Bruggraber et al. ⁽Reference Bruggraber, Chapman and Thane¹⁾ recently reported and corrected⁽Reference Bruggraber, Chapman and Thane²⁾ a new analysis of the Fe content of plant-based foods from the UK. Comparing the new analysis with historical data, they showed evidence for apparent declines averaging − 41 and − 16 % for twenty fruits from the 1930s to the 1980s and 2000s, respectively, and declines of − 19 and − 10 % for twenty-two vegetables and legumes.

Based on Bland–Altman plots of these differences, the authors conclude that ‘there was remarkable consistency between analytical data for foods spanning the 70 years’. This conclusion seems subjective and limited in context to the Bland–Altman plots. The authors' apparent declines are not necessarily small in the context of nutrition, and they are not small compared with the following prior reports of historical declines in the Fe content of foods.

Expressed as median changes, with adjustment for moisture differences (in one case added by Davis)⁽Reference Davis³⁾, the following apparent declines are reported: − 19 % in twenty fruits and − 12 % in twenty vegetables from the 1930s to 1980s in the UK⁽Reference Mayer⁴⁾; − 15 % in forty-three garden crops from 1950 to 1999 in the USA⁽Reference Davis, Epp and Riordan⁵⁾; − 2 % in thirty-eight fruits and − 4 % in twenty-six vegetables from the 1930s to 1980s in the UK⁽Reference White and Broadley^{6, 7)}; − 43 % in forty-one botanical fruits in the USA⁽Reference Davis, Jenks and Bebeli⁸⁾.

All these findings, including those of Bruggraber et al. ⁽Reference Bruggraber, Chapman and Thane¹⁾, are ‘apparent’, because they depend on historical data from different laboratories in different eras. They also often have large uncertainties due to inadequate numbers of samples⁽Reference Davis, Epp and Riordan⁵⁾ to cope with large natural variations among samples of the same food. (The median CV for Fe is an extraordinary 53 % in forty-three US vegetables and fruits⁽Reference Davis, Epp and Riordan⁵⁾.) Bruggraber et al. ⁽Reference Bruggraber, Chapman and Thane¹⁾ had available only three modern samples of each food, and their analyses have potential confounding between eras due to small differences in the moisture content of high-water foods⁽Reference Davis, Epp and Riordan⁵⁾, illustrated by an example in Davis⁽Reference Davis³⁾. Further, their statistical analysis is based on means and parametric CI, which are questionable, because most of their distributions of changes have large deviations from normality (probable outliers, skewing and kurtosis). Deviations from normality do not affect the cited distribution-free analyses of median declines.

Recent side-by-side plantings and analyses of new and old varieties of the same crop eliminate all uncertainties associated with historical data and also avoid the need to average over a large number of foods. In four such studies, clear evidence is found for genetic declines in Fe and other minerals in wheat cultivars released between 1919 and 2000⁽Reference Garvin, Welch and Finley⁹⁾ and between 1950 and 1992⁽Reference Monasterio and Graham¹⁰⁾, and in broccoli cultivars from 1950 to 2004⁽Reference Farnham, Keinath and Grusak¹¹⁾, but not in potatoes⁽Reference White, Bradshaw and Dale¹²⁾. We and others suggest that these declines are primarily attributable to ‘dilution effects’ from increasing yields associated with selective breeding⁽Reference Davis, Epp and Riordan^5, Reference Fan, Zhao and Fairweather-Tait¹³⁾. However, for several reasons noted, these declines can be difficult to prove with historical data.

Unfortunately, declines in the Fe content of plant foods bolster arguments for the use of fortification Fe, an understandable concern of Bruggraber et al. ⁽Reference Bruggraber, Chapman and Thane¹⁾ and others.