*3.2. Assessment of Dietary Habits*

Supplemental Tables S2 and 3 show the average intake of food and beverages for the study participants. No subjects reported a special dietary regimen for medical purposes. The total energy intake was slightly higher in cases compared to controls. In regards to cereal products, cases were shown to have a higher intake compared to controls, in particular EO-FTD subjects. Cases had a slightly higher intake of meat products compared with controls, including red and white meat, and a higher intake of dairy products, particularly milk and yogurt, especially in EO-AD compared to EO-FTD cases. Cases also showed a substantially similar intake of fish and seafood, but a higher consumption of preserved and tinned fish particularly for EO-FTD cases, a lower intake of piscivorous fish and crustaceans/molluscs, especially in EO-FTD cases. Cases also showed a lower intake of overall vegetables, with similar results for all vegetable types but cabbage, and a lower intake of potatoes and legumes. Overall, fresh fruit intake was somewhat higher in cases than controls, although the former showed a lower intake of citrus fruits yet a higher intake of all other fruits, with similar results across EO-AD and EO-FTD. Dry fruit intake was lower in cases than in controls, particularly for nuts and seeds and especially in EO-FTD, characterized by a much lower intake of both overall dry fruit, nuts, and seeds. Sweet intake was higher in cases, particularly due to a higher consumption of ice-cream, biscuits, and dry cakes, while consumption of chocolate-based products was lower. Cases consumed less oils and fats, with a lower intake of both vegetable oils and olive oil. Concerning beverages, cases had a lower intake of coffee and tea, but results were the opposite in the subgroup analysis according to an EOD diagnosis, with a higher intake in EO-AD cases and a much lower one in EO-FTD cases, who showed very low tea consumption. Wine consumption was substantially similar in cases and controls, with EO-AD cases reporting a lower intake and EO-FTD a higher one, mainly due to the higher consumption of white wine. A higher intake of aperitif wines and beers was reported for controls. The overall alcohol intake was comparable in cases and controls, although EO-FTD cases showed a much higher intake compared to EO-AD cases. Concerning non-alcoholic beverages, cases reported a higher intake of fruit juices and a lower consumption of soft drinks.

### *3.3. Assessment of the Relation between Dietary Habits and Dementia Risk*

In Supplemental Table S4, we report risk estimates associated with increasing tertiles of food and beverage intake. Overall, about cereals products, we found an inverse association with EOD risk, although a positive association can be noted for subjects in the third tertile for both pasta and rice consumption. For meat products, we found a direct association with EOD, with a higher risk in the second tertile compared to the third one. In the subgroup analyses, however, red, white, and processed meat showed a very imprecise inverse association in the second tertile, but a null/positive association in the third tertile. On the other hand, offal intake seemed inversely associated with EOD risk, especially for subjects in the third tertile. Milk and dairy products were in general positively associated with EOD, particularly in the third tertile of exposure. Both piscivorous and non-piscivorous fish seemed inversely associated with EOD risk. For preserved and tinned fish, crustaceans and molluscs, conversely, we found an inverse association in the second tertile, but a positive one in the third tertile. In general, vegetables were inversely associated with EOD risk, particularly leafy, root, and other vegetables as well as mushrooms and potatoes, and an almost substantial null association for legumes. In regards to fruits, we found an inverse association for fresh fruits, especially citrus fruits, as well as dry fruits. Conversely, the intake of all other fruits showed a negative association in the second tertile and a positive one in the third tertile. For overall sweets, we found a positive association with EOD risk, mainly due to ice-cream, cakes, pies/pastries, and biscuits/dry cakes, while chocolate and other (non-chocolate) confectionery showed an inverse association. Overall, oils and fats showed an inverse association, mainly due to the intake of olive and non-olive oils, while we found a positive association with butter and other animal fats. Concerning beverages, coffee and tea, aperitif wines/beers, and spirits/liqueurs showed an inverse association with EOD risk. Red wine showed an inverse association for subjects in the second tertile, and a positive one in the third tertile. Conversely, white wine and soft drinks showed a higher risk in the second tertile and lower one in the third tertile. Finally, fruit juices seemed positively associated with disease risk, especially in the second tertile.

