How should I read a sequence of three DEXA scans over nine months of semaglutide?

I have three scans from the same centre, same technician, morning fasted each time, and I want a sanity check on how to read them, because the second scan made me cut my deficit and I am no longer sure that was the right call.

Context: 41, female, started at 96.4 kg. Semaglutide, titrated to 1.7 mg and held there. Protein went from unmeasured to tracked at roughly 105-120 g/day from month three. Two full-body resistance sessions a week from month four, nothing before that.

Between scan 1 and 2 I lost 2.4 kg of lean, which panicked me. Between 2 and 3 I only lost 0.6 kg. Is that a real improvement caused by the protein and lifting, or is it just that the first interval was faster? And should I read the bone number as anything?

Your scans look good, your panic at scan 2 was understandable but probably unnecessary, and the change between intervals is partly real and partly an artefact of loss rate. Work the numbers explicitly.

Interval 1: Apr to Aug (about 17 weeks)

weight change   = 84.7 - 96.4 = -11.7 kg
fat change      = 32.6 - 41.9 =  -9.3 kg
lean change     = 49.7 - 52.1 =  -2.4 kg
lean share of tissue lost = 2.4 / 11.7 = 0.205 = 20.5%
rate of loss    = 11.7 kg / 17 wk = 0.69 kg/wk = 0.71% of body weight per week

Interval 2: Aug to Dec (about 17 weeks)

weight change   = 78.9 - 84.7 =  -5.8 kg
fat change      = 27.4 - 32.6 =  -5.2 kg
lean change     = 49.1 - 49.7 =  -0.6 kg
lean share of tissue lost = 0.6 / 5.8 = 0.103 = 10.3%
rate of loss    = 5.8 kg / 17 wk = 0.34 kg/wk = 0.40% of body weight per week

Two things follow. First, both intervals are better than the 25% benchmark from the trial substudies, and the second is dramatically better. Second, the rate of loss halved between intervals, and rate of loss is itself a determinant of lean fraction. So you cannot attribute the whole improvement to protein and lifting; some of it is simply that you were losing more slowly.

How much of the improvement is attributable to what? An honest apportionment is not possible from n=1. But note that the interventions and the slowdown are confounded in the direction that favours your interpretation, and both plausibly helped. The one thing I would push back on is the framing that scan 2 was bad news that required a correction. At 20.5% lean share while losing 0.7% of body weight per week with no resistance training and untracked protein, you were already doing better than the trial average. The corrective action was still correct; it just was not a rescue.

The precision problem

Your interval-2 lean change is 0.6 kg. Ask the centre for its precision error, but a typical in-vivo figure for lean mass gives a least significant change around 1.5-2.0 kg. That means 0.6 kg is inside the noise. The statistically honest reading of interval 2 is "no detectable change in lean mass while losing 5.8 kg", which is a better result than "lost 0.6 kg of lean", and it is the one you should tell yourself.

The bone number

2.40 to 2.40 to 2.39 kg is flat. Bone mineral content is slow, and the 0.01 kg move is far inside precision error. What you cannot see from BMC alone is site-specific density change, and the hip and lumbar spine are where meaningful loss shows up first during large weight reduction. If your centre ran a lumbar and femoral neck density alongside the whole-body composition, look at those T-scores and Z-scores across the series instead. If it did not, request them on the next scan; whole-body BMC is a poor screen. Rapid, large weight loss does reduce hip bone density in the trial literature, and it is one of the few body-composition findings on these drugs with a plausible long-term clinical consequence.

What I would actually do next

Nothing dramatic. Your current configuration is producing near-total lean preservation. Hold protein where it is, and if anything add a third resistance session or add load progression to the two you have, because two sessions a week maintained at the same weights eventually stops being a growth or even a full maintenance stimulus at 78.9 kg. Re-scan in four to six months, not sooner; you cannot resolve real change faster than that against 1.5-2 kg of noise.

edited 18 Apr 2026 by s_kalniete — reworded for clarity after a comment

Adding the calculation OP did not do, which is the one that answers "am I actually losing muscle or getting leaner": look at lean mass as a proportion of total mass, and look at the fat-to-lean ratio.

Scan 1: lean 52.1 / weight 96.4 = 54.0%   fat:lean = 41.9/52.1 = 0.80
Scan 2: lean 49.7 / weight 84.7 = 58.7%   fat:lean = 32.6/49.7 = 0.66
Scan 3: lean 49.1 / weight 78.9 = 62.2%   fat:lean = 27.4/49.1 = 0.56

Fat mass index and fat-free mass index are worth computing too if you know your height, because absolute kilograms of lean tissue mean nothing without a frame reference. FFMI = lean soft tissue in kg divided by height in metres squared, and it is the number that tells you whether 49.1 kg is a lot or a little for you.

The trend across all three of these is monotonic and in the right direction. What OP has done is convert a body with 41.9 kg of fat into a body with 27.4 kg of fat while giving up roughly 3 kg of lean soft tissue, some of which is water and glycogen rather than tissue. That is a good outcome by any standard in the literature, and if the trend continues the absolute lean number will flatten on its own as the deficit narrows.

One thing to watch that these numbers cannot show: the difference between losing lean soft tissue that was structural support for excess mass, which you no longer need, and losing contractile capacity you do need. Function tests separate those. Strength going up while lean mass goes slightly down is the signature of the good version.

Methodological caution on comparing your three intervals, since nobody has raised it: the intervals are not equally well controlled.

In interval 1 you were untracked on protein and not training. In interval 2 you were tracking protein and training twice weekly. But you also changed a third thing that nobody has mentioned: you have been at a stable 1.7 mg dose across interval 2 rather than titrating through it. Titration weeks tend to come with the worst intake, because that is when nausea and aversion peak and total food, not just calories, falls off. Protein is often the first macronutrient to be dropped in those weeks, because meat and eggs are exactly what people find unappealing.

So a fourth candidate explanation for interval 1's larger lean loss is simply that the escalation phase involved several weeks of unintentionally very low protein intake. That is worth knowing for anyone reading this who is about to titrate again, because it is the most actionable item in the whole thread: escalation weeks are the weeks to be deliberate about protein, and they are precisely the weeks when it is hardest.

There is a related confound in the other direction that also deserves stating. OP started resistance training in month four, which is inside interval 2. New or resumed training produces several weeks of fluid retention in the loaded tissue, and DEXA scores that fluid as lean soft tissue. So part of interval 2's apparently excellent lean retention may be training-induced water sitting on top of whatever the real tissue change was.

The net of the two confounds is that interval 1 is probably flattered by nothing and penalised by the titration, while interval 2 is probably flattered slightly by the training fluid. Both push in the direction of the improvement being real but smaller than the raw numbers suggest. That is not a reason to change anything OP is doing. It is a reason to treat the fourth scan, taken after both protein and training have been stable for six months, as the first genuinely clean comparison in the series.