We observed a significant association between regular high-fat dairy consumption and the chance of regression to normal glycemia in a 9-year follow-up of Pre-DM subjects. Each 200 g/d of high-fat dairy corresponded to an elevated chance of returning to normal glycemia by 69%; higher intakes of high-fat dairy were also related to a lower 9-years average of 2h-SG among Pre-DM adults. A median consumption of 0.5 and 1.9 serving of milk per day, compared to its daily intake of <0.2 servings, was related to a better post-prandial glycemia over time. We further noted a significant positive association between daily consumption of yogurt and the chance of Pre-DM regression.

The causality and underlying mechanisms of the observed relations between dairy intake and the risk of developing T2D in cohort studies remain unclear [24]. Both protective and neutral effects are documented [15, 25,26,27,28,29]. Regular dairy consumption in relation to the risk of Pre-DM progression to T2D has been less established, and no evidence is available connecting dairy products to the chance of regression from Pre-DM to normal glycemia. High-fat dairy showed evidence of a dose-response and an inverse association with incident T2D in a 12-year follow-up of Pre-DM subjects (70% reduced risk, in relation to ≥14 vs. < 1 serving/week [25]. In a recent global observational study among 21 countries (low, middle, and high-income nations), the protective effect of whole-fat compared with low-fat dairy against the risk of T2D was numerically more strong [26]. Likewise, the Framingham Heart Study Of spring Cohort reported that higher high-fat dairy consumption decreased the risk of Pre-DM progression by 70% [25]. A part of the protective effects of dairy products in relation to the risk of T2D is attributed to their fatty acids profiles (medium-chain, odd, very long-chain SFAs, and trans-palmitoleic acid) [30]; however, other bioactive components, including probiotics, menoquinones, and milk fat globule membrane have also received a significant attention [24]. Biomarkers of dairy fat consumption, including C15:0, C17:0, and transC16:1n7, were significantly related to reduced risk of T2D (0.80, 95% CI = 0.73–0.87; 0.65, 95% CI = 0.59–0.72; 0.82 95% CI = 0.70–0.96) [31]. A dose-response meta-analysis of seventeen cohort studies, however, reported a neutral association between high-fat dairy and developing T2D (0.98, 95% CI = 0.94, 1.03) per 200 g high-fat dairy products/d) [28].

We did not find a significant association between low-fat and total fat dairy consumption and the chance of Pre-DM regression or progression. A recent report from the TLGS research group also indicated that a 3-years change in dairy intakes might modify the risk of developing T2D among Pre-DM adults. A decreased consumption of total dairy (> 0.5 servings/day) compared with a remaining stable state was related to an elevated chance of Pre-DM progression (OR = 1.56, 95%CI = 1.02–2.41), while increasing low-fat dairy consumption (especially milk and yogurt) by 0.50 serving/d was associated with a lower risk of T2D (OR = 0.56, 95% CI = 0.35–0.90) [15]. A pooled relative risk of 0.93 (95% CI = 0.87, 0.99) per 400 g total dairy products/d and 0.91 (95% CI = 0.86, 0.96) per 200 g low-fat dairy products/d was reported with a non-linear trend flattening of the curve at higher intakes [28].

Similar controversies are observed in observational studies connecting different dairy products with the risk of T2D. Non-linear inverse associations were found for daily consumption of yogurt (80 vs. 0 g/d, RR = 0.86, 95% C = 0.83, 0.90) and ice cream (for each 10 g/d, RR = 0.81, 95% CI = 0.78, 0.85) [32]; these inverse associations were not linear and no incremental benefits were found at a higher intake [32]. Pooled estimated relative risks of seventeen cohorts indicated a protective effect for cheese (RR = 0.92, 95% CI = 0.86–0.99, per 50 g cheese/d), and neutral effects for either milk (RR = 0.87, 95% CI = 0.72–1.04, per 200 g milk/d) and yogurt (RR = 0.78, 95% CI = 0.60–1.02, per 200 g yogurt/d) [28]. An updated meta-analysis also reported that higher yogurt consumption was significantly associated with decreased risk of T2D (OR = 0.83, 95% CI = 0.73–0.94) [29]. A 12-year follow-up of Pre-DM subjects showed that cheese consumption has a dose-response and an inverse association with incident T2D (63% reduced risk, in relation to ≥4 vs. < 1 serving/week for cheese) [25], while a neutral association between cheese intake and risk of T2D (RR = 1.00 per 10 g/d, 95% CI = 0.99–1.02) was obtained by a more recent meta-analysis [32]. A possible adverse effect of high cheese intake on glucose metabolism is supported by evidence indicating that dietary patterns with a high load of cheese consumption increase the risk of gestational diabetes, obesity, and abdominal obesity [33, 34]. In our study, the amount of yogurt consumption was meaningfully higher from a dietitian’s point of view in subjects who returned to normal glycemia compared to those who remained Pre-DM (1.73 vs. 1.1 serving/d), and higher intake of yogurt was related to reverting Pre-DM to normal glycemia.

