Effect of an Internet-based, personalized nutrition randomized trial on dietary changes associated with the Mediterranean diet: the Food4Me Study

At baseline, high MD scorers had a lower BMI than low scorers (0.5 kg/m2 lower, P=0.007).

“At baseline, high MD scorers had 0.5 kg/m 2 lower BMI (P=0.007) and 0.03 higher PAL (P=0.003)than low scorers.”

• Also stated in: § Socio-demographic, anthropometric and health-related characteristics by MD score · plain

  “High MD scorers weighed 2.3 kg less (P=0.003), had 0.5 kg/m 2 lower BMI (P=0.007) and 1.9 cm lower WC (P<0.001) than low scorers (Table 1).”

At month 6, participants in L3 (diet, phenotype, and genotype advice) had higher MD scores than those in L2 (diet and phenotype advice) (5.63 ± 0.10 vs 5.38 ± 0.10, P=0.029).

“However, differences in MD scores at month 6 were greater in L3 vs. L2 (L3: 5.63 ± 0.10 vs. L2: 5.38 ± 0.10 respectively; P=0.029).”

• Also stated in: § Differences in MD scores following intervention · plain

  “However, MD scores at month 6 for participants receiving PN advice in L3 (diet + phenotype + genotype) were greater than in participants in L2 at month 6 (L3: 5.63 ± 0.10 vs. L2: 5.38 ± 0.10, respectively, P=0.029; Table 3).”

At month 6, MD scores did not differ significantly between participants given PN advice based on current diet alone (L1) and those given advice based on diet plus phenotype (L2) or diet plus phenotype plus genotype (L3).

“MD scores at month 6 in participants receiving PN advice based on current diet alone (L1) were not significantly different from those randomized to L2 and L3 (who received advice based on current diet + phenotype (L2) and diet + phenotype + genotype (L3); Table 3).”

At month 3, MD scores were lower in L2 than in L3 (P=0.010).

“MD scores at month 3 between interventions arms were lower in those randomized to L2 compared with L3 (P=0.010;Supplemental Table 3).”

In the Netherlands only, L3 participants had higher MD scores than L2 participants (P=0.013).

“For the Netherlands only, MD scores was higher for L3 participants than for L2 participants (P=0.013;Supplemental Table 4).”

For the FTO and MTHFR genes, the month 6 MD score was higher in PN than Control only among risk carriers.

“For FTO and MTHFR genes, MD score at month 6 was higher in individuals randomized to PN compared with the Control in riskcarriers only.”

When countries were grouped into Mediterranean (Greece and Spain) and non-Mediterranean, the PN (mean L1, L2, L3) versus Control effect on month 6 MD scores was significant only in non-Mediterranean countries (PN: 5.31 ± 0.09 vs Control: 5.02 ± 0.06; P=0.007).

“When Mediterranean (Greece and Spain) and non-Mediterranean countries (the UK, Ireland, the Netherlands, Germany and Poland) were grouped, the effect of PN (mean L1, L2 and L3) vs.Control on MD scores at month 6 was significant in non-Mediterranean countries only (PN: 5.31 ± 0.09 vs. Control: 5.02 ± 0.06; P=0.007; data not shown).”

At month 6, participants assigned to personalized nutrition (PN; L1, L2, L3 combined) had higher MD scores than those in Control (L0) (P=0.002).

“MD scores at month 6 were greater in individuals randomized to PN (L1, L2 and L3) compared with Control (PN: 5.20 ± 0.05 vs. Control: 5.48 ± 0.07 respectively; P=0.002).”

• Also stated in: § Differences in MD scores following intervention · plain

  “After 6 months intervention, improvements in MD scores were greater in individuals randomized to PN (mean L1, L2 and L3) compared with Control (L0) (PN: 5.20 ± 0.05 vs. Control: 5.48 ± 0.07, respectively, P=0.002; Table 3).”

At baseline, high MD scorers had a lower waist circumference than low scorers (1.9 cm lower, P<0.001).

“High MD scorers weighed 2.3 kg less (P=0.003), had 0.5 kg/m 2 lower BMI (P=0.007) and 1.9 cm lower WC (P<0.001) than low scorers (Table 1).”

At baseline, a greater proportion of low MD scorers than high scorers wanted to lose weight (49% vs 45%, P=0.041).

“More low MD scorers wanted to lose weight than high scorers (49 vs. 45%; P=0.041; Table 1), whereas more high scorers reported being on a restricted diet (9 vs. 6%; P=0.014; Table 1).”

