Weight loss

Weight Loss

Do Online Weight-Loss Programmes Actually Work? What the Evidence Says

Dr. Himalay Agarwal

Dr. Himalay Agarwal

Consultant Internal Medicine

Yes, modestly, and not automatically. Online weight-loss programmes add roughly one to three kilograms of loss compared with minimal care, the gain shrinks with time, and weekly in-person care beat them by 2.1 kg in one trial. A meta-analysis of 20 trials covering 2,318 adults with obesity found that mobile-delivered programmes produced 2.35 kg more weight loss than control groups. The size of the gain depends mainly on what the programme was compared against, and it narrows as the months pass.

Key Takeaways

  • Across 20 trials and 2,318 adults, mobile-delivered programmes produced 2.35 kg more loss than controls (95% CI 1.87 to 2.84).
  • The same analysis put the difference at 2.25 kg at three to four months, 2.66 kg at six months and 1.23 kg at 12 months.
  • In the POWER trial (n=415), remote coaching produced 6.1 kg at six months and 4.6 kg at 24 months, against 1.4 kg and 0.8 kg in self-directed care.
  • Seven of 11 web-based trials lost at least 20% of their participants, so results based only on those who finished deserve less trust.
  • India's 12-week mDiab trial found 1.0 kg against 0.3 kg with usual care, and both Indian trials aimed at preventing diabetes rather than treating obesity.

How much weight do these programmes actually add?

Between roughly one and three kilograms more than the comparison group, and which end of that range you get depends on what the comparison was. That 2019 meta-analysis of mobile interventions found 2.35 kg more loss than controls. A 2012 Cochrane review of computer-based programmes found a similar pattern against minimal care, and a reversal against in-person care. These figures read two ways.

A digital programme compared against a pamphlet or a waiting list looks effective, while the same kind of programme compared against weekly face-to-face appointments did 2.1 kg worse at six months in the single trial that tested it. Side by side, the figures are:

Mobile programme versus control

  • 2.35 kg more (95% CI 1.87 to 2.84), pooled across follow-ups, from 20 trials and 2,318 adults

Computer programme versus minimal care

  • 2.5 kg more at three months (95% CI 1.6 to 3.4), five trials, 430 people, moderate variation between trials (I²=44%)
  • 1.5 kg more at six months (95% CI 0.9 to 2.1), two trials, 511 people
  • 1.8 kg more at 12 months (95% CI 0.5 to 3.1), one trial of 474 people
  • 1.1 kg more at 12 months (95% CI −0.3 to 2.6) once the arm that included in-person contact was excluded, 280 people

Digital versus weekly in-person care

  • 2.1 kg worse at six months (95% CI 0.8 to 3.4), one trial of 319 people

And once the Cochrane reviewers removed the arm that included in-person contact, the 12-month advantage fell to 1.1 kg with a confidence interval from −0.3 to 2.6, which includes no difference at all.

What happens to the effect over time?

It shrinks. The 2019 analysis reported 2.25 kg at three to four months, 2.66 kg at six months and 1.23 kg at 12 months, so more than half the advantage seen at six months had gone by the one-year mark. The 2012 Cochrane figures show the same pattern, with the largest difference at three months and the 12-month estimate resting on a single trial.

The strongest long-term evidence comes from the 2011 POWER trial, which followed 415 participants for two years. Its remote arm combined telephone coaching, a study website and email rather than an app.

Remote coaching produced 6.1 kg of loss at six months and 4.6 kg at 24 months, against 1.4 kg and 0.8 kg in self-directed care. The maintenance figures matter more than the averages: at 24 months, 38.2% of the remote-coaching group were still at least 5% below their starting weight, against 18.8% of those managing alone and 41.4% of a group that also had in-person sessions. Adding in-person sessions to the remote arm was not significantly better than the remote arm by itself.

Does dropout change what the numbers mean?

Yes, and it is the main reason to read any programme's published results carefully. In a 2019 review of web-based trials, seven of 11 lost at least 20% of their participants, and pooled heterogeneity across those trials reached I²=94%. This was driven mainly by the different comparators each programme used. The type of control group alone explained about four-fifths of the variation between trials.

Where a trial reports only the people who finished, the result describes a self-selected group of persisters rather than everyone who signed up, and it deserves correspondingly less trust.

