Association between changes in adherence to the planetary health diet and adiposity and body fat distribution: a cohort study

Summary

Background

The longitudinal relationship between planetary heath diets and adiposity remains understudied in Europeans. We investigated the association between changes in adherence to the EAT-Lancet planetary health diet (PHD) and changes in adiposity and body fat distribution.

Methods

We evaluated 6845 adults (mean age 49.4 years, 51% women), with repeat assessments in the Fenland Study at baseline (2005–2015) and follow-up (2014–2020). Diet was assessed with food frequency questionnaires and scaled to the PHD Index (PHDI) as a measure of adherence to the PHD (range: 0–140 points). Anthropometry, total fat, visceral, subcutaneous fat and lean mass, and liver fat score was measured at both visits. Multivariable-adjusted linear regression, adjusted for sociodemographic, lifestyle and clinical variables, assessed the association between changes in the PHDI and concurrent changes in adiposity.

Findings

The mean (standard deviation) change in PHDI was 2.1 (±9.1) points. An increase in the PHDI (per 10 points, equivalent to ∼1 standard deviation), was associated negatively with changes in adiposity indices: total body weight (adjusted beta-coefficient: −0.47 kg; 95% confidence interval: −0.63, −0.31), waist circumference (−0.55 cm; −0.73, −0.37), fat mass (−0.43 kg; −0.56, −0.30), subcutaneous fat (−24.5 g; −34.0, 15.0), visceral fat (−40.0 g; −51.5, −28.5), liver fat score (−0.12; −0.15, −0.09), and lean mass (−0.08 kg; −0.13, −0.02). After further adjustment for changes in total body weight, inverse associations persisted for waist circumference, visceral fat, and liver fat. We found stronger associations among those with a BMI ≥30 kg/m² for body weight, fat mass, lean mass, subcutaneous fat, visceral fat, and liver fat score (p for interaction<0.05).

Interpretation

Increasing adherence to the PHD in midlife may reduce or slow gaining adiposity and improve body fat distribution, with greater magnitude in those already obese. These findings suggest a beneficial impact of adherence to the PHD on adiposity in Europeans.

Funding

Funded by the UK Medical Research Council, NIHR Cambridge Biomedical Research Centre and the Independent Research Fund Denmark (1057-00016B) and Danish Diabetes Association.

Research in context.

Evidence before this study

We searched Pubmed including all studies published from 2019, when the EAT-Lancet Commission report was released, to 18 September 2025, for observational studies investigating the association between adherence to the EAT-Lancet Planetary Health Diet (PHD) and measures of adiposity using the search terms: “(planetary health diet OR planetary diet OR healthy reference diet OR EAT Lancet∗ OR EAT-Lancet∗) AND (Index OR score OR adherence), also registered in PROSPERO (CRD42021286597). Reference lists of included studies were scanned. Most identified studies were cross-sectional and reported a lower body weight, body mass index, and waist circumference among those with higher adherence to the PHD compared to lower adherence. Prospective cohort studies showed that higher adherence to the PHD, measured at baseline only, was associated with lesser gain of body weight, waist circumference or incidence of obesity. No study investigated changes in adherence to the PHD and change adiposity, particularly more detailed assessments of body fat distribution, including visceral fat and liver fat, or lean mass.

Added value of this study

In this cohort of 6845 UK adults with repeated dietary, lifestyle and adiposity assessments, including anthropometry, total fat, visceral fat, subcutaneous fat and lean mass, and liver fat, increased adherence to the PHD was inversely associated with changes in adiposity, including visceral fat and liver fat, and the associations were stronger in those with obesity. For the first time, the study also found that the inverse associations were independent of body weight, though the magnitude of the association was reduced, with associations persisting for waist circumference, visceral fat and liver fat following adjustment for changes in total body weight.

Implications of all the available evidence

These findings, together with previous evidence, strongly support the need for urgent adoption of a dietary pattern in line with the PHD as part of the strategy for the prevention of obesity in European populations that also aligns with sustainability goals for the Region. The finding of changes in visceral and liver fat independent of weight change warrant follow-up in human intervention studies to explore the underlying weight-loss independent mechanisms of how the PHD reduce central adiposity.

