The Zutphen Elderly Study And Cocoa, Explained

The Zutphen Elderly Study is an observational cohort of elderly Dutch men, and the cocoa paper drawn from it is a 2006 analysis of 470 of them. It is observational, which means it can describe what went together in that population and cannot show that one thing caused another. The authors say as much, and name residual confounding as a major concern.

It gets cited constantly and read badly. What follows is the design, the numbers, the fifteen-year follow-up and the specific confounding the authors could not remove, because the limitations are the most useful part of the paper and the part that never makes it into the citing article.

What the study was, and who was in it

Buijsse B, Feskens EJM, Kok FJ and Kromhout D, Archives of Internal Medicine, 2006 analysed 470 men from the Zutphen Elderly Study, a cohort recruited in the Dutch town of Zutphen in 1985. The men were aged 65 to 84 at recruitment, with a mean age in the tertiles of about 71 to 72. All 470 were free of chronic disease at baseline and none were taking antihypertensive medication.

The design is two analyses stitched together. Blood pressure was examined cross-sectionally, at one point in time. Mortality was followed prospectively for fifteen years, from 1985 to 2000, covering 4,908 person-years, over which 314 of the men died, 152 from cardiovascular causes. Men only. One small Dutch town. Born, in most cases, before 1920.

What “cocoa intake” meant in practice

Nobody was given cocoa. Habitual diet was assessed by the cross-check dietary history method in 1985, 1990 and 1995, and cocoa intake was then estimated by multiplying each man’s reported consumption of 24 cocoa-containing foods by the cocoa content of those foods. Chocolate accounted for roughly two thirds of the total. The cocoa content itself came from food composition tables and labels rather than from any measurement of what the men ate.

The method is worth understanding before trusting the exposure. As the Seven Countries Study research group, sevencountriesstudy.com, 2019 describes it, a dietician interviewed each man at home for one to two hours, with whoever prepared the food usually present, and cross-checked a week of household purchases against what was reported. In subsamples of about 50 men, that method estimated energy intake roughly 200 kcal a day higher than a seven-day written record. Reproducibility on repeat surveys was above 0.70 for bread, milk products, sugar products and alcohol, but 0.51 or lower for meat and vegetables.

The amounts involved were small. About a third of the men used no cocoa at all at baseline, and among users the median was 2.11 g of cocoa a day.

Tertile Blood pressure analysis Mortality analysis
Lowest Under 0.36 g cocoa/day Under 0.50 g cocoa/day
Middle 0.36 to 2.30 g cocoa/day 0.50 to 2.25 g cocoa/day
Highest Over 2.30 g cocoa/day Over 2.25 g cocoa/day

What the paper reported, and what it could not rule out

Comparing highest with lowest tertile in the fully adjusted models, the paper reported systolic blood pressure 3.7 mm Hg lower (95% CI −7.1 to −0.3, P for trend .03) and diastolic 2.1 mm Hg lower (95% CI −4.0 to −0.2, P for trend .03). Swap the food-based model for a nutrient-based one and the systolic estimate falls to 3.1 mm Hg and stops reaching significance at P = .07. A result that moves when the statistical model changes is a result to hold loosely.

For mortality the adjusted relative risks, highest against lowest tertile, were 0.50 for cardiovascular death (95% CI 0.32 to 0.78, P for trend .004) and 0.53 for death from any cause (95% CI 0.39 to 0.72, P under .001). The models adjusted for age, body mass index, smoking, physical activity, alcohol, socioeconomic status, prescribed diet, total calorie intake and calorie-adjusted food intake.

And the authors still put residual confounding first among the limitations. The men who used cocoa were not otherwise comparable: they ate less meat and less coffee and more dairy, sugar confectionery, biscuits, alcohol and nuts. The paper states it could not rule out that residual confounding by physical activity and diet partly explains the findings. Causes of death in elderly cohorts are hard to establish, so secondary and tertiary cardiovascular causes were counted. Reverse causation was argued against rather than excluded by design.

