Search Weight Loss Topics:

Home » Uncategorized » Lipidome changes due to improved dietary fat quality inform cardiometabolic risk reduction and precision nutrition – Nature.com

Study designs and populations DIVAS trial

Lipidomics analysis was performed in a subset of participants (n=113 of 195) from the DIVAS trial, a 16-week, single-blind randomized controled parallel trial (registered at http://www.clinicaltrials.gov under accession number NCT01478958). The DIVAS trial was conducted according to the guidelines of the Declaration of Helsinki, and favorable ethical opinion for conduct was given by the West Berkshire Local Research Ethics Committee (09/H0505/56) and the University of Reading Research Ethics Committee (09/40). All individuals provided written informed consent before participating. This study recruited men and women aged between 21 and 60 years and with estimated moderate CVD risk who were randomized to one of three isoenergetic diets: rich in SFAs, rich in MUFAs or rich in mixed UFAs including both MUFAs and omega-6 PUFAs. The target compositions (percent total energy of total fat:SFA:MUFA:PUFA) were 36:17:11:4 for the SFA-rich diet (n=38), 36:9:19:4 for the MUFA-rich diet (n=39) and 36:9:13:10 for the mixed UFA-rich diet (n=36). We collapsed the MUFA-rich and mixed UFA-rich diets into one UFA-rich diet arm for the generation of the MLS.

In the DIVAS dietary intervention trial, all participants diets were isoenergetic and provided 36% of total energy (percent total energy) from fats. Nonfat macronutrient intake and sources were consistent between the intervention and control diets. However, different spreads, oils, dairy products and snacks were used to modify the diets SFA:UFA ratio. The control diet was high in saturated fat (SFA-rich diet; 17% of total energy from SFAs and 15% of total energy from UFAs; n=38 with lipidomics data). In the intervention diet, 8% of total energy from SFAs was substituted for 8% of total energy from UFAs (UFA-rich diet; 9% of total energy from SFAs and 23% of total energy from UFAs; n=75 with lipidomics data). The analysis of 4-day weighed diet diaries indicated successful implementation of these dietary targets over the intervention period (Fig. 2a)29,35. The SFA:MUFA:omega-6-PUFA content in percent total energy in the control group was 17:11:4 and was either 9:19:4 or 9:13:10 in the intervention group arms with different MUFA:PUFA ratios. The omega-3-PUFA content was standardized across all diet groups. Extensive sensitivity analyses revealed that our analysis workflow yielded highly consistent results in the two intervention arms. Therefore, we present comparisons between the control group (high SFA intake) and a pooled intervention group (high UFA intake). We collapsed the MUFA-rich and mixed UFA-rich diet into one UFA-rich diet arm to generate the MLS.

All participants were nonsmokers; were not pregnant or lactating; had normal blood biochemistry and liver and kidney function; did not take dietary supplements or medication for hypertension, raised lipids or inflammatory disorders; had no prior diagnosis of MI, stroke or diabetes; did not consume excessive amounts of alcohol (males: less than 21U per week; females: less than 14U per week) and performed fewer than three 30-min sessions of aerobic exercise per week. The trial was single blinded, and randomization was conducted by a study researcher using minimization stratified for sex, age, BMI and estimated CVD risk. The participants were unaware of the assigned intervention diet and were asked to replace habitually consumed sources of exchangeable fats with study foods (spreads, oils, dairy products and commercially available snacks) of specific fatty acid composition provided free of charge.

Dietary guidance was provided at baseline and throughout the study via 1:1 verbal and written instructions. Compliance was monitored through weighed 4-day diet diaries (weeks 0, 8 and 16), records of study food intake and plasma phospholipid fatty acids as short-term biomarkers of intake (weeks 0 and 16). Observed fatty acid intake compositions were largely in line with the defined target fatty acid compositions35. Body weight, which was to remain constant, was monitored every 4 weeks, and changes were addressed with advice to the participants to adapt study food or carbohydrate consumption and/or activity levels. Fasting blood samples were taken at baseline and after 16 weeks at a similar time of day, and blood fractions were immediately separated and stored at 80C.