At baseline, percentage energy from monounsaturated fatty acids was higher in high MD scorers than in low scorers (P=0.009).

“As expected, high MD scorers had lower percentage energy intakes from total fat (P<0.001) and SFA (P<0.001) and higher percentage energy intakes from MUFA (P=0.009) and PUFA (P<0.001) than low scorers (Table 2).”

At baseline, a greater proportion of high MD scorers than low scorers reported following a restricted diet (9% vs 6%, P=0.014).

“More low MD scorers wanted to lose weight than high scorers (49 vs. 45%; P=0.041; Table 1), whereas more high scorers reported being on a restricted diet (9 vs. 6%; P=0.014; Table 1).”

At baseline, percentage energy from carbohydrates was 0.8% lower in high MD scorers than in low scorers (P=0.042).

“Percentage energy intakes from protein and sugars were 1.2 and 1.7% higher in high MD scorers than low scorers (P<0.001),whereas percentage energy intakes from carbohydrates were 0.8% lower (P=0.042).”

At baseline, a greater proportion of high MD scorers than low scorers met fruit and vegetable recommendations (41% more, P<0.001).

“More high MD scorers met the recommendations for oily fish (36% more; P<0.001), red meat (7%; P=0.006) and fruit and vegetables (41%; P<0.001) than low scorers (Table 2).”

At baseline, the energy intake to basal metabolic rate ratio was higher in high MD scorers than in low scorers (1.72 ± 0.70 vs 1.62 ± 0.63, P=0.012).

“Although energy intakes did not differ, EI: BMR ratio was higher in high MD scorers than low MD scorers (1.72 ± 0.70 vs. 1.62 ± 0.63); P=0.012; Table 2).”

At baseline, high MD scorers had higher moderate-to-vigorous physical activity (MVPA) than low scorers (P<0.001).

“High MD scorers spent less time in sedentary behaviors (P=0.005), had higher PAL (P=0.003) and MVPA (P<0.001) and met more PA recommendations (P=0.022)than low scorers (Table 1).”

At baseline, high MD scorers weighed less than low scorers (2.3 kg less, P=0.003).

“High MD scorers weighed 2.3 kg less (P=0.003), had 0.5 kg/m 2 lower BMI (P=0.007) and 1.9 cm lower WC (P<0.001) than low scorers (Table 1).”

At baseline, percentage energy from polyunsaturated fatty acids was higher in high MD scorers than in low scorers (P<0.001).

“As expected, high MD scorers had lower percentage energy intakes from total fat (P<0.001) and SFA (P<0.001) and higher percentage energy intakes from MUFA (P=0.009) and PUFA (P<0.001) than low scorers (Table 2).”

At baseline, sedentary time was lower in high MD scorers than in low scorers (P=0.005).

“High MD scorers spent less time in sedentary behaviors (P=0.005), had higher PAL (P=0.003) and MVPA (P<0.001) and met more PA recommendations (P=0.022)than low scorers (Table 1).”

At baseline, percentage energy from sugars was 1.7% higher in high MD scorers than in low scorers (P<0.001).

“Percentage energy intakes from protein and sugars were 1.2 and 1.7% higher in high MD scorers than low scorers (P<0.001),whereas percentage energy intakes from carbohydrates were 0.8% lower (P=0.042).”

At baseline, high MD scorers met more physical activity recommendations than low scorers (P=0.022).

“High MD scorers spent less time in sedentary behaviors (P=0.005), had higher PAL (P=0.003) and MVPA (P<0.001) and met more PA recommendations (P=0.022)than low scorers (Table 1).”

At baseline, percentage energy from total fat was lower in high MD scorers than in low scorers (P<0.001).

“As expected, high MD scorers had lower percentage energy intakes from total fat (P<0.001) and SFA (P<0.001) and higher percentage energy intakes from MUFA (P=0.009) and PUFA (P<0.001) than low scorers (Table 2).”

At month 6, MD scores did not differ significantly between Control and PN (mean L1, L2, L3) when stratified by country.

“MD scores at month 6 when stratified by country were not significantly different for Control vs. PN (mean L1, L2 and L3).”

At baseline, percentage energy from saturated fatty acids was lower in high MD scorers than in low scorers (P<0.001).

“As expected, high MD scorers had lower percentage energy intakes from total fat (P<0.001) and SFA (P<0.001) and higher percentage energy intakes from MUFA (P=0.009) and PUFA (P<0.001) than low scorers (Table 2).”

At baseline, percentage energy from protein was 1.2% higher in high MD scorers than in low scorers (P<0.001).