Introduction

The world currently faces a global syndemic of obesity, undernutrition, and climate change.¹ There is especially a concern of a rapid rise in obesity in Europe.² Diet is unquestionably central in promoting this syndemic and it has been suggested that transitioning to a flexitarian dietary pattern such as the EAT-Lancet planetary health diet (PHD), promoting greater consumption of plant-based foods and moderate consumption of animal-based foods, offers a possible solution.³

Evidence to date has been limited, however. Cross-sectional studies have reported a lower body weight, body mass index (BMI), and waist circumference among those with higher adherence to the PHD compared to lower adherence.^4,5 Prospective cohort studies showed that higher adherence to the PHD, measured at baseline only, was associated with lesser gain of body weight and waist circumference.^6,7 However, none of these studies investigated if within-person changes in the dietary adherence over time were associated with changes in adiposity.⁸

Moreover, past evidence is particularly sparse for any link between a dietary pattern such as the PHD and measures of body fat distribution, including visceral fat and liver fat, that are more strongly related to development of diseases such as type 2 diabetes and cardiovascular disease.⁹ Some short-term intervention studies suggest that hypocaloric diets and diet composition both play a role in reducing liver fat.¹⁰ However, there is no evidence from a study on the potential effect of increasing such dietary adherence over years in a non-clinical trial setting. Thus, we conducted the current study with an observational design to test the hypothesis that improving the dietary adherence to the PHD would reduce or slow gains in adiposity and improve body fat distribution. Specifically, we aimed to investigate the associations of changes in adherence to the PHD with concurrent changes in overall and regional adiposity (i.e., a change-on-change analysis) in a general adult population in the United Kingdom (UK). The change-on-change approach has previously been suggested to be optimal for investigating the impact of changes in an exposure and changes in the outcome.¹¹ This method directly investigates within-individual dietary change that would allow closer approximation of an interventional effect than a dietary change inferred from a between-individual difference, while we still assumed that the dietary change preceded the change in adiposity.

Methods

Study population

The Fenland Study recruited 12,435 participants born between 1950 and 1975, via general practices from 3 centres (Cambridge, Ely, and Wisbech) in Cambridgeshire, UK, between 2005 and 2015.¹² As the aim was to investigate adults at risk of developing diabetes, people with known diabetes were excluded. Other exclusion criteria included pregnancy, an inability to walk unaided, psychosis, and terminal illness. All participants attended a clinic visit during which they completed a food frequency questionnaire (FFQ) and a general health and lifestyle questionnaire, and had anthropometric measures taken by trained staff.

Those who had participated in the baseline assessment of the Fenland Study and had consented to be re-contacted were invited to participate in a follow-up assessment, conducted over 2014 to 2020. Participants had to have at least 4 years between visits, and the assessment procedures in the follow-up were identical to those at the baseline. A total of 7831 individuals participated in the follow-up. We excluded participants with prevalent undiagnosed diabetes with measurements of HbA1c ≥ 48 mmol/mol (6.5%), fasting plasma blood glucose ≥7.0 mmol/L or a 2-h glucose ≥11.1 mmol/L during an oral-glucose tolerance test, as the study protocol was to inform the participants’ family doctor and participants may have changed their health behaviours including their diet as a consequence of this new knowledge.¹³ We further excluded participants missing the follow-up assessment, without dietary data at baseline, with implausible energy intakes (≤2092 kJ/day (500 kcal/day) or >14,644 kJ/day (3500 kcal/day) for women; ≤3347 kJ/day (800 kcal/day) and >17,573 kJ/day (4200 kcal/day) for men),¹⁴ or with missing information on dietary data at follow-up and covariates. Therefore, the analytical sample in the current study included 6845 participants (Supplemental Figure S1). For each specific adiposity outcome, those with missing outcome data were additionally excluded.