In general, we found similar results in the analysis based on the EOD diagnosis, with a few exceptions (Supplemental Table S4). Bread intake was inversely associated with EOD risk and this was also true for EO-AD. However, a direct/null association is present in EO-FDT. Fresh cheese intake showed a positive association with EO-FTD, but a null association for EO-AD. Conversely, aged cheese intake did not show such an association with EO-FTD and a slight positive association emerged for EO-AD in the third tertile of exposure. With regard to fish intake, EO-FTD showed a positive association with preserved and tinned fish, while an inverse one can be noted for EO-AD. In regards to fresh fruit intake, both EO-AD and EO-FTD showed an inverse association, but we found a positive association for citrus fruit in EO-FTD and a negative one in EO-AD. In relation to sweet intake, the increased risk for biscuits and dry cakes was confirmed in EO-AD, while we found a decreased risk for EO-FTD. Finally, alcohol intake was associated with a slightly increased risk in EO-FTD, but not in EO-AD.

#### *3.4. Assessment of the Relation between Dietary Patterns and Dementia Risk*

Moving on to the analysis of adherence to the investigated dietary patterns, cases (overall and both EO-AD and EO-FTD) generally showed slightly lower mean scores (Table 2). In the risk analysis according to the tertile distribution (Table 3), we found a lower EOD risk at increasing adherence in all three dietary patterns, especially for high adherence to MIND dietary patterns.

**Table 2.** Adherence to dietary patterns (Greek-Mediterranean (GM) diet; DASH: Dietary Approaches to Stop Hypertension (DASH) diet); Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet) in the study population (Early-onset dementia (EOD), early-onset Alzheimer's dementia (EO-AD), and early-onset frontotemporal dementia spectrum (EO-FTD). Values reported as mean and standard deviation.


**Table 3.** Odds ratio (OR) and 95% confidence intervals (CI) of early-onset dementia (EOD), early-onset Alzheimer's dementia (EO-AD), and early-onset frontotemporal dementia spectrum (EO-FTD) for increasing tertiles of adherence to dietary patterns. Greek-Mediterranean (GM) diet, Dietary Approaches to Stop Hypertension (DASH) diet, and Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet. A linear trend for 1-unit increase is reported.


In Figures 1–10, we present data based on the spline regression analysis adjusted for sex, age, educational attainment, and total energy intake. Due to the very high number of non-consumers, the spline analysis was not feasible for some offal and several beverages, namely tea, red and white wine, aperitif wines and beers, and soft drinks. Overall, cereal products showed a U-shaped relation with EOD, with a lower risk at around 200 g/day, yet a higher one above 350 g/day (Figure 1). A similar relation was found for bread and rice intake. Conversely, pasta and other grains showed a linear relation with decreased risk at low intake levels and an increasing one from 50–60 g/day. On the other hand, pizza, crackers, and other salty snacks showed increased risk in the case of null exposure, with decreased risk from 40 g/day after which a plateau was reached. Overall meat consumption did not seem to be associated with EOD (Figure 2), although an increase in risk can be noted for a high intake of both red (>100/120 g/day) and white meat (>40/50 g/day). In regards to dairy products, we found null/decreased risk for all products, except for an increase in risk at very high intake levels, namely >400 g/day for overall dairy products or >350 g/day for milk and yogurt (Figure 3). Similarly, we found null risk in association with the intake of eggs (Figure 3) and overall fish and seafood (Figure 4). Interestingly, we found a U-shaped relation with fish intake, with increased risk for null consumption and above 20 g/day for preserved and tinned fish (Figure 4). Conversely, a high intake (>20 g/day) of other fish, especially piscivorous fish, appeared to the decreased EOD risk. All vegetables showed an inverse association with EOD, with increased risk in the case of null intake and decreasing risk starting from 100 g/day, above which a plateau seems to have been reached (Figure 5). Such pattern of association can also be noted in all vegetable subgroups, especially leafy

