Objectives National dietary guidelines were introduced in 1977 and 1983, by the US and UK governments to reduce coronary heart disease (CHD) mortality by reducing dietary fat intake. Our 2016 systematic review examined the epidemiological evidence available to the dietary committees at the time; we found no support for the recommendations to restrict dietary fat. The present investigation extends our work by re-examining the totality of epidemiological evidence currently available relating to dietary fat guidelines.
Methods A systematic review and meta-analysis of prospective cohort studies currently available, which examined the relationship between dietary fat, serum cholesterol and the development of CHD, were undertaken.
Results Across 7 studies, involving 89 801 participants (94% male), there were 2024 deaths from CHD during the mean follow-up of 11.9±5.6 years. The death rate from CHD was 2.25%. Eight data sets were suitable for inclusion in meta-analysis; all excluded participants with previous heart disease. Risk ratios (RRs) from meta-analysis were not statistically significant for CHD deaths and total or saturated fat consumption. The RR from meta-analysis for total fat intake and CHD deaths was 1.04 (95% CI 0.98 to 1.10). The RR from meta-analysis for saturated fat intake and CHD deaths was 1.08 (95% CI 0.94 to 1.25).
Conclusions Epidemiological evidence to date found no significant difference in CHD mortality and total fat or saturated fat intake and thus does not support the present dietary fat guidelines. The evidence per se lacks generalisability for population-wide guidelines.
- Public health
- Heart disease
- Cohort study
Statistics from Altmetric.com
US public health dietary advice was announced by the Select Committee on Nutrition and Human needs in 19771 and was followed by UK public health dietary advice issued by the National Advisory Committee on Nutritional Education in 1983.2 Dietary recommendations in both cases focused on reducing dietary fat intake; specifically to (1) reduce overall fat consumption to 30% of total energy intake and (2) reduce saturated fat consumption to 10% of total energy intake.
The recommendations were intended to address mortality from coronary heart disease (CHD). In 2015, we published a systematic review and meta-analysis,3 which reported that evidence from randomised controlled trials (RCTs), available to the dietary guideline committees, did not support the introduced dietary fat recommendations. In 2016, we published a systematic review and meta-analysis,4 which reported that RCT evidence currently available does not support the introduced dietary guidelines. In 2016, we published a systematic review,5 which reported that epidemiological evidence, available to the dietary guideline committees, did not support the introduced dietary fat recommendations. The aim of these systematic reviews has been twofold: to examine the epidemiological and RCT evidence base for the dietary fat guidelines to assess if they were justified at the time of their introduction and to review if the evidence currently available supports the extant recommendations. This systematic review and meta-analysis completes this work by examining the totality of epidemiological evidence currently available.
This systematic review and meta-analysis uses the Meta-analysis Of Observational Studies in Epidemiology methodology (MOOSE).6 MOOSE uses parts of the Cochrane methodology for systematic review and meta-analysis of RCTs, Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA), for example, the figure for presenting search methodology.7 To examine best practice evidence available, this review has focused on prospective cohort studies.8
A search was undertaken to identify prospective cohort studies that examined the relationship between dietary fat intake, serum cholesterol and mortality from CHD. Exclusion criteria were: clinical trials, cross-sectional studies, case–control studies. Inclusion criteria were: prospective cohort studies, participants were human adults, primary study outcome was CHD mortality, data related to dietary fat consumption were available, data on CHD mortality and serum cholesterol measurements were available.
The search was undertaken on 30 September 2015. No date limitations were set, enabling the earliest possible papers to be included. Searches of the literature were performed using MEDLINE, Embase and the Cochrane Library (figure 1).
Selection of studies
Of 669 identified articles, 607 were rejected on review of the title and abstract. Of these, 55 were rejected for being review articles. Fifty-five were commentaries, editorials or letters. Thirty-nine were clinical trials, 9 were cross-sectional or case–control studies. Seventy-five related to conditions other than heart disease, primarily cancer, diabetes, hypertension and obesity. There were 53 studies in which animals or children/adolescents were the primary focus. Ninety-eight were articles about pharmacology/blood analysis. One hundred and twenty-three were rejected for being related to a particular food, nutrient or supplement, rather than dietary fat. A further 40 papers were educational material and 36 reviewed dietary compliance in nutritional studies. Twenty-four articles were mathematical modelling exercises, including meta-analysis. Sixty-two papers remained. Thirteen were rejected on closer inspection of the full paper: 12 were reviews, commentaries or expert opinion pieces and 1 was educational material. Where an abstract was unavailable, the publication type, journal name and metatags were reviewed to assess if the article should be rejected (eg, ‘letter’, ‘clinical trial’, ‘paediatric’). Copies of the remaining articles were obtained from university libraries or the British Library. Forty-nine papers, prima facie, met the inclusion criteria.