Assessment of dietary patterns

In the Fenland Study, diet was assessed using a 130-item semi-quantitative FFQ reflecting habitual intake in the previous year.¹⁵ There were 9 possible categories of intake frequencies, ranging from “never or less than once per month” to “6+ per day” for a standard food portion size. The reported frequency of food intakes were converted into quantity of dietary intakes and nutrients as previously described.¹⁵ The validity of the FFQ was previously assessed against four 4-day weighed diet records over a 1-year period and recovery biomarkers from six 24-h urine samples over a 9-month period.^16,17

Based on the estimated dietary intake, we calculated adherence scores to the PHD using the planetary health diet index (PHDI).¹⁴ The PHDI originally included 15 components, but we included 14 components after excluding soybean/soy foods due to their high non-consumption in the Fenland study population. Each of the 14 dietary components was assigned 0 to 10 points based on each participant's reported diet (Supplemental Table S1 for detail). For seven components for which the PHD recommends higher consumption (whole grains, vegetables (not including starchy vegetables), fruit, fish/shellfish, nuts, legumes, and unsaturated oils), a score of zero was assigned for non-consumption and a maximum score of 10 was assigned if consumption was equal to or above the recommended target amount. Intakes greater than 0 g/d but below the recommended target amounts were scored proportionally. For seven dietary components for which moderate intakes were recommended (starchy vegetables, dairy foods, red/processed meat, chicken and other poultry, eggs, saturated fats, and added sugar), reverse scoring was applied: a score of 10 was awarded for intakes below or equal to the target intake limit while intakes above or equal to the maximum intake limit were given 0 points. Intakes between the target and maximum intake were given proportionately from 0 to 10 points. The sum of each of the 14 components provided the score for the PHDI for each individual.

Assessment of measures of adiposity

Measures of adiposity included body weight, body mass index, waist circumference, fat mass, visceral fat, subcutaneous fat, liver fat, and lean mass. Participants were invited to a clinic visit for physical examination and testing at baseline and follow-up.¹² Body weight (TANITA model Bc-418 MA; Tanita, Tokyo, Japan), height (SECA 240; Seca, Birmingham, UK) and waist circumference (D-loop tape; Chasmors Ltd, London, UK) were measured using standardised procedures.¹² Body fat, lean mass, visceral fat mass, and subcutaneous fat mass were quantified (in grams) using dual-energy X-ray absorptiometry (DEXA), acquired on GE Lunar Prodigy, Prodigy Advanced, and iDXA scanners (GE Healthcare, Bedford, UK), and analysed using GE enCORE software with advanced algorithms (Versions 14.10.022 and 16.00.158).¹² Liver fat was assessed using ultrasonography (LOGIQ Book XP ultrasound system; GE Healthcare, Bedford, UK) and images scored according to standardised criteria by two trained operators who were unaware of other study measures.¹⁸ Scoring criteria were increased echotexture of the liver parenchyma, decreased visualisation of the intra-hepatic vasculature, and attenuation of the ultrasound beam, each scored on a 4-point scale, summing to a liver fat score (range: 3–12). The ultrasound assessment of hepatic steatosis has previously been validated against magnetic resonance spectroscopy with a sensitivity of 96% and specificity of 94%.¹⁹

Assessment of covariates

At clinic visits, participants completed health and lifestyle questionnaires on age, sex, smoking status, education, income, occupation, marital status, and medication use including drugs for hypertension, dyslipidaemia and depression. Physical activity was measured objectively using individually calibrated combined heart rate and movement sensing (Actiheart, CamNTech) for 6 days and 6 nights,²⁰ from which an estimate of activity energy expenditure was derived.²¹

Statistical analysis

Changes in the PHDI and adiposity measures were calculated by subtracting each baseline measure from the follow-up measure. Baseline characteristics were described across four categories of the PHDI changes stratified by their quartile values and also compared between those who were included in the current study and those who were not, to check for potential systematic bias.

The association between changes in the PHDI and changes in adiposity was modelled using multivariable linear regression using a change-on-change approach as previously described.¹¹ No violation of linearity using spline regression analysis and homoscedasticity, using Bland–Altman plots, was observed (data not shown). We used the change in the outcome and adjusted for baseline adiposity to take regression-to-the-mean into account and because we were interested in the exogenous change in adiposity related to differences in adherence to the PHD. Model 1 further adjusted for baseline PHDI (score), age (years), sex, test site (Cambridge, Ely, Wisbech) and follow-up time (years); Model 2 further adjusted for education levels (none, compulsory, further, higher), occupation (professional, clerical, senior manager, technical, semi-routine, mid-management, traditional profession), and household income (<20,000, 20,000–40,000, >40,000 UK pounds/year) at baseline and marital status (single, living as married, widowed/separated/divorced, undisclosed), smoking status (never, former, current), physical activity (kJ/kg/day), alcohol intake (g/day), binary variables of medication use (nonsteroidal anti-inflammatory drugs, lipid-lowering drugs, anti-hypertensive drugs, antidepression drugs, and, in women, use of hormone replacement therapy) at baseline and follow-up, and anti-diabetes medication use (yes, no) at follow-up. We also estimated the yearly change in each measure of adiposity as annualised changes and tested the cross-sectional associations between PHDI and measures of adiposity at both baseline and follow-up.

