Our cross-sectional study suggests that healthy dietary pattern is associated with lower plasma glucose concentrations and lower risk for hyperglycemia in middle-aged and elderly Finnish men. These results remained statistically significant after adjusting for confounding factors.

When the participants were stratified by their polygenetic risk score, healthy diet was still associated with lower plasma glucose concentrations in both low and high risk groups. The significance remained even after adjusting with BMI for all variables except for FPG in the low PRS group. After adjusting with age and other lifestyle factors (exercise, smoking and alcohol consumption), the healthy dietary pattern was associated with lower 2-h PG, gluc AUC, and higher Matsuda in the low PRS group, and with higher Matsuda and DI in the high PRS group. Both PRS groups seem to benefit from a healthy diet. There was no association between the unhealthy pattern and plasma glucose concentrations.

Healthy dietary pattern was associated with lower risk of hyperglycemia in both low and high genetic risk groups in an unadjusted model. A significant association was, however, lost after the adjustment for BMI. This may imply that the effects of a healthy dietary pattern in the hyperglycemia risk are mediated by BMI. Since adjusting with lifestyle factors beyond diet, such as smoking, exercise, and alcohol consumption, erases the statistical significance, it may be that the benefits of a healthy diet are confounded by an overall healthier lifestyle. The lack of statistical significance after the adjustments can also be due to lack of power when the study population has been divided into smaller groups. Our results suggest that focusing on adding healthy food choices is more essential than avoidance of unhealthy choices.

We found no interaction between the dietary patterns and the PRS in plasma glucose concentrations or in the risk of hyperglycemia. This suggests that the genetic risk does not modify the possible effects of diet.

Similarly to our data-driven pattern, foods such as fruits and berries, vegetables, non-tropical vegetable oils, whole grain products, fish, and low-fat dairy products, load into many established patterns such as Mediterranean Diet Score (MDS), Dietary Approaches to Stop Hypertension (DASH), Healthy Eating Index (HEI), Healthy Nordic Food Index (HNFI) and Baltic Sea Diet Score (BSDS) [47,48,49,50,51]. The two dietary patterns derived with PCA explained 17.5% of the food consumption variance. This level of explanatory value is in line with some similar data-driven dietary pattern studies [16, 37]. Since the principal component analysis aims to distinguish differences in dietary habits, there can be dietary components that are either beneficial or harmful for glucose metabolism that are not shown in the patterns. The dietary patterns identified by principal component analysis only explain a portion of the variance in food consumption. There may be individual dietary components with potential benefits or harmful effects on glucose metabolism that the PCA did not capture.

Our results are in line with previous studies showing beneficial effects of healthy dietary patterns and improved plasma glucose concentration [10], decreased risk for prediabetes [14, 16, 18], and decreased risk for T2D [13,14,15, 17, 52,53,54,55]. There are also some studies showing that unhealthy dietary pattern is associated with inferior plasma glucose concentration [22] and greater risk of prediabetes [18] and T2D [13, 21]. Our study implies similar results, but the results were not statistically significant.

Some previous studies support our findings showing that PRS does not affect the impact of diet on plasma glucose levels [29, 30, 36] or risk of T2D [31,32,33,34]. A few studies, however, report that the dietary impact on T2D risk may have interactions with the genetics [23, 28, 35, 37, 38], and that especially those in the highest genetic risk might benefit from healthy diet [39].

The strengths of the study include a rather large study population with oral glucose tolerance test enabling plasma glucose measurements beyond fasting levels, polygenetic risk score, and many variables measured that we were able to use to adjust for confounders.

The limitation of this study is its cross-sectional design. This allows us to make conclusions on associations without the possibility for interpreting causal relationships. Since our study was conducted in middle-aged Finnish men of Caucasian origin, the findings may not be extrapolated to other populations due to e.g. cultural, genetic, and lifestyle differences. Similar studies need to be conducted in females and participants of different ages and ethnicities. Our PRS stratified sample sizes may be inadequate to detect statistical differences across genetic groups. A retrospective food frequency questionnaire is based on the recall of the study participants, and self-reported dietary assessment methods are prone to misreporting [56]. Apart from bread consumption where the consumption was asked with quantitative measures, the qualitative questionnaire captures consumption frequencies of certain foods instead of measuring total food, energy, or nutrient intakes. We have no data on the amounts of foods consumed and there can be aspects of healthy diet that are not captured in this paper due to this limitation. The questionnaire covers a range of typical foods consumed in Finland and can be used to rank participants’ dietary habits. It also measures the usual diet from the past 12 months, which is a rather long period covering seasonal changes in the diet. The questionnaire is, however, a non-exhaustive list of foods consumed, which limits the components included in the dietary patterns. Even with adjustment with covariates such as BMI, age, exercise, smoking, and alcohol consumption, we cannot rule out possible residual confounders that may mediate the results found. We acknowledge the potential risk of multiple testing error, which is always present when more than one endpoint is being tested.

Our results support the existing literature on the protective effect of healthy dietary pattern against hyperglycemia. In addition to reduced risk for T2D, healthy dietary patterns have been shown to reduce risk for cardiovascular diseases, cancer, bone health, mental health, obesity, and premature death [57].