vegetables showing a slight continuous decrease, while cabbages showed a null association at all intake levels. Mushrooms showed a U-shape relation, with a lower risk at around 4 g/day. On the other hand, we found a slight inverse association for potatoes and legumes, albeit a very imprecise one (Figure 6). In spite of the substantially linear inverse correlation of citrus fruit intake with EOD risk, we found a U-shaped relation with fresh fruit and all other fruit intake, with a lower risk at approximately 250 and 200 g/day, respectively (Figure 7). Conversely, dry fruit intake showed an increased risk for subjects reporting null consumption and a decreased risk from 4 g/day of dry fruits and nuts and from 1 g/day of dry fruits, above which a plateau was reached (Figure 7). Sweet products showed an inverse-U relation with EOD risk, mainly driven by the lower risk associated with a high intake of chocolate and other chocolate-based products (Figure 8). However, sugar and other confectionery showed a null association, while ice-cream, biscuits, and dry cakes showed a positive correlation, with null risk for non-consumers and an increased risk for intake above 20–30 g/day for both ice-cream and dry cakes, respectively (Figure 8). In general, oils and fats showed that a linear inverse association with increased risk starting from null intake, soon reversed above 20 g/day especially for olive oil and other vegetable oils and fats (Figure 9). For beverages, we found a U-shaped relation with coffee consumption, with a lower risk at 70 g/day (Figure 10). Conversely, we found almost a null relation with wine and alcohol intake, while risk increased with fruit juice consumption exceeding 100 g/day (Figure 10).

**Figure 1.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of cereals and cereal products. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 2.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of meats and meat products. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake. Note: spline analysis was not possible for offal due to a few subjects reporting consumption which is different from a null value.

Spline regression analyses stratified by the EOD clinical type are reported in Supplemental Figures S1–S10 for EO-AD and in Supplemental Figures S11–S20 for EO-FTD. We generally found comparable results across EOD forms. Nonetheless, there was evidence of a U-shaped relation between red meat intake and EO-AD (Supplemental Figure S2), and between processed meat and EO-FTD (Supplemental Figure S12), with the lowest risk ratio occurring at 60 and at 25 g/day, respectively. In regards to dairy products, the shape of the relation with cheese intake seemed opposite for EO-AD and EO-FTD, with an inverse U-shaped relation for EO-AD, especially for fresh-cheese intake (Supplemental Figure S3), and a U-shaped relation for EO-FTD, especially for overall cheese intake (Supplemental Figure S13). Moreover, egg intake seemed to establish an inverse-U relation for EO-AD only (Supplemental Figure S3). In addition, mushroom intake showed a linear inverse association for EO-AD (Supplemental Figure S6), and a U-shaped one with EO-FTD (Supplemental Figure S16).

**Figure 3.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of milk, dairy products, and eggs. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 4.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of fish and seafood. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 5.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of vegetables. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 6.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of mushrooms, legumes, and potatoes. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 7.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of fresh and dry fruits. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 8.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of sweets products. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

**Figure 9.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of oils and fats. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake.

The spline regression analysis carried out for dietary pattern adherence showed no substantial association between EOD risk and both GM and DASH diets, with a reduction in risk from a score of ≥6 and ≥28, respectively (Figure 11). Conversely, a substantially linear inverse association emerged for the MIND diet, with a higher risk at very low adherence and decreasing risk above a total score around 8/9. We found comparable results in the analyses performed by the EOD subtype (Supplemental Figures S21 and 22), with the exception of adherence to the DASH diet. This dietary pattern showed an inverse U-shaped relation with EO-FTD risk, with a decreased risk for subjects at very low and very high adherence levels (Supplemental Figure S22), while no association emerged for EO-AD (Supplemental Figure S21).

**Figure 10.** Spline regression analysis of early-onset dementia (EOD) risk for increasing intake of beverages. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with black spikes indicates the distribution of participant intake. Note: Spline analysis was not possible for most of the beverages (namely tea, red, white and aperitif wines and beers, spirits and soft drinks) due to a few subjects reporting consumption which is different from a null value.

**Figure 11.** Spline regression analysis for dietary patterns. Greek-Mediterranean (GM) diet, Dietary Approaches to Stop Hypertension (DASH) diet, and Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) diet. The black line indicates odds ratios for dementia risk; dash gray lines are 95% confidence limits; the reference line at 1.0 with gray bars shows the dietary-pattern score distribution.

