Among 6,525 US adults in NHANES 2017-2020, ultra-processed foods that met moderation criteria (limited added sugar, sodium, refined grains, saturated fat) made up 48% of energy intake, while poor-quality ultra-processed foods made up only 6%. Each 1% of energy from moderation ultra-processed food replacing Nova 1-3 foods was linked to higher BMI, waist circumference, body fat, insulin and C-reactive protein. Non-moderation ultra-processed intake showed null associations with those same markers, and moderation minimally processed food tracked with higher adiposity and glucose than its non-moderation counterpart.
These findings are counterintuitive: the "healthier" ultra-processed foods looked worse, which complicates both the processing-is-the-problem camp and the nutrient-profile camp. Reverse causation is the likeliest explanation. People with higher adiposity or insulin resistance may preferentially choose foods that look nutritionally acceptable, such as diet products and fortified items. Two 24-hour recalls, self-reported intake and a single cross-sectional snapshot cannot separate that from true harm, and the tiny coefficients (0.03 BMI units per 1% energy) are modest, though a 10-point energy swap would scale them tenfold. The unusually small non-moderation category also limits statistical power. The authors' policy conclusion, that nutritional quality should anchor guidelines, is reasonable but not proven by these data. Prospective cohorts and feeding trials, such as the 2019 NIH study showing ultra-processed diets drive excess calorie intake, remain the stronger evidence. This is a useful, incremental caution against assuming processing level alone, or a good nutrient score alone, predicts metabolic health.