The POWER trial shows what the stricter approach looks like. It used weights measured by staff rather than figures participants supplied, modelled outcomes for everyone randomised rather than only completers, and still assessed 94.5% of participants at 24 months. Even there, disengagement from the programme itself reached 13.0% in the remote arm, so attendance and outcome assessment are separate things.

The weights people report themselves are the other correction. In a 2015 analysis of the POWER data covering the 191 coached participants who reported their weight at 24 months, participants' own figures understated their measured weight by 1.1 kg at 24 months and misclassified 9% of them on whether they had reached the 5% target.

Which components change the result?

Human contact adds a little, and no single ingredient has turned out to be essential. A 2025 component network meta-analysis of 68 trials and 151 treatment arms estimated that specialist contact added 1.08 kg of loss at six months (95% CI 0.17 to 1.99), and an appraisal of those 68 trials rated 23 of them at high risk of bias. No individual component was consistently essential across the analysis, and the "best-bet" combination it identified, pairing information with specialist contact and incentives, was exploratory and has never been tested intact as a programme.

The head-to-head evidence on coaching is more deflating than the component analysis suggests. A 2020 randomised trial of 1,298 participants compared feedback from a human dietitian against automated personalisation and found that the human feedback increased engagement but not weight loss at 24 weeks.

Two details soften that result. The trial's main outcome was weight that participants reported themselves, the measure shown above to understate real weight. The authors also found that dietitian feedback kept people using the platform for longer and was linked to a higher chance of losing a clinically meaningful amount.

More frequent self-monitoring has been associated with greater loss, but in analyses observed inside trials rather than randomised comparisons of monitoring frequency, so they cannot establish that monitoring more often causes greater loss.

Evidence is sparse on several design choices:

  • Synchronous versus asynchronous contact
  • Group versus individual delivery
  • Tailoring
  • Programme duration and intensity

Specialist contact is the one component with a measured effect attached to it, so who provides that contact is the detail worth checking.

What has been tested in Indian adults?

Two controlled trials, both aimed at preventing diabetes rather than treating obesity:

  • The 12-week mDiab trial, published in 2019, combined an app with weekly coaching and found 1.0 kg of loss against 0.3 kg with usual care, with 15% reaching 5% loss against 9% in the control group.
  • A 2025 Tamil Nadu pilot enrolled 610 participants and analysed 567 of them, reporting intervention-control differences of 2.40 kg in urban participants and 1.19 kg in rural participants, with 7% dropout.

Both results are consistent with the international picture of a modest advantage, and neither establishes durability, because both lasted a few months. No completed ICMR-linked randomised trial of a digitally delivered obesity-treatment programme was identified.

Both Indian trials also recruited people at risk of diabetes rather than people seeking obesity treatment, so what they tested is closer to blood-sugar risk than to weight management. An Indian adult paying for a weight-loss programme is reading evidence generated either in that population or abroad.

What to ask before paying for a programme

Four details separate a programme with trial evidence behind it from one with marketing behind it, and all four are answerable before any money changes hands:

  • Are weights measured by staff, or entered by members themselves? In the POWER trial, people's own figures understated their measured weight by 1.1 kg.
  • Is the published result based on everyone who enrolled, or only on those who finished? Seven of 11 web-based trials lost at least a fifth of their participants.
  • Who provides the human contact, and how often? Specialist contact is the one component with a measured effect.
  • What happens when the paid contact ends? The advantage in the pooled trials fell by more than half between six and 12 months, and little evidence follows people after contact stops.

None of those four answers needs a trial to interpret, and a provider unwilling to give them has answered a different question.

Cost raises its own questions, starting with what a three-month care fee should cover. If the reason for shopping around is a metabolic result rather than a number on the scale, a clinician-led assessment with Voy answers something no programme comparison can: what your own markers need before a plan is chosen at all.

Dropout appears across these trials more consistently than any component that works, so the stretch worth preparing for is the one where progress slows and logging lapses, which is where weight-loss plateaus begin.

Weighing up a programme because of a blood test result? A Voy clinician can review your markers before you choose a plan. Book an assessment.

This article is for general information and education only and is not medical advice, diagnosis, or treatment. GLP-1 and other medications referenced are prescription-only and are appropriate only for certain people under the supervision of a qualified clinician. Do not start, stop, or change any medication based on this article. Please consult a registered medical practitioner about your individual circumstances. Information reflects what was available at the time of review and may change.

Related articles

4.7 / 5 (28,937+ reviews)

Still confused?

App Store
Download on theApp Store
Google Play
Get it onGoogle Play