“Percentage energy intakes from protein and sugars were 1.2 and 1.7% higher in high MD scorers than low scorers (P<0.001),whereas percentage energy intakes from carbohydrates were 0.8% lower (P=0.042).”

At baseline, a greater proportion of high MD scorers than low scorers met oily fish recommendations (36% more, P<0.001).

“More high MD scorers met the recommendations for oily fish (36% more; P<0.001), red meat (7%; P=0.006) and fruit and vegetables (41%; P<0.001) than low scorers (Table 2).”

At baseline, a greater proportion of high MD scorers than low scorers met red meat recommendations (7% more, P=0.006).

“More high MD scorers met the recommendations for oily fish (36% more; P<0.001), red meat (7%; P=0.006) and fruit and vegetables (41%; P<0.001) than low scorers (Table 2).”

Among participants with two copies of a risk allele for any of the 5 genes, month 6 MD scores were higher in PN (mean L1, L2, L3) than Control (5.69 ± 0.11 vs 5.14 ± 0.08; P<0.001).

“Stratifying analyses by carriage of a risk allele for any one of the 5 genes studied showed that in participants with two copies of a risk allele of any of the 5 genes, MD scores at month 6 were greater between participants randomized to PN (mean L1, L2 and L3) than those randomized to Control (5.69 ± 0.11 vs. 5.14 ± 0.08; P<0.001; data not shown).”

In Food4Me, although not statistically significant, the low MD score group had a higher percentage of routine and manual occupations than the high score group.

“The PREDIMED study found that low-economic status was associated with low-MD adherence and, although not statistically significant in the Food4Me study, we observed higher percentages of individuals in routine and manual occupations in the low MD score group compared with the high score group.”

In Food4Me, higher MD scores were associated with greater sugar intake.

“In Food4Me, higher MD score was associated with higher intakes of sugar, although this may be due to higher fruit juice intake.”

Among TCF7L2 risk carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.016).

“Risk (n=725) | 5.19 ± 0.07 | 5.49 ± 0.10 | 5.38 ± 0.14 | 5.49 ± 0.14 | 5.61 ± 0.13 | 0.016 | 0.23 | 0.45”

Among FADS1 non-risk carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.019).

“Non-risk (n=839) | 5.24 ± 0.07 | 5.62 ± 0.10 | 5.59 ± 0.13 | 5.54 ± 0.13 | 5.73 ± 0.13 | 0.019 | 0.75 | 0.21”

Among ApoE non-risk carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.028).

“Non-risk (n=1078) | 5.15 ± 0.06 | 5.38 ± 0.08 | 5.38 ± 0.11 | 5.34 ± 0.11 | 5.43 ± 0.11 | 0.028 | 0.94 | 0.48”

At month 6, the L1 arm did not differ significantly from the combined L2+L3 arms in Mediterranean diet score (P=0.46).

“MD score at month 6 | 5.20 ± 0.05 | 5.48 ± 0.07 | 5.43 ± 0.10 | 5.38 ± 0.10 | 5.63 ± 0.10 | 0.002 | 0.46 | 0.029”

Among ApoE risk carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.040).

“Risk (n=386) | 5.33 ± 0.10 | 5.72 ± 0.15 | 5.56 ± 0.19 | 5.47 ± 0.20 | 6.13 ± 0.20 | 0.040 | 0.24 | 0.006”

Among FTO risk allele carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (5.57 vs 5.21, P=0.002).

“Risk (n=1002) | 5.21 ± 0.06 | 5.57 ± 0.09 | 5.46 ± 0.12 | 5.48 ± 0.12 | 5.74 ± 0.12 | 0.002 | 0.21 | 0.06”

Among ApoE risk carriers, L3 had a higher month 6 Mediterranean diet score than L2 (P=0.006).

“Risk (n=386) | 5.33 ± 0.10 | 5.72 ± 0.15 | 5.56 ± 0.19 | 5.47 ± 0.20 | 6.13 ± 0.20 | 0.040 | 0.24 | 0.006”

At month 6, the fruit component score was higher in the combined personalized nutrition arms than in control (0.67 vs 0.58, P=0.001).

“Fruit | 0.58 ± 0.01 | 0.67 ± 0.02 | 0.67 ± 0.03 | 0.66 ± 0.02 | 0.69 ± 0.03 | 0.001 | 0.99 | 0.33”

At month 6, the L3 arm (advice personalized using diet, phenotype, and genotype) had a higher mean Mediterranean diet score than the L2 arm (advice personalized using diet and phenotype), 5.63 vs 5.38, P=0.029.