Of the 49 papers included in qualitative synthesis, 18 were found prima facie to meet the inclusion criteria and 31 were duplications of these 18 studies. On detailed examination of the 18 studies, 6 were the studies included in Harcombe et al,5 which were available to the UK National Advisory Committee in 1983. The Western Electric study,9 ,10 The Puerto Rico Heart Health Program,11–13 The Seven Countries Study,14 The Framingham Heart Study,11 ,15 The Honolulu Heart Program11 ,16 ,17 and a study conducted in London and the South East.18 None of these had data conducive to meta-analysis. Twenty-four papers were additional publications related to these 6 studies; the majority of these were about The Seven Countries Study. The additional publications were examined for data published after the introduction of the dietary fat guidelines to assess if any Harcombe et al5 studies could be updated. A 1984 paper about The Honolulu Heart Program19 and a 1991 paper about The Framingham Heart Study20 were reviewed, but neither reported CHD mortality.
A total of 12 new studies were found.21–32 Seven papers were additional publications related to these studies. Four studies were excluded on closer inspection for not reporting CHD mortality data22 ,23 ,25 ,26 and another one contained CHD mortality data that were not separable from CHD events.24 These five studies also failed to meet inclusion criteria for other reasons: no saturated fat data,22 ,26 no total fat data,25 ,26 no serum cholesterol data.22–24 ,26 Mann et al26 was the only study to review animal fat, not all dietary fat. This study, examining the benefits of a vegetarian diet, reported consumption of total animal fat and saturated animal fat, thus excluding the fat and saturated fat in many non-animal foods.33
Six studies did not report serum cholesterol data,21 ,27 ,28 ,30–32 which left one study meeting the full inclusion criteria.29 The serum cholesterol component part of the diet-heart hypothesis was set aside as an inclusion criterion, which enabled data for dietary fat (total and saturated) and CHD mortality to be examined for seven studies.21 ,27–32 The UK Health survey21 presented data for men and women separately and The Strong Heart Study of American Indians31 presented data for those aged 47–59 years and those aged 60–79 separately. Table 1 presents the extracted data for total and saturated dietary fat and CHD deaths for these seven studies; nine information sets.
To ascertain the validity of eligible observational studies, a pair of reviewers (ZH and BD) worked independently to determine which studies met the inclusion criteria. The same one additional study was agreed on as meeting the original diet-heart hypothesis criteria and the same seven studies were agreed on for the revised search. One of the authors (ZH) examined contemporary systematic reviews of prospective cohort studies34–36 to ensure that no studies had been omitted or included inconsistently. Correspondence was entered into with one author for thorough verification (Personal communication with Siri-Tarino. Harcombe Z, ed, 2015). Risk of bias was assessed using the Cochrane Collaboration assessment tool for component parts relevant to observational studies, defined as follows: selection bias (cohort appropriately reflected wider population characteristics), detection bias (blinding of outcome assessment), attrition bias (incomplete data outcome) and reporting bias (selective reporting) (figures 1 and 2).37
Table 1 details data extraction of: study name, year of key publications, participant characteristics, whether free from CHD at study entry, years of follow-up, CHD deaths, outcomes relating to total dietary fat and saturated dietary fat for CHD deaths.
The overall pooled effect was calculated using random effects meta-analysis. This takes into account between-study variance when assigning weights in meta-analysis. The weighting of studies is more balanced with random effects than fixed-effects methodology; larger studies have less impact and smaller studies have more impact than with fixed effects. Heterogeneity and bias were quantified using the I2 and T2 calculations, I2=100%×(Q−df)/Q, where Q is Cochran's heterogeneity statistic and df the degrees of freedom. Funnel plot methodology and Egger's regression intercept38 ,39 were calculated to assess risk of publication bias. Analyses were performed using Comprehensive Meta-Analysis.40
Meta-analysis was possible for all but one study in table 1.27 Kushi et al reported fat intake as a percentage of overall energy intake for those who died from CHD and those who did not. However, no SDs or SEs were presented, which are necessary for inclusion in meta-analysis. Forest plots for the associations between CHD deaths and total fat and saturated fat are presented in figures 3 and 4, respectively.