We further tested the association of dietary intake with specific adipose compartments, such as visceral fat. We conducted post hoc adjustment for the changes in total body weight (continuous) in the model treating each of waist circumference, specific fat-mass variables, and liver fat as an independent variable. In the models examining visceral and liver fat changes, we also adjusted for changes in subcutaneous fat and lean body mass and used the residual adjustment method.

We examined if the association between changes in adherence to the PHD might differ depending on participants’ baseline characteristics, as it was shown previously that BMI reduction due to a plant-based diet intervention was more pronounced among those with higher body mass.²² Analyses were stratified for age (<50, ≥50 years at baseline), sex (men, women), body mass index (<30 kg/m², ≥30 kg/m²), physical activity (<53, ≥53 kJ/kg/day) and smoking (ever, never). To test the between-strata heterogeneity, we fitted separate linear models to include an interaction term between the change in PHDI and age, sex, BMI, physical activity, or smoking, respectively.

In sensitivity analyses we removed outliers, defined as the top and bottom 1% of the distribution of the adiposity variables. To minimise potential selection bias due to loss to follow-up or non-response, we used inverse probability weights to weight the regression models to the baseline population. Stabilised weights were calculated using logistic regression model with participation in the follow-up assessment as the binary outcome and available exposure and covariates as described above, which might predict participation. To assess potential bias from missing covariates, multiple imputation with five datasets was conducted with chained equations. The role of individual components was investigated by constructing new PHDIs without each of the individual components, adjusting for Model 2 covariates. In a separate analysis, we adjusted our models for total energy intake. We also used the EAT-Lancet reference diet index (Supplemental Table S2) as an alternative measure of adherence to the PHD as this ordinal scoring system (score range between 0 and 42 points) was different to our proportional scoring. In addition, we also conducted longitudinal change-analysis, which was meant to relate adherence to the PHDI measured at baseline with the change in the measure of adiposity, adjusted for the baseline adiposity measure. The same covariates were included in the regression model as above, using their baseline measures.⁸

R (version 4.1.3) was used for all analyses and a p-value of less than 0.05 was used to discern statistical significance.

Role of the funding source

The funders had no role in study design, data collection, data analysis, interpretation or writing of the report.

Ethics approval

The study was approved by the Cambridge Local Research Ethics Committee (NRES Committee–East of England, Cambridge Central, ref. 04/Q0108/19), and all participants provided written informed consent.

Results

Study population characteristics

The mean change (standard deviation) among those with the highest increase in the PHDI was 13.4 (4.9) points whereas those with greatest reduction reduced the PHDI by 9.2 (4.8) points (Table 1), Supplemental Figure S2. Participants with the greatest increase in adherence to the PHD between the two study phases had a lower adherence at baseline, were more likely to: have higher education, higher income, use more lipid-lowering medication at follow-up and less likely to: use antidepression medication at follow-up, compared to those with the greatest decrease in adherence (Table 1). The mean (standard deviation) time between assessments was 6.6 (1.9) years. Body weight, BMI, waist circumference, fat mass, visceral fat, and liver fat increased in the overall population whereas lean mass decreased over time (Supplemental Table S3). Subcutaneous fat decreased among those with an increase in adherence to the PHD while it increased slightly among those whose adherence decreased (Supplemental Table S3).

Table 1.