The EPIC-Potsdam cohort study is a prospective cohort study that recruited 27,548 participants (16,644 women and 10,904 men of primarily Middle European ancestry, age range: 3565 years) from the general population of Potsdam, Germany, and the surrounding geographical area from 1994 to 1998. Follow-up occurred every 23 years by mailed questionnaires and, if necessary, by telephone. Response rates ranged between 90% and 96% per follow-up round. The study protocol was approved by the ethics committee of the Medical Society of the State of Brandenburg, Germany, and all participants provided a statement of written informed consent before enrollment.

Incident CVD was defined as incidence of primary nonfatal and fatal MI and stroke (International Statistical Classification of Diseases and Related Health Problems (ICD)-10 codes: I21 for acute MI, I63.0 to I63.9 for ischemic stroke, I61.0 to I61.9 for intracerebral hemorrhage, I60.0 to I60.9 for subarachnoid hemorrhage and I64.0 to I64.9 for unspecified stroke). Incidence of CVD was captured by participants self-reports or based on information from the death certificates, which were validated by contacting the treating physicians. Inquired information included ICD-10 code, date of occurrence and further information on symptoms and diagnosis criteria. For MI, diagnostic criteria included clinical symptoms, electrocardiograms, cardiac enzymes and known coronary heart disease. For stroke, diagnosis was based on anamnesis, clinical symptoms, computed tomography/magnetic resonance tomography, angiogram, lumbar puncture, echocardiogram, Doppler and electrocardiogram plus imaging techniques if available. Participants with silent cardiovascular events that had not been documented within 28days after occurrence were excluded as nonverifiable cases from all analyses.

Information on incidence of T2D was systematically acquired through self-report of a diagnosis, T2D-relevant medication or dietary treatment due to T2D diagnosis during follow-up. Additionally, death certificates and information from tumor centers, physicians or clinics that provided assessments for other diagnoses were screened for indication of incident T2D. For participants who were classified as potential cases based on that information, a standard inquiry form was sent to the treating physician. Only physician-verified cases with a diagnosis of T2D (ICD-10 code E11) and a diagnosis date after the baseline examination were considered confirmed incident cases of T2D.

Nested casecohorts were constructed for efficient study of molecular phenotypes. From all participants who provided blood at baseline (n=26,437), a random sample (subcohort, n=1,262) was drawn, which served as a common reference population for both endpoints. For each endpoint, all incident cases that occurred in the full cohort until a specified censoring date were included in the analysis. After excluding prevalent cases of the respective outcomes, the analytical sample for T2D comprised 1,886 participants, including 775 incident cases (26 cases in the subcohort), and the analytical sample for CVD comprised 1,671 participants, including 551 incident cases (28 cases in the subcohort). Follow-up was defined as the time between enrollment and study exit determined by diagnosis of the respective disease, death, dropout or final censoring date, whichever came first. Endpoint-specific censoring dates were 30 November 2006 for stroke and MI and 31 August 2005 for T2D.

Anthropometric and blood pressure measurements were conducted according to a standardized protocol65,66. Information on lifestyle and education was obtained using computer-assisted personal interviews. These included information on recreational physical activity, smoking status, average alcohol intake and educational attainment. Participants were categorized as hypertensive at study baseline if they had a systolic blood pressure of 140mmHg, diastolic blood pressure of 90mmHg, reported prior diagnosis of hypertension or current antihypertensive medication use. At baseline, trained study personnel obtained 30ml of peripheral venous blood from each participant. Blood was partitioned into serum, plasma (with 10% of total volume citrate) and blood cells and was subsequently separately stored in tanks of liquid nitrogen at 196C or in deep freezers at 80C until the time of analysis. Plasma samples, from which aliquots were drawn for the lipidomics measurements in 2016, were never or only once thawed and refrozen during storage (93 samples were defrosted and refrozen once for aliquoting for unrelated analysis).

