The DHD-index is capable of ranking participants according to their adherence to the Dutch Guidelines for a Healthy Diet by reflecting variation in the components that constitute the index, except for the component ADF. This component showed a low variation and is consequently not discriminative in ranking subjects according to their adherence to the guidelines. Furthermore, the index score is positively associated with ’following a diet regime’ and inversely associated with energy intake, which were not included in the index. Additionally, the DHD-index showed to be a good measure of nutrient density of diets.
The components of the DHD-index were based on three different documents about the guidelines: the guidelines as communicated by the health council of the Netherlands , the background document describing the guidelines and the evidence in more detail  and the information provided by the NNC . The NNC communicates the guidelines in a more understandable way and provides food-based examples of the dietary guidelines to the general Dutch population and subpopulations. These three documents were more or less comparable to each other and we decided to stay as close as possible to the guidelines, with three exceptions. For the component dietary fiber, the background document indicated an energy-dependent recommendation which was more specific than the range of 30–40 gram mentioned in the guidelines. For the fish component, the background document had a specified recommended amount of fish fatty acids instead of consuming two portions of fish. The third exception was the fruit component, which was based on the recommendations of the NNC . The NNC communicates that 100 grams of fruit can be replaced by all fruit juices complying to the criteria of naturally containing vitamin C and folate . The sensitivity analysis showed that the total scores increased by an average of 0.86 after the inclusion of fruit juices.
For the threshold values of the moderation components, the 85th percentiles of the current population were used, as was done by others . Although we used the 85th percentiles of the 2-day average, the HEI-2005 used the 1-day distribution . Also other indices, such as the heart disease prevention eating index  and the Mediterranean diet score , used the distribution of intake of the population under study for determining cut-off values. However, because of the use of the 85th percentiles of the distribution of the 2-day averages of 19-30-year-olds, the results of the DHD-index cannot be compared with other Dutch subpopulations, as the cut-off values will differ. An evidence-based threshold value for all moderation components, like the binge-drinking threshold values for the alcohol component, would be the most preferred. However, for the other moderation components these do not exist. Yet, a more appropriate solution would be to use 85th percentiles of usual or long-term intakes of a reference dataset representative of the total Dutch population for all future use.
All ten components of the DHD-index have similar weights, as mentioned in the guidelines . However, some components were correlated, which indicates an overlap in dietary behaviors which causes indirectly more weight to that dietary behavior. The components vegetables and fruit were correlated to the dietary component fiber (r=0.36 and r=0.32, respectively), which can be explained by the fact that fiber represents consumption of vegetables and fruit in addition to wholegrain products. The correlation between the component SFA and TFA was 0.29, which is plausible as these fatty acids appear partly in the same products [15, 31]. These correlations should be studied in future research to explore the effect of the additional weight on diet-disease relations. If judged necessary, differential weighting of the components could be applied.
We hypothesized that participants who adhered to a higher degree to the Dutch Guidelines for a Healthy Diet, have both higher absolute intakes of micronutrients and a more nutrient-dense diet. However, only vitamin C intake increased across quintiles of the DHD-index when energy was not taken into account. The intake of the micronutrients folate, iron, magnesium, potassium, thiamin, riboflavin and vitamin B6 only showed a positive association across quintiles of the DHD-index after adjustment for energy intake. This latter result indicates that participants in the higher quintiles of the DHD-index have a more nutrient-dense composition of the diet. However, they have a lower absolute intake of these micronutrients, because of the inverse association of energy intake across quintiles of the DHD-index. The intake of calcium, riboflavin, and vitamin E showed a decline across the quintiles. Nevertheless, the mean average intake in all quintiles was still acceptable compared to the recommended average intakes , which made the lower intakes less worrisome for public health practices. The inverse association of these three micronutrients disappeared after energy adjustment.
In contrast to energy intake, BMI was not inversely associated with the DHD-index score. This result may be due in part to the self-reported nature of the dietary data, which could invoke underreporting . It can also be caused by specific subject characteristics like restrained eating in the higher quintiles of the DHD-index score. This hypothesis can be confirmed by the increasing percentage of participants following a diet regime in the higher quintiles of the DHD-index score. Unfortunately, no data on other subject characteristics as eating behavior or true energy intake was available in the DNFCS-2003. In the HEI-2005, energy intake from solid fats, alcoholic beverages, and added sugars is included as component of the index . For the Dutch situation, no operational guideline for energy intake is available. The health council states that the guidelines are meant for the apparently healthy population with a healthy and stable weight. Consequently, no component is constructed for energy intake in the DHD-index. Energy adjustment should be therefore applied when examining diet-disease associations.
The adherence to the physical activity criterion was quite high compared to previously described physical activity levels in the Netherlands . This may be due to a possible over-reporting by using the SQUASH , although it is a validated questionnaire for estimating usual physical activity . It was suggested by Ocké et al.  that the population under study was slightly different compared to the general Dutch population in the same age category, which may partly explain the high level of physical activity.
The average score of the component ADF ranged from 8.9 to 9.9 across quintiles, consequently, the variation of this component was low (SD = 1.8). Therefore this component is not that discriminative in ranking subjects according to their adherence to the guidelines. The component was included in the Dutch guidelines because it is important for the prevention of teeth erosion, which is quite different from the aims for prevention of chronic diseases and nutrient deficiencies of the other recommendations . We advise to adapt or delete the component ADF from the index in future research, if variation in the component appears to be low in other studies as well.
Data on sodium intake is expected to be underestimated through lacking information on sodium added at the dinner table and during cooking. We have tried to correct for this by lowering the guideline by 30%. However, the variation in intake of sodium within the population was ignored by this method, which could have biased the results. Preferably, sodium intake is measured in 24-hour urine samples, which is considered the standard for measuring sodium intake .
The estimation of the components of the DHD-index was based on the 2-day average of dietary intake. Although two non-consecutive 24-hour recalls are acceptable for assessing dietary intake on group level , the 2-day average will not be a good estimate to assess usual intake distributions for some components, e.g. fish and alcohol, due to a low frequency of consumption. A FFQ designed to assess usual intake could give better estimates for intake of episodically consumed foods. A FFQ, however, is designed for ranking participants according to their intake and not for estimation of absolute intakes . Moreover, a FFQ cannot be used to estimate the component ADF. Statistical models as the Multiple-Source-Method or the National Cancer Institute method can be used to estimate usual intake distributions or individual usual intakes [40–45]. However, these statistical models have their limitations as well. Altogether, dietary assessment methods are prone to errors which will be reflected by the estimates of the DHD-index. Therefore, care should be taken when comparing DHD-index scores based on different dietary assessment methods.