Baseline characteristics across changes in adherence to the planetary health diet (PHD) among participants in the Fenland Study.^a

Characteristic	Overall N = 6845	Q1 (−32.2, −3.6) N = 1695	Q2 (−3.5, 2.0) N = 1708	Q3 (2.1, 7.7) N = 1727	Q4 (7.8, 51.9) N = 1715
PHDI (change), points	2.1 (9.0)	−9.2 (4.8)	−0.7 (1.6)	4.8 (1.7)	13.4 (4.9)
PHDI (baseline), points	82.5 (10.5)	87.2 (9.7)	84.6 (10.0)	82.0 (9.4)	76.5 (9.8)
Age, years	49.4 (7.3)	49.4 (7.5)	49.5 (7.2)	49.5 (7.3)	49.2 (7.2)
Women, n (%)	3502 (51%)	877 (52%)	873 (51%)	906 (52%)	846 (49%)
Follow-up time, years	6.6 (1.9)	6.6 (1.9)	6.5 (1.9)	6.5 (1.9)	6.7 (2.0)
Higher education, n (%)	201 (2.9%)	35 (2.1%)	53 (3.1%)	46 (2.7%)	67 (3.9%)
Managerial/professional occupation, n (%)	3485 (51%)	882 (52%)	876 (51%)	875 (51%)	852 (50%)
Income > £40,000/year	3804 (56%)	892 (53%)	953 (56%)	1004 (58%)	955 (56%)
Marital status, single, n (%)	446 (6.5%)	114 (6.7%)	127 (7.4%)	90 (5.2%)	115 (6.7%)
Smoking, current, n (%)	613 (9.0%)	152 (9.0%)	141 (8.3%)	151 (8.7%)	169 (9.9%)
Alcohol intake, g/day	10.6 (13.4)	10.5 (13.3)	10.5 (13.3)	10.7 (13.6)	10.6 (13.5)
Activity energy expenditure, kJ/kg/day	54.8 (21.7)	54.9 (21.8)	54.5 (20.9)	54.6 (21.8)	55.1 (22.3)
Energy intake (change), kcal/day	−81.5 (540)	−89.9 (583)	−93.2 (518)	72.9 (511)	−70.1 (546)
Energy intake (baseline), kcal/day	1950 (576)	1967 (580)	1937 (571)	1948 (573)	1947 (579)
Use of hormone replacement therapy,b yes, n (%)	198 (5.4%)	57 (6.5%)	45 (5.2%)	44 (4.9%)	43 (5.1%)
Use of NSAIDs, n (%)	238 (3.5%)	59 (3.5%)	58 (3.4%)	53 (3.1%)	68 (4.0%)
Use of lipid medication, n (%)	253 (3.7%)	63 (3.7%)	68 (4.0%)	60 (3.5%)	62 (3.6%)
Use of blood pressure medication n (%)	448 (6.5%)	100 (5.9%)	119 (7.0%)	119 (6.9%)	110 (6.4%)
Use of antidepression medication, yes, n (%)	373 (5.4%)	100 (5.9%)	89 (5.2%)	98 (5.7%)	86 (5.0%)

PHDI = planetary health diet index; NSAID = non-steroidal anti-inflammatory drug.

^aValues are frequency (%) for categorical variables and mean (SD) for continuous variables.

^bOnly among women (n_overall = 3502, n_Q1 = 877, n_Q2 = 873, n_Q3 = 906, n_Q4 = 846).

Association between adherence to the planetary health diet and adiposity

In change-on-change analysis, changes in the PHDI were inversely associated with changes in body weight, waist circumference, fat mass, subcutaneous fat, visceral fat, liver fat, and lean mass (Table 2, Supplemental Table S4). Each 10-point increase in PHDI was associated with reduced body weight by 0.47 kg (95% confidence interval: −0.63, −0.31) and visceral fat mass by 40 g (−51.5, −28.5) (Table 2). After adjustment for changes in total body weight, the associations persisted, though attenuated for waist circumference, visceral fat, and liver fat, but were no longer significant for fat mass, subcutaneous fat, and lean mass (Table 2). For instance, visceral fat was lower by 12.8 g (−20.5, −5.2) (Table 2). In a sensitivity analysis, after adjustment for changes in fat mass and subcutaneous fat, a negative association was still present for visceral fat mass (−7.9 g per 10 points PHDI increase, 95% CI: −13.5, −2.2, n = 6362) and liver fat score (−0.09 per 10 points increase, 95% CI: −0.12, −0.06, n = 5785). Analyses without adjustment for medication use were also performed but showed very similar associations as analyses with adjustment (results not shown). Similarly, annualised changes and cross-sectional analyses at baseline and follow-up were also similar to the main findings and are shown in Supplemental Tables S5 and S6.