“MD score at month 6 | 5.20 ± 0.05 | 5.48 ± 0.07 | 5.43 ± 0.10 | 5.38 ± 0.10 | 5.63 ± 0.10 | 0.002 | 0.46 | 0.029”

Among TCF7L2 non-risk carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.036).

“Non-risk (n=742) | 5.20 ± 0.07 | 5.49 ± 0.10 | 5.52 ± 0.14 | 5.32 ± 0.14 | 5.64 ± 0.14 | 0.036 | 0.75 | 0.044”

Among MTHFR risk allele carriers, the combined personalized nutrition arms had a higher month 6 Mediterranean diet score than control (P=0.006).

“Risk (n=809) | 5.59 ± 0.10 | 5.59 ± 0.10 | 5.52 ± 0.13 | 5.50 ± 0.13 | 5.75 ± 0.13 | 0.006 | 0.47 | 0.10”

At month 6, the tomato sauce component score was higher in L3 than L2 (0.030 vs 0.014, P=0.040).

“Tomato sauce | 0.011 ± 0.003 | 0.020 ± 0.005 | 0.017 ± 0.007 | 0.014 ± 0.007 | 0.030 ± 0.007 | 0.15 | 0.51 | 0.040”

Among TCF7L2 non-risk carriers, L3 had a higher month 6 Mediterranean diet score than L2 (P=0.044).

“Non-risk (n=742) | 5.20 ± 0.07 | 5.49 ± 0.10 | 5.52 ± 0.14 | 5.32 ± 0.14 | 5.64 ± 0.14 | 0.036 | 0.75 | 0.044”

In the Food4Me trial, at month 6, the mean Mediterranean diet score was higher in the combined personalized nutrition arms (L1-L3) than in the control arm (L0), with adjusted means 5.48 vs 5.20 and P=0.002.

“MD score at month 6 | 5.20 ± 0.05 | 5.48 ± 0.07 | 5.43 ± 0.10 | 5.38 ± 0.10 | 5.63 ± 0.10 | 0.002 | 0.46 | 0.029”

At month 6, the olive oil intake component score was lower in the combined personalized nutrition arms than in control (0.002 vs 0.012, P=0.039).

“Olive oil intake | 0.012 ± 0.003 | 0.002 ± 0.004 | 0.002 ± 0.005 | 0.005 ± 0.005 | 0.001 ± 0.005 | 0.039 | 0.99 | 0.31”

These findings confirm that dietary intake changes linked to PN advice also produce significant improvements in dietary patterns measured by the 14-point PREDIMED MD score.

“The present findings confirm that changes in dietary intakes associated with PN advice also result in significant improvements in dietary patterns, as estimated from the 14-point PREDIMED MD score.”

The 0.5-unit PREDIMED score advantage for PN in this study suggests the potential health benefit may be modest.

“The 0.5 unit advantage in PREDIMED score (14 point scale) for PN in the present study indicates that the potential health benefit may be relatively modest.”

PN improved MD scores more than generalized dietary advice (Control).

“We observed that PN was more effective than generalized dietary advice (Control) in improving MD scores.”

• Also stated in: § Conclusions · plain

  “Following a 6-month RCT, MD score were greater in individuals who received PN advice, compared with those who received non-personalized advice.”

MD score improvements were greater with PN based on diet, phenotype, and genotype than with advice based on diet and phenotype alone.

“Furthermore, improvements in MD score were greater in individuals who received PN based on diet, phenotype and genotype compared with advice based on diet and phenotype alone.”

We showed robustness by finding the same result pattern with three recommended analytical approaches for RCTs with missing outcome data (LMM, LOCR, and CC).

“Furthermore, we confirmed the robustness of our findings by showing the same pattern of results when using three recommended analytical approaches for RCTs with missing outcome data (LMM, LOCR and CC analyses).”

PN is more effective than generalized dietary advice for improving MD score.

“PN is a more effective approach for improving MD score than generalized dietary advice.”

MD was estimated using the PREDIMED 14-point score, a validated and widely used MD score.

“Our estimation of MD was based on the PREDIMED 14-point score, which is a validated and widely-used MD score.”

• cites [11] Estruch, 2013, Primary prevention of cardiovascular disease with a Mediterranean diet, N Engl J Med sourceNot checked yet

  Not checked in the free run.

• cites [16] Martínez-González, 2012, Cohort profile: design and methods of the PREDIMED study, Int J Epidemiol sourceNot checked yet

  Not checked in the free run.

A study limitation is that internet-based self-measured and self-reported data may have introduced measurement error.