Participants and study design
The seven prospective cohort studies in table 1 included 89 801 participants. Eighty-four thousand two hundred and seventy of the participants were men (94%). All but one study excluded people with previous heart disease.27 This study could not be included in the meta-analysis, as it did not contain suitable data. The age ranges varied from 30 to 79 years at baseline. Xu et al31 grouped participants from 47 to 59 years and from 60 to 79 years. No explanation was given for this grouping. Esrey et al29 grouped participants from 30 to 59 years and from 60 to 79 years. The rationale for this arbitrary classification was reported as to maximise deaths in each group.
Participants were followed for a minimum of 6 years and a maximum of 20 years. The mean duration of the seven cohorts was 11.9±5.6 years. The weighted mean duration (person years by participants) was 8.3±1.1 years.
All trials had complete outcome data, avoiding attrition bias (figure 2).37 One study was judged unclear for selection bias, as participants excluded non-smokers.30 Four studies were judged unclear for detection bias, as blinding of outcome assessment was unclear.27 ,30–32 Risk of reporting bias was judged low for four of the studies: it was judged unclear for The Strong Heart Study and The Lipid Research Study, as the rationale for the age group divisions were not clear;29 ,31 and it was judged high risk for the Japanese Study as CVD by fat intake was researched, but not reported, for women32 (figure 2).
For total fat and CHD deaths, Esrey et al29 carried the greatest weight, 31.76% for the 30–59 age group and 29.45% for the 60–79 age group (figure 3 random effects methodology). The two Boniface and Tefft21 observations, for men and women, carried a combined weight of 30.04%. The two Xu et al31 observations carried a weight of 1.02% between them. The risk ratio (RR) for the meta-analysis of the eight data sets available for total fat and CHD deaths was 1.04 (95% CI 0.98 to 1.10). The overall effect measurement lies on the line of no effect. There was no statistically significant difference between total fat and CHD deaths.
The meta-analysis for total fat and CHD deaths was tested for sensitivity analysis of the exclusion of any one study. There were no circumstances in which the exclusion of any one study made the overall effect size significant.
There was evidence for between-study heterogeneity. The Q-value was 18.218 (8 df) and this was statistically significant p=0.020. I2 was 56.087 and T2 was 0.002, indicating difference in true effects.
Visual inspection of the funnel plot revealed that one study was outside the SE funnel for the meta-analysis of total fat and CHD deaths. This was the Xu et al31 study for people aged 47–59, which was substantially different from other RRs. The Egger's regression test indicated no statistically significant asymmetry for total fat and CHD deaths. The Egger's regression intercept was 0.640 (95% CI, two-tailed, −0.959 to 2.238), but this was not statistically significant (one-tailed p=0.188; two-tailed p=0.376).
For saturated fat and CHD deaths, Esrey et al29 carried the greatest weight, 19.83% for the 30–59 age group and 19.28% for the 60–79 age group (figure 4 random effects methodology). The two Boniface and Tefft21 observations, for men and women, carried a combined weight of 28.96%. The two Xu et al31 observations carried a weight of 5.10% between them. The RR for the meta-analysis of the eight data sets available for saturated fat and CHD deaths was 1.08 (95% CI 0.94 to 1.25). The overall effect measurement lies on the line of no effect. There was no statistically significant difference between saturated fat and CHD deaths.
The meta-analysis for saturated fat and CHD deaths was tested for sensitivity analysis of the exclusion of any one study. There were no circumstances in which the exclusion of any one study made the overall effect size significant.
There was evidence for between-study heterogeneity. The Q-value was 37.080 (8 df) and this was statistically significant p<0.001. I2 was 78.425 and T2 was 0.025, indicating difference in true effects.
Visual inspection of the funnel plot revealed that three studies were outside the SE funnel for the meta-analysis of total fat and CHD deaths28 ,30 and the Xu et al31 study for people aged 47–59. The Egger's regression test indicated no statistically significant asymmetry for saturated fat and CHD deaths. The Egger's regression intercept was 0.686 (95% CI, two-tailed, −1.978 to 3.351), but this was not statistically significant (one-tailed p=0.281; two-tailed p=0.562).