Table 2.

Association between changes in adherence to the planetary health diet using the planetary health diet index (PHDI) and changes in adiposity measures in the Fenland Study (n = 6845).

Outcome	N	Mean (SD) change of the outcome	Multivariable-adjusted modela	Additionally adjusted for change in body weightb
			Change in outcome per 10 points PHDI increase (95% CI)	Change in outcome per 10 points PHDI increase (95% CI)
Body weight, kg	6845	0.6 (5.8)	−0.47 (−0.63, −0.31)
Body mass index, kg/m2	6845	0.3 (2.0)	−0.16 (−0.22, −0.11)
Waist circumference, cm	6187	2.0 (6.2)	−0.55 (−0.73, −0.37)	−0.12 (−0.23, 0.00c)
Fat mass, kg	6611	0.9 (4.6)	−0.43 (−0.56, −0.30)	−0.02 (−0.06, 0.02)
Subcutaneous fat, g	6582	0.7 (344)	−24.5 (−34.0, −15.0)	−2.07 (−8.65, 4.51)
Visceral fat, g	6362	147 (399)	−40.0 (−51.5, −28.5)	−12.8 (−20.5, −5.19)
Liver fat, score	5885	0.2 (1.2)	−0.12 (−0.15, −0.09)	−0.09 (−0.12, −0.06)
Lean mass, kg	6611	−0.6 (1.9)	−0.08 (−0.13, −0.02)	0.02 (−0.03, 0.06)

^aMultivariable adjusted model: Linear regression model with change in PHDI as exposure and change in measure of adiposity as outcome adjusted for: measure of adiposity, PHDI, age, sex, test site, education, occupation and income at baseline and marital status, smoking, physical activity, alcohol, hormone replacement therapy (in women), use of non-steroidal anti-inflammatory drugs, lipid lowering medication, blood pressure lowering medication and antidepressant medication at baseline and follow-up, diabetes medication at follow-up and follow-up time.

^bMultivariable adjusted model with further adjustment for change in body weight.

^cp = 0.04.

Stratified analyses

Stratification by baseline BMI indicated that among those with obesity, for each 10-point increase in the PHDI, there was a greater reduction or lesser gain in body weight, fat mass, subcutaneous fat, visceral fat, liver fat and lean mass (p < 0.05 for each interaction) (Fig. 1, Supplemental Figure S3). Similarly, greater reductions or lesser gain was observed among never smokers, except for liver fat score (Fig. 1, Supplemental Figure S3). Stratification or interaction by sex, age, and physical activity did not indicate different associations.

Fig. 1.

Subgroup analysis of the association between changes in adherence to the planetary health diet index (per 10 points increased score) and changes in visceral adipose tissue in grams (a) and liver fat scores (b) in the Fenland study. Linear regression model with change in PHDI as exposure and change in the measure of adiposity as outcome adjusted for: measure of adiposity, PHDI, age, sex, test site, education, occupation and income at baseline and marital status, smoking, physical activity, alcohol, hormone replacement therapy (in women), use of non-steroidal anti-inflammatory drugs, lipid lowering medication, blood pressure lowering medication and antidepressant medication at baseline and follow-up, diabetes medication at follow-up and follow-up time. Variables used for stratification were not adjusted for in the respective strata. p for interaction (using a Wald-test) was estimated in separate linear models including an interaction term between the change in PHDI and the stratum-variable.

Sensitivity analyses

Removing outliers and analyses with multiple imputation did not alter the observed patterns of associations between increased adherence to the PHD and changes in adiposity (Supplemental Table S7). Though participants in the current analyses (with data available at both baseline and follow-up) were slightly older, more likely to have a managerial/professional occupation and have a higher income, be less likely to smoke, be single and use lipid or antidepression medication than those who only participated at baseline (Supplemental Table S8), adjustment for non-response probability did not alter the results. Adjustment for total energy intake showed similar patterns of associations as the main analysis (Supplemental Table S4). For analyses removing individual components from the PHDI, we generally found similar associations as the main results (Supplemental Figure S4). The exception was for lean mass, where the association was attenuated towards the null when not including red and processed meat or potatoes. Because there could be strong correlations between changes in body weight and other measures of adiposity, we estimated the correlation between these variables and generally found no strong correlations (Supplemental Figure S5). Also, alternative adjustment using residuals from a linear regression of changes in body weight and changes in adiposity measures found similar patterns of associations (Supplemental Table S9). When applying an alternative PHD index, the EAT-Lancet diet Index, the association patterns did not change substantially (Supplemental Table S10). In the longitudinal change-analysis, higher adherence to the PHD (at baseline) was associated with reduced or lesser gaining of waist circumference and liver fat score (Supplemental Table S11). Estimates indicated a reduction or lesser gain in body weight, fat mass, lean mass, subcutaneous fat, and visceral fat though the confidence intervals were wide and included the null.

