When I got my PCOS diagnosis at 23, my total testosterone was 72 ng/dL. Not dramatically high, but my free testosterone was elevated at 1.8 pg/mL and my SHBG was low, which meant more of that testosterone was actually circulating. imo Nobody explained the distinction at the time. They noted “slightly elevated androgens” and moved on. Nine mos into tirzepatide, I had labs drawn again. Free testosterone is now 0.9 pg/mL. Total testosterone barely moved. SHBG increased by about 40%. My acne, which has been a constant since I was 19, has genuinely improved since around month four. The mechanism I keep coming back to: as insulin resistance drops, LH levels stabilize, which means the ovaries stop overproducing androgens. The tirzepatide isn’t acting directly on testosterone. It’s downstream of the insulin problem being addressed. My fasting insulin went from 22 to around 8, and I think that’s the actual driver here, not anything else. What I wish someone had told me earlier is how much work SHBG does in this picture. Low SHBG from chronic hyperinsulinemia means even a “normal” total testosterone creates high-androgen symptoms, because so little is bound and inactive. When insulin improves, SHBG rises and starts buffering properly. The free testosterone number tells you far more than total. My cycles went from 45+ days to around 32, which tracks with the androgen change in a way that finally makes sense to me. Still on metformin alongside the tirz. The real test will be whether these numbers hold if I drop the metformin next year. That’s what I’m watching.
The SHBG picture you’ve drawn is the part most people never get walked through, and it’s genuinely the right frame here, low SHBG meaning even a “normal” total reads as high-androgen. Where I’d gently push is “I think that’s the actual driver, not anything else.” You’re still on metformin, and that’s an insulin-sensitiser doing its own work across the whole nine months, so the drop from 22 to 8 isn’t cleanly attributable to the tirz, the two are tangled until you actually drop one. The other thing worth asking before you lean too hard on the free T figure: what platform ran it? Direct immunoassay free testosterone gets unreliable down in the low female range, and a calculated free T off SHBG and albumin is usually the steadier number to track changes against.
The assay critique actually lands for me – direct immunoassay free T in the low female range is noisy, and if I’m being honest about the 0.9 pg/mL number, the absolute value should carry a wider confidence interval than I implied. Same lab, same method both draws, so the directional change is still meaningful, but the specific number deserves skepticism. Where I’d push back is the “tangled” framing on the insulin drop. The case for metformin contributing is real in principle, but I was on 1500mg metformin for roughly two years before starting tirz and my fasting insulin was sitting at 22 the whole time. Metformin had the field to itself and didn’t move that number. The drop from 22 to 8 happened after tirz was added. That’s not a clean RCT, obviously, but “tangled” implies roughly comparable contributions and I don’t think that’s what the timeline shows. The reason I want to drop metformin next year isn’t to retroactively assign credit – it’s to see whether tirz alone holds the floor, which is genuinely unknown to me right now.
The two years of metformin sitting at insulin 22 is genuinely persuasive, and I’d concede the “tangled” framing doesn’t fit a timeline where one drug had the field and didn’t budge the number. Where I’d gently push: dropping the metformin next year won’t be a clean read on tirz-alone either, because you’ll be removing it from an already remodelled system, not the hyperinsulinaemic one it was failing to move at 22. Worth deciding in advance what number would actually count as “holding.”
The fasting insulin drop from 22 to 8 is where this becomes mechanistically coherent, and what you’ve worked out about SHBG doing the buffering work is something most people on this journey never get explained properly, so it’s worth having it written down clearly like this. The part I’d flag, though: you mention you’re still on metformin alongside the tirz, and metformin’s effect on hepatic glucose output and SHBG upregulation is not negligible. Your phrase “the real test will be whether these numbers hold if I drop the metformin” is actually the central interpretive question right now, not a footnote. The androgen signal you’re seeing is probably a combined one, and I wouldn’t give tirz sole credit for that SHBG change until there’s a cleaner separation
fasting insulin 22 to 8 while running both tirz and metformin makes it hard to attribute the SHBG rise cleanly to either one. metformin has its own hepatic effect on SHBG in PCOS, separate from the GLP-1 downstream mechanism you’re describing. so “the free testosterone number tells you far more than total” is right and worth saying louder, but the experiment that actually isolates the tirz contribution is the one you already named at the end.
edit: forgot to add
the part that jumps out: total testosterone barely moved but SHBG went up 40%. free T is calculated from those two numbers, so a 40% SHBG rise mechanically drops your free T even if your ovaries didn’t change output at all. which means the halving you’re seeing is at least partly just the binding math, not reduced production. that doesn’t break your LH/ovarian story. the cycle change from 45+ to 32 days is real and points at production. but it does mean the free T number is doing double duty here: it’s reading both “less androgen made” and “more of the same androgen bound up.” with total flat, the data leans harder on the buffering side than the production side, and you can’t cleanly separate the two from a free T calc alone. if you wanted to actually split them, an LH/FSH ratio or a DHEA-S / androstenedione draw would tell you whether ovarian or adrenal output dropped independent of the SHBG shift. fasting insulin 22 to 8 is a big move and I’d bet production did fall some, but the labs you’ve posted can’t prove how much. the SHBG point is the one I’d underline for anyone reading: it’s not a passive marker, it’s part of the mechanism, so watching it across the metformin taper next year matters as much as the free T itself. ymmv but I’d draw SHBG and LH together each time, not free T in isolation.
“The free testosterone number tells you far more than total” is what should be explained at PCOS diagnosis, not buried as a footnote. The caveat: metformin independently affects hepatic SHBG production and insulin sensitization. With both running simultaneously, the androgen shift can’t be cleanly attributed to tirz’s mechanism. Dropping the metformin next year will be the real separation experiment.
the SHBG read is the part you’ve got dead right, and it’s the part most PCOS threads skip. low SHBG from chronic hyperinsulinemia means a “normal” total T still reads as high-androgen because so little is bound, so the free number is the one that actually tracks symptoms. no argument there. where I’d push back is “I think that’s the actual driver here, not anything else.” you’re on metformin the whole nine months. metformin independently lowers insulin and nudges SHBG up, so the fasting insulin drop from 22 to 8 isn’t cleanly attributable to tirz. on top of that, SHBG rises with weight loss on its own, partly independent of the insulin axis, and tirz almost certainly moved your weight over nine months. so you’ve got at least three things pulling the same direction: tirz, metformin, and fat loss. naming one as “the driver” when all three are confounded the entire window is the same attribution problem that bites every n=1 stack post. the clean version of your own experiment is exactly the one you flagged at the end: drop the metformin, hold tirz and weight roughly steady, redraw free T and SHBG. that isolates the metformin contribution. it won’t isolate weight from tirz, but it’ll at least tell you whether metformin was carrying part of the SHBG rise. the order matters too, because if you drop metformin and lose more weight in the same stretch you’ve just rebuilt the confound. one thing I’d add: free T at a single timepoint is noisy, especially calculated free T off SHBG and albumin rather than measured directly. was your 0.9 measured or calculated? if calculated, a 40% SHBG bump mechanically drops the free number even before any real androgen change, so part of that 1.8 to 0.9 could be the formula doing the work, not the ovaries. worth checking which assay your lab ran before you read too much into the exact delta.