Vegan Nutrition

Tofu on top, cheddar at the bottom: the index that redrew the protein map

A researcher at the University of Cambridge folded emissions, water pollution, land use, sodium, saturated fat, cholesterol and price into a single score from 0 to 40. When the numbers landed, the ten plant proteins assessed held ten of the top twelve places — and the first six were all plant-based.

Ask ten people which protein source is best and you will hear ten answers — nearly all of them anchored to some label, some nutrition table, some green seal on the front of a pack. The problem is not a shortage of information. It is that each of those instruments measures one thing only. The British traffic-light system sees fat, sugar and salt, but is blind to deforestation. The carbon footprint label sees the climate, but cannot tell a food loaded with sodium from one that carries none. And a label that measures a single thing, read in isolation, can be technically correct and still misleading.

That discomfort is where the BPI Score begins. It was set out by Chris Macdonald, of the University of Cambridge, in a paper published in July 2026 in Frontiers in Sustainable Food Systems. The proposal is easy to state and hard to execute: instead of twenty partial labels, one score, built from eight indicators that cut across health, environment and household budget.

It is worth being clear from the outset about what this work is and is not. It is a Perspective article, an editorial category intended to set out and argue for a conceptual framework. It is not a clinical trial, not a systematic review, not a meta-analysis. What Macdonald does is combine established, existing datasets according to a method of his own, stating openly that this is a first version. He also discloses that he founded the Better Protein Institute (BPI), the organisation the index is named after — a transparency readers should weigh when judging the conclusions.

One score, built in two halves

Twenty high-protein foods went into the assessment: ten of plant origin (tofu, beans, lentils, chickpeas, quinoa, chia, sunflower seeds, walnuts, Brazil nuts and peanuts) and ten of animal origin (eggs, chicken breast, salmon, tuna, cod, pork, lamb, beef, mozzarella and cheddar).

The unit of comparison is the most important methodological decision in the whole paper — and the most honest. Everything is calculated per 100 grams of protein, not per 100 grams of food. It is the ruler that defenders of animal agriculture usually ask for, because it neutralises the obvious advantage of comparing a steak with a lettuce leaf. Here every item delivers the same quantity of protein, and the question becomes: what else comes in the package?

The score splits in two. The Planet Score is worth up to 16 points and draws on four indicators taken from the dataset published by Clark and colleagues in PNAS in 2022, built from 57,000 food products: greenhouse gas emissions, water pollution, land use and blue water use — the freshwater actually consumed, rather than the rain that falls on the field. The People Score is worth up to 24 points and mixes health (sodium, saturated fat, cholesterol and the question of complete protein) with accessibility (price and retail availability). Maximum possible total: 40 points.

Each indicator is scored by quintiles on a 0-to-4 scale: the best 20% of each list take 4 points, the worst 20% take none. It is a deliberately simple system — and, as we will see, that is precisely where one of the limitations acknowledged by the author sits.

What happened when the numbers landed

The result is more lopsided than anyone used to the nutrition debate would expect. There was no technical draw, no grey zone, no familiar "it depends". There was an almost clean separation between two blocs.

Taken separately, each half repeats the pattern. On the Planet Score, tofu scored 16 out of 16 — the absolute ceiling, the only food to take full marks on all four environmental indicators. Beef finished on 1 point out of 16. On the People Score, tofu made 20 out of 24; lamb, 10.

The scale of the gap

The figures behind that separation are not marginal differences. On several metrics the distance runs to two orders of magnitude — around a hundredfold.

The water case deserves attention precisely because it is the counterpoint that lends the rest its credibility. If every indicator pointed the same way without exception, it would be reasonable to distrust the method. That is not what happens: nuts drag the plant average down on water consumption, and the paper records it plainly.

What travels with the protein

On the health side, Macdonald made a decision worth explaining, because it is the opposite of what campaigning rankings usually do. He chose not to penalise foods for lacking micronutrients. Marking beef down for being low in vitamin C would be misleading, he argues, since nobody eats steak expecting vitamin C — fruit and vegetables cover that role in the meal. The focus therefore falls only on what comes with the protein and may do harm: sodium, saturated fat and cholesterol, the factors linked to obesity and cardiovascular disease, the world's leading cause of death. The values were drawn from the Nutritionix database.

On sodium, the plant-animal split reappeared brutally: beans register 11.11 mg per 100 g of protein; mozzarella, 2,549.55 mg — 229 times more. Beans, walnuts, Brazil nuts and lentils form the lowest group; eggs, tuna, cheddar and mozzarella the highest.

On saturated fat the picture is more nuanced, and again this works in the index's favour: the best and worst performers are both animal foods. Cod records 0.47 g per 100 g of protein; cheddar, 84.74 g. Among the plants, beans sit in the good quintile (1.11 g) and Brazil nuts in the bad one (78.32 g).

On cholesterol there is no softening the picture: every plant food registers zero. Absolute zero, without exception. To avoid turning the indicator into an unfair on-off switch, the author built an internal scale across the animal products, recognising that tuna (152.94 mg) sits a long way from eggs (2,960.32 mg).

And the price?

