Methods

Diet intervention

Participants received two approximately iso-energetic, study diets (LFHA: 9121 kJ; HFHA 8955 kJ): one containing 23 % of energies from fat and the other with 42 % of energies from fat.

The fatty acid distribution of both diets was 2:1:1 (MUFA:PUFA:SFA). High-AGE cooking methods (broiling, roasting, baking and frying) were used for the preparation of both diets. (see Table 1).

Table 1 Macronutrient and advanced glycation end products (AGE) content of study diets low-fat, high-AGE and high-fat, high-AGE dietFootnote * ^, Footnote †

AT, α tocopherol.

* Numbers in AGE column (kU) represent the calculated AGE value of the food item from tables in the article by Uribarri et al.⁽ Reference Uribarri, Woodruff and Goodman ^{14 )}, which employs ELISA.

† Numbers in AGE column (mg) represent the estimated calculated AGE value of the food item from tables in the article by Hull et al.⁽ Reference Hull, Woodside and Ames ^{13 )}, which employs HPLC.

Breakfast and lunch were consumed in the laboratory. Dinner, afternoon and bedtime snacks were consumed in the laboratory or boxed and given to the participant to be consumed at home, according to participants’ need.

Dietary analysis

Dietary macronutrient content was calculated initially using Nutritionist Pro (Axxya) and then reanalysed using Nutrition Data System for Research 2014. Then, 100-g frozen samples of the diets were sent for proximate analysis by an independent laboratory (Pope Labs). Analysis of the diet sample is shown in Table 2.

Table 2 Comparison of test diets: calculated macronutrient values v. experimentally determined values

LFHA, low-fat, high-advanced glycation end products (AGE) diet; HFHA, high-fat, high-AGE diet.

Participants were allowed to eat only the study foods during each 1-d intervention period.

Results

Anthropometric measurements and biochemical analyses of the study participants at baseline are listed in Table 3. Some individual participants showed fluctuations in body weight over the course of the study, but there were no overall significant changes in weight, BMI or percentage of body fat over the 2–4-week study period. Anthropometric and biochemical variables were compared between males and females enrolled in the study. Males and females were different with regard to body fat composition (males: 27 (sd 4) %; females: 41 (sd 4) %; P<0·0001) but were otherwise similar. Data from males and females were grouped together for the remaining analyses.

Table 3 Anthropometric measurements and selected biochemical analyses of study participants at baseline: all participants, males only, females only (Mean values and standard deviations; medians and 25th–75th percentiles)

TC, total cholesterol; CML, N ^ε-carboxymethyl-lysine; sRAGE, soluble receptor for advanced glycation end products; hsCRP, high sensitivity C-reactive protein; HMW, high molecular weight.

The effect of diet sequence and the interaction between diet effect and sequence were tested, and no carryover effect was observed.

Participants tolerated both the HFHA and LFHA diets well and consumed all of the study foods in most cases. One participant reported indigestion, which she attributed to consumption of diet cola, and another participant reported vomiting a small amount following consumption of cashew nuts. Otherwise, both study diets were well tolerated.

CML levels following consumption of the LFHA breakfast meal rose from a fasting median of 518 (25th–75th percentile 425–803) ng/ml to a peak of 553 (25th–75th percentile 425–799) ng/ml at 2 h postprandial, then decreased to 495 (25th–75th percentile 391–682) ng/ml by 3 h postprandial and rose to 538 (25th–75th percentile 403–740) ng/ml the next day (see Table 4). By contrast, CML levels following consumption of the HFHA breakfast meal rose from a fasting median of 463 (25th–75th percentile 428–664) ng/ml to a presumed peak of 578 (25th–75th percentile 474–865) ng/ml at 3 h, before declining slightly to 565 (25th–75th percentile 426–769) ng/ml the next day. The pattern of change in CML following the LFHA meal was generally more complex with greater variability from person to person. At 3 h postprandial, only the HFHA diet showed a pattern of significant change in CML compared with baseline (interaction P=0·05). By contrast, the change in CML following the LFHA diet was not significant at 3 h (P=0·36) or even at the numerical peak at 2 h (interaction P=0·32). However, the response to the two diets over time was significantly different (interaction P=0·01) (see Table 4 and Fig. 1). sRAGE levels following consumption of LFHA breakfast meals decreased from a fasting median of 556 (25th–75th percentile 402–676) pg/ml to 515 (25th–75th percentile 379–642) pg/ml by 1 h, to 492 (25th–75th percentile 330–623) pg/ml by 2 h and to 484 (25th–75th percentile 313–580) pg/ml by 3 h (time effect P=0·02). The changes at 2 and 3 h were statistically significant (P<0·01). Following the HFHA breakfast meal also sRAGE decreased significantly from baseline to 1 to 2 to 3 h postprandially (time effect P=0·008) and then rose the next day, though not significantly. The response pattern in sRAGE to the diets differed significantly (interaction P=0·05) (see Table 4 and Fig. 2). AUC for sRAGE at 0, 1, 2 and 3 h did not differ significantly between LFHA and HFHA diets.

