Should I use a glucose monitor to decide which foods are right for me?

Neither side is talking about people with diabetes, where continuous glucose monitoring has an established clinical job that nothing here touches, and losing that qualifier is dangerous. Means is addressing people whose standard tests come back normal, which is the population where no outcome trial exists in either direction.

Hall's participants ate duplicate presented meals in a metabolic ward, which is the strictest possible test of reliability and the least like an ordinary week. Means is describing people eating their own food at home, where the confound she is actually fighting is that nobody knows what is in their food.

More than the framing suggests, and it is the same agreement Hall and Spector reached. Both accept that people differ substantially in their responses to identical meals. Both hold that limiting ultra-processed food is the single highest value change available and that this is a property of the food rather than of the eater's willpower. Both reject calorie counting as the organising frame.

The dispute is narrower than it looks: it is whether one person's response to one food is stable enough over time to act on, which is a reliability question rather than a variability question.

A repeat-measures reliability study of the same foods in the same free-living people over weeks, and separately a trial randomising adults without diabetes to glucose-guided food choice against a general dietary pattern, with outcomes rather than glucose as the endpoint.

Hall's own paper names the remedy rather than rejecting the idea: aggregated repeated measurements. Her position would survive a demonstration that food scores are stable on retest; his objection is met by the same evidence.