Why cocoa flavanol studies are not about chocolate bars

Read the methods section of a cocoa trial and the intervention is rarely chocolate.

It is typically a STANDARDISED COCOA EXTRACT in a capsule, or a specially manufactured high-flavanol cocoa DRINK, prepared so that the flavanol dose is known and constant. That is not a criticism — it is exactly what a trial should do, because an intervention whose active content varies between servings cannot support a conclusion.

The consequence is that the study is about a defined quantity of a defined compound class. It is not about chocolate, and the paper generally does not claim to be. The claim that it is about chocolate is added afterwards, by somebody else.

So the honest reading of a positive cocoa finding is: this quantity of these compounds, delivered this way, produced this measured change in this population over this period. Every one of those clauses is doing work, and the one that gets dropped first is the first.

Cocoa flavanols — principally epicatechin, catechin and the procyanidins they form — are abundant in the raw bean and are progressively destroyed by almost everything done to it.

FERMENTATION reduces them substantially. It is also non-negotiable: an unfermented bean does not taste of chocolate, and the flavour precursors the whole subject depends on are made in that step.

DRYING and ROASTING reduce them further. Higher roast temperatures and longer roasts cost more.

ALKALISATION — Dutching — is the most destructive single step. Treating cocoa with alkali darkens the colour, softens the acidity and improves dispersion, and it removes a large proportion of the flavanols. A heavily Dutched cocoa powder can retain a small fraction of what the bean started with.

AND DILUTION does the rest. A 70% bar is 30% sugar and its cocoa content includes cocoa butter, which contains essentially no flavanols. The flavanols live in the non-fat cocoa solids, which are a smaller fraction of the bar than the percentage suggests.

The result is that two bars showing the same number on the front can differ severalfold in flavanol content, and nothing on either wrapper distinguishes them.

Three reasons, and none of them is concealment.

IT IS EXPENSIVE TO MEASURE. Flavanol quantification requires chromatography, per batch, and cocoa is an agricultural product whose composition varies by harvest.

IT IS NOT STABLE. Flavanols degrade over time and with storage conditions, so a figure measured at manufacture is not a figure at the point of eating.

AND IN MOST JURISDICTIONS DECLARING IT WOULD BE A HEALTH CLAIM, which is regulated. The EU has an authorised claim relating cocoa flavanols to normal blood-vessel elasticity, with a specified daily intake and specified conditions of use — and that authorisation is precisely the point: the claim is tied to a defined intake of a defined preparation, not to chocolate generally.

Which means a reader cannot tell from a wrapper how much of the relevant compound they are getting. Not approximately, not within a factor of two. That is the state of it.

The question underneath all of this is "how much dark chocolate should I eat to get the benefit", and there is no honest answer.

It would require knowing the flavanol content of a specific bar, which nobody declares and which varies by batch; knowing how much of that survives digestion, which varies between people and is affected by what else is eaten; and knowing what dose produces what effect over what period, which the trials establish for their preparations and not for a bar.

Three unknowns multiplied together do not produce a number. They produce a number-shaped object, and this catalogue has no field anywhere in its data model that can hold a recommended dose — which is a structural decision rather than a stylistic one.

WHAT CAN BE SAID: less-processed cocoa products retain more flavanols than heavily processed ones, so natural cocoa powder generally retains more than Dutch-process, and cocoa nibs more than either. Higher cocoa content means more non-fat cocoa solids and therefore more flavanols in principle. And none of that tells you how much is in the bar in your hand.

Four questions, in order, and the first two dispose of most of them.

WHAT DID THEY GIVE PEOPLE? If it was an extract, a capsule or a specially prepared high-flavanol drink, the study is not about chocolate. This alone accounts for the majority of misleading coverage.

WHAT DID THEY MEASURE? A biomarker, or an outcome? Flow-mediated dilation, blood pressure at one visit and a lipid fraction are biomarkers. Heart attacks, strokes and deaths are outcomes. The distance between them is the whole game.

HOW MANY PEOPLE, FOR HOW LONG? Cocoa trials are frequently small and short. A three-week study in forty people is a reason to do a larger study.

WHO PAID? A substantial share of cocoa flavanol research is funded by confectionery manufacturers. That does not make it wrong and it is material, which is why this catalogue records funding on its source register rather than leaving a reader to find out.

Ask those four and almost every chocolate health story resolves into something considerably smaller and rather more interesting than its headline.