Better Science, Better Pet Food: Why Study Design Matters More Than Headlines

As pet food consultants, we continuously review new research as it’s published to aid our clients in formulations and marketing decisions. Every few months, a new study makes headlines claiming one type of pet food or another—raw, fresh, extruded, or otherwise—offers unique health or metabolic advantages. The titles, buzzwords, or study conclusions sound convincing: Research finds canine metabolism responds better to fat than carbs, New Study Shows That Fresh Food Can Support Healthier Aging in Senior Dogs, or “high-carbohydrate kibble diet was associated with changes often linked to adverse metabolic health.” But when you look closer, the language often outpaces the science.  

Too often, we find studies that change everything at once: processing method, macronutrient profile, feeding duration, and calorie intake. That means we’re not testing “raw vs. kibble” or “fresh vs. kibble.” We’re testing everything but the kitchen sink. 

For example, a recent study had two treatment groups: a kibble group and a raw meat-based diet (RMBD) group. Dogs in the RMBD group were fed diets that contained 0% carbohydrates, while those in the kibble group were fed diets that contained 48% carbohydrates. On the surface, this sounds like a comparison between carbohydrates and no carbohydrates. But in reality, the nutrient profiles were so different that it was comparing two entirely different nutritional systems and how the food was made, not just a single nutrient class. 

So, how do we know if research is done right or done wrong? Let’s try to break it down. 

Within Treatment Variability 

In the above study, the RMBD group could choose one or both of the two raw formulas with different protein sources and fat levels, whereas the kibble group had only one option. That introduced uncontrolled variability into the treatment; thus, researchers are unable to properly interpret the results. Any marketing claims a brand wants to make about raw now simply can’t be supported by the results in this study.  

Instead, use a single defined formula per treatment group, or treat each formula as its own group. All diets being compared should be analytically characterized so that nutrient composition is known, consistent, and interpretable. 

No Washout or Neutral Control 

Without a washout period or a reference control group, researchers can’t determine whether measured changes are due to the new diet, residual effects from the previous one, or simply due to time and handling. Going from a high carbohydrate diet to a high protein raw diet will elicit metabolic changes. If metabolic measurements were taken after only one week, are we really seeing the effect of diet, or are we seeing the body’s adaptation response?  

Including a washout period gives the body time to reset and let prior diets, microbiota, and metabolic effects return to baseline. A well-defined control group helps determine whether an observed effect is specific to the diet or reflects normal physiological drift over time.  

If a study’s treatment diets differ drastically in nutrients, energy density, or feeding management, those differences will overshadow the hypothesis, and the variable you’re claiming to test isn’t isolated. Even studies that use advanced techniques like metabolomics, microbiome sequencing, or continuous glucose monitoring can fall into this trap. Sophisticated measurements can’t fix weak design—if the baseline isn’t controlled, the conclusions won’t hold. 

Macronutrient Differences Dominate Physiology 

The RMBD group received 0% carbohydrates and an inverse protein-to-fat ratio in their diet, which fundamentally shifts metabolism, gut fermentation patterns, and hormone regulation. The body moves from glycolysis toward fat oxidation and ketone production. That shift alone can change glucose, triglycerides, and ketone concentrations, even if both diets were raw or both were extruded.  

In other words, if one diet drives lipid oxidation and the other promotes glucose turnover, the metabolic differences you measure are expected—even if both diets were made using the same process. When fat, carbohydrate, and micronutrient levels vary that widely, any metabolic change could come from any of those factors. That’s like comparing orange juice to steak and concluding the difference is the juicing process. 

Processing (extrusion, baking, freeze-drying, etc.) can influence digestibility and compound formation, but to truly test its effect, diets must be compositionally matched. Otherwise, nutrient differences will overshadow any processing effect. Sophisticated tools like metabolomics or microbiome sequencing can’t rescue a weak design.  

Unequal Feeding Intake and Time 

Allowing unequal feeding durations on treatment groups and uncontrolled intake means you’re measuring different adaptation stages and masking real effects.  

Use fixed, equal feeding durations across all treatments. Define the adaptation and measurement phases in advance so that data represent the same physiological window for every subject. Studies should feed to the same calculated energy target with regular weight checks and adjustments to maintain body weight. Intake and leftovers should always be recorded to verify compliance and enable interpretation of metabolic outcomes. 

Why This Matters for Industry Professionals 

For brands, formulators, veterinarians, and marketing teams, study design directly influences how products are developed, claims are made, and recommendations are formed. When research compares diets that differ in multiple uncontrolled ways, the conclusions can misrepresent what an ingredient, processing method, or diet style actually does for pets. That leads to shaky marketing claims, unclear formulation direction, confusing clinical takeaways, and unnecessary regulatory risk. Strong, well-controlled studies, on the other hand, give the industry data that can be trusted, applied, and defended, ensuring innovations are evidence-based rather than headline-driven. 

What Better Science Looks Like  

A good study design isolates the variable being tested. When it doesn’t, even the most advanced measurements can’t save the conclusions. Strong nutrition research is built on clarity and not necessarily complexity. 
 

A high-quality study should: 

• Ask one clear question. 

• Use compositionally matched diets. 

• Randomize and balance groups. 

• Feed to maintain body weight. 

• Define equal exposure times. 

• Include washouts or stable control groups. 

• Use statistics that reflect the design (mixed models or repeated measures). 

Crossover or parallel designs can both work—what matters is control and clarity, not complexity. 

A Call to Action 

When key variables aren’t controlled, data lose meaning. Misinterpreted results confuse scientists, mislead consumers, and dilute trust in real innovation. Our industry doesn’t need more half-baked marketing masquerading as research; it needs better studies for understanding. If we want to advance evidence-based nutrition, it starts with a stronger design and shared accountability. 

If you’re a researcher, design trials that actually test the hypothesis; Reviewers, demand clarity, transparency, and defined controls. And for brands and funders, invest in rigor, not rhetoric. At BSM Partners, we help our clients produce sound and solid research that we stand behind. Because better design produces better data, better data produces better diets, and better diets improve lives. 

Follow us on LinkedIn for the latest updates on all things happening here at BSM Partners.

About the Author

Dr. Sydney McCauley is a Board-Certified Companion Animal Nutritionist and earned both her bachelor’s and doctoral degrees at Virginia Tech in Animal and Poultry Sciences. McCauley’s research was in nutritional physiology with a focus on understanding the effects of low birth weight on glucose, fatty acid, carbohydrate, and amino acid metabolism in skeletal muscle and overall metabolic homeostasis during neonatal development.