Plasma concentrations of standard blood lipids (total cholesterol, HDL-C, triglycerides, HbA1c, glucose and hsCRP) were measured at the Department of Internal Medicine, University of Tbingen, with an automatic ADVIA 1650 analyzer (Siemens Medical Solutions) in 2007. All biomarker measurements conducted in plasma, including the lipidomics measurements (detailed below), were corrected for the dilution introduced by citrate volume to improve comparability with concentrations measured in EDTA-plasma reported in the literature. Laboratory measurements were conducted by experienced technical personnel following the manufacturers instructions. Single imputation based on linear regression was used to impute missing covariate information (participants with missing data for: waist circumference, n=2; BMI, n=12; standard blood lipids (triglycerides, HDL-C and triglycerides), n=82; and blood pressure, n=148).

The NHS recruited 121,701 female nurses aged 3055 years in 1976 (ref. 67). A subset of 32,826 nurses provided blood samples in 1989 or 1990, of whom 18,743 provided a second blood sample in 2000 or 2001. The NHSII cohort was established in 1989 and recruited 116,429 female nurses aged 2542 years. In NHSII, blood samples from 29,611 participants were collected between 1996 and 1999. The standardized blood collection procedure is described elsewhere37. Participants reported their usual intake of a standard portion of each item in the FFQ (frequency ranging from never to more than six times per day) during the past year every 4 years. The reproducibility and validity of the FFQ has been extensively documented68,69,70. The NHSs were approved by the Human Research Committee at the Brigham and Womens Hospital, Boston, MA, and participants provided written informed consent.

We computed the intake of individual nutrients by multiplying the frequency of consumption of each food by the nutrient content of the specified portion based on food composition data from the US Department of Agriculture and data from manufacturers. Intake of carbohydrate, fat and protein was expressed as nutrient densities (that is, percent energy)71. In a validation study comparing energy-adjusted macronutrient intake assessed by the FFQ with four 1-week diet records, the Pearson correlation coefficients were 0.61 for total carbohydrates, 0.52 for total protein and 0.54 for total fat70.

Participants who reported a stroke were asked for permission to review their medical records. For both nonfatal and fatal strokes, available medical records related to the clinical event, such as imaging and autopsy reports, were reviewed by physicians who were blind to participant risk factor status. Strokes were defined according to the National Survey of Stroke criteria and were classified as ischemic or hemorrhagic72,73. The ischemic stroke lipidomics casecontrol study in the NHS/NHSII cohorts used in our analyses included 968 participants with lipidomics data to construct the rMLS (484 casecontrol pairs). Matching factors included age, fasting, smoking status, race, ethnicity and season of blood collection.

In NHS/NHSII cohorts, T2D incidence was detected based on self-reported diagnosis and was confirmed by a validated supplementary questionnaire74. Before 1998, confirmation of T2D incidence relied on the National Diabetes Data Group criteria and from 1998 onward relied on the American Diabetes Association diagnostic criteria. Validation studies in the NHS have demonstrated the validity of the supplementary questionnaires to adjudicate T2D diagnosis, showing that more than 97% of participants with self-reported T2D detected by questionnaires were reconfirmed through medical record review by endocrinologists blinded to questionnaire information74,75.

We also designed a 1:1-matched nested casecontrol study for lipidomics and T2D. Matching factors were age, race, ethnicity and season of blood collection. The T2D casecontrol study in NHS included 1,456 participants (728 matched casecontrol pairs) with baseline lipidomics data to construct the rMLS. A subset of casecontrol pairs had repeated lipidomics data approximately 10 years apart to construct the rMLS based on fasting (8h) blood samples from both times (1989/1990 and 2000/2001). In the repeated blood sampling study, all participants remained diabetes free until after the second blood collection, and all incident T2D cases occurred between 2002 and 2008.

The study protocols were approved by the Institutional Review Boards of Brigham and Womens Hospital and Harvard T.H. Chan School of Public Health. Participants completion of questionnaires was considered as implied consent.