Here the index usefully contradicts the received wisdom on both sides of the debate. Converted to the same standard — cost per 100 g of protein, with prices collected at Aldi, Britain's cheapest chain — the top and bottom accessibility quintiles each contain plant and animal foods. The cheapest product in the whole sample is plant-based: lentils. So is the most expensive: quinoa.

The lesson is that "plant food is expensive" and "plant food is cheap" are equally false as generalisations. What exist are expensive and cheap foods within each category — and pulses sit consistently among the cheapest protein sources available to any consumer, anywhere.

What was left out — and why it matters

The most revealing part of the paper is not the results but the list of what this first version could not measure. Because everything left out would push the result in the same direction.

Animal welfare and ethics do not enter BPI Score V1. Macdonald discusses the absence explicitly and argues that including them, even modestly, would widen the gap between the two groups further still. He notes that more than 80 billion land animals are killed every year for meat production, most of them in industrial systems, and that in some sectors of animal agriculture the freedom to roam is incompatible with the product itself. The paper contains a cool, methodical description of the dairy cycle in the United States: artificial insemination, pregnancy, separation of the calf shortly after birth, mechanised milking on concrete floors, re-insemination, the cycle repeating until yield falls — and then slaughter.

He also identifies a dilemma that rarely surfaces in the debate: the space trap. More room improves welfare for confined animals — there are some 8.3 billion laying hens worldwide, a figure close to the human population, and most live in cages offering less space than a sheet of A4 paper. But expanding the footprint of an already resource-hungry industry shrinks the space available to wildlife. Pasture expansion for cattle is the main driver of deforestation in the Amazon. It is a game in which someone always loses: either farmed animals or wild ones. Macdonald's conclusion is blunt — the only fair way to score ethics in animal agriculture may be a blanket deduction across the sector.

Then there is the carbon opportunity cost of land, possibly the most striking figure in the entire paper. If humanity moved entirely to plant proteins, global agricultural land would fall from more than 4 billion hectares to around 1 billion. If that released land were returned to nature, the potential carbon sequestration would exceed 500 billion tonnes of CO₂. For scale: annual fossil fuel emissions for the whole planet total less than 40 billion tonnes. That is more than a decade of global fossil emissions — and it excludes the recovery of biodiversity, which does not even enter the calculation.

"our protein choice is a referendum on the compassion we are willing to show" Chris Macdonald, University of Cambridge

Where the index is fragile

No serious reading of a scientific paper stops at the favourable results. Macdonald devotes a substantial part of the text to listing the limitations of his own creation, and that self-criticism is what separates an index from a piece of advertising.

The most serious limitation is geographical. Prices came from a single British supermarket chain and availability was defined by a binary test — whether the product is on sale in UK retail. The author classifies both indicators as a localised proof of concept, not a globally generalisable model. Agricultural subsidies, import tariffs, socioeconomic inequality and food deserts change the relative cost of every food dramatically from country to country. For readers outside the UK, that caveat is the one to keep in mind.

The availability criterion, in fact, turned out redundant: all twenty items scored "yes". The author kept it because it will be decisive in future, when precision fermentation proteins and cultivated meat move from laboratory to shelf.

The second limitation is the scale. A five-point ruler flattens the extremes. Beef is so far outside the distribution on emissions that "zero points" fails to express what it actually represents — awarding it the same minimum as a food ten times less polluting distorts the reading. A wider scale is planned for V2, already in development, alongside regionally weighted data or a purchasing-power-parity framework.

What this means in an ordinary kitchen

An index built from British retail prices arrives at a conclusion that grandmothers across Latin America, the Caribbean, South Asia and West Africa reached long ago. Beans sit in the top quintile of the overall ranking, register the lowest sodium level in the entire sample, consume almost no freshwater (0.07 L per 100 g of protein), contain zero cholesterol and rank among the cheapest protein sources in the world.

The only disadvantage beans carry in the BPI Score is the complete-protein criterion — precisely the criterion that the rice-and-beans plate has resolved for centuries, effortlessly and without planning, simply by putting a grain alongside the pulse. What nutrition calls amino acid complementarity, most of the world calls lunch.

There is a useful irony there. While much of the debate about sustainable protein organises itself around products that are new, expensive and ultra-processed, the food this index crowns is tofu — invented two thousand years ago — and the runner-up group is led by the most ordinary pulse on the planet. For most people, the protein transition so widely discussed requires buying nothing that is not already in the cupboard.

What to do with a score

Macdonald is clear about the objective: the BPI Score is not meant to become another front-of-pack label. The ambition is different — to build literacy, the capacity to see the whole system behind a food rather than react to a single isolated indicator, chosen by industry precisely because it is the most flattering.

It is a modest and an ambitious goal at once. Modest because it does not promise to change behaviour on its own — the author himself acknowledges, citing earlier research, that environmental labels produce mixed and discreet effects on purchasing decisions. Ambitious because it changes the question. It is no longer about which food carries the most protein, but about what each 100 grams of it costs the body, the land and other beings.

For that question, on the evidence of this first version, the answer has already emerged: the six best available answers are all plant-based.