Manage weight

The short trials and the long follow-ups measure different things, weight on one side and disease on the other, so neither one checks the other. The long-term half only watched what people already drank, and people already at higher risk switching to diet drinks would produce the same pattern. This is a joint position from the Korean Diabetes Association and Korean Nutrition Society rather than a new analysis.

The 2025 review rated its own certainty very low on every outcome, because the experiments differed widely and most carried design weaknesses that could tilt a result either way. The older review's ranges were wide enough to hide a real effect: on body weight it ran from a 1.6kg loss to an 11.6kg gain. Not finding an effect is not the same as ruling one out.

The lean-mass figure is the solid one; the fat-loss figure ran anywhere from 1.34kg down to 0.11kg, close enough to nothing at one end that the effect there is unsettled. Everyone pooled here was under 50. In a small share of the trials, how people were assigned or assessed left room for the result to tilt.

A null result is not proof of equivalence. Eight small trials, with moderate inconsistency on three of the five outcomes, are underpowered to rule out a real difference between the two. No risk-of-bias assessment or certainty grading is reported.

Eight trials and 342 people in the metabolic pool, so a modest real effect could be missed. Participants were mostly healthy and not diabetic, which is the population least likely to show a metabolic improvement and leaves the question open in people who already have metabolic disease. Several pooled trials ran fourteen days, far too short for HbA1c, which reflects about three months of blood sugar, to move at all.

The trial cannot separate how much of the benefit came from the weight loss and how much from the drug's other effects, since the treated arm got both. Adverse events forced 16.6% off the drug permanently, against 8.2% on placebo, in a group with strong reason to persist. Novo Nordisk funded it and employed several of the authors.

Both trials were funded by Novo Nordisk. Participants stayed on varying background diabetes medication, which is part of what separates the two groups rather than diabetes on its own. Gastrointestinal side effects reached 63.5% here against 34.3% on placebo.

This was a prespecified secondary analysis, not what the trial was built to answer. The 10.2% is lower than the obesity trials' figures because this group was older, already had heart disease and started at a lower body mass index, not because the effect wears off. People came off the drug more often than off placebo.

Persistence was inferred from gaps between prescription fills, which cannot tell someone who chose to stop from someone who lost their insurance or could not fill during the shortage. Only 53.6% of the initiators found met the continuous-enrolment rule, and people who keep coverage are probably the more persistent ones, so the real figures are likely lower still. The analysis was run by a pharmacy benefit manager. Commercially insured Americans only.

Novo Nordisk funded the trial and the drug is theirs. The cohort was largely female and largely white, so it says less about everyone else. Sixty-eight weeks is what was measured, and nothing here describes what happens after the last injection.

Three hundred and four people is small for a phase 3 trial, and 93% were white and 78% women. Gastrointestinal effects, mostly mild, reached 82.2% against 53.9% on placebo. Nobody was followed after the last injection, so this describes two years on the drug rather than two years of results.

A network meta-analysis compares drugs that were never tested against each other, so this ranking is inferred rather than measured. Fourteen of the 56 trials were semaglutide trials and most came from one company's own programme, which pooling does not correct for. Every figure describes people still taking the drug.

Everyone had already tolerated 20 weeks of dose escalation before being randomised, so the people most likely to quit were gone before the clock started. Lifestyle support carried on in both arms, which is not the same as stopping everything. Four in five participants were women.

Most of these trials enrolled men by age rather than by a low testosterone reading, so many participants were not deficient. A gain in lean mass is not by itself a gain in anything a man would notice. In the 16-trial pooling, whose men averaged 77 years old, no musculoskeletal difference from placebo was left by 12 months, which could mean the effect fades or only that fewer trials reported that far out.

Men only, all obese and insulin-resistant, so this does not describe healthy or older users, and twelve weeks is not years. The mitochondrial half is a sub-study of the same forty men rather than an independent replication, and the two count once. The trial did not report whether NAD actually rose in these participants, so a failure of delivery cannot be fully separated from a failure of the mechanism.

A label is the manufacturer's own trial programme summarised by the regulator, not an independent analysis, and trial populations exclude many people who take the drug in practice. The label carries a separate table for children with different figures. The copy read here is dated March 2024 and the agency marks it as possibly not the current version.

Report counts have no denominator, so 990 cannot be turned into a risk per patient and a bigger number does not by itself mean a higher rate. Most compounding pharmacies are not required by federal law to report at all, which both understates the true total and makes any comparison against the approved products unreliable in either direction. The salt concern is stated as missing equivalence information, not as evidence of harm. This is about specific products, not about compounding, which is lawful and often necessary.

