Letter to the Editor / Reply

Caries experience was assessed at D 1 /d 1 level (the authors called it C1) by one experienced examiner. However, caries diagnosis at this initial level requires much experience by the examiner, which must be well documented [see e.g. Jablonski-Momeni et al., 2008], while in the present case no information was provided on how the examiner was trained and which was the reproducibility of caries and non-caries diagnoses. With a statistical analysis using presence/absence of caries as outcome variable and in a sample characterised by high caries experience, such potential unreliability of diagnoses is likely to produce an excessively high proportion of false negatives, with detrimental consequences on the results of the statistical analysis investigating variables associated to caries diagnosis [Shoukri and Pause, 1999]. Food intake was assessed by parents without the help of their children using a self-administered food frequency questionnaire (FFQ), which was based on another FFQ ‘used in a similar population’ [García-Closas et al., 1997]. However, FFQs (including the FFQ cited by the authors) refer to a precise period, such as the previous year, month or week and not merely to average consumption, require pictures of the portion sizes, not only of the food items; they must be administered by a trained and calibrated interviewer, generally a nutritionist or a dietician, and not self-administered, and, if used in schoolchildren, require additional information concerning food/beverage intake consumed at school, which is generally provided by the teacher [Livingstone and Robson, 2000; Fumagalli et al., 2008]. In addition, the authors, who sought to assess the effect of sugar only, starch only and sugar plus starch on caries, made their nutritional investigation assuming that these food categories were consumed alone, not in association with foods from other categories, but this is very unlikely in real life. For example, how did the authors classify a breakfast based on milk (category 4 or 5), industrial bread (category 8) with honey or jam (category 1)? This information bias is enough to invalidate the results of the study concerning the effect of sugar alone and starch alone on caries, because the FFQ used in this study did not allow to assess whether sugar or starch were consumed alone or in association. A serious limit of FFQs is that they are subjected to reporting bias. For example, obese subjects are more likely to under-report energy intake than their normal-weight counterpart [Champagne et al., 1998]. Selective misreporting of certain types of foods affects reporting accuracy and is particularly detrimental to understanding the role of nutrition in health [Heitmann and Lissner, 1995; Johansson et al., 1998]. For instance, subjects who underreport energy intake also report eating less sugar and carbohydrate, particularly from biscuits/pastries/puddings and sugar/ confectionery [Pryer et al., 1997]. The non-significant association between confectionery intake and dental caries found by the authors cited by Llena and Forner [2008], which inspired them to design their FFQ, also was explained by selective under-reporting Letter