The PREDIMED study was a multicenter dietary intervention trial with 7,447 participants in three intervention arms and demonstrated cardiometabolic risk reduction by a Mediterranean diet intervention (www.predimed.es; ISRCTN registry: ISRCTN35739639)33,76. The PREDIMED trial received ethical approval from the Institutional Review Board of the Hospital Clinic in Barcelona, Spain, 16 July 2002. The PREDIMED trial inclusion criteria were either prevalence of T2D or prevalence of three or more major cardiovascular risk factors (smoking, dyslipidemia, hypertension and adiposity). Besides the low-fat diet control group, the Mediterranean diet intervention included two arms (one particularly high in extra virgin olive oil and the other particularly high in tree nut intake) that we pooled into one Mediterranean diet group for our primary analyses. Preintervention blood samples were taken after an overnight fast by trained study personnel according to a standard protocol and fractioned, and the EDTA-plasma was stored at 80C in deep freezers.

The PREDIMED T2D casecohort study with available lipidomics data comprised 694 randomly selected participants (approximately 20% of participants) who fulfilled inclusion criteria, that is, no prevalent T2D at recruitment and available plasma samples and all incident T2D cases during a median of 3.8 years of intervention (n=251; per casecohort design 53 incident T2D cases were randomly included in the subcohort). The analytical sample was restricted to participants with complete data on lipid metabolites in the rMLS (n=678, including 211 participants with incident T2D). Of those, 468 participants (including 148 participants with subsequent T2D incidence) had additional plasma samples and lipidomics profiles from 1 year after recruitment.

The PREDIMED CVD casecohort study with lipidomics data comprised 791 randomly selected participants with available plasma samples at recruitment (approximately 10% of the eligible participants) and all incident CVD cases during a median of 3.8 years of intervention (n=231). After excluding participants with missing rMLS lipid metabolite values, the analytical sample comprised 871 participants, including 215 participants with incident CVD. Of those, 736 participants (including 136 participants with subsequent CVD incidence) had additional plasma samples and lipidomics profiles from 1 year after recruitment. The study protocols were approved by the Institutional Review Boards at all study locations (PREDIMED) and the Harvard T.H. Chan School of Public Health (PREDIMED casecontrol subproject). All participants gave written informed consent.

The LIPOGAIN-2 study was a 12-week, double-blind, parallel-group randomized trial focusing on overweight individuals. In this manuscript, only the first phase of the trial, consisting of an 8-week overfeeding period, was used.

Participants aged between 20 and 55 years with a BMI ranging from 25 to 32kgm2 were eligible. Exclusion criteria were diabetes (fasting glucose of >7mM on two occasions) or liver disease, pregnancy, lactation, alcohol abuse, claustrophobia, abnormal clinical chemistry test results, use of drugs influencing energy metabolism, use of omega-3 supplements or extreme diets, regular heavy exercise (>3h per week), intolerance to gluten, egg or milk protein and implanted metals. Participants were required to fast overnight for 10 to 12h and avoid physical exercise and alcohol for 48h before measurements were taken.

The trial took place at Uppsala University Hospital in Uppsala, Sweden, from August 2014 to June 2015. Participants were assigned to groups through a computer-generated list, which was prepared by a statistician not involved in the study, and stratified for sex, age and BMI. This study is registered on http://www.clinicaltrials.gov under the identifier NCT02211612 and was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent before inclusion, and the study was approved by the Regional Ethical Review Board in Uppsala (Dnr 2014/186).

In total, 61 participants were randomized to receive muffins made with either refined sunflower oil, which is high in PUFAs (specifically linoleate 18:2n-6), or refined palm oil rich in SFAs (mainly palmitate 16:0) for 8 weeks. Participant body weight was monitored weekly when they visited the clinic to receive their muffins, which were prepared in large batches under controlled conditions in a metabolic kitchen at Uppsala University. These muffins, identical in composition except for the type of fat, were added to the participants regular diets to be eaten at any time of the day. Their number was adjusted weekly by plus or minus one muffin per day based on the rate of weight gain, with the goal being an average weight gain of 3% (equivalent to about 2.90.5 muffins or approximately 40g of oil per day). The muffins comprised 51% fat, 44% carbohydrates and 5% protein by energy percentage. One participant was removed due to missing sphingolipid measurements.

