The NHANES dietary data are physiologically implausible and inadmissible as scientific evidence

Dear Editor:

The recent report by Gu and Tucker (1) presented data from the NHANES collected via memory-based dietary assessment methods (M-BMs; e.g., 24-h recalls). Before their report, M-BM data were shown to be “physiologically implausible” (i.e., meaningless numbers) (2), “incompatible with survival” (2), “incompatible with life” (3), and “inadmissible as scientific evidence” (4–6). In 2013 and 2015, my colleagues and I empirically and theoretically refuted the validity of all M-BMs (i.e., self-reported) data and showed that no human could survive on the average amount of foods and beverages reported in the NHANES (2, 4). We further showed that the self-reported proxy-estimates for energy intake from 26,975 of 63,369 NHANES participants (42.5%) were below the level of energy intake needed to support a comatose patient’s survival (2). This rigorous body of work suggests that it is highly unlikely that the implausible NHANES dietary data can be used to create valid “healthy eating indexes” or used in valid examinations of dietary patterns and dietary quality. This conclusion is supported by ≥60 y of research that shows that epidemiologic M-BM data and concomitant conclusions lack scientific support. For a review of this extensive body of research, inclusive of a publication in this Journal on the unscientific nature of self-reported dietary data, see reference 4.

NONQUANTIFIABLE MEASUREMENT ERRORS RENDER M-BM DATA NONFALSIFIABLE (i.e., PSEUDOSCIENTIFIC)

M-BM data consist of pseudoquantified dietary anecdotes. Or stated more precisely, M-BM data are not measurements of actual dietary intake but are data created by investigators via the assignment of nutrient and energy (caloric) estimates to reported memories of past dietary intake. It is well established that reports based on human memory suffer from numerous intentional and nonintentional distorting factors (e.g., false memories, confabulations, forgetting, encoding failures, and lying) (4). Thus, as we stated in our previous work, “without objective corroboration it is impossible to quantify what percentage of the recalled foods and beverages are completely false, grossly inaccurate, or somewhat congruent with actual consumption” (4; p. 919) and “neither the researchers nor the participants know the validity or reliability of the reported food and beverage consumption” (4; p. 918). Therefore, because the participants’ actual past dietary intake is unknown and unknowable, the measurement error associated with M-BM data is nonquantifiable. Given that the distinction between quantifiable and nonquantifiable measurement error is the demarcation between falsifiable data and pseudoscientific data, the M-BM data that Gu and Tucker analyzed (1) were clearly neither scientific nor admissible.

ESTABLISHED PROTOCOLS FOR REMOVING IMPLAUSIBLE DIETARY DATA

Given the ubiquity of implausible dietary data, numerous validated protocols were developed to determine the extent of these data in any given data set (e.g., see references 2 and 7). If any of these validated protocols were used, >20,000 participants would have been excluded from Gu and Tucker’s sample of 38,487. Thus, readers unfamiliar with the extensive literature on the implausibility and refuted validity of the NHANES M-BM data (2, 4, 5, 8, 9) would be misled by the results and conclusions presented by Gu and Tucker (1).

SUMMARY AND CONCLUSIONS

The purpose of our research (2, 4–6, 8–10) was to end the use of physiologically implausible data in the formation of the USDA’s National Evidence Library, the US Dietary Guidelines for Americans, and public health policy (9). Our work shows that if the population-level estimates from M-BMs are implausible and the measurement error of the estimates for each respondent is nonquantifiable and nonfalsifiable, then M-BM data, results, and conclusions are “inadmissible” as scientific evidence (4, 5, 9). To date, no researchers have presented data to refute our results or challenge our conclusions. There were rhetorical responses, but mere rhetoric cannot refute our data. As such, our uncontested results and conclusions (2, 4, 5) represent an existential threat to nutrition epidemiology and challenge the public health recommendations from all studies over the past ≥60 y that used M-BM data, inclusive of the recent report by Gu and Tucker (1).