Obesity: fewer injections to keep the weight off? $VKTX, $AMGN, $LLY and $NVO

1. What Viking actually studied

VK2735-102 was randomized, double-blind and placebo-controlled, involving approximately 180 otherwise healthy adults with obesity. Its objectives concerned safety, tolerability and pharmacokinetics; weight measures were exploratory. The announcement describes transition from weekly treatment to maintenance regimens, including placebo. Viking’s study description.

This design is useful for identifying doses and intervals to investigate further. It should not be treated as the final demonstration of a commercial regimen. The existence of a Phase 3 program for a molecule does not automatically turn every study of that molecule into a pivotal test of every possible use.

Induction matters because it establishes the starting point for maintenance. Participants reaching that stage have already received treatment for a meaningful period. The result therefore does not describe someone starting directly on widely spaced injections. Nor does it establish that a patient stabilized on a different molecule would respond identically after switching to VK2735.

Preservation of a previous reduction must also be distinguished from additional weight loss. The first asks how much of an existing result remains; the second asks how much more weight is lost. Their reference points differ. An arm continuing weekly treatment may lose additional weight, but its total weight reduction does not become a maintenance percentage simply because it is reported alongside maintenance arms.

Why “up to” changes the interpretation

The best-performing group may help identify what to test next. When describing the experiment to readers, however, it belongs alongside the pooled result. Showing only the maximum hides how much the headline depends on selecting a particular dose and a small sample.

It is also inappropriate to infer a firm ranking of higher and lower doses. Individual differences and statistical uncertainty can have a large effect on small-group means. A credible dose-response relationship requires a planned analysis and a coherent body of evidence, rather than placing the reported numbers in descending order.

One detail deserves attention: Viking excludes one participant with an outlying value from the monthly 22.5 mg efficacy analysis. Including that participant would produce 99% maintenance for that group. The monthly safety population is consequently 66, while the reported efficacy population is 65. Footnote to Viking’s table.

A disclosed exclusion is not evidence of misconduct. It does demonstrate why the statistical plan, exclusion rule and sensitivity analyses matter. One observation can move a small-group mean substantially. Replacing the headline’s 90% with the footnote’s 99% and presenting it as stronger confirmation would change the analysis being described.

2. The table that comes before any comparison

The useful question is which experiment answers the question at hand, rather than which medicine displays the highest percentage. The rows below are not head-to-head comparisons, and they do not share a common endpoint that can be ranked.

On small screens, scroll the table horizontally.

Sources: Viking, MariTide design, SURMOUNT-MAINTAIN, The Lancet, ATTAIN-MAINTAIN, Nature Medicine, SURMOUNT-4, JAMA, STEP 4, JAMA, STEP 1 extension.

Five checks that prevent a false ranking

Population. People who have already tolerated and responded to a medicine answer a different question from people starting treatment. Selecting responders can be appropriate for maintenance research, but it limits generalization.

Time. Twelve weeks and one year are not interchangeable. A favorable initial result could persist, improve or deteriorate. Extending a line beyond the observed data is a hypothesis, not an observation.

Starting point. Weight before any treatment and weight at the switch are different reference values. Percentage changes calculated from those different bases cannot simply be added together.

Statistical question. An analysis may estimate the effect under continued adherence, or account for treatment discontinuation and subsequent interventions. The estimand defines which effect is being measured. It is part of the clinical question, not statistical decoration.

Handling regain. When a protocol permits rescue therapy, the treatment of subsequent observations matters. Weight measured after rescue no longer reflects only the originally assigned strategy.

3. MariTide: longer follow-up, selected participants

MariTide, or maridebart cafraglutide, is Amgen’s investigational candidate combining GLP-1 receptor agonism with GIP receptor antagonism. That differs from dual GLP-1/GIP agonism. Amgen’s August 2026 update also describes Phase 3 studies switching participants from weekly semaglutide or tirzepatide to MariTide every eight weeks or quarterly. These are ongoing programs, not established outcomes. Amgen’s second-quarter 2026 update.

Entry into Part 2 of Phase 2 required at least 15% weight loss after one year and continued use of study treatment. The design included another year of monthly regimens, including lower doses, or dosing every 12 weeks, with placebo control. Monthly arms were 70, 140 or 420 mg, so not every dose was below the maximum studied in the first year. Amgen subsequently reported qualitatively that a large majority maintained their weight loss, with a low incidence of nausea and vomiting and no new observed safety signals. Design disclosed in June 2025; February 3, 2026 update.

The longer follow-up offers a different perspective from Viking’s new experiment. Selection, however, prevents translating the findings into a forecast for everyone starting MariTide. People who had not achieved that response or had not stayed on treatment are not represented by the same maintenance population.