Fig 1 (a) N ^ε-carboxymethyl-lysine (CML) 3-h time-course: , low-fat, high-AGE diet v. , high-fat, high-AGE diet; (b) fasting CML levels before diet (baseline day 1) and after diet (day 2). Values are geometric means and 95 % CI.

Fig 2 (a) Soluble receptor for AGE (sRAGE) 3-h time-course: , low-fat, high-AGE diet v. , high-AGE diet; (b) fasting sRAGE levels before diet (baseline day 1) and after diet (day 2). Values are geometric mean and 95 % CI.

Table 4 N ^ε -carboxymethyl-lysine (CML) and soluble receptor for advanced glycation end products (sRAGE) response to low-fat, high-advanced glycation end products (AGE) diet (LFHA) v. high-fat, high-AGE diet (HFHA) diets at baseline, 1, 2, 3 and 24 h: CML, sRAGE (Medians and 25th–75th percentiles)

* Statistically significant (P≤0·05).

† Represents the interaction between diet and time (i.e. a difference in diet response over the time-course) in CML from 0 to 3 h postprandial.

‡ Represents the interaction between diet and time (i.e. a different in diet response over the time course) in sRAGE from 0 to 3 h postprandial.

Other biochemical variables changed little from baseline to fasting the next day; however, TC decreased significantly for participants consuming the HFHA diet (HFHA: baseline: 5·83 (sd 1·24) mmol/l, post: 5·52 (sd 1·13) mmol/l, P=0·02; LFHA: baseline: 5·28(sd 0·75) mmol/l, post: 5·34 (sd 1·04) mmol/l). The response in cholesterol to the diets was different (P=0·03). High sensitivity CRP (hsCRP, a pro-inflammatory indicator) and HMW adiponectin (an anti-inflammatory indicator) did not differ following consumption of either diet (P=0·61 and 0·92, respectively) (see Table 5). The 0 v. 24 h changes were not statistically different within either diet.

Table 5 Response to low-fat, high advanced glycation end products (AGE) diet (LFHA) v. high fat, high AGE diet (HFHA) diets at baseline and 24 h: glucose, cholesterol, HDL, TAG, high molecular weight (HMW) adiponectin and C-reactive protein (CRP) (Medians and 25th–75th percentiles)

TC, total cholesterol; hsCRP, high sensitivity CRP

* Statistically significant (P<0·05).

† Represents the interaction between diet and time (i.e. a difference in diet response over time) from 0 to 24 h postprandial.

Discussion

This study is the first to compare consumption of a low-fat and high-fat diet, each high in AGE, in terms of serum AGE, sRAGE and indicators of inflammation. Our study had some limitations, including small sample size, the presence of many participants with dyslipidaemia despite the screening to identify healthy participants and the short duration of the study.