The inconsistent findings for cheese and yogurt in the literature have been attributed to their complex and heterogeneous nature, a large variety of dairy products in different countries, and to differences between how these foods are eaten in diverse populations (e.g., hard cheeses consumed with fruit and nuts, plain whole-fat yogurt in the Spanish people, versus the use of processed cheese on pizza and deli meat sandwiches, or sugar-sweetened low-fat yogurt in US population). Our FFQ did not differentiate various types of cheeses (e.g., Lighvan, Koozeh, Gouda, Cumin, Feta, Cheddar, Iranian white cheese) and they were just categorized as regular and cream cheese, while previous reports attributed the different metabolic effects to multiple kinds of cheeses (e.g., fermented vs. non-fermented cheese, Dutch vs. curd cheese) [35]. Furthermore, we believed that the high-sodium content of cheese (~377–600 mg of sodium per serving, considered a significant contributor to total daily sodium intake [36, 37]) might explain some adverse effect of cheese, since high-sodium intake is a risk factor for developing T2D [38, 39]. Possible contamination of cheese with histamine-producing bacteria during cheese processing and storage [40], may also be a potential risk for developing T2D [41].

Due to the different underlying pathophysiology of IGT and IFG, the effects of diet on pre-DM progression and regression may differ for these subgroups. IFG is coincidental with reduced hepatic insulin sensitivity, β-cell dysfunction and mass reduction, altered glucagon-like peptide-1 secretion, and elevated glucagon secretion. In contrast, IGT is accompanied by a reduced peripheral insulin sensitivity with almost normal hepatic insulin sensitivity, a progressive loss of β-cell function, decreased secretion of the glucose-dependent insulinotropic polypeptide, and elevated glucagon secretion [7, 42]. Because of the relatively low sample size and lack of enough power, we could not differentiate between the potential effects of dairy intakes on the chance of regression/progression of Pre-DM in the isolated-IGT, isolated-IFG, or combined IFG-IGT subgroups. Because of distinct etiologies of isolated-IFG and isolated-IGT (relevance to genetic factors, smoking, and male sex, vs. physical inactivity, unhealthy diet, and short stature [42]), one can speculate that our findings would be more related to IGT rather than IFG state. Strong correlations between 2h-SG may support this idea compared to FSG with dairy intakes, especially high-fat dairy products over time.

Some strengths and limitations should be considered to interpret the study findings. Data collection use of a valid and reliable semi-quantitative 168-FFQ reduced the possibility of reporting biases. The well-known risk factors of T2D were detected and controlled in our analyses, however, due to existing of other possible unknown risk factors, complete controlling for confounders was not possible in our models. Repeated measurements of glycemic parameters and other covariates during the study follow-up, at the 3-year intervals, enabled us to monitor the study participants’ glycemic changes more precisely over time and detect the occurrence of the outcomes at mid-interval periods.

To sum up, our findings in the TLGS cohort provide some evidence against previous claims on adverse effects of whole-fat dairy products on cardiometabolic risk factors and support the last reported protective effect of high-fat dairy against the development of T2D, as increased high-fat dairy consumption over the course of follow-up were related to a considerably increased chance of Pre-DM regression to normal glycemia. Inversely, higher consumption of cheese was associated with the risk of developing T2D. These findings further support that regular consumption of dairy may attenuate the risk of developing T2D or the chance of returning to normal glycemia, and various dairy products may affect these pathways differently. Furthermore, our findings may indicate that the effects of regular dairy intake on the risk of developing T2D or the chance of returning to normal glycemia in Pre-DM subjects are mediated through improving glucose tolerance and insulin sensitivity, as dairy intake was associated with repeated measures and overall mean of 2h-SG rather FSG, over time.