#### **4. Discussion**

In this study, we found detrimental effects of consumption of high amounts of cereals, dairy products, and some types of sweets on EOD risk, while the intake of vegetables, dry fruits, and chocolate appeared to be beneficial, as was a specific dietary pattern, the MIND diet.

The role of dietary habits on cognitive decline has been investigated in previous studies [39], but the real association between diet and dementia, along with the mechanisms involved in beneficial and detrimental effects, are still unclear and thoroughly debated. In addition, it should be noted that previous studies were generally performed among individuals with late-onset dementia, while the present study specifically focused on younger subjects. Consistent with previous findings, we found a protective effect on EOD particularly with a higher consumption of leafy vegetables and fresh fruit. This protective effect on cognitive decline could be linked to the consumption of foods rich in bioactive substances such as polyphenols, highly present in fruits, vegetables, cereals, coffee/tea, cacao, and wine [40–42]. In the studies suggesting that vegetable intake showed an inverse association with cognitive decline [43,44], an apparent beneficial role of consumption of leafy and root vegetables has been reported [17,45]. Conversely, fruit intake showed contrasting results with null [17,44] or an inverse association [42], with the latter mainly emerging from the analysis evaluating dry fruit intake [17,40,46].

Similarly, the intake of omega-3 fatty acids due to fish consumption has been linked to improved cognition capacities [41,43,47–50]. Interestingly, we found a null association when overall fish and seafood intake was considered, whereas a slight decrease in risk could be found when considering fish, particularly fresh fish, only. The association we detected between a high daily intake of preserved and tinned fish and EOD is of interest, since such an intake in Italy is mainly due to the consumption of canned tuna [51], a source of heavy metals such as mercury, lead, cadmium, and a metalloid of potential neurotoxicity such as selenium [51–56].

Contrasting results with detrimental [57,58] or null effects [31,59] have been reported for dairy products. However, a preventive role has not yet been entirely ruled out, especially considering the different types of dairy products [60,61]. In particular, we found a dose-dependent association between dairy products and EOD risk. In addition, a null to negative association was observed for subjects reporting an intake of up to 350 g/day, when EOD risk starts to increase, thus suggesting adverse effects only at very high intake levels. A possible beneficial role in the cognitive function of substances released during fermentation processes, such as oleamide and dehydroergosterol, has been suggested, following suppression of microglial inflammation, as well as promotion of synaptic extension and neuronal survival [62].

A high cereal intake seems associated with a sharper cognitive decline [58]. In our study, we found an optimum cereal intake at approximately 200 g/day with reference to EOD risk. This is not surprising in that cereals (particularly bread) are fundamental components of the Mediterranean diet [63]. This mainly relates to a beneficial effect of whole grain consumption [64], possibly due to high concentrations of dietary fiber, resistant starch, and oligosaccharides, as well as several phytochemicals (phytates and phenolic compounds) characterized by antioxidant activities [65].

Sweets are generally associated with adverse cognitive effects [66,67]. Possible biological mechanisms of that association may involve generally high fat contents, especially high in saturated and trans-fats and lower in polyunsaturated and monounsaturated fats leading to blood brain barrier dysfunction and increased amyloid beta protein aggregation [68]. However, we observed an indication of beneficial effects from moderate consumption of chocolate products, consistent with previous studies [69,70]. Such beneficial effects, if real, may be due to the cocoa polyphenol intake [71] that might slow MCI progression to dementia [72].

We found a clear inverse association between coffee consumption and EOD risk. A recent dose-response meta-analysis of prospective studies [73] showed a statistically imprecise U-shaped relation between coffee intake and all-cause dementia risk, with the lowest dementia risk at two cups/day (RR = 0.90, 95% CI 0.95 to 1.08) and increased risk at and above five cups/day (RR = 1.11, 95% CI 0.94 to 1.30). Estimate imprecision might be due to the heterogeneity of exposure assessment methods (i.e., cups), which may correspond to different coffee amounts depending on country. Nonetheless, we also found a U-shaped association with a lowest risk at 70 g/day, corresponding to approximately two small (i.e., 'espresso type') cups, according to typically Italian habits as assessed in the EPIC-FFQ [74,75]. Contrary to our expectations based on the epidemiologic literature, we found no clear and meaningful relation between wine or alcohol intake and disease risk. No consumption and excessive consumption of wine or alcohol have both been associated with increased dementia risk [76,77]. In particular, a J-shaped relation has been proposed where low-to-moderate intake was associated with reduced risk, while both null and elevated intakes were correlated with increased risk [78–80].