Another potential limitation of this study is the use of circulating CML as a proxy measure of AGE and a primary end point. There remains controversy within the realm of AGE research regarding the following: whether CML is the most significant AGE physiologically; whether it negatively impacts health; the extent to which circulating CML is affected by dietary CML; and the validity of ELISA as a method to measure it⁽ Reference Poulsen, Hedegaard and Andersen ^{19 ,} Reference Uribarri, Cai and Sandu ^{20 ,} Reference Ames ^{21 ,} Reference Semba, Gebauer and Baer ^{42 ,} Reference Ames ^{51 –} Reference Piroddi, Palazzetti and Quintaliani ^{55 )}. Because multiple human studies suggest an association between consumption of heat-treated foods rich in CML and risk factors for chronic illness such as endothelial dysfunction and inflammation⁽ Reference Vlassara, Cai and Crandall ^{25 ,} Reference Negrean, Stirban and Stratmann ^{28 –} Reference Uribarri, Negrean and Sander ^{30 ,} Reference Harcourt, Sourris and Coughlan ^{32 ,} Reference Uribarri, Cai and Ramdas ^{33 ,} Reference Stirban, Negrean and Stratmann ^{35 ,} Reference Luevano-Contreras, Garay-Sevilla and Wrobel ^{41 ,} Reference Birlouez-Aragon, Saavedra and Tessier ^{56 )}, we felt it important to assess how a low-fat v. a high-fat meal impacts the postprandial response to CML. In addition, because several studies by the same group, which use ELISA to measure CML, have found a rise in circulating CML following consumption of a high-AGE diet⁽ Reference Koschinsky, He and Mitsuhashi ^{10 ,} Reference Negrean, Stirban and Stratmann ^{28 ,} Reference Uribarri, Negrean and Sander ^{30 ,} Reference Stirban, Negrean and Stratmann ^{35 )}, we wanted to attempt to replicate this finding using the same methodology. The AGE content of the diets was calculated using the tables created by Uribarri et al.⁽ Reference Uribarri, Woodruff and Goodman ^{14 )}. However, we separately estimated the AGE content of the diet using tables created by Hull et al.⁽ Reference Hull, Woodside and Ames ^{13 )}, which use HPLC, choosing similar foods from that database when exact matches were not available, and found little difference in AGE content between the two diets using this method (8·3 mg for the LFHA diet v. 8·9 mg for HFHA diet (a 6 % difference) (see Table 1). In spite of these limitations, the findings are important in several ways.

Even though both diets were calculated to be the same in AGE and used almost the same foods with fruit juices added to the low-fat diet in order to compensate for decreased energy due to less fat, the measured AGE content of the total LFHA diet, which was quantified using ELISA, was about 15 % higher in AGE compared with the AGE content in the high-fat diet (data not shown). It is possible that this measurement was inaccurate due to the limitations of ELISA in quantifying CML in food substrates⁽ Reference Ames ^{51 )}. The calculated AGE content of the test meals using the tables of Hull et al.⁽ Reference Hull, Woodside and Ames ^{13 )} showed the HFHA test meal to be about 17 % higher in CML compared with the LFHA test meal; however, the food lists of Hull were significantly limited in scope and often did not contain exact matches to the foods used in this study; thus, these calculations may not be accurate. The foods served in the breakfast test meals were designed to be almost identical, only differing in fat content. The meat offering in the LFHA test meal was smoked deli ham, and the HFHA meal contained bacon. Whole milk was used in place of skimmed milk in the HFHA meal. Slightly more margarine was offered in the HFHA meal. Finally, the breakfast cereals were made from different grains, with the HFHA meal containing an oat and maize-based cereal, whereas the LFHA cereal was rice-based; however, both cold cereals were prepared from moistened grains, which were subsequently cooked at high temperatures. Because of the strong similarity between the breakfast meals and the methods of preparing them, it is probable that the meals were very similar in AGE content. However, whereas consumption of the LFHA diet was associated with a non-significant postprandial decrease in serum CML with no rise in CRP or reduction in adiponectin at 24 h, the HFHA diet was associated with a significant rise in postprandial serum CML with no rise in CRP or reduction in adiponectin at 24 h.

Because we used fruit juices to balance the energy intake between the diets while attempting to hold protein steady and change only the quantity – not type – of fat, we may have introduced more inhibitors of AGE formation into the total diet. Inhibitors of AGE formation in cooked food include plant compounds such as flavonoids, phenolic acids, terpenes and vitamins⁽ Reference Poulsen, Hedegaard and Andersen ^{19 ,} Reference Wu, Huang and Lin ^{57 )} as well as acid content, moisture and cooking times and temperatures⁽ Reference Poulsen, Hedegaard and Andersen ^{19 )}. However, because we did not cook the foods together with the fruit juices, it is unlikely that these compounds inhibited formation of AGE in the food products used for the test diet. In addition, the fruit juice type and amount was identical at the breakfast test meal. Thus, the reason for the postprandial rise in CML with the HFHA diet is not explained by differences in fruit juice consumption.

