Meal replacements for weight loss and the return to regular food

Meal replacements can help within a weight loss program. Research also reveals weight regain and limits to diet quality after regular food returns.

日本語で読む（Japanese version）

Your lunch break is short, and buying a meal feels like another task. A bottle on the shelf offers an appealing possibility: could replacing lunch make both the day and weight management simpler? Convenience matters. But if you eat lunch again later because you are hungry, it becomes harder to say what the drink actually replaced. The useful question starts with the whole eating plan, rather than the bottle.

Some weight loss programs using meal replacements have produced greater weight loss than programs based on regular food. Those results include the way the diet was designed and the support offered alongside it.[1][3] They do not establish a weight loss effect for an ordinary protein drink by itself. A product, a replacement meal and a supported dietary program are different things to evaluate.

Losing weight, keeping it off after a program ends and improving diet quality are also separate questions. Long-term follow-up found weight regain, while a trial examining diet quality found greater improvement in the regular-food group.[7][17] Before deciding what to buy, it helps to look beyond the first replacement meal to the day you return to the family table. The research is most useful when the entire sequence remains in view.

Meal replacements and protein drinks ask different questions

In these studies, meal replacement means replacing part or all of the diet. An older meta-analysis of partial replacement examined programs that used products fortified with vitamins and minerals while retaining meals made from regular food.[16] A liquid format alone does not make another product equivalent to that intervention. The nutritional role of the product and its place in the day’s meals were part of what the researchers tested.

That review brought together randomized controlled trials in adults with overweight or obesity. In a randomized trial, chance determines which of the methods being compared a participant is assigned to follow. Both the partial-replacement group and the conventional weight loss diet group were prescribed the same energy intake.[16] The comparison therefore concerned different ways of arranging an energy-restricted diet. It was not a comparison between drinking something and making no dietary changes.

The practical details matter. The partial-replacement plans used one or two vitamin- and mineral-fortified replacement meals and retained at least one meal of regular food.[16] This leaves a different daily task from replacing everything you eat. Some meals are specified by the program, while others still require food choices. To understand whether a study is relevant to your plan, you need to know which meals it replaced and which meals remained.

Someone saying “I drink protein” has not yet described a meal replacement program. The drink might replace lunch, come before dinner, or be added after exercise. Those arrangements do not tell us the same thing about the rest of the diet. The evidence discussed here evaluates organized replacement programs, rather than the general role of protein or the effect of an additional drink.[1][16] A label on a bottle cannot supply the missing details of a day’s eating.

A useful starting point is to write down what the product would replace. Alongside it, write down the meals you would continue eating. It also makes the comparison with a paper more concrete: does the trial describe partial replacement, or does it describe an entire diet made from formula products?

Total diet replacement has its own structure. In the trial discussed below, a period using formula food alone was followed by reintroducing regular food, with scheduled behavioral support.[3] That design is different from skipping meals because work is busy and relying on drinks without a plan. The research result belongs to the tested sequence. The return to regular food was already part of it, rather than an extra step invented once weight had fallen.

The weight difference depends on the comparison

First, separate the amount of weight lost from the additional amount lost relative to another method. A 2019 systematic review and meta-analysis included 23 studies with 7,884 adults with overweight or obesity and compared weight change at one year.[1] Here, the review gathered eligible replacement-diet studies and combined their one-year weight results. The resulting figure describes the studies that met those conditions, rather than every possible use of a meal replacement.

This review excluded total diet replacement and diets providing less than 800 kcal per day. For diets incorporating meal replacements compared with alternative diets, the mean difference favored the replacement group by −1.44 kg, with a confidence interval of −2.48 to −0.39 kg.[1] For this comparison, the interval describes uncertainty in the estimated difference between diets, rather than the spread of individual weight losses.

The −1.44 kg figure is therefore neither a promise that everyone will lose that amount nor a statement that replacement users lost only that amount in total. Both groups were undertaking weight loss diets. The number describes the difference between them.[1] An advertisement can make a result sound much larger or smaller by leaving out the comparison. Asking “compared with what?” is a practical way to restore the part of the finding that gives the number meaning.

When a meal replacement diet with support was compared with another diet also offering support, the mean difference was −2.22 kg, with a confidence interval of −3.99 to −0.45 kg. When a replacement program with enhanced support was compared with an alternative diet and regular support, the difference was −6.13 kg, with a confidence interval of −7.35 to −4.91 kg.[1] In the latter comparison, the intensity of support differed as well as the dietary approach.