Discussion

In this UK population, we found that increasing the adherence to the planetary health diet in midlife was associated with a reduction or lesser gain of adiposity and improved body fat distribution. For visceral and liver fat, the association was independent of changes in body weight or total fat and subcutaneous fat mass. The strongest associations were observed among people with obesity at baseline.

A strength of the study was that we used repeated detailed assessments of both dietary intake and adiposity. This enabled us to investigate simultaneous changes in adherence to the PHD and changes in adiposity (i.e., change-on-change analysis), approaching an intervention design. However, a potential response to a dietary intervention (e.g. education, dietary counselling) of the PHD may not be the same as we observed in our observational context. Also, this design does not ensure that the change in dietary intake preceded the change in adiposity, a key consideration for drawing causal inferences, or rule out mediation from covariates in the statistical model. The results align with our sensitivity analysis using only baseline dietary intake and change in adiposity (i.e., longitudinal change analysis). The effect sizes were smaller when using baseline diet only which may be due to the baseline diet being further removed in time from the adiposity measure at follow-up, misclassification of dietary intake or reverse causation in the change-on-change analysis or a combination.¹¹ Hence, our analyses suggest a potential benefit in adults in the UK when shifting their diets towards the PHD in midlife, on adiposity, particularly among those with obesity, which is a major public health problem. Moreover, when adjusting for changes in body weight or total fat and subcutaneous fat mass, we still observed reductions or lesser gaining in central and metabolically detrimental adipose tissue compartments per estimated increase in adherence to the PHD. This indicates a metabolic benefit beyond lesser weight gain in the population. The detailed information on participants’ characteristics, health behaviours and medication use at both time points allowed prospective investigations of dietary changes and changes in adiposity measures. Further, this allowed for comprehensive adjustment for diverse potential confounding factors, including energy intake and physical activity that impact energy balance and energy storage. Nevertheless, a limitation of the study is that residual confounding cannot be ruled out, impeding strong causal conclusions. In addition, habitual diet and covariates were likely to be measured with errors and likely to bias the results in an unknown direction. The study included both DEXA and ultrasound assessments at both baseline and follow-up to capture changes in different adiposity compartments. Though assessments of adiposity were standardised and carried out by trained staff at both visits, measurement error could exist and make association estimates imprecise. Some participants did not attend the follow-up visit which might induce selection bias by non-response, but our analysis with inverse probability weighting did not indicate this bias to be substantial. The observed associations between the PHDI and measures of adiposity were not driven by a single component as indicated in the analyses without individual components. However, for lean mass the association was no longer present when not including red and processed meat or potatoes. This may be due to the association between these foods and changes in body weight, as the association with lean mass was also attenuated after adjustment for changes in body weight. The conceptualisation of the PHD into a dietary score has been performed in several ways in the literature,²³ but our results appeared not sensitive to the scoring algorithm used. Lastly, as this study only included solely a UK population, our findings may not be generalisable to all UK or other European populations.

No prior studies are available for directly comparing our findings that suggest that the more metabolically detrimental adiposity depots, namely visceral fat and liver fat, are reduced or lesser increased with increased adherence to the PHD. In an indirect comparison, a randomised controlled trial comparing a low-fat vegan diet with no dietary change over 16 weeks in overweight adults found that the intervention reduced body weight, and also liver fat, though not significantly so when adjusted for changes in body weight.²⁴ This has also been shown with hypocaloric diets.²⁵ Some studies have suggested that a diet high in monounsaturated fatty acids instead of saturated fat and/or diets high in dietary fibre reduce liver fat.¹⁰ Our findings supported such specific effects and would stimulate further mechanistic investigations to identify biologically causal agents in the PHD.