It is possible that consumption of the high-fat meal resulted in greater peroxidation of unsaturated fatty acids, which led to higher serum CML levels. However, the HFHA meal contained more vitamin E; thus, one might expect that lipid peroxidation of the fats ingested would have been minimised, and postprandial CML levels would have been similar between the two diets.

Antioxidant content of the diets might also have affected serum CML. The antioxidant content was necessarily different between the two diets. Specifically, the α-tocopherol content of the HFHA diet was 43 % higher than that of the LFHA diet (16 v. 9 mg). The opposite relationship was true of the vitamin C content of the two diets: the LFHA diet had a 46 % higher content compared with the HFHA diet (138 v. 75 mg). Therefore, although the antioxidant content of the diets differed, whereas the HFHA diet was lower in vitamin C, it was higher than the LFHA diet in vitamin E. Higher α-tocopherol was present in the HFHA breakfast test meal compared with LFHA, but vitamin C content was identical between the two meals. It seems unlikely that these antioxidant differences resulted in higher circulating CML following consumption of the HFHA diet.

It is possible that the study results were confounded by previously ingested AGE compounds, which were gradually degraded by intestinal microbiota in the large intestine and later contributed to a rise in serum CML. However, the participants were fasting and did not report previous digestive problems; thus, they might be supposed to have typical gut transit times, which should have allowed minimal presence of these compounds present from previous meals. In addition, one might expect the effect to be similar between diet phases.

Finally, it is possible that the participants in this study were aware of their group assignment and altered their dietary pattern accordingly. Participants were not told of their group assignment; however, in a feeding study of this nature, true blinding is not possible. Because the foods served in both arms of the trial seem very similar to the lay person (2 % Cheddar v. whole milk Cheddar on a grilled cheese sandwich with the same bread; one slice of toast with 2·5 or 5 g of margarine; whole milk v. skimmed milk), several participants wondered aloud to the primary researcher regarding their study diet, seemingly unaware of their group assignment. Thus, although it is possible that participants altered their home diets based on group assignment, we do not think this occurred to any appreciable or consistent extent. Because all study foods were provided during the day of study and energy intake was substantial, we do not think there was likely to be much eating of foods outside of the study diet. Most participants reported feeling that it was ‘a lot of food’. Although it is possible that recent AGE intake could influence CML levels during the study, we attempted to limit this effect by requiring fasting and asking participants to keep a 3-d food record before beginning the study.

Thus, the mechanism for the rise in CML following the HFHA meal remains unclear. However, a possible explanation is that the higher fat content of the breakfast test meal caused delayed gastric emptying, allowing longer time for hydrolysis of the AGE-containing proteins, making CML more available for absorption. With documented low absorption of dietary AGE for reasons that are not entirely clear but may relate to the resistance of AGE to break down by enzymes or acid⁽ Reference Koschinsky, He and Mitsuhashi ^{10 )}, absorption in two diets and meals that are similar in AGE content may relate more to other characteristics of the meal affecting digestion and absorption, rather than solely the AGE content of the diet.

