1-year obesity treatment starts still rest on incomplete evidence
Choosing an initial obesity medication for the full first year looks straightforward on paper and uneven in the evidence base. This Bayesian longitudinal network meta-analysis asked a more practical question than most obesity drug syntheses: which treatment-initiation strategies actually have usable 52-week data for decision-making, and where do the trade-offs between weight loss, dose escalation, titration, and discontinuation begin to appear?
The answer is both encouraging and sobering. Across 180 reports from 178 trials, the literature contained 196 randomization units and 81,014 participants, yet only 78 units reached the 1-year evidence layer. Even fewer fed the primary longitudinal model in adults without diabetes: 50 units. For clinicians trying to translate trial evidence into a year-long treatment plan, the gap is not abstract; it means many tempting strategies still depend on extrapolation rather than direct observation.
The analysis was built to align with treatment-policy decisions rather than drug-to-drug comparisons alone. Investigators searched eight databases and two trial registries through October 2026, included randomized trials of obesity pharmacotherapy initiation in adults with overweight or obesity, and excluded maintenance, withdrawal, and switching designs. The study was registered with PROSPERO, and the analytic framework estimated treatment-policy-compatible weight change at 52 weeks using Bayesian longitudinal network meta-analysis.
The largest weight reductions came from higher-dose incretin-based strategies
At 52 weeks, the most pronounced placebo-adjusted weight reductions were seen with retatrutide 12 mg at -19.80 percentage points, 95% CrI -21.81 to -17.78, followed by tirzepatide 15 mg at -16.87, -18.07 to -15.51, and cagrilintide-semaglutide 2.4/2.4 mg at -16.37, -18.10 to -14.70. Those are large effects by any standard, and they fit the biology of multi-receptor appetite suppression: the more complete the satiety signaling and gastric-emptying restraint, the more likely weight loss extends across a full year rather than stalling early.
Only tirzepatide 15 mg and semaglutide 2.4 mg carried high-certainty evidence. The contrast matters because effect size alone does not define clinical confidence. A dramatic reduction supported by sparse or indirect data is still less useful than a slightly smaller reduction backed by stronger trial evidence and clearer treatment-policy reporting.
Even among the 160 target strategies examined, only 45 had non-extrapolated 52-week estimates. In plain clinical terms, most of the decision tree was still only partially observed. Dose, timing, and the path from initiation to maintenance remain underdescribed in the trials that are shaping everyday prescribing.
Discontinuation risk rose with several active strategies, but not all-cause stopping
Safety did not fall neatly in line with efficacy. Thirteen of 32 strategies increased adverse-event-related permanent discontinuation versus placebo, yet none increased all-cause discontinuation. The pattern suggests a tolerability burden that is real enough to drive treatment stopping when adverse effects are counted specifically, but not so overwhelming that patients abandoned therapy overall at higher rates across the board.
This distinction has bedside consequences. For obesity pharmacotherapy, nausea, vomiting, diarrhea, and other dose-related adverse effects can determine whether a patient stays on therapy long enough to reach the weight-loss trajectory seen in trials. A treatment can be potent and still fail in routine care if escalation outpaces tolerance.
Five of 22 adjacent dose steps had at least an 0.80 probability of adding more than 2 percentage points of weight reduction. In other words, incremental up-titration sometimes appears worth the added burden, but not uniformly. The data support dose sensitivity, not blind escalation.
Titration and lifestyle support remain far less evidence-rich than the drugs themselves
The trial map was especially thin where real-world prescribing often becomes most uncertain. Investigators identified eight 1-year titration contrasts and only three trials of randomized lifestyle-support increments. Those numbers are small enough to leave large gaps in day-to-day decision-making, particularly when clinicians are trying to decide how aggressively to escalate doses or how much structured lifestyle support should accompany pharmacotherapy.
This is where trial design and clinical practice diverge. Patients do not experience obesity medication as a static choice between branded agents; they experience it as a sequence of dose adjustments, counseling intensity changes, side-effect management, and adherence negotiations. A trial literature that mostly compares end-point drugs without fully testing those decisions leaves prescribers to improvise.
The authors also examined diabetes-related attenuation, and the reduction in adults with type 2 diabetes was 62% of that seen without diabetes, ratio 0.62, 95% CI 0.58-0.66. That finding is biologically plausible. The metabolic and hormonal milieu of type 2 diabetes may blunt weight-response magnitude, and background therapies, insulin resistance, and disease-related physiology can all compress the ceiling of achievable loss.
What this evidence can and cannot support at the bedside
The strongest conclusion is not that one drug clearly dominates every other for every person. It is that the evidence supporting 1-year obesity treatment-initiation decisions remains incomplete, even while several newer agents produce substantial weight loss. For clinicians, the most relevant choices are increasingly about balancing depth of weight reduction against tolerability and certainty of evidence, rather than simply asking which medication lowers weight the most in a vacuum.
The current dataset also makes a reporting problem impossible to ignore. Better treatment-policy reporting would make it easier to distinguish initiation from maintenance, titration from steady dosing, and medication effect from support-intensity effect. Without that structure, network meta-analysis can estimate a great deal, but it still cannot manufacture the missing trial layer.
For adults with obesity who do not have diabetes, the evidence base is now broad enough to suggest real differences between strategies, yet uneven enough to keep some of the most relevant questions open. The report stops short of endorsing any one pathway as universally preferred, and the preprint status matters: these findings have not been peer reviewed and should not be used as stand-alone guidance for routine care.
The clearest next step is not more extrapolation. It is more randomized evidence on titration, more directly observed 52-week initiation trials, and cleaner reporting of the treatment policies that patients actually follow.