“A large majority of participants” is also not equivalent to a particular mean percentage of previous weight loss preserved. Without a common quantitative definition, threshold and denominator, placing Amgen’s statement next to Viking’s 85% cannot identify a winner.

Switching studies are especially relevant to future competition because they address people who already achieved a result with one product and may follow a different treatment path. The development strategy has a recognizable commercial rationale. Its actual usefulness remains dependent on efficacy, safety and any eventual authorized indication.

4. Lilly: lowering the dose and switching formulations are separate experiments

SURMOUNT-MAINTAIN isolates the first question. After 60 weeks of tirzepatide, eligible participants were assigned to the maximum tolerated dose, a reduction to 5 mg, or placebo for another 52 weeks. Injections remained weekly. Of 441 participants entering the initial period, 378 were randomized to maintenance. Original report, The Lancet, online May 12, 2026.

In Lilly’s efficacy-estimand summary, the group continuing the maximum tolerated dose moved from 89.5 to 88.7 kg during maintenance; the 5 mg group moved from 89.0 to 94.6 kg. The latter retained a substantial portion of its initial reduction while regaining an average 5.6 kg. These are group means, not individual forecasts or a comparison with Viking. Lilly’s May 12, 2026 tables.

The distinction affects the wording of any conclusion. Performing better than placebo does not mean retaining every kilogram lost. A treatment can be clinically useful while preserving less weight loss than continuation of a higher dose. Assessing the acceptability of that trade-off requires the individual’s circumstances and health goals.

ATTAIN-MAINTAIN addresses switching to oral treatment. It randomized 376 participants following 72 weeks of SURMOUNT-5 to daily orforglipron or placebo. The cohorts previously receiving tirzepatide and semaglutide were analyzed separately. The primary endpoint concerned participants who had reached a plateau; from week 24, placebo participants regaining at least half of their previous loss could start orforglipron as rescue therapy. Those already on active treatment could instead re-escalate to the higher dose if they were taking the lower dose of the investigational formulation. Nature Medicine, May 13, 2026.

Lilly reports average regain of 0.9 kg after semaglutide and 5.0 kg after tirzepatide at one year in its efficacy analysis of mean weights. Earlier losses differed, so those values do not establish a universally superior sequence. A ratio calculated from aggregate means also does not necessarily equal the mean of individual ratios used for a maintenance endpoint. Lilly’s detailed results.

The publication additionally reports the primary endpoint in the plateau subgroup under the modified treatment-regimen estimand: 74.7% of previous loss preserved after tirzepatide versus 49.2% with placebo, and 79.3% after semaglutide versus 37.6%. This analysis did not credit the originally assigned treatment with any improvements after rescue therapy. It is a different view of the data: the population, measure and treatment of subsequent events must be specified. Those figures cannot be replaced by ratios calculated from the release’s mean weights. Aronne and colleagues, Nature Medicine, 2026.

ATTAIN-MAINTAIN did not include an arm continuing the injectable treatment. It therefore cannot establish equivalence between switching to oral treatment and remaining on the previous injection. Rescue therapy also complicates interpretation of placebo beyond week 24. The authors discuss both limitations. ATTAIN-MAINTAIN publication.

The FDA approved orforglipron, branded Foundayo, on April 1, 2026, for weight management in eligible adults alongside diet and physical activity. Describing it as entirely unapproved in the United States would therefore be outdated. Product approval does not turn every investigational switching sequence into advice to change treatment. FDA, April 1, 2026.

When a percentage really does count people: SURMOUNT-4

An earlier trial provides a particularly clear illustration of this article’s central distinction. After 36 weeks of initial tirzepatide treatment, SURMOUNT-4 randomized 670 participants to continued therapy or placebo for another 52 weeks. A secondary endpoint counted people retaining at least 80% of their initial weight loss: 89.5% of participants continuing tirzepatide met that threshold, compared with 16.6% on placebo. Aronne and colleagues, JAMA, online publication 2023.

Two percentages have different jobs here: 80% is the threshold for an individual’s retained result, and 89.5% is the share of the group reaching it. That figure cannot be numerically ranked against Viking’s 90% mean. Its value here is to show why the endpoint definition must come before the number. SURMOUNT-4, JAMA.

5. Novo Nordisk: lessons from continuing and withdrawing treatment

STEP 4 provides a useful reference. After 20 weeks of initial semaglutide treatment, 803 participants were randomized to continue weekly treatment or switch to placebo for 48 weeks. From randomization, mean weight fell 7.9% with continuation and increased 6.9% with placebo. Both groups continued lifestyle intervention. These percentages measure change from week-20 weight, not the share of previous weight loss preserved. STEP 4, JAMA, 2021.

The STEP 1 extension followed a subset of 327 participants for a year after both medication and structured lifestyle intervention were discontinued. Participants previously receiving semaglutide regained approximately two-thirds of their earlier reduction. This was an exploratory analysis, not a new randomization between maintenance regimens. STEP 1, Diabetes, Obesity and Metabolism, 2022.

