Down 14 pounds in 11 weeks, but the gap in the research is what I notice most. All the long-term GLP-1 data is in people with type 2 diabetes. I’m postpartum, non-diabetic, running on broken sleep and whatever’s left of postpartum metabolism. My OB called it a tool, not a miracle. So I’m tracking what the trials don’t: sleep fragmentation, A1C creeping, which doses flatten my mood. tbh The honest part - I’m basically my own trial, an n-of-1 logging data because my demographic isn’t what they studied. It works, but it’s terrifying to be the unmeasured population
The case for “n-of-1 logging” being the rational response to a research gap is real, and postpartum non-diabetic women are genuinely understudied in the GLP-1 literature. But the framing that “all the long-term GLP-1 data is in people with type 2 diabetes” isn’t quite accurate, and the distinction matters for how you interpret your own tracking. STEP 1, STEP 3, and STEP 4 enrolled non-diabetic adults with obesity or overweight, with follow-up out to 68 weeks, and SELECT ran roughly 3.5 years in non-diabetic patients with established cardiovascular disease. The gap you’re describing is more specific, lactating or recently postpartum women on semaglutide, where the data really is thin because pregnancy and lactation were exclusion criteria in the registration trials. The other thing I’d push on, A1C creeping in a non-diabetic person on semaglutide is an unusual signal worth flagging to your OB or PCP rather than just logging. Semaglutide generally lowers A1C, so an upward trend is the kind of pattern that wants a clinician’s eyes, not a spreadsheet’s. Your sleep fragmentation and mood-by-dose tracking is genuinely useful data for a real prescriber to work with.
STEP 1 enrolled non-diabetics, but not postpartum ones with two young kids on four hours of sleep. The tracking served its purpose - I spotted the A1C pattern and flagged it w/ my OB. That’s exactly why logging matters when the research gap leaves you as your own comparison group.
The A1C flag is the part worth dwelling on, because that’s exactly the kind of signal the trial endpoints weren’t powered to catch in your subgroup. Postpartum thyroid shifts and cortisol from broken sleep can both nudge fasting glucose independent of the drug, so the value of your log is partly that it preserves the timing. Worth asking your OB whether a fasting insulin or HOMA-IR alongside the next A1C would help separate signal from noise.
“the unmeasured population” framing is fair and I’d actually steel-man it harder than your OB did. the diabetic trial pop isn’t just a cohort gap, they were also generally older, less metabolically flexible, and not running a 4-month sleep deficit. so the read-across is weaker than the headline numbers suggest. where I’d push back gently is on which gap matters most. the postpartum signal you’re tracking (sleep fragmentation, mood floor by dose, A1C drift) is doing more work than the diabetic-vs-nondiabetic axis imo. broken sleep on its own raises fasting glucose and tanks insulin sensitivity in healthy adults, I’ve seen numbers like 0.2-0.3 A1C drift attributable to sleep restriction alone in non-GLP1 populations (a study I read, can’t dig up the cite rn). so A1C creeping at 11 weeks postpartum on fragmented sleep doesn’t automatically indict the sema, it might just be sleep doing what sleep does. disentangling those two is the actual hard part of your n-of-1, and I’m not sure weekly weigh-ins or dose logs help much there. second thing, from my own postpartum window not a study: data you collect in months 0 to 6 postpartum has a shorter half-life than you’d think. with my second I had a tracking system I was proud of and by week 5 the gaps in the log were bigger than the data itself. not because I was lazy, but because newborn-state fragmentation makes retrospective patterns illegible. a food noise score of “4” on day 3 after a 90-min sleep window is a fundamentally different number than the same “4” after a 4-hour stretch. the journal-note-per-entry layer matters more than the metric does, because in 8 months when you read this back, the bare number won’t carry the context. ask me how I know. also worth naming and the trials really do gloss this: cycle return, lactation status, and any birth control reintroduction are going to scramble the curve in ways diabetic pops don’t capture at all. flag those dates in the log now while you remember them, ime they’re impossible to reconstruct later. what’s your dose schedule looking like right now, are you titrating up or holding steady while you stabilize the postpartum baseline?
One angle nobody’s named yet that’s worth tracking in your spreadsheet: postpartum lymphatic clearance. There’s bariatric post-op literature on how rapid fat loss in women specifically can lag the lymphatic system, you get this period where adipose is mobilizing faster than the clearance pathways can keep up, and it shows up as ankle puffiness, finger rings getting tight at the end of the day, that kind of thing. I don’t know if anyone’s looked at it specifically in postpartum GLP-1 users, but the underlying physiology is the same and postpartum bodies are already doing a lot of fluid remodelling. Worth a column if you’ve got the bandwidth. The other thing your demographic adds that the diabetic cohorts don’t, breastfeeding status if it applies, and where you are in the return-of-cycle window. Estrogen affects GI motility and that’s already a moving target on sema. If you’re still in the postpartum amenorrhea phase your nausea pattern and food noise pattern might look completely different from what they’ll look like once cycles resume, and that’s a confounder the trial pop just doesn’t have. I noticed my own food noise scores read differently across cycle phases once I started cross-referencing, and I’m not even postpartum anymore, just perimenopausal-adjacent in my tracking. The “unmeasured population” framing your OB pushed back on is actually the strongest argument for keeping the log granular. Trial means tell you nothing about your specific n-of-1, but a clean personal log over 6-12 months is genuinely useful data, especially if you’re tracking the things the RCTs didn’t bother to. The sleep fragmentation column is the one I’d protect hardest, that’s where I think the real signal lives for postpartum specifically.
the postpartum bit is the real gap, and you’re naming it correctly. lactation, the hormonal recalibration of the first year out, and the way sleep fragmentation interacts with appetite signaling, none of that is well studied with these drugs. that part of your framing holds. where i’d push back, gently, is on the “all the long-term data is in type 2 diabetics” line. the STEP program (1 through 5, plus SELECT) was specifically in people with overweight or obesity and no diabetes, and the follow-on extension data isn’t trivial. SELECT in particular ran a median of around three years on cardiovascular outcomes in a non-diabetic cohort. so the non-diabetic question has actually been studied in a way the postpartum question hasn’t, and it’s worth separating those two. if you conflate them you end up more anxious than you need to be about the parts that are well measured, and possibly less anxious than you should be about the parts that genuinely aren’t. the n-of-1 instinct is right for the actual gap. one thing from people who’ve done self-tracking well: pick your variables before you start logging, not after. retrofitting which doses flattened mood is much harder than tracking a daily mood score with the same prompt from week one. same for sleep, which is the noisiest of the three you mentioned. you’ll want a consistent metric (a wearable score, or even just a one-line rating) rather than reconstructing from memory at the GP appointment six months in. and on the A1C creep, remember the intra-individual CV on that assay is small but non-zero, so a single draw moving 0.1 or 0.2 isn’t necessarily signal yet. the unmeasured population part is real. but you have more company in the non-diabetic cohort than you think.