Lipidomics analysis was performed with Metabolons Complex Lipid Panel for the EPIC-Potsdam cohort and the DIVAS trial separately. In brief, the platform generates concentrations of molecular species and nearly complete fatty acid composition per lipid class in plasma. The lipid fraction is extracted with methanol:dichloromethane, concentrated under nitrogen and reconstituted in ammonium acetate dichloromethane:methanol (BUME extraction). Extracts are then infused into the ionization source of a Sciex SelexION-5500 QTRAP mass spectrometer operated in multiple reaction monitoring mode with positive/negative switching. Lipid classes are subsequently separated by differential mobility spectrometry. Using 1,100 multiple reaction monitorings, lipid mass and characteristic fragments are determined with the help of more than 50 isotopically labeled internal standards that are simultaneously introduced with the biological sample. Molecular species are quantified by taking the ratio of the signal intensity of each target compound to that of its assigned internal standard and multiplying by the concentration of internal standard added to the sample77.

The Complex Lipid Panel produced measurements for 14 lipid classes (cholesteryl esters, monoglycerides, ceramides, dihydroceramides, lactosylceramides, hexosylceramides, sphingomyelins, lysophosphatidylethanolamines, lysophosphatidylcholines, diglycerides, triglycerides, phosphatidylcholines, phosphatidylethanolamines and phosphatidylinositol). For phosphatidylethanolamines, species from the two subclasses phosphatidylethanolamine ether and phosphatidylethanolamine plasmalogen were detected. Measured concentrations of molecular species were used to calculate within-class fatty acid sums (summing all concentrations of molecular species containing a specific fatty acid within a lipid class). Within-class fatty acid sums are synonymous with molecular species level in lipid classes containing only one reported variable fatty acid per molecule (one-fatty-acid-containing classes: cholesteryl esters, monoglycerides, ceramides, dihydroceramides, lactosylceramides, hexosylceramides, sphingomeylins, lysophosphatidylethanolamines and lysophosphatidylcholines).

For comparability with the species-level lipidomics in the PREDIMED trial and NHS/NHS2 cohorts (see below), we further calculated the species level for those classes with more than one fatty acid per molecule (that is, diglycerides, triglycerides, phosphatidylcholines, phosphatidylethanolamines and phosphatidylinositol) by summing all species with the same total atomic mass and degree of saturation of the contained fatty acids (that is, isobaric species). We used the updated shorthand notations from the LIPIDMAPS initiative where applicable78. We only refer to the shorthand notations of fatty acids for brevity. According to the manufacturer, the median coefficient of variation of species at a 1M concentration in serum or plasma was approximately 5%. Several lipid species had higher percentages of missing values because they were likely below the lower limit of quantification. Lipid species with more than 70% missing values were excluded, while missing values in the remaining lipid species were imputed using the Quantile Regression Imputation of Left-Censored data approach from the R package imputeLCMD (https://CRAN.R-project.org/package=imputeLCMD).

At the Broad Institute, plasma polar and nonpolar lipids were identified using a Shimadzu Scientific Instrument Nexera x2 U-HPLC system, which was linked to a Thermo Fisher Scientific Exactive Plus Orbitrap mass spectrometer. Lipids were extracted from the plasma (10l) using 190l of isopropanol that had 1,2-didodecanoyl-sn-glycero-3-phosphocholine as an internal standard, supplied by Avanti Polar Lipids. After centrifugation (10min, 9,000g, room temperature), the supernatants (2l) were directly injected onto a 1002.1mm ACQUITY BEH C8 column (1.7m) from Waters. The column was flushed isocratically at a flow rate of 450lmin1 for 1min at 80% of mobile phase A (95:5:0.1 (vol/vol/vol) of 10mmoll1 ammonium acetate:methanol:acetic acid), succeeded by a linear gradient to 80% of mobile phase B (99.9:0.1 (vol/vol) methanol:acetic acid) for 2min and a linear gradient to 100% mobile phase B over 7min and finally maintained at 100% mobile phase B for 3min.