Why this cannot be read across to a tube of chocolate

Four features of the study block any such reading. The exposure was estimated from interviews, not administered. The amounts were around two grams of cocoa a day, which is a biscuit habit rather than a portion of dark chocolate. The participants were elderly Dutch men recruited forty years ago. And the design was observational, so association is all it can deliver. Anyone converting those relative risks into a statement about a product is doing something the paper gives no warrant for.

For completeness, the composition of ours. Chaski Cacao is a 600 g tube of 77% dark chocolate buttons made with single-origin Peruvian and Ecuadorian cacao and coconut sugar, with Lion’s Mane as a fruiting body extract and Cordyceps sinensis as mycelium, and the Ginkgo Biloba component is being replaced by 200 mg of Tremella fuciformis (10:1 fruiting body extract) per daily dose. No emulsifiers, no seed oils, no palm oil, no added cane sugar. Four buttons is the daily dose printed on the pack, and every batch is tested for lead, cadmium and pesticides.

The 2006 paper was not about this product, was not about dark chocolate specifically, and says nothing about what any chocolate does for anyone. It is on this site because the phrase gets searched and the honest version of the answer is a study design with its caveats attached. If the practical question underneath was about forms of cacao, compare nibs, powder and buttons or raw cacao against cocoa powder, and the rest of the library is indexed at our guides index.

Frequently Asked Questions

What was the Zutphen Elderly Study?

An observational cohort recruited in 1985 in the Dutch town of Zutphen, part of the wider Seven Countries Study. The 2006 cocoa analysis by Buijsse, Feskens, Kok and Kromhout in Archives of Internal Medicine used 470 men aged 65 to 84 at recruitment, all free of chronic disease at baseline and none taking antihypertensive medication. Blood pressure was analysed cross-sectionally and mortality prospectively over fifteen years to 2000, covering 4,908 person-years, during which 314 men died, 152 of them from cardiovascular causes. Being observational, it can describe associations within that population and cannot establish that one thing caused another.

How much cocoa did the men in the study consume?

Very little, and nobody was given any. Intake was estimated by the cross-check dietary history method in 1985, 1990 and 1995, multiplying reported consumption of 24 cocoa-containing foods by their cocoa content taken from food tables and labels. About a third of the men used no cocoa at baseline, and the median among users was 2.11 grams of cocoa a day. The highest tertile was defined as more than 2.30 grams a day in the blood pressure analysis and more than 2.25 grams a day in the mortality analysis. Chocolate made up roughly two thirds of the estimated total.

What are the main limitations of the Zutphen cocoa paper?

The authors name residual confounding first. Cocoa users in the cohort differed in other ways, eating less meat and coffee and more dairy, sugar confectionery, biscuits, alcohol and nuts, and the paper states it could not rule out that residual confounding by diet and physical activity partly explains the findings. The systolic blood pressure estimate also moved from 3.7 to 3.1 mm Hg and lost significance when a nutrient-based model replaced the food-based one. Causes of death are hard to establish in elderly populations. The cohort was men only, from one small town, recruited in 1985.

Does the study tell you anything about eating dark chocolate today?

Not in any direct way. The exposure was an interview-based estimate of around two grams of cocoa a day across 24 mixed foods, not a measured portion of dark chocolate, and chocolate in the Netherlands in 1985 is not the category sold now. The participants were elderly Dutch men recruited forty years ago, the design was observational, and the authors flag confounding they could not remove. Reading a relative risk from that cohort as a statement about a particular modern product is a step the paper gives no warrant for, and this page is not going to take it.

Sources

  1. Buijsse B, Feskens EJM, Kok FJ and Kromhout D, Archives of Internal Medicine, 2006
  2. Seven Countries Study research group, sevencountriesstudy.com, 2019
Liquid error (sections/pagefly-section line 6): Could not find asset snippets/pf-6bdf23d6.liquid