Together, these studies support addressing continuity of treatment. They do not mean every person will regain the same amount, or that every possible dose reduction must fail. Complete withdrawal and continuation under a different regimen are separate experimental questions.

Novo’s position has also evolved since the early discussion of injectables alone. FDA documentation includes daily Wegovy tablets; in March 2026 the agency also approved 7.2 mg Wegovy HD for eligible adults. Historical doses used in STEP 4 and SURMOUNT-5 should be identified as trial-specific, rather than presented as the entire current portfolio. Oral Wegovy, FDA; Wegovy HD approval, March 19, 2026.

The practical distinction remains: oral does not mean occasional. The existence of an oral formulation of the same molecule does not authorize patients to substitute products, doses or schedules themselves. Indications and availability must be checked in the relevant jurisdiction.

6. Tolerability includes the people who do not finish

During Viking’s monthly maintenance phase, three of 66 participants discontinued early, including one because of an adverse event; vomiting was reported in five of 66, compared with two of 27 on placebo. These are small counts over 12 weeks, not proof of equivalent safety. VK2735-102 safety table.

Maintenance begins after previous exposure. People reaching it do not necessarily represent everyone who would start therapy. Favorable tolerability in the second period does not erase the first period’s experience or establish the practicality of the entire treatment journey.

A complete assessment requires three different counts: adverse-event discontinuations, discontinuations for any reason, and participants lacking usable final measurements. Looking only at the first may miss information about feasibility. Strategies used to manage nausea, vomiting or regain are also part of the treatment experience.

Approved labels contain product-specific warnings and contraindications, including serious gastrointestinal events, pancreatitis and other risks requiring clinical assessment. A known product’s profile cannot automatically be transferred to a different investigational candidate. Nor can failure to observe a signal in a small sample establish the absence of risk. FDA Wegovy information; Foundayo and Zepbound safety information.

Weight alone is not the only relevant outcome for an individual. Future evidence should also document physical function, body composition, quality of life and preservation of cardiometabolic benefits. A favorable signal on the scale does not automatically establish a benefit on clinical events.

7. From convenience to commercial advantage: the missing steps

For Viking, extended intervals create a potential avenue for differentiation. For Amgen, a long-acting design needs confirmation through advanced development. Lilly and Novo also have to provide credible options for people already on treatment. This is an industrial interpretation of the programs described, not a market-share forecast or equity valuation.

The step from theoretical convenience to adoption is not automatic. Some patients might prefer a less frequent schedule; others may find a daily routine easier. Establishing which approach improves persistence requires actual-use evidence, rather than an abstract preference alone.

“Savings” also needs qualification. Fewer administrations do not necessarily imply a lower annual price. The price per dose, payer contract, associated services and number of patients remaining on treatment all matter. Active ingredient requirements, devices, manufacturing capacity and logistics can affect producers differently. None of the clinical findings discussed here establishes a reduction in overall costs by itself.

A payer may ask whether the treatment pathway’s cost is justified by retained benefits and health outcomes. A company may ask how long it can continue serving the patient and at what margin. Those perspectives are connected but not identical. Greater persistence could support revenue without reducing total expenditure; a lower price could broaden access without automatically guaranteeing profitability.

Competition may therefore extend beyond maximum initial weight loss into induction, maintenance, comorbidity management and switching between formulations. This remains a possible development of the market. Available data do not already allocate those segments to the four stocks.

8. Evidence that would materially change the assessment

The first requirement is duration: whether early preservation lasts and how often patients need intensification or a different strategy. The second is distribution: how many retain a clinically meaningful threshold, alongside the group average.

The third is an appropriate comparator. To establish a strategy as an alternative to effective treatment, comparing it with that treatment is useful, under objectives and margins specified before analysis. Beating placebo and demonstrating noninferiority to active treatment answer different questions.

The fourth is statistical transparency: protocol, analysis plan, missing-data handling, outlier rules and sensitivity analyses. The fifth is clinical relevance: statistical significance must be interpreted together with effect size, uncertainty and implications for health.

The sixth is generalizability across clinical conditions, previous therapies and ability to follow the regimen. The seventh is execution: regulatory development, production, access and economic sustainability. A favorable exploratory result is one step in that process, not the completion of every step.

The question often raised in retail discussions — what would make a maintenance readout convincing? — is therefore a useful one. Those conversations do not measure scientific quality, clinical consensus or acquisition probability: the answer needs to come from the trials and their limitations.

Viking’s result makes a consequential development question more concrete: how much treatment is needed to preserve an established benefit? Today, the answer still depends on the molecule, prior treatment, patient and observation period. These studies help make those questions measurable without turning a promising average into a universal promise.