These findings suggest that the contribution of dietary AGE to circulating AGE may be modulated by dietary factors such as phytochemicals, dietary fat or other unidentified factors. These various dietary factors may affect circulating AGE by modulating absorption and/or metabolism. Our breakfast test meal had a lower calculated AGE content than that of the low-fat, high-AGE dinner test meal evaluated by Poulsen⁽ Reference Poulsen, Bak and Andersen ^{36 )}. However, like Poulsen, we also found no significant change in serum CML following consumption of the LFHA diet. We did find an increase in CML following the HFHA meal, but we found no difference in CML by 24 h. Uribarri et al.⁽ Reference Uribarri, Negrean and Sander ^{30 )} demonstrated increases in serum AGE at 90 min following a single AGE-rich beverage challenge, whereas we found no such effect at 1 or 2 h for either high-AGE dietary challenge. Stirban et al. ⁽ Reference Stirban, Negrean and Stratmann ^{35 )} found significant decreases in adiponectin following a single high-AGE meal, but we found no such effect. In contrast, we found that CRP and adiponectin were very stable in the presence of both diets. Because CRP has a half-life of about 19 h and has stable concentrations over time, which are not thought to be influenced by fasting v. eating⁽ Reference Pepys and Hirschfield ^{58 )}, one might not expect changes in CRP within such a short time period. However, synthesis rates of CRP can increase rapidly, thus producing strong, rapid shifts in CRP in response to inflammatory stimuli⁽ Reference Pepys and Hirschfield ^{58 )}. Finally, Negrean et al.⁽ Reference Negrean, Stirban and Stratmann ^{28 )} reported an increase in CRP with the administration of a single very high-AGE test meal. AGE are theorised to increase inflammation via up-regulation of NFκB and TNFα ⁽ Reference Brownlee ^{7 ,} Reference Ramasamy, Vannucci and Yan ^{59 )}; thus, one might theorise that these molecules would up-regulate low-grade inflammation rapidly in the presence of high AGE intake. However, we did not find any effect of either diet on hsCRP.

In addition, sRAGE decreased in response to both study diets. sRAGE is thought to act as a decoy receptor and bind to excess AGE. However, the decrease in sRAGE does not suggest an up-regulation of sRAGE in response to a higher AGE dietary load.

It is not clear why with healthful fats included in both test diets, very similar micronutrient profiles and almost identical foods being purchased whenever possible, there would be a trend towards decrease in CML in the presence of the LFHA diet and a rise in CML after consumption of the HFHA diet. We suggest that this result could relate to delayed gastric emptying with the higher fat content of the HFHA diet. This finding suggests that serum CML may not be substantially influenced by dietary AGE content in healthy adults at all given such low absorption. Our results are supported by the findings of Semba et al.⁽ Reference Semba, Ang and Talegawkar ^{37 )} who reported that estimation of dietary AGE using dietary records did not correlate with serum AGE levels. In this study, intake of foods thought to be high in AGE also did not correlate with serum CML. Finally, Semba et al. ⁽ Reference Semba, Gebauer and Baer ^{42 )} recently published a longer trial of a well-controlled high-AGE diet in healthy adults and found no effect of dietary CML on inflammatory mediators or endothelial dysfunction.

This study is limited in its size, scope and methods. To uncover whether dietary AGE influences short-term circulating levels of AGE, feeding of radioactively labelled CML and other dietary AGE could be performed followed by several hours of postprandial monitoring. In addition, such a feeding protocol would need to be carried out in the context of various dietary modifications: high-fat, low-AGE; HFHA; LFHA; and low-fat, low-AGE. Beyond the effect of macronutrient content on AGE absorption, the presence of dietary inhibitors of AGE formation in the diet would need to be evaluated for its impact on serum CML. How dietary AGE influences circulating AGE and markers of oxidative stress and inflammation in the context of diabetes as well as renal impairment also needs to be more closely assessed. Ideally, measurement of CML or other AGE in study diets should be performed via HPLC. The work of food scientists assessing how CML and other Maillard reaction products are influenced in food by cooking temperature, time, macronutrients and inhibitors indicates that the AGE content of foods is highly complex and is affected by the amount of oil and carbohydrate in a cooked food, heating temperature, heating time, acidity and the presence or absence of polyphenols⁽ Reference Delgado-Andrade, Seiquer and Haro ^{15 ,} Reference Srey, Hull and Connolly ^{16 ,} Reference Poulsen, Hedegaard and Andersen ^{19 )}. As complex as this reaction may be in foods, we should not suppose that the digestion, absorption and physiology of AGE in the body is any less complex.

Our results suggest that at least in the very short term, dietary CML alone is not very important to serum CML, sRAGE, CRP and HMW adiponectin. Thus, there is insufficient evidence at this time to recommend the public to modify their intake of AGE to improve health status. Before such recommendations could be made, we need reliable methods of estimating AGE content of foods, knowledge of which AGE have the greatest physiological impact on health outcomes (if any) and how cooking and food preparation methods affect formation of AGE.