Mass spectrometry analyses were performed in positive ion mode using electrospray ionization and full scan analysis over m/z 2001,100 at a resolution of 70,000 and a data acquisition rate of 3Hz. The following other mass spectrometry parameters were used: ion spray voltage at 3.0kV, capillary and probe heater temperature at 300C, sheath gas at 50, auxiliary gas at 15 and S-lens RF level at 60. Progenesis QI software (NonLinear Dynamics) was used to process raw data for feature alignment, nontargeted signal detection and signal integration. Targeted processing of a subset of lipids was conducted using TraceFinder software (version 3.2; Thermo Fisher Scientific). Lipids were characterized by their headgroup, overall acyl carbon content and total acyl double bond content79. The Broad Institute metabolomics data in NHS/NHSII were measured in several casecontrol studies. Within each casecontrol study, lipid species with more than 70% missing values were excluded, whereas missing values in remaining lipid metabolites were imputed with half the minimal measured value. Due to the platform evolution in the NHS/NHSII cohorts, some metabolite levels were not measured in specific casecontrol studies. For calculation of the rMLS, nonmeasured values of specific metabolites in specific casecontrol studies were substituted with the median of all measured values across the whole dataset (only applicable to the rMLS diet substitution models in the NHS/NHSII cohorts).

Sphingolipids from serum were extracted using butanolmethanol methods80,81. Sphingolipids were detected and quantified using ultraperformance liquid chromatography/tandem mass spectrometry, as previously described82.

All lipidomics variables in all study samples were log transformed.

We assessed the difference in postintervention within-class fatty acid sum concentrations between the SFA-rich and UFA-rich diets via linear regression models with trial arm coded as an indicator variable (SFA-rich diet as a reference) and adjusted for respective baseline concentrations in addition to age, BMI and sex. All lipids that were statistically significantly different between the diets after controlling for an FDR83 at 5% were used for calculating the MLS (Supplementary Tables 10 and 11). Using the estimated intervention effects as weights, we calculated the MLS in the DIVAS trial and, again, used linear regression to estimate baseline-adjusted differences in MLS between the diets. For the analyses of sphingolipids, sphingolipid score and apolipoprotein B in the LIPOGAIN-2 trial, we used the same approach as in the DIVAS trial. The models were similarly adjusted for age, sex and BMI.

Using the estimated intervention effects as weights, we calculated the MLS in the EPIC-Potsdam cohort and divided the score by the observed diet effect on the MLS in the DIVAS trial so that one unit increase in the MLS corresponds to the magnitude of the DIVAS diet intervention effect. Like the above approach, we estimated the diet effect on other risk biomarkers (HbA1c, fasting glucose, total triglycerides, HDL-C, non-HDL-C and hsCRP) and applied the respective observed effect as a scale for the hypothetical DIVAS intervention effect in the EPIC-Potsdam cohort.

We assessed the association between MLS and incident CVD and T2D with Cox proportional hazards models. The casecohort design was accounted for by assigning weights as proposed by Prentice. Age was the underlying time variable, with entry time as age at baseline and exit time as age at event or censoring. The fully adjusted model included age (years), sex, waist circumference (cm), height (cm), leisure-time physical activity (average h per week), highest achieved education level (three categories: primary school, secondary school/high school and college/higher education degree), fasting status at blood draw (three categories: overnight fast, only drink and unfasted), total energy intake (gday1), blood pressure (systolic and diastolic; mmHg), smoking status (four categories: never, former, current smoker (<20Uday1) and current smoker (20Uday1)), alcohol intake (six sex-specific categories: none, low, moderately low, moderately high, high and very high), antihypertensive medication (yes/no), lipid-lowering medication (yes/no) and acetylsalicylic acid medication (yes/no) as covariates. Models for CVD were additionally adjusted for prevalent T2D. To check if the presentation of stratified results was warranted, we tested the potential for effect measure modification by sex by including MLSsex interaction terms into the multivariable-adjusted model.