Everyone counted here had already stopped, so these are shares of people who quit and not of people who started. The reason came from what a clinician typed into the record rather than from asking the patient, and 11.1% had no reason recorded at all. Semaglutide and tirzepatide were counted together. One US health system, over a period that included the shortage.

These are label figures from the manufacturer's own pharmacokinetic studies, measured in trial participants rather than in general use. The label flags the interaction with oral medicines as something to watch rather than a quantified effect on any particular drug.

Four cases against one is too few to size the risk; it gives the order of magnitude and nothing more. The rodent finding is why anyone with a personal or family history of medullary thyroid cancer, a rare tumour of the thyroid, is told not to take it. The label lists further warnings the trials were never built to quantify, among them gallbladder disease, acute kidney injury and suicidal thoughts.

Both pools followed what people already chose to drink, so these are links rather than proof that either drink causes anything. Someone told to watch their weight or blood sugar tends to switch to diet, which would draw this same pattern with no effect from the drink at all. The pooled studies varied widely in their results for several of these outcomes. And these are drink categories, not Coca-Cola and Coca-Cola Zero specifically.

This pools the whole drug class rather than semaglutide alone. Randomised trials screen out higher-risk patients and are mostly too short to catch rare events, so finding no signal for pancreatitis is weaker than showing there is none. The subgroups on weight-loss doses did show larger effects, but not significantly so, which leaves the dose question open.

Whether semaglutide raises rare serious gut harms is unsettled. Pooled trials find little or no rise in pancreatitis, obstruction or a stalled stomach but mostly ran too short for rare events; the cohort that found rises cannot size them.

The pooled trials were short and produced only 3.55kg of loss, so those kilograms do not scale up to someone losing a seventh of their body weight; the proportion is what carries over. Lean mass as a share of body weight did not change. Not one trial in the pool measured strength, mobility or falls, so whether the loss costs anything a person would notice is untested.

Whether losing lean mass on semaglutide costs anything a person would notice has not been measured either way. Trials rank it weak at holding lean mass; one study with no comparison group says it was preserved.

This followed what people already consumed rather than assigning it, so the arrow could run the other way: someone watching their blood sugar has good reason to switch to sweeteners well before a diagnosis arrives. Intake came from participants' own repeated diet diaries. And these are the same participants and the same diaries behind the cancer finding, so the two results are not independent of each other.

Thirteen people is very small even for a crossover design, which makes the two positive results the fragile half and the long list of nulls the sturdy one. The fat-free mass change was one of many outcomes tested, exactly the setup in which a chance finding appears. Six weeks, in healthy overweight and obese volunteers.

Spontaneous reports have no denominator, so none of these ratios is a risk to a patient and none supports saying the products caused anything. Only 707 of the 81,078 reports involved compounded products, and some ratios rest on a handful of events, which is why the preparation-error interval runs from 12.63 to 189.6. State-licensed pharmacies mostly do not have to report at all, which biases the comparison in a direction the analysis cannot correct.

Compounded semaglutide was named less often in reports of dosing errors, against the regulator's own warning about overdoses. Uneven reporting is likelier than safer products, but nothing in the data can settle it.

The extension was exploratory and powered for nothing. Its 327 people came from high-recruiting sites rather than a random slice of the 1,961 randomised, and they had lost more than the trial average to begin with. Drug and lifestyle support were withdrawn together, so the regain cannot be pinned on the drug alone. Nobody was randomised after week 68, so there was no control arm left to compare against.

It is observational, so people prescribed these drugs may already differ from the comparison group in ways no adjustment removes. It was published as a research letter, so the methods are described only briefly. The authors say they cannot be certain every prescription was for weight loss. Reported incidence in the semaglutide group was 4.6 pancreatitis cases per 1,000 person-years.

The ninefold figure cannot stand alone: its confidence interval is wide enough to hold almost any answer, so this cohort cannot say how large the pancreatitis risk is. The randomised trials are much the stronger evidence.

There was no control group and no randomisation, so nothing separates the drug from the specialist nutrition care these patients also received, or from people simply moving more as they got lighter. Grip strength rising during weight loss is partly expected because a lighter body is easier to move, and the study did not adjust for that. Mean body mass index was 46.3, far heavier than the trial populations. Calling 3kg preserved is the authors' reading of a real loss.

What this study can show is how lean mass fell over time, stopping by month seven. With nobody off the drug measured alongside, it cannot show that semaglutide, rather than anything else, preserved it.