Other previous observational studies have primarily focused on body weight, BMI and waist circumference.4, 5, 6, 7 None have investigated changes in adherence to the PHD and changes in adiposity. An inverse association between adherence to the PHD and change in waist circumference and risk of obesity has been reported in a Danish cohort,⁶ and with risk of obesity in a French cohort.⁷ In the Danish cohort,⁶ baseline BMI-stratified analyses were also performed. Here, an inverse association was only observed among those with a BMI <25 kg/m² at baseline, whereas we found the association to be stronger in those with BMI >30 kg/m² at baseline; this calls for further investigation to understand this difference. In contrast, findings from a Finnish cohort reported no associations between the PHD and any of these adiposity measures.²⁶ Our finding of an interaction between the PHDI and smoking status suggested a blunted response among current or former smokers vs never smokers, in all adiposity markers except for liver fat. This could potentially be related to greater weight gain among those who stopped smoking between assessments as smoking cessation is acknowledged to be related with weight gain²⁷ and/or altered antioxidant and oxidative stress metabolism in current and former smokers.²⁸ Similar blunted associations were observed among smokers vs non-smokers in a European cohort investigating the relationship between plant-based protein intake and risk of cardiovascular disease²⁹ and a US cohort assessing the association between adherence to the PHDI and cardiovascular disease-specific mortality.³⁰ However, in another US study including three cohorts the association between PHDI and total mortality did not vary by smoking status.¹⁴ Nonetheless, the potential of a lesser benefit among current or former smokers we observed underlines the importance of interventions to limit smoking though further studies are needed to increase understanding of the mechanisms involved. It should be noted that although the p-value for interaction estimated in the linear model was statistically significant, the subgroup analyses had overlapping confidence intervals in stratified analyses, which may be due to a loss of power.

Amid a global syndemic of obesity, health-behaviour related diseases and climate change, our findings support the adoption of a dietary pattern such as the PHD in midlife. The effect sizes of the changes in body weight in the linear model suggest modest benefit. Using the results from our models, assuming linearity, increasing the PHDI from 80 points (the mean in the population) to 110 (the max score in the population was 116 points) would be expected to reduce body weight by 1.4 kg, on average, and by 2.3 kg in people with obesity, mostly comprising fat mass, including visceral and liver fat.

It is worth noting that of a maximum achievable PHDI of 140 points, the mean score in the highest decile of adherence at baseline was 100 points, highlighting that few free-living individuals fully adhere to the planetary health diet. Also, we did not evaluate the diet-related environmental impact in this study, though a recent cohort study in another UK population showed lower greenhouse gas emissions and land use with greater adherence to the PHD.³¹ Intervention studies are needed to investigate the possibility of more complete adherence to the planetary health diet and its effects on nutrient intakes, adiposity and multiple environmental indicators compared with current Western diets in an isocaloric setting, particularly in people with obesity.

In this UK cohort of middle-aged adults with repeated assessment of dietary intake and adiposity, we found that increased adherence to the planetary health diet was associated with reduced or lesser gain of both overall and visceral adiposity, particularly in people with obesity. These findings suggest a beneficial impact of improving adherence to the planetary health diet on adiposity in a European population, though randomised controlled trials are needed to further support potential causal conclusions.

Contributors

DBI and NGF designed the analysis, with input from FI, SAS and EDLR. SB, SJG, NGF and NJW acquired the data. DBI performed the statistical analyses and drafted the manuscript. All authors interpreted study data and revised and approved the final version of the manuscript. NGF and DBI are the guarantors of this work and, as such, had full access to all the data in the study, verified the data and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data sharing statement

The Fenland Study data underlying the findings of this manuscript are available to bona fide researchers for health-related research that is in the public interest. A request for data can be made via an email to datasharing@mrc-epid.cam.ac.uk.

Declaration of interests

NGF received reimbursement for travel expenses for attending meetings as a Commissioner of the EAT-Lancet Commission. SB has been invited as speaker at an academic symposium on climate change impact and metabolism and has a pending patent on assessment of cardiorespiratory fitness from heart rate response to exercise. All other authors declare that there are no relationships or activities that might bias, or be perceived to bias, their work.