Plant-based diets and weight loss: what the trials really show

Plant-based diets show weight-loss benefits in some trials. The size of the difference depends on the diet, its comparison, support, and follow-up.

日本語で読む（Japanese version）

You reach for tofu instead of meat while planning dinner. There will be more vegetables, but the rice and cooking oil will stay much the same. The shopping basket looks different. Whether the number on the scale will change is a separate question. Calling the meal plant-based does not answer it.

Vegetarian and vegan diet trials do include weight-loss results. Yet many tested more than the removal of animal foods: a low-fat menu, dietary instruction, and support over a defined period often came together. For a household planning meals in Tokyo, those details are more useful than the label on the diet. They show what participants actually changed, and what a similar change at home would involve.[2][3]

The question concerns how weight changes when people change their diets, rather than which group of eaters deserves approval. A substantial difference in a short trial may not persist in a longer comparison. The evidence includes pooled results favoring vegetarian diets and a longer trial that found no significant advantage over its comparison diet. Both belong in the picture when deciding what to put on the table.[4][16][17]

A plant-based label does not define the whole menu

Researchers use the term “plant-based” to describe several different eating patterns. A vegan diet excludes animal foods, while a lacto-ovo-vegetarian diet includes eggs and dairy. Other trials have compared diets that include fish or reduce animal foods without excluding them. Reading all of these as studies of the same menu loses a distinction that matters for applying their findings.[2][4]

The same problem appears in everyday cooking. A dinner with plenty of vegetables is not necessarily a dinner without animal foods. Neither description tells you how the rest of the meal is prepared. Before borrowing a study’s result, decide whether you are trying to follow its dietary intervention or simply looking for an alternative main dish. Those are different projects, and the research cannot provide the same prediction for both.

A randomized trial in adults with overweight compared vegan, vegetarian, fish-eating vegetarian, semi-vegetarian, and omnivorous diets. Randomized means participants were assigned to their groups by chance. The shared basis of all the diets was low fat and a low glycemic index, an indicator relating to the rise in blood glucose after eating. Calorie restriction was not emphasized, but food choices still followed a specified plan.[2]

That distinction changes how the trial should be read. Not emphasizing calorie restriction does not mean that every possible menu was tested. The intervention changed the composition of meals as well as their animal-food content. A result obtained under those conditions cannot tell us what would happen if someone removed meat while leaving every other part of their eating pattern untouched.[2][3]

At home, it helps to look separately at the main dish, vegetables, staple food, and cooking oil. Tofu and beans can appear in the space previously occupied by meat, but the rest of the plate still needs a decision. This is a way of translating a whole-diet intervention into a practical question. It is not a claim that a particular tofu dish or bean recipe has a proven weight-loss effect.

A familiar meal can make that distinction easier to see. Imagine replacing the meat in dinner but keeping the same side dishes and preparation method. You have made an identifiable change, though not necessarily the full change used in a low-fat trial. Imagine instead reconsidering the main dish and its cooking method together. That plan resembles a broader dietary intervention more closely, without making it equivalent to the study.

The neighboring topics of Mediterranean diets and nuts concern their own comparisons. Here the focus is the combination of foods described as vegetarian or vegan and the diet used as its comparator. An ingredient may appear in several eating patterns, but that does not make the questions interchangeable. A trial of an entire diet cannot be reduced to a verdict about every ingredient it contains.

Average differences are useful, but they are not promises

A meta-analysis combines results from several studies to estimate an overall effect. One analysis of randomized trials comparing vegetarian and nonvegetarian diets included 12 trials and 1,151 participants. The median intervention duration was 18 weeks. This provides a broader view than a single small trial, while still reflecting the particular studies that were available to combine.[4]

The pooled difference in weight change favored the vegetarian groups by −2.02 kg. Its 95% confidence interval was −2.80 to −1.23 kg. A confidence interval expresses uncertainty around the estimate; it is not the range of weight losses experienced by individual participants. The estimate itself is a difference between diets, not the total amount lost by the vegetarian participants.[4]

That last point is easy to miss in a headline. Saying that people lost about that amount on a vegetarian diet removes the comparison group from the sentence. A comparison group might also have lost weight. To understand the advantage estimated by a trial or meta-analysis, keep the question intact: how much did weight change with this diet relative to the other diet? The scale reading within a group and the difference between groups answer separate questions.

The vegan subgroup in the same analysis had a pooled difference of −2.52 kg, with a 95% confidence interval of −3.02 to −1.98 kg. For diets including eggs and dairy, the estimate was −1.48 kg, with an interval of −3.43 to 0.47 kg. That second interval includes zero, the point representing no difference. Although the abstract describes weight reduction, the interval does not establish a clear advantage for that subgroup.[4]

It would be misleading to smooth those findings into a statement that every vegetarian pattern works equally well. It would also be misleading to conclude that diets including eggs and dairy cannot lead to weight loss. The subgroup estimate is uncertain. It describes the included comparisons rather than setting an absolute rule about what any household can achieve with a particular menu.[4]

Another meta-analysis focused on vegan diets lasting at least 12 weeks in people with overweight or type 2 diabetes. It included 11 trials and 796 participants. Compared with control diets, the pooled weight difference was −4.1 kg, with a 95% confidence interval of −5.9 to −2.4 kg. This is a larger estimate, but its eligibility criteria, participants, and diet comparisons differed from those in the earlier analysis.[16]

