The claim, in numbers
Cadmium in cacao is a trade problem before it is a soil-science problem: much of Latin America’s fine-flavour crop grows on soils naturally or diffusely enriched in cadmium, and the metal follows the bean into chocolate. A new open-access field study in Agronomy (Alvarado-Alva et al. 2026) tests the obvious in-row fix: plant two fast, weedy cadmium accumulators, black nightshade (Solanum nigrum) and beggarticks (Bidens pilosa), between the cacao trees and let them mine the topsoil.
The trial ran in four cacao plots in the Amazonas region of northeastern Peru (La Concordia 01 and 02, La Lluhuana, La Chorrera), with the two species set at three planting distances (30, 50 and 70 cm) around 72 trees, rainfed, over a single roughly 150-day season. The headline numbers are striking. Soil cadmium fell significantly in every plot (Wilcoxon p < 0.001), most dramatically at La Chorrera, from 0.81 to 0.03 mg kg⁻¹, a 96% drop. Foliar bioconcentration factors reached 6.97, and translocation factors exceeded 1 in every plot, the textbook signature of an accumulator moving metal to its shoots. Read quickly, it looks like a win.
Read as a peer reviewer, it is a cautionary tale about what those numbers can and cannot mean.
The mass balance does not close
Start with that 96% soil drop, because it is not physically plausible as a phytoextraction result. The soil was sampled to 30 cm; a hectare-metre of that topsoil at a bulk density of about 1.3 g cm⁻³ is roughly 390 kg per square metre. La Chorrera’s 0.78 mg kg⁻¹ loss therefore corresponds to about 300 mg of cadmium removed per square metre. The extractive plants stood at a density of eight plants around each tree over ~189 m² plots, about 0.76 plants m⁻². Using the study’s own most generous cited offtake, 3.53 mg Cd per plant (from Ji et al.), that is roughly 2.7 mg Cd m⁻² — under one per cent of what left the soil. Using the tissue concentrations actually measured here (leaves around 1 mg kg⁻¹ on small annuals), the plants account for a few hundredths of a milligram per square metre, a gap of two to four orders of magnitude.
Three independent reviewers ran this mass balance and reached the same verdict: the plants removed on the order of 0.01–1% of the cadmium that disappeared from the soil. And this is a near-total (EPA 3050B) digest, so it is not a labile-fraction artefact. The reductions are better explained by everything the design failed to control for. There was no unplanted control plot, so every before-and-after change is confounded with season, rainfall and leaching. The final mean at La Chorrera, 0.03 ± 0.09 mg kg⁻¹, sits essentially at the instrument’s quantitation floor. Standard deviations are large (0.81 ± 0.44 to 0.03 ± 0.09). And the authors’ own 24-plot pre-screen reported these soils higher than the experimental baseline (La Lluhuana 1.27 vs 1.08 mg kg⁻¹), a reminder of how spatially heterogeneous a smallholding’s cadmium is. The honest reading is that the soil-cadmium reductions, headline included, cannot be attributed to the phytoextractors.
Soil and beans move independently
The finding that survives all of this is a decoupling, and it is the useful result. The soil and the bean move almost independently of each other. At La Chorrera, soil cadmium fell 96% while the beans fell only about 35% (1.65 to 1.08 mg kg⁻¹) and cacao roots 27%. At La Lluhuana, soil fell 53% while the beans did not move at all (1.98 to 1.97 mg kg⁻¹, p = 0.76). Across the four plots, bean cadmium fell significantly in only two (La Concordia 02, 1.12 to 0.69; La Chorrera, 1.65 to 1.08), and after treatment the beans still ranged from 0.69 to 1.97 mg kg⁻¹.
Two structural facts force that decoupling. First, cacao’s own root architecture: the authors note that its lateral roots reach up to 4.8 m and its fine roots spread homogeneously through the topsoil, which is why planting distance made no difference and why the tree forages a soil volume the intercrop’s rhizosphere never touches. Second, timing: a cacao pod takes about five to six months to mature, so the “before” and “after” beans are different pods, and the harvested ones largely formed before the weeds could have altered anything. Bean cadmium simply cannot equilibrate to a one-season soil change, which means even the two significant bean reductions are as plausibly coincidence as consequence. For a grower, the compartment that decides market access is the slowest one to respond and the least coupled to the soil you can actually reach.
The hyperaccumulator that never switched on
There is a quieter lesson in the tissue numbers. In the field, the extractors’ leaves held at most 1.27 mg Cd kg⁻¹ and their roots at most 0.80 — roughly two orders of magnitude below both the 100 mg kg⁻¹ hyperaccumulation threshold and the pot-study values the paper itself cites for these species (about 125 mg kg⁻¹ in S. nigrum leaves at 25 mg kg⁻¹ soil, Han et al. 2021; about 192 mg kg⁻¹ in B. pilosa at 100 mg kg⁻¹ soil). On cacao soils carrying 0.4–2 mg Cd kg⁻¹, the hyperaccumulation machinery never engages: these are low-tonnage extractors here, whatever their ceiling in a spiked pot. The eye-catching foliar BCF of 6.97 is a small-denominator artefact — a large ratio over a tiny soil pool, not a large harvest — the same trap we flagged in the amaranth and sunflower work. (One root BCF, at the highest-cadmium site, is actually below 1, which fits the low-uptake picture.) Because species and planting distance were both statistically null, the real lever is not the plant or its spacing but site-level soil chemistry — pH, organic matter, cadmium speciation — none of which were measured, and which this design cannot isolate.
What a credible cacao-remediation trial has to show
None of this means the idea is wrong. Both weeds clear BCF > 1 and TF > 1, so they genuinely move cadmium into their shoots, and in-row phytoextraction on perennial tree farms is worth pursuing. What the study demonstrates is the standard such work has to meet to be believed: an unplanted control to separate plant uptake from season and leaching; a closed mass balance in grams per hectare (biomass yield times tissue concentration, not concentrations and ratios alone — biomass was not measured here); soil chemistry, because bioavailability rather than the accumulator is the binding constraint; and a timescale honest about the arithmetic, since removing a few milligrams per square metre a year from a pool of hundreds is a decades-to-never proposition without amendments — the same kinetic wall the sunflower work hit.
The stakes make the rigour worth it, and here the overlay is ours, not the paper’s, which makes no regulatory comparison. Cacao-bean cadmium is the number that gates the crop into Europe. The EU sets no limit on the raw bean, but Commission Regulation (EU) 2023/915 caps finished products — 0.80 mg kg⁻¹ for chocolate with at least 50% cocoa solids, 0.60 for cocoa powder sold to consumers, tightening to 0.10 for milk chocolate. Beans still carrying roughly 0.7–2.0 mg Cd kg⁻¹ after a season of intercropping map onto product that mostly fails those thresholds; the bean-to-product transfer is not one-to-one, but it runs the wrong way, concentrating cadmium in the high-solids products that carry the strictest limits. A remediation that halves soil cadmium on paper while leaving the beans over the line has not yet solved the farmer’s problem.
Provenance: based on the full open-access text (CC BY) of Alvarado-Alva et al. (2026), Agronomy 16(16):1535, read via a reader proxy because MDPI blocks direct fetches; every soil, root, bean, leaf, BCF and TF figure above was checked verbatim against the paper’s Tables 2–7 by three independent model vendors (Claude, GPT and Kimi), which also independently reproduced the mass-balance shortfall; the EU maximum levels were confirmed against Commission Regulation (EU) 2023/915. The mass balance, the dose-dependence reading and the EU-limit comparison are our analysis, not claims made by the authors. See our methodology for how we verify.