“A limitation of our study is that data were self-measured and self-reported via the internet, which may have introduced measurement error.”

Intervention studies increasingly show that randomization to the MD reduced CVD risk in primary and secondary prevention.

“There is also accumulating evidence from intervention studies that randomization to the MD reduced CVD risk in both primary and secondary prevention studies (9,12).”

• cites [9] Eguaras, 2015, Does the Mediterranean diet counteract the adverse effects of abdominal adiposity?, Nutr Metab Cardiovasc Dis sourceNot checked yet

  Not checked in the free run.

• cites [12] de Lorgeril, 1999, Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study, Circulation sourceNot checked yet

  Not checked in the free run.

PN advice intended to improve dietary intake produced dietary behavior changes consistent with the MD.

“The main findings from our secondary analysis in the Food4Me PoP study show that PN advice aiming to improve dietary intakes brought about changes in dietary behaviors that were in line with the MD.”

The small sample size limited power to examine individual gene effects in this study.

“Small sample size limited our power to investigate the effect of individual genes in the present study.”

Adding genotypic information to PN advice improved MD scores relative to PN advice based on diet and phenotype alone.

“Furthermore, the addition of genotypic information to PN advice improved MD scores compared with PN advice based on diet and phenotype alone.”

• Also stated in: § Conclusions · plain

  “Furthermore, improvements in MD score were greater in individuals who received PN based on diet, phenotype and genotype compared with advice based on diet and phenotype alone.”

This was the largest study of PN in European adults to date to estimate MD score changes.

“We estimated changes in MD score in the largest study of PN in European adults to date.”

Adjustment for under- and over-reporters did not change the pattern of results.

“Adjustment for under-and over-reporters did not change the pattern of results (data not shown).”

This study suggests that giving people more detailed, tailored recommendations based on diet, phenotype, and genotype is advantageous.

“This study suggests that providing individuals with more detailed, tailored recommendations based on a combination of their diet, phenotype, and genotype is advantageous.”

Earlier PN interventions improved sodium intake in people at higher genotypic risk.

“Previous PN interventions have achieved improvements in sodium intake in individuals at higher genotypic-based risk (20), however, the Food4Me PoP study was the first to examine the effect of including genotype-based PN on overall patterns of healthy eating.”

• cites [20] Nielsen, 2014, Disclosure of genetic information and change in dietary intake: a randomized controlled trial, PLoS One sourceNot checked yet

  Not checked in the free run.

After a 6-month RCT, MD scores were higher in people given PN advice than in those given non-personalized advice.

“Following a 6-month RCT, MD score were greater in individuals who received PN advice, compared with those who received non-personalized advice.”

Unlike the main PN intervention analysis, this secondary analysis of MD score differences between arms suggests genotype-based advice adds benefit beyond diet-and-phenotype PN advice.

“In contrast to the main analysis of the PN intervention, our secondary analysis of difference in MD scores between treatment arms suggest that the provision of genotype-based advice offers added benefit compared to PN advice based on diet and phenotype only.”

The FFQ was developed and validated for Food4Me.

“This FFQ, which was developed and validated for the Food4Me Study (30,31), included 157 food items consumed frequently in each of the 7 recruitment countries.”

• cites [30] Forster, 2014, Online dietary intake estimation: the Food4Me food frequency questionnaire, J Med Internet Res sourceNot checked yet

  Not checked in the free run.

• cites [31] Fallaize, 2014, Online dietary intake estimation: reproducibility and validity of the Food4Me Food frequency questionnaire against a 4-day weighed food record, J Med Internet Res sourceNot checked yet

  Not checked in the free run.

Our study confirmed the accuracy of internet-based self-reported anthropometric measurements.

“Nonetheless, the accuracy of internetbased, self-reported anthropometric have been confirmed in our study (37).”

• cites [37] Celis-Morales, 2015, How reliable is internet-based self-reported identity, socio-demographic and obesity measures in European adults?, Genes Nutr sourceNot checked yet

  Not checked in the free run.

RCTs show MD-based interventions lower cardiovascular disease risk in primary and secondary prevention.

“In addition, randomized controlled trials (RCTs) show that MD-based interventions reduce risk of cardiovascular disease in both primary and secondary prevention studies (11,12).”

• cites [11] Estruch, 2013, Primary prevention of cardiovascular disease with a Mediterranean diet, N Engl J Med sourceNot checked yet

  Not checked in the free run.

• cites [12] de Lorgeril, 1999, Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction: final report of the Lyon Diet Heart Study, Circulation sourceNot checked yet

  Not checked in the free run.