In a moderate-sized elderly sample from northern Germany, we assessed the associations between three a priori dietary scores and all-cause mortality over an 11-year follow-up period (93 deaths). While we did not detect overall associations between DASH score and MMDS and all-cause mortality, a 1-SD increment in adherence to HNFI was associated with a 21% reduction in hazard of all-cause mortality after multivariable adjustment.

Association of HNFI with lower all-cause mortality

Our results are in line with previous studies reporting inverse associations of adherence to HNFI with all-cause mortality in three population-based Scandinavian cohorts [19,20,21] and in a large sample of the EPIC cohort [22]. A recent meta-analysis reported that the Healthy Nordic Diet positively affects various cardiovascular risk factors, including lipid levels, and also reduces the risk for the development of various cardiometabolic outcomes, such as diabetes and clinically overt cardiovascular disease [40]. These associations are an important explanation for the observed association of the HNFI with all-cause mortality. Of the six food groups comprising the HNFI, we identified oats and cereals intake to be the potential driver of this association as a higher consumption of oats and cereals was independently associated with lower all-cause mortality. When we excluded oats and cereals from HNFI, the association between all-cause mortality and adherence to this modified dietary score failed to show statistical significance, further emphasizing the relevance of oats and cereals as a HNFI score component.

Lack of association of DASH score and MMDS with all-cause mortality

As opposed to HNFI, in our sample from northern Germany, the MMDS was not associated with all-cause mortality. This is in contrast to previously reported analyses in Western cohorts that showed reduced mortality risks with greater adherence to a MD score [7,8,9,10,11,12,13, 21, 22, 37, 41,42,43,44]. At that, reduced mortality hazards were even more pronounced in Mediterranean populations [16], whereas results from studies in Western non-Mediterranean regions sometimes yield conflicting results [14, 15]. To the best of our knowledge, this is the first study analyzing adherence to a MD score in relation to all-cause mortality risk in a population-based sample from Germany outside the German EPIC cohort [22, 37]. Interestingly, when only considering the German sub-sample of the EPIC cohort, Trichopoulou et al. failed to show an inverse association of adherence to MMDS and all-cause mortality; while in the total EPIC sample, greater adherence to this score was associated with reduced mortality risks [37].

The various published MD scores differ in scoring methods used, energy-adjustments applied, and food groups considered [10, 16, 45] which hinders comparisons between different studies. When comparing the associations between four differently generated MD scores and mortality risk in a UK population-based sample, results from the MD score based on Trichopoulou et al. [8] differed from other scores considered [11]. This score—as well as its modified version [37] applied in our study—uses sex-specific sample medians of food group consumption to classify the intake of individuals as “more adherent” vs. “less adherent” to the MD. Importantly, this scoring method does not consider the absolute intake of these food groups. Thus, individuals considered as adherent to the MD based on this relative classification scheme might still have rather low absolute intakes of typical Mediterranean foods.

Energy-adjusted consumption of various food groups differ considerably between our study and some studies in which an inverse association between adherence to MD scores and all-cause mortality was found [9, 43, 44]. In comparison, consumption of beneficial score components like vegetables, fruits, and legumes was considerably lower in our study sample. As intake of these food groups was independently associated with lower mortality risks in a meta-analysis [2], their relatively low consumption in our sample—which is, on average, also considerably lower than the German recommendations [46]—might have impeded an association between adherence to MMDS and mortality risk.

In addition, no association of DASH score with all-cause mortality was observed in our sample, which is in contrast to previously reported analyses in population-based studies [7, 10, 13, 24, 41]. Like the MMDS, the DASH score is based on the sample-specific relative intake of food groups instead of considering food group intake in absolute terms. Amongst others, the DASH score also considers vegetable and fruit intake, which was shown to be considerably lower in our study sample compared to studies showing improved survival with greater adherence to MD scores [9, 43, 44].

Scoring high on dietary scores based on sample-specific relative intakes of food groups does not necessarily implicate high absolute consumption of food groups and therefore a true strong adherence to the underlying dietary pattern [11, 16]. The lack of association of adherence to DASH score and MMDS and all-cause mortality risk in our sample might, therefore, be explained by the important principle of generating the dietary scores using sample-specific relative intake of food groups instead of considering absolute consumption.

