22% on one drug and 49% on two stopped a study medication, mostly because of side effects. Combine the two stages and the share still taking both at the end sits far below what a guideline recommending the combination would assume. That is the practical reason a prescription and a filled prescription are different measurements.
The money version
Two prescriptions cost more than one, and nobody on this roster sells SGLT2 inhibitors, so this is not a purchase available here. The transferable lesson is that when the evidence for a more expensive option comes from subgroups nobody randomized, the extra cost is certain and the extra benefit is not.
If a prescriber proposes both, that is a legitimate clinical judgment and this analysis is one input to it. The honest unit for weighing it is how many people must be treated for one event to be prevented. None of these three papers reports it. The same caution applies to swapping a GLP-1 in for insulin, where the trial evidence is randomized and the price comparison still is not.