Those estimates do not isolate the contribution of the product from everything else in the program. The stronger-support comparison tested a different package of care.[1] For someone considering a program, this makes access to support part of the decision. A large published difference should prompt a look at what participants were offered, including the help around the diet, rather than a search for a bottle that seems to resemble the one they used.

The quality of the evidence also deserves space beside the headline result. Of the 23 studies, six were judged to have a low risk of bias across all assessed domains, while five had a high risk in at least one domain.[1] The average direction favored replacements, but the included studies did not all carry the same level of reassurance about bias. A collection of studies is useful evidence without making every estimate equally secure.

A newer meta-analysis, published in 2025, combined 70 studies including both total and partial replacement. It reported a weight difference of −3.35 kg compared with controls, with a 95% confidence interval of −4.28 to −2.42 kg.[2] This is larger than the earlier partial-replacement estimate, but the eligibility conditions differ. It should not be read as an updated value for an otherwise identical question. Total replacement and partial replacement were both included in the newer analysis.

Replacing more is not a personal prescription

A 2021 meta-analysis examined 22 studies involving 1,982 people with overweight or obesity. Comparing low-energy diets using meal replacements with low-energy diets based on regular food, it found a small effect favoring replacement. The standardized effect size was 0.261, with a 95% confidence interval of 0.156 to 0.365.[4] The authors’ description of the effect as small belongs with the result, even when the direction favors replacement.

A standardized effect size places differences from separate studies on a shared scale. The value 0.261 is not a number of kilograms and cannot simply be presented as one.[4] This distinction matters when several reviews appear in the same article or search result. A standardized effect and a weight difference answer related questions, but they are expressed differently. Comparing the raw digits would lose the meaning of the measurements.

In the same analysis, studies in which meal replacements provided at least 60% of total daily energy intake showed a medium effect. The effect size was 0.545, with a 95% confidence interval of 0.260 to 0.830.[4] This was an analysis of studies grouped by a feature of their diets. It does not prescribe a threshold that an individual should reach, or show that increasing your own replacement proportion to that point will reproduce the pooled effect.

The 2025 analysis likewise found larger effects on body measurements for total replacement than for partial replacement in its subgroup analyses.[2] Yet total replacement and replacing lunch alone differ in more than the share of meals replaced. Formula foods, the energy plan, support and duration are features to examine alongside that share.[2][3] A stronger average result is a reason to understand the fuller intervention before considering whether it fits your circumstances.

Other outcomes did not all move together. In the 2025 analysis, the difference in fat mass was −2.77 kg, with a 95% confidence interval of −3.59 to −1.6 kg. Waist circumference differed by −2.82 cm, with a 95% confidence interval of −3.51 to −2.12 cm. There was no significant difference in the hormone leptin.[2] Weight change therefore cannot stand in for a conclusion that every process or measure inside the body improved.

The people included in a review also shape what its estimate means. A review of ten trials involving 1,254 people with prediabetes and features of metabolic syndrome found a weight difference of −1.38 kg compared with conventional food-based weight loss diets. Its 95% confidence interval was −1.81 to −0.95 kg, and the certainty of the evidence was rated low to moderate.[5] These health characteristics were part of the comparison, rather than incidental details to omit.

In that review, the waist circumference difference was −1.17 cm, with a 95% confidence interval of −1.93 to −0.41 cm. The authors described the waist change as trivial.[5] A statistically detectable difference and a difference that feels substantial in daily life are not the same judgment. Keeping the authors’ assessment of magnitude helps avoid turning a favorable direction into a larger expectation than the evidence supports.

When comparing results, put the participants, replacement approach, comparator and measurement time beside each number. A one-year partial-replacement analysis and a pooled analysis including total replacement have different scopes.[1][2] If you keep a result as a reminder, its conditions deserve a place alongside it. Those details help you recognize whether the study resembles the program you are considering and prevent the largest number from becoming the whole story.

Numbers: meal replacements in context

23 studies and 7,884 adults: the size of the one-year review that excluded total diet replacement.[1]

−1.44 kg: the mean difference between diets incorporating replacements and alternative diets; confidence interval −2.48 to −0.39 kg.[1]

−2.22 kg: the mean difference when both dietary approaches included support; confidence interval −3.99 to −0.45 kg.[1]

−3.35 kg: the mean difference across 70 studies including total and partial replacement; 95% confidence interval −4.28 to −2.42 kg.[2]

−3.3 kg: the adjusted difference at three years after a total diet replacement program; 95% confidence interval −5.2 to −1.5 kg.[17]