Where the Evidence Is Strongest

Most of what this register examines is evidence with something wrong with it: heterogeneity, confounding, exploratory subgroups, thresholds barely cleared. This one is different. In people with chronic kidney disease, a synthesis of 13 trials and 97,428 participants rates the cardiovascular and kidney benefits at high certainty by GRADE^[1] — the top of the scale, and something we have not been able to write on this site before.

What the evidence shows

That last row matters more than it looks. The obvious mechanistic worry with these drugs in kidney disease is dehydration — vomiting and reduced fluid intake are exactly how someone with impaired kidneys ends up with acute injury. The concern is sensible and it did not materialize.

The question every CKD patient actually has

Almost everyone with chronic kidney disease and type 2 diabetes is now taking an SGLT2 inhibitor, which has its own strong kidney evidence. So the practical question is not whether a GLP-1 works — it is whether it adds anything on top of what someone is already taking.

It is worth appreciating how much work that one number does. We covered a case where an interaction P of .06 left a clean-looking split unestablished, in the statistic that asks if groups differ, and a mirror case where values of 0.62, 0.40 and 0.99 showed subgroups behaving alike, in how many people to treat to help one. This is the same test answering a question worth money and inconvenience: yes, both.

What 'high certainty' means, and why it is unusual here

GRADE rates evidence on four levels — very low, low, moderate, high — based on risk of bias, consistency, directness, precision and publication bias. High means further research is very unlikely to change the estimate.

Almost nothing this register has covered reaches it. The BMJ synthesis of 262 trials rated the newest drugs very low to low. The aspiration meta-analysis was low. The dose-withholding finding was very low. Set against that background, two high-certainty outcomes is a genuinely different kind of result.

A register that only ever discounts becomes as useless as one that only ever promotes.

The reason this particular question is well answered is not luck. It was studied in dedicated randomized trials, in a defined population, against hard endpoints that are counted rather than measured on a scale — dialysis, transplantation, death. That is what produces high certainty, and it is what most questions in this field have not had.

The one figure to hold loosely

The high-certainty finding is about the class. The ranking within it is not, and reading the SUCRA figure as “semaglutide is the best one for kidneys” would be taking the weakest number in the paper as its strongest.

For the trial that anchors this whole area, and the numbers needed to treat it produced, see how many people to treat to help one; for the broader kidney picture, GLP-1s and your kidneys.