These figures should therefore sit beside their descriptions, rather than become competing promises. The larger estimate does not show that the effect of vegan eating grew over time or that the later analysis offers a better personal target. The analyses assembled different evidence. Comparing the numbers without their populations and conditions would give a precision that the comparison itself does not support.[4][16]

The second analysis found no effect on blood pressure, high-density lipoprotein cholesterol, or triglycerides. A favorable weight result did not mean that every measured health indicator improved. Its authors also noted that differences in macronutrient composition and energy intake could have contributed to the effects. Macronutrients are the major nutrients considered in a diet’s composition; the authors’ point was that the vegan label did not fully explain the intervention.[16]

For meal planning, the useful question is therefore broader than whether a diet ranks above another by name. Ask which foods changed, which comparison was used, and how uncertain the estimate remains. An average can help frame expectations while leaving room for a person’s actual circumstances. It should not become a required loss against which to judge the success of a dinner plan.

Short trials measured more than the scale

A 16-week randomized trial assigned 75 participants with overweight to a plant-based vegan diet or a control diet. Researchers measured body composition as well as body weight, using dual-energy X-ray absorptiometry. Body composition concerns the components of body weight, including fat. The trial therefore looked beyond the single total displayed by a household scale.[5]

The reported treatment effect on weight was −6.5 kg, with a 95% confidence interval of −8.9 to −4.1 kg. For fat mass, it was −4.3 kg, with an interval of −5.4 to −3.2 kg. Keeping these measurements separate matters: the total change in body weight cannot be described as if it were entirely a change in fat mass. The study reported distinct estimates for distinct outcomes.[5]

Increased plant protein intake was associated with a reduction in fat mass, and decreased animal protein intake was also associated with that reduction. These associations do not isolate the effect of plant protein. Assignment to a dietary group and an association between nutrient changes and an outcome are different pieces of evidence. Participants were not assigned to change plant protein alone while every other feature of the diet remained fixed.[5]

This can inform how you think about replacing a main dish without supplying a protein prescription. The task is to consider which foods will take over the role of the previous main dish within the meal. The abstract does not provide a personal intake target or a food-by-food serving guide. Detailed questions about protein amounts and muscle belong to their own discussion, rather than being answered by stretching this trial’s weight findings.

A newer trial followed 90 adults with overweight or obesity for 12 weeks. It compared vegan, lacto-ovo-vegetarian, Mediterranean, and EAT-Lancet diets with a control diet based on World Health Organization recommendations. Of the participants, 85 completed the intervention. Each plant-based group showed reductions in body weight, waist circumference, fat mass, and energy intake compared with the control group. The vegan group’s reported weight reduction was −6.7%.[1]

However, the trial found no substantial between-group differences in blood glucose, insulin, lipid levels, blood pressure, or appetite-related traits. Weight loss did not establish that hunger or every metabolic measure had improved. The abstract also did not give a confidence interval for the reported percentage weight loss. A precise-looking percentage does not, by itself, tell a reader how precisely the effect was estimated.[1]

The reduction in energy intake is part of the result, not a detail to remove from the explanation. If the dietary pattern and energy intake changed together, the trial does not show that the pattern produced the same weight response with unchanged intake. It tested the assigned diets as participants followed them. Separating what was measured from what we would like to infer keeps the result useful.[1]

In the smaller trial comparing several low-fat diets, participants were followed for 6 months. Each group began with 12–13 participants; 50 people, or 79% of those enrolled, completed the study. Weight change was −7.5% in the vegan group and −3.1% in the omnivorous group. The difference between these groups was statistically significant. These are percentage changes over the study period, rather than kilogram differences.[2]

Support was part of that intervention too. Initially, the groups other than the omnivorous control group attended weekly meetings. The control group attended monthly meetings supplemented by weekly email lessons. Later, all groups attended monthly meetings. The early form of contact was therefore not identical across groups, which makes it harder to attribute the entire difference simply to the presence or absence of animal foods.[2]

When imagining how a trial might translate into a busy week, retain that support in the description. Participants did not merely receive a dietary label and disappear into their kitchens. Meetings and instruction were part of the setting in which the reported changes occurred. The abstract does not separate how much of the result came from the menu and how much came from the accompanying contact, so neither contribution should be invented.

Numbers: plant-based diets and weight change

−2.02 kg: the pooled between-group difference from 12 vegetarian-versus-nonvegetarian trials; 95% confidence interval, −2.80 to −1.23 kg.[4]

−4.1 kg: the pooled difference from 11 vegan diet trials in people with overweight or type 2 diabetes; 95% confidence interval, −5.9 to −2.4 kg.[16]

−4.3 kg: the fat-mass treatment effect in a 16-week trial of participants with overweight; 95% confidence interval, −5.4 to −3.2 kg.[5]

−7.5% and −3.1%: weight changes at 6 months in the vegan and omnivorous groups of a small trial comparing low-fat diets.[2]

−2.39 kg and −2.03 kg: weight changes at 12 months in another trial; there was no significant difference between the groups.[17]