The rMLS was constructed with the same weights as were used in the EPIC-Potsdam cohort; however, those lipids that were not available in the Broad Institute lipidomics data in the NHS/NHSII cohorts and PREDIMED trial were either skipped or, where possible, imputed using regression weights from the EPIC-Potsdam cohort. In detail, the Broad Institute lipidomics datasets available in the NHS/NHSII cohorts and the PREDIMED trial offer species-level lipidomics in those lipid classes that contain more than one fatty acid residue per molecule, whereas the platform used in the EPIC-Potsdam cohort and the DIVAS trial generated resolution down to the molecular species level, indicating all fatty acid residues per molecule (with the exception of triglycerides). We calculated species levels in the EPIC-Potsdam cohort and used these to predict within-class fatty acid sums. These lipid species-specific weights were then applied to generate a predicted value of the missing lipid variable in the PREDIMED trial and the NHS/NHSII cohorts, where possible.

Diet and lipidomics profiles were available from 10,894 women in the NHS (n=7,479) and NHSII (n=3,415) cohorts. For macronutrient substitution modeling, we used the average of the macronutrient intakes derived from the two FFQs closest to the blood collection that was used in the dietary substitution analyses (NHS cohort: 1986 and 1990; NHSII cohort: 1995 and 1999). We then included all dietary macronutrient variables (as percent total energy) except for saturated fat in a linear model with the variance standardized MLS as outcome, adjusting for total energy intake excluding alcohol (kcalday1), alcohol intake (gday1), BMI (kgm2), age (years) and diet quality (AHEI without alcohol points). Macronutrient intake was scaled to 8% of total energy. Therefore, effect estimates from this linear model can be interpreted as the association of substituting 8% of total energy from SFAs with 8% of total energy from other macronutrients. Conditional logistic regression models were used to assess the associations of the rMLS with the risk of developing stroke and T2D.

We further assessed the correlation of the rMLS with established diet quality indices, including LCDs84, the aMed85 and the AHEI86. For the general LCD, participants were divided into 11 strata based on percentage of energy from each total fat, protein and carbohydrates. Points were assigned descending from 10 for the highest stratum in fat and protein to 0 for the lowest. For carbohydrates, scoring was reversed, with the lowest intake receiving 10 points and the highest receiving 0. We applied the same methodology to compute two additional LCD scores: one animal based and one vegetable based. The animal-based LCD score was based on the percentage of energy derived from carbohydrates, animal protein and animal fat. Conversely, the vegetable-based LCD score was calculated from the energy percentages from carbohydrates, vegetable protein and vegetable fat84.

The aMed score, adapted from Trichopoulou et al.87, includes vegetables (excluding potatoes), fruits, nuts, whole grains, legumes, fish and the ratio of monounsaturated to saturated fats along with red and processed meats and alcohol. Participants scoring above the median intake in these categories received 1 point, except for red and processed meats where scoring below the median earned a point; all others received 0. Alcohol intake scoring awarded 1 point for daily consumption between 5 and 15g. The aMed score ranges from 0 to 9, with higher scores indicating greater adherence to the Mediterranean diet85.

The AHEI was developed after a comprehensive literature review and consultations with nutrition researchers to identify dietary factors consistently linked with a reduced risk of chronic diseases in clinical and epidemiological research. Beneficial AHEI components include vegetables, fruits, whole grains, nuts, legumes, long-chain omega-3 PUFAs and total PUFAs, whereas adverse components comprise sugar-sweetened beverages, red and processed meats, trans-fats and sodium. Moderate alcohol consumption scores highest, with high consumption scoring lowest. Each AHEI component is rated from 0 (worst) to 10 (best), resulting in a total score ranging from 0 (no adherence) to 110 (perfect adherence)86.