Concerning dietary patterns, our results for EOD are fully consistent with previous studies, although these were generally carried out in older individuals, suggesting a beneficial effect on the overall cognitive function at higher adherence levels [19,35,81–86]. In particular, we found an indication of protective effects on EOD risk only at high adherence levels to the Greek-Mediterranean Diet (≥6) and DASH index (≥28). On the other hand, the only diet which was strongly and linearly associated with EOD risk was the MIND pattern. This was not entirely unexpected, since the MIND pattern has been shown to better predict an incidence of cognitive impairment or cognitive decline as compared to the Mediterranean Diet and DASH index [87,88]. In addition, the MIND diet has shown beneficial effects on other neurological diseases [89–91], while a clinical trial about the role of the diet on cognitive decline and brain neurodegeneration is ongoing [92].

The three dietary patterns under investigation have some similarities, since they emphasize natural plant-based foods and limit the intake of animal-derived products and high saturated-fats. Nevertheless, MIND has a distinctive pattern. It was developed as a hybrid of the DASH and Mediterranean diets with modifications, by reflecting the most compelling scientific evidence on foods and nutrients that protect the brain [93]. The MIND pattern attributes beneficial effects to the intake of cheese (<1 serving/week), green leafy vegetables (≥6 servings/week), berries (>1 serving/week), and fast fried food (<1 time/week) [35,83]. In addition, a high adherence to the investigated dietary patterns is generally associated with high levels of physical activity and other diet-related lifestyle factors [63] having beneficial effects on cognitive function [94,95]. These factors were not taken into account in our study. They might have contributed to a lower EOD risk and to masking the effect of the diet, which becomes evident only at high adherence levels.

To the best of our knowledge, our study is the first to investigate the role of dietary factors on early-onset dementia risk. In fact, previous studies focused on late-onset dementia or did not implement a specific restriction to EOD cases. The use of the validated EPIC-FFQ also allowed for a comprehensive assessment of dietary habits, including the investigation of single food categories and overall dietary-pattern adherence. In addition, we included recently-diagnosed EOD cases, thus limiting the risk of bias related to a long duration of the disease due to disease progression and increasing disability. Another strength of the present study was the ability to apply stratified analyses to the investigation of possible selective associations between diet and EOD risk according to the clinical dementia type.

Some limitations of our study should also be noted. First, the small samples of EOD participants, mainly due to the low incidence of the disease, affected the statistical precision of the risk estimates, especially for those food categories characterized by a large number of non-consumers. As a consequence, such a limited number of subjects hampered our ability to reliably identify weak associations. This was the case especially in subgroup analyses, which suggest caution in the interpretation of estimates characterized by a high imprecision, as well as the need to plan larger studies on the relations between diet and EOD. Secondly, although the EPIC-FFQ was designed and validated to collect dietary information at least over the previous year, we cannot rule out a possible misclassification due to changes of dietary habits of study participants, especially with reference to life-long habits. However, a recent survey comparing adherence to the Mediterranean diet over a long

time period (1997–2012) showed similar values for the Emilia Romagna population [96], suggesting a general stability over time of dietary habits. In addition, the use of caregivers as controls could have led to a risk of overmatching and downplayed the real associations between dietary factors and disease risk. As a matter of fact, some controls (particularly cases' family members) may have shared lifestyle habits including dietary patterns with EOD cases, thus reducing the strength of the associations we estimated. Finally, a genetic status was unavailable for almost all study participants. In particular, no data about the referent population could be retrieved, thus hampering the ability to assess their role in modifying EOD risk in relation to dietary factors.
