Cassava, sweet potato, potato, rice: what head-to-head studies show about carbohydrate sources for pet food

Why these ingredients are compared at all

“Grain-free” says what a recipe leaves out, not what it puts in. The carbohydrate still has to come from somewhere, and in most grain-free dry foods it comes from tubers (potato, sweet potato, cassava) or pulses (peas, chickpeas, lentils). Rice appears in this article as the cereal reference against which the tuber sources were tested, not as a grain-free ingredient.

These ingredients differ in starch structure, fibre, protein, digestibility and how they cook in an extruder. The trials below compare some of them under controlled conditions. Each is summarised on its own terms, with the species, the design and the result, because the design decides how far the result can be carried.

Study 1 — six carbohydrate sources in dogs (Carciofi et al. 2008)

Design. Dogs were fed six diets, each with a single carbohydrate source: cassava flour, brewer’s rice, corn, sorghum, peas or lentils [3].

Results. Brewer’s rice and cassava flour had the highest digestibility of dry matter, organic matter and gross energy, followed by corn and sorghum, with peas and lentils lowest. Starch digestibility was above 98% for every diet but was still greater for brewer’s rice and cassava flour than for lentils and peas. The early glucose and insulin responses (up to 30 minutes after the meal) were greater for the brewer’s rice, corn and cassava flour diets, while the later responses (from 30 minutes on) were greater for sorghum, lentils and peas [3].

What it supports. In this design, cassava flour behaved like brewer’s rice: highly digestible, with a fast early glucose rise. It does not say anything about sweet potato or potato starch, which were not tested.

Study 2 — seven starch sources in dogs (Kaelle et al. 2023 and 2024)

Design. The same research group tested seven starch sources — corn, brown rice, sorghum, potato starch, sweet potato flour, chickpea and pea — as the sole starch source in diets formulated to about 300 g/kg starch. The digestibility, fermentation and glycaemic trial used 14 adult Beagles [2]; the extrusion and palatability work was reported separately, with 16 adult Beagles in the paired preference tests [5].

Results — nutrition [2]. The sweet potato flour diet had the lowest total-tract apparent digestibility of nutrients; brown rice and pea were the most digestible. Corn and potato starch produced a greater rise in peak glucose than the other sources; sweet potato, sorghum, pulses and brown rice gave a smaller increment. The sweet potato diet gave higher faecal short-chain fatty acid concentrations than the potato and brown rice diets, a higher abundance of Faecalibacterium prausnitzii and Blautia, and, like the pulses, greater faecal volume and moisture. The authors conclude that sweet potato and pulses “improve indicators of gastrointestinal functionality”.

Results — processing [5]. Potato starch and sweet potato flour gave more expanded, more porous and less dense kibble than the cereals and pulses; pea gave the densest, least expanded kibble; potato starch kibble was the hardest of the seven. In paired tests against the corn diet, dogs preferred every diet except the sweet potato diet, whose intake did not differ significantly from corn. The extrusion results are discussed in more detail in our article on sweet potato flour in extruded kibble.

What it supports. Within one set of diets at a fixed starch level, sweet potato flour traded digestibility for a smaller glucose peak and more fermentation, and potato starch raised glucose like corn while expanding strongly in the extruder. The potato starch result is consistent with an earlier dose–response study by the same group, in which replacing corn with up to 300 g/kg potato starch improved nutrient digestibility and metabolisable energy [8]. Cassava was not in this trial.

Study 3 — three complete diets in dogs (Vastolo et al. 2023)

Design. Fifteen adult dogs were fed three complete diets in rotation: a grain-free diet with sweet potato, a grain-free diet with pea starch, and a cereal-based diet with spelt and oats [4]. These were full commercial-type recipes, so the diets differed in more than the carbohydrate source.

Results. The sweet potato diet gave the lowest glucose area under the curve over six hours, and, together with the pea-starch diet, a lower insulin area under the curve than the cereal diet. The authors note that the glycaemic response can be modulated by the carbohydrate source and composition of the diet, and that this may be worth considering when choosing a maintenance diet [4].

What it supports. A sweet potato-based complete diet gave a flatter glucose and insulin curve than a spelt-and-oat diet in these dogs. Because the whole recipes differed, the effect cannot be attributed to sweet potato alone.

