Reply to E Archer and SN Blair

Dear Editor,

We write in response to a recent letter to the Editor entitled “Implausible Data, False Memories, and the Status Quo in Dietary Assessment” by Archer and Blair (1). Although we disagree with some other aspects of their letter, we confine ourselves here to the portion of the letter in which the authors cite data from our recently published article (2). The authors claim that our data “demonstrate the futility” of self-report dietary data methods. They cite estimates of squared average correlation between true usual energy intake and self-reported energy intake of between 0.04 and 0.10, stating that these values “provide unequivocal evidence that self-report dietary data offer an inadequate basis from which to draw scientific conclusions.”

We strongly disagree with their conclusions. It does not follow logically that because energy intake is poorly estimated by self-reporting methods, self-report dietary data can never be used to establish scientifically valid conclusions. Archer and Blair ignore 2 of our findings. First, FFQ-reported protein density (protein intake divided by energy intake) has a far higher correlation with true usual intake than does protein itself. This same finding was also evident for potassium and sodium, as well as for the sodium-potassium ratio (3). Increased correlations are also seen with 24-h recall reported intake after forming densities. These findings indicate that self-report instruments are more suited to the elicitation of a person’s dietary composition than his or her absolute intake. This has long been recognized within the nutritional epidemiology community, and it has led to the common practice of energy adjustment (4) when analyzing self-reported intake of nutrients and food groups. In a similar vein, recent versions of the Healthy Eating Index (5) have been based on energy-adjusted intake. Second, we found that the averages of 2 and 3 24-h recall protein reports had substantially higher correlations with true usual intake than a single recall. This also was evident for potassium (3). Thus, the use of repeated 24-h recalls is another device that can be used to improve the quality of self-report dietary data.

Throughout their letter, Archer and Blair claim that their arguments are logical and empirically supported. In fact, their conclusions are far too sweeping.