That stopping reverses things is established, and this site has covered it in the withdrawal evidence and in what comes back besides the weight. What this analysis adds is a gradient [1]. The damage is not a switch that flips when you stop. It tracks how much of the weight comes back.
The authors took the 308 SURMOUNT-4 participants who had lost at least 10% on tirzepatide by week 36 and been randomized to placebo, then sorted them by how much of that loss had returned by week 88. Read the bands carefully: they are shares of the weight lost, not shares of body weight. Someone in the 75%-or-more band regained three quarters of what they had taken off.
Non-HDL cholesterol shows the pattern most cleanly. It moved -0.4% in the smallest-regain band, 1.6% in the next, then 8.4% and 10.8% in the two largest. The first two confidence intervals cross zero and the last two do not. So the lipid change appears only once regain passes roughly half of what was lost.
One limit governs how far any of this can be pushed. The four bands were observed, not assigned. Everyone in this analysis received placebo; nobody was randomized into a regain category. People who regain little may differ from people who regain a lot in ways the drug never touched — that is exactly the problem this site set out in what happens in ordinary care. This is an association measured inside a randomized arm, not a comparison between arms.
For a reader deciding whether to stop, the useful version is this. Stopping is not a single outcome with a single price. Where you land on the regain curve appears to set how much of the cardiometabolic benefit you keep, and the blood pressure component came back regardless — which is also the pattern behind the composition finding in what is lost and what returns.