Risk associations with stroke and T2D in the nested 1:1-matched casecontrol studies were assessed with conditional logistic regressions adjusted for age, BMI, alcohol intake, diet quality and smoking. Analyses on 10-year change in MLS were further adjusted for status after 10 years of these variables (except age).

We used Prentice-weighted Cox proportional hazards regression to assess the association between the rMLS and the risk of incident disease endpoints in PREDIMED. The interaction analyses were performed in the subsamples with two lipidomics profiles (preintervention and 1-year into the intervention). The interaction model contained a three-way interaction term between Mediterranean diet intervention and the repeated rMLS measurements (preintervention rMLSMediterranean diet intervention1-year intervention rMLS) along with the main effect terms and were adjusted for age and sex. The results of the interaction analyses informed the subsequent stratified analyses according to the Mediterranean diet intervention. The Cox models in the intervention strata were adjusted for age, sex and preintervention BMI.

We estimated a network model of conditional dependencies, where edges represent covariance between two lipids that could not be explained by adjustment for any subset of other lipids. To this end, we applied an order-independent implementation of the causal structure learning PC algorithm88. The resulting network graphically encoded the family of causal models that could have generated the observed conditional independence structure, that is, the skeleton of the data-generating directed acyclic graph. Within this network, we identified clusters of lipids using the Louvain modularity detection algorithm. The Louvain method is a fast heuristic algorithm for detecting communities in large networks by optimizing modularity. It iteratively merges nodes into communities to maximize within-community links compared to between-community links38.

We then calculated cluster-specific lipid scores using the same weights as for the full MLS and associated the resulting scores with risk of cardiometabolic diseases in the same way as the full MLS. We furthermore applied the NetCoupler algorithm (netcoupler.github.io/NetCoupler/) to identify those lipiddisease connections that could not be attributed to the influence of related MLS lipids. The algorithm uses the conditional independence network to detect links between individual lipids and disease incidence that could not be explained by confounding influences through other lipids. By definition, at least one subset of direct neighbors is sufficient to block confounding from the whole network. However, sufficient adjustment sets cannot be unambiguously read from the graph because the edges are not directed. Therefore, the NetCoupler algorithm iterates for each lipid through adjustment for all possible combinations of direct network neighbors. A lipid is then only classified as a direct effector if the association with disease incidence is robust across all these submodels39,40.

All analyses were performed using R (version 4.3.0). Further information on used R packages is reported in Supplementary Table 13.

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

More here:
Lipidome changes due to improved dietary fat quality inform cardiometabolic risk reduction and precision nutrition - Nature.com

Your Full Name
Your Email
Your Phone Number
Select your age (30+ only)	30313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120
Select Your US State	Select Your StateAlabamaAlaskaArizonaArkansasCaliforniaColoradoConnecticutDelawareFloridaGeorgiaHawaiiIdahoIllinoisIndianaIowaKansasKentuckyLouisianaMaineMarylandMassachusettsMichiganMinnesotaMississippiMissouriMontanaNebraskaNevadaNew HampshireNew JerseyNew MexicoNew YorkNorth CarolinaNorth DakotaOhioOklahomaOregonPennsylvaniaRhode IslandSouth CarolinaSouth DakotaTennesseeTexasUtahVermontVirginiaWashingtonWashington,DCWest VirginiaWisconsinWyoming
Program Choice	Select Your ProgramInjectable HGHInjectable TestoteroneInjectable SermorelinInjectable Weight LossInjectable HCG DietInjectable Vitamins
Confirm over 30 years old	Yes
Confirm that you resident in USA	Yes
This is a Serious Inquiry	Yes
Message:

Home » Uncategorized » Lipidome changes due to improved dietary fat quality inform cardiometabolic risk reduction and precision nutrition – Nature.com