Reply to LS Freedman et al.

Dear Editor:

We thank Freedman et al. (1) for their reply to our letter entitled, “Implausible Data, False Memories, and the Status Quo in Dietary Assessment” (2), and appreciate the continuance of the scientific discourse our previous work (3) engendered. Herein we address the claims from Freedman et al. that post hoc data manipulations that improve correlations between biomarkers and physiologically implausible data establish the scientific utility of those improbable data, and that additional 24-h recalls (24HRs) “improve the quality of self-report dietary data.”

Science is the pursuit of lawful relations between natural phenomena (i.e., knowledge of objective reality), and not mere statistical correlations between numbers (i.e., simple abstractions). It is an oft ignored fact that mathematical and/or statistical representations of natural phenomena per se do not demonstrate the validity of the allied data collection protocols or the representations themselves. As such, numeric representations often mislead those who lack a fundamental understanding of the phenomena under examination. For example, well over 100 y ago Pearson demonstrated that ratios such as those used by Freedman et al. often produce correlations that are a function of the “arithmetic” (i.e., spurious) and are not necessarily indicative of an “organic relation” (4). Given this fact and our previous work (3), we think no amount of mathematical or statistical machinations can establish lawful relations between objective reality and numbers generated by data collection protocols demonstrated to lack validity (i.e., 24HRs and FFQs) (3).

It is well established that nutrients, specific foods, beverages, and food groups (e.g., protein, fat, carbohydrate, sugar, alcohol, and vegetables) are subject to differential misreporting that significantly affects subsequent estimates of energy intake (EI) (5–8). For example, dietary composition is significantly altered in under-reporters, with reported fat and/or carbohydrate consumption lower, and reported protein, fruit, and vegetable intake higher (6, 7).Therefore, patterns of dietary consumption will inevitably be differentially and unpredictably misreported when total reported EI is misreported, and this phenomenon will be exacerbated as reported EI becomes increasingly physiologically implausible. The differential misreporting occurs because participants qualitatively and quantitatively misreport dietary composition (e.g., type and amount of foods and beverages) because of intentional and nonintentional factors (e.g., misleading statements because of health-related perceptions, forgetting, and false memories) (7, 8). The nonuniformity of misreporting leads to macro- and micronutrient-specific errors that alter nutrient-to-energy intake ratios in an unpredictable and nonquantifiable manner (8). Stated more simply, it is impossible to discern and quantify the type or amount of dietary components that led to the misreported energy. Thus, the nonquantifiable nature of altered nutrient-to-energy intake ratios via misreporting renders energy adjustments fallacious (5) and demonstrates that data derived from 24HRs and FFQs cannot be used to examine patterns of diet or dietary composition.

Finally, the statement from Freedman et al. that additional 24HRs “improve the quality of self-report dietary data” is contrary to empirical evidence. The studies examined in their article (9) exhibit greater misreporting of energy and macronutrients with additional administrations of 24HRs. For example, the Energetics Study demonstrated a statistically significant decreasing trend over 8 24HRs for energy (−535 kcal/d) that varied by macronutrient (protein: −22g/d and −88 kcal/d; fat: −27g/d and −243 kcal/d; and carbohydrate: −54g/d and −216 kcal/d) (10). Given these results, Freedman et al. are proposing the paradoxical argument that improved correlations coupled with decreasing nutrient-specific validity demonstrate that multiple 24HRs “improve the quality of self-report data.” We find their position to be contrary to extant evidence and therefore not logically valid.

If the essence of science is the ability to discern fact from fiction (i.e., falsifiability), and it is impossible to quantify what percentage of the recalled foods and beverages are completely false reports (e.g., misleading statements and/or false memories), grossly inaccurate (e.g., misestimation), or somewhat congruent with actual consumption, then dietary recall data constitute unscientific evidence (8). As such, we think statistical machinations and the demonstrated lack of validity of 24HRs and FFQs are driving the contradictory nature of recent dietary guidelines (8).