Study 4 — five carbohydrate sources in cats (Zhang et al. 2023)

Design. Thirty young cats were fed, for 28 days, diets based on potato, cassava, sweet potato, rice or wheat [1].

Results. Apparent starch digestibility was 99.0% for rice, 98.7% for wheat, 98.6% for cassava, 98.2% for sweet potato and 98.0% for potato, with no significant difference between diets. The cassava group had the numerically lowest mean and peak glucose and the lowest serum lipids; the sweet potato diet measured the highest degree of starch gelatinisation (90.2%) and had the numerically highest overall gut microbial diversity. Glucose, insulin, serum lipids and overall microbiome diversity did not differ significantly between groups (P > 0.05); some individual bacterial genera did differ [1]. The authors nevertheless concluded that cassava and sweet potato diets were worth developing for cat food.

What it supports. In cats, on glucose, serum lipids, starch digestibility and overall microbiome diversity over 28 days, none of the five sources was shown to differ from the others. The study is useful as a demonstration that all five were well digested, and as a source of trends worth testing, not as evidence that any one of them is better.

Study 5 — what is in commercial dog foods (Perry et al. 2022)

Design. Twenty commercial dry dog foods were analysed: ten traditional formulas (rice, barley, oats, corn) and ten non-traditional formulas (cassava, chickpeas, lentils, potato, pumpkin) [6].

Results. Non-traditional formulas contained less total starch (about 20–22% versus 35–36%) but a higher proportion of gelatinised starch (about 95–97% versus about 86%) [6].

What it supports. Describes what is on the market. Because these are different products made under different conditions, it does not show which ingredient caused the difference in gelatinisation.

Why the results cannot be stacked

Putting these studies side by side shows why a single ranking would mislead:

• Species. Four studies are in dogs, one in cats. Results do not transfer between them.
• Diet design. Two trials used single starch sources at a fixed inclusion in otherwise matched diets [2, 3, 5]; one used complete commercial-type recipes [4]; one analysed finished products [6]. Only the first type isolates the ingredient.
• Which sources were present. In dogs, cassava appears only in the 2008 trial, while sweet potato and potato starch appear in the Kaelle trials and (sweet potato) in Vastolo; no single dog study contains all four. The only study that did test all four together, in cats, found no significant differences between them in glucose, serum lipids or starch digestibility.
• Starch level, fibre and protein. Each diet’s fibre and protein moved with its starch source, and these drive digestibility and glucose response as much as the starch itself.
• Processing. Degree of gelatinisation, expansion and kibble density differed with the source and the extruder settings, and these affect digestibility and glucose response in their own right.

So a sentence such as “cassava is more digestible than sweet potato” is not supported by any study here, because the two were never compared in the same dog trial, and in the cat trial their starch digestibility did not differ significantly.

The DCM question

Any article on grain-free carbohydrate sources has to mention the investigation by the US FDA into reports of dilated cardiomyopathy (DCM) in dogs eating certain diets. The agency’s last public update, dated 23 December 2022, states that the reports it has received are not sufficient to establish a causal relationship between these diets and DCM, and that it does not intend to issue further public updates until there is meaningful new scientific information. Among the diets named in the reports it analysed, 93% contained peas and/or lentils and 42% contained potatoes or sweet potatoes [7]. Those percentages describe the reported diets; they are not a measure of the risk of any ingredient, and the absence of an established link is not proof of safety either. The question remains open, and a formulator should follow the FDA page and the peer-reviewed literature.

Questions to take into your own formulation work

Rather than choosing a source from the literature, use the studies to decide what to measure in your own trial:

1. Digestible energy. Measure or calculate it for the actual ingredient and inclusion level; the literature shows digestibility can differ by source [2, 3], not what your formula will deliver.
2. Stool quality. Score it in the trial; higher fermentation went with higher faecal volume and moisture in dogs [2].
3. Glucose response, if the product has a target for it. Measure it with the finished diet, because complete-diet composition changes the curve [4].
4. Extrusion and kibble quality. Expect expansion, density and water demand to change with the starch source [5]; confirm on your own line.
5. Species. Do not carry a dog result into a cat product or the reverse.
6. Ingredient data. Ask the supplier for starch, moisture and fibre content, particle size and whether the root was peeled; these affect how the ingredient performs in all of the above.