RRR Dietary Pattern Tracks With Lower Visceral Adiposity in Tehran Cohort
A cross-sectional study in 850 Iranian adults adds another layer to the long-running effort to connect everyday eating patterns with metabolic risk. Investigators compared dietary patterns derived from reduced rank regression RRR and principal component analysis PCA in relation to insulin resistance and adiposity indices, and only one pattern reached statistical significance: the RRR-derived profile was associated with a lower odds of visceral adiposity index VAI across tertiles.
The finding is narrow, but clinically relevant. Visceral fat is not simply excess weight stored in the trunk; it is metabolically active tissue that drives hepatic lipid flux, inflammatory signaling, and insulin resistance far more aggressively than subcutaneous fat. When a dietary pattern appears to track with VAI rather than with broad body-size measures alone, it points toward a potential pathway through abdominal fat biology rather than crude calories or body mass index alone.
How the Study Built Its Dietary Signals
The cohort came from health centers in Tehran in 2020 and included adults aged 20 to 59 years. Mean age was 44.76 years and mean body mass index was 27.77 kg/m², placing the group, on average, in the overweight range. Dietary intake was measured with a validated 168-item semi-quantitative food frequency questionnaire, a familiar tool in nutritional epidemiology but one that still captures only habitual intake rather than the day-to-day timing and postprandial physiology that shape insulin action.
The analytic approach matters here. PCA and RRR both compress complex diets into interpretable patterns, but they do so for different reasons. PCA identifies clusters of foods that tend to be eaten together, while RRR is anchored to outcomes or intermediary biomarkers, making it more likely to extract a pattern tied to metabolic physiology rather than social eating habits alone. In this study, PCA yielded two patterns labeled “Healthy” and “Mixed,” while RRR produced a single pattern.
That difference in method is central to the result.
A Pattern Rich in Contrasts, Not Simple Virtue
The RRR pattern was characterized by high intake of fruits, organ meat, eggs, sweets, nuts, and refined grains. On first glance, that combination is difficult to reduce to a single nutritional label. It mixes foods often viewed as favorable, such as fruit and nuts, with items commonly linked to metabolic burden, including sweets and refined grains. Organ meat and eggs add another layer, bringing dense micronutrient content, cholesterol, and protein into the picture.
The study did not show that this pattern was universally protective. Instead, it was significantly associated with a lower odds ratio for VAI in the highest tertile compared with lower intake categories, with an odds ratio of 0.53, 95% confidence interval 0.28 to 0.97, and a trend value of 0.041. No significant associations emerged for triglyceride glucose index TyG, lipid accumulation product LAP, body roundness index BRI, body adiposity index BAI, or the PCA-derived patterns.
The odds ratio is clinically small in one sense and meaningful in another. A lower VAI signal suggests a pattern may be influencing central adiposity or its lipid-linked metabolic footprint rather than inducing a broad shift in all obesity indices at once. In metabolic disease, that kind of selectivity often reflects the biology of hepatic triglyceride handling, insulin sensitivity, and visceral fat partitioning more than total weight change.
Why RRR Outperformed PCA Here
The absence of significant associations for the PCA-derived patterns is not a failure of nutrition research so much as a reminder of how pattern discovery changes the question being asked. PCA finds common eating behaviors, but commonality is not the same as metabolic relevance. A “Healthy” pattern can still contain enough heterogeneity to dilute an association with insulin resistance biomarkers, especially in a cross-sectional sample where current diet, long-term habits, and existing metabolic status may all be misaligned.
RRR, by contrast, is designed to maximize the proportion of variation in selected response variables. In a setting where insulin resistance and abdominal adiposity are the targets, this makes RRR a more biologically focused tool. The tradeoff is familiar to clinical epidemiology: a method that is more tightly tethered to physiology can produce a pattern that is less intuitive to interpret in the clinic.
One sentence from this dataset stands out: only the RRR-derived dietary pattern reached significance, and only for VAI.
What VAI Adds to the Insulin Resistance Conversation
The study used several surrogate markers: TyG and LAP for insulin resistance, and BRI, VAI, and BAI for adiposity. That panel matters because insulin resistance and adiposity overlap without being identical. A person can carry excess adiposity without the same degree of metabolic derangement, while another can show marked insulin resistance before overt obesity becomes obvious by conventional metrics.
VAI is especially relevant because it incorporates waist circumference, BMI, triglycerides, and HDL-related assumptions to estimate visceral adiposity and lipid dysfunction. When VAI moves and TyG does not, the signal may reflect a narrower effect on fat distribution or lipid trafficking rather than a broad change in glucose handling. The biology is plausible: visceral fat drains into the portal circulation, delivering free fatty acids to the liver and promoting hepatic insulin resistance, triglyceride synthesis, and dyslipidemia.
What This Means for Nutritional Risk Assessment
The practical question for clinicians and researchers is not whether one pattern “works” in a universal sense, but whether it can sharpen risk stratification in adults whose diet and body composition do not line up neatly with standard screening tools. This study suggests that method choice changes the answer. A pattern extracted by RRR may capture a metabolic signature that PCA misses, especially when the phenotype of interest is visceral adiposity rather than general obesity.
Still, the design imposes hard limits. Cross-sectional data cannot determine whether the diet preceded the adiposity pattern or whether existing metabolic changes influenced food choice. Self-reported intake from an FFQ can blur the true exposure, and a single-center regional sample from Tehran may not translate cleanly to other populations with different food availability, cultural patterns, or cardiometabolic risk profiles.
Funding was not provided externally, and the investigators reported no competing interests.
The result is therefore useful, but not definitive. It points to a dietary signal worth testing prospectively in populations where visceral adiposity is rising faster than routine screening can keep up, and where the difference between body size and metabolic risk remains clinically consequential.