Dietary score adherence and survival, stratified by prevalence of diabetes, smoking, and BMI categories

In analyses stratified by defined risk factors, improved survival with greater adherence to HNFI was particularly pronounced in individuals with diabetes and failed to reach statistical significance in those without the condition. As one of the characteristics of diabetes, hyperglycemia is associated with a higher risk for adverse health outcomes, including all-cause mortality [47]; whereas, improved glycemic control by certain pharmacological regimens was associated with reduced risks for all-cause mortality in individuals with type 2 diabetes [48]. β-glucan, a soluble dietary fiber in oat, is known to lower postprandial glycemic responses [49]. Thus, individuals with diabetes might benefit from postprandial glucose-lowering effects of β-glucan [50] by intake of oats, which partly comprise the food group oats and cereals. Even though oats and cereals intake and, therefore, also β-glucan intake is very low in our cohort, this adds to the possible explanation how adherence to HNFI might be particularly beneficial in this sub-sample.

Furthermore, in contrast to results in the overall study sample, in individuals with diabetes also greater adherence to MMDS was inversely associated with lower all-cause mortality risk. The MMDS does not consider oats and cereals as a score component but includes cereals, a heterogeneous food group comprising different wholegrain and refined grain products as well as oats and cereals. Still, the effect on blood-glucose control by β-glucan in oats included in the food group cereals might account for the beneficial impact on mortality risk in individuals with diabetes and greater adherence to MMDS, even though the very low intake of oats and cereals in our study sample needs to be considered in this context.

Apart from this, to the best of our knowledge, we are the first to report greater adherence to DASH score and HNFI to be significantly associated with a reduction in all-cause mortality risk in current smokers. Amongst others, smoking contributes to enhanced oxidative stress which increases the risk for adverse health outcomes [51]. Beneficial food groups, like fruits and vegetables, are rich in bioactive compounds with antioxidant capacities that counteract oxidative stress responses [52]. High HNFI or DASH scores implicate a higher relative intake of such beneficial food groups. This is one of various biological mechanisms possibly explaining an especially protective effect of such a healthful diet in current smokers. The same might apply for individuals with obesity as excessive body fat also contributes to inflammation caused by oxidative stress [53]. When stratifying by BMI, we found individuals with obesity to benefit from adherence to HNFI in terms of improved survival. However, as no data on antioxidative capacity of specific foods were available in our cohort, this point is speculative and requires further investigation.

Overall, our findings indicate that adherence to health-promoting dietary habits, represented by scoring higher on DASH score, MMDS, or HNFI, might be even more beneficial for individuals at risk, than for people without such health risk factors. However, as only the interaction between diabetes and HNFI adherence showed statistical significance, interpretation of these findings warrant caution and require further investigation in future studies.

Strengths and limitations

Strengths of this study include its prospective and population-based design, the comprehensive assessment of diet and potential covariates using established instruments and methods, as well as the small number of participants lost to follow-up.

Still, some limitations merit consideration. Even though the FFQ used in this study is a well-established dietary assessment tool, its questions are not perfectly designed to assess the intake of food groups required for the different scores applied. For example, as the intake of oatmeal is not specifically asked in the FFQ, we had to use the food group oats and cereals that results from a question that also comprises other breakfast cereals besides oats. Similar limitations also apply for the MMDS and the DASH score. As we only considered data about dietary intake from the second examination cycle of the “popgen controls”, potential changes in dietary habits over time influencing mortality risk could not be taken into account. Furthermore, a certain heterogeneity considering healthy lifestyle factors was observed in the study sample when comparing individuals scoring highest on any of the three dietary scores with those scoring lowest. In quartile 4, individuals were more likely to be physically active, less likely to be current smokers, and tended to have a lower BMI (data not shown). Even though we considered such covariates in our multivariable-adjusted model, we cannot entirely rule out residual confounding by other factors correlating with a healthy lifestyle. From our original sample, we had to exclude a total of 99 individuals due to missing data. The excluded individuals were slightly younger and more likely to be current smokers as compared to the analytical sample. Finally, since our sample comprises elderly adults from a specific German region, the generalizability of our results is limited.