Examination of the dietary guidelines
None of the seven cohort studies examined either of the introduced dietary guidelines: a total fat consumption of 30%, or a saturated fat consumption of 10%, of energy intake. Two studies examined the total fat intake and the saturated fat intake, as a percentage of calorie intake, for participants who died from CHD compared with those who did not.27 ,29 Esrey et al also reported the RR for an additional one unit of total or saturated fat as a percentage of energy intake. Three studies measured total and saturated fat intake in quintiles and reported data for CHD mortality as RRs comparing the lowest and highest quintiles of intake.28 ,30 ,32 The Strong Heart Study followed this method, but for quartiles, not quintiles.31 The remaining study reported the RR for an additional 100 g of total or saturated fat per week for men and women separately.21
Across 7 studies, involving 89 801 participants, there were 2024 deaths from CHD during the period of follow-up. The death rate from CHD was 2.25%, during the mean follow-up of 11.9±5.6 years. The death rate was influenced by the large Finnish study, which included only participants who smoked. The death rate for this study was 2.9%.
The death rate contrasts with the 30% death rate for the six RCTs reviewed in Harcombe et al,3 which reinforced the high death rate in secondary studies. The one study that included men with previous heart disease reported a death rate of 11%.27
Significance reported by the studies
Three of the seven studies reported some participants separately: two by age groups29 ,31 and one by gender.21 This enabled 10 sets of data to be examined (nine in meta-analysis) (table 1). Six found no significant relationship between total fat and CHD mortality,27 ,30 ,32 men in the UK health and lifestyle survey,21 men and women aged 60–79 years in The Lipid Research Study29 and the participants aged 60–79 years from The Strong Heart Study.31 The remaining four found a significant relationship between total fat and CHD mortality: Ascherio et al;28 women in the UK health and lifestyle survey;21 and the participants aged <60 years from The Lipid Research and The Strong Heart Study.29 ,31 Five found no significant relationship between saturated fat and CHD mortality;27 ,32 men in the UK health and lifestyle survey;21 and participants aged 60–79 years from The Lipid Research and The Strong Heart Study.29 ,31 The remaining five found a significant relationship between saturated fat and CHD mortality;28–30 women in the UK health and lifestyle survey;21 and participants aged <60 years in The Lipid Research and The Strong Heart Study.29 ,31
Meta-analysis confirmed that the pooled data were not significant for CHD mortality and total fat intake or CHD mortality and saturated fat intake.
Serum cholesterol levels
The only study to meet the full systematic review inclusion criteria was Esrey et al.29 This study found a significant relationship between baseline serum cholesterol and subsequent mortality from CHD in the age group 30–59 years, but not in the age group 60–79 years.
The main findings of this systematic review and meta-analysis are that the epidemiological evidence currently available to the dietary committees provides no statistically significant retrospective support for the introduction of dietary fat guidelines.
Most studies reported other significant findings: Kushi et al found a significant association with higher intake of carbohydrate and fibre and lower CHD mortality. Esrey et al found significant positive associations with age, male gender, body mass index (BMI), smoking and CHD deaths. Two studies reported significant associations with trans-fat intake and CHD mortality.28 ,30 Pietinen et al additionally reported significant associations with the number of cigarettes smoked per day, the number of years of smoking, blood pressure, BMI and CHD mortality. Pietinen et al found that high-density lipoprotein cholesterol, education and physical activity were inversely related to the risk of CHD. Ascherio et al found that associations found between dietary fat and heart disease were greatly attenuated by adjusting for fibre intake. Boniface and Tefft reported that smoking, not exercising and being socially disadvantaged were related to higher saturated fat intake and CHD death, which meant that confounding variables could have played a part in findings. Xu et al found that age, being male, having diabetes and smoking were all significantly associated with CHD deaths.
As was found with the review of RCT evidence in Harcombe et al,3 ,4 the fundamental design limitation of epidemiological evidence currently available is that it lacks generalisability. RCT evidence available to the dietary guideline committees had studied men with pre-existing heart disease;3 epidemiological evidence available to the dietary guideline committees had studied men mostly free from heart disease.5 One RCT currently available included men and women, without previous heart disease, and contained all component parts of the diet-heart hypothesis for examination.41 One prospective cohort study currently available included men and women free from heart disease and contained all component parts of the diet-heart hypothesis for examination.29
The method of extracting dietary information was a limitation. Dietary recall is generally unreliable and 24-hour recall may not be representative of usual diet.42 ,43 Dietary surveys, where food is weighed at the time of being recorded, are also unreliable.44 ,45
Two studies concluded that evidence was strong for an association between dietary fat and death from CHD. Boniface and Tefft21 reported: “Strong evidence was found for the within cohort relationship of dietary fat and CHD death in women while no evidence was found for a relationship in men” (p. 786). Xu et al31 found that: “Total fat, saturated fatty acid, and monounsaturated fatty acid intake were strong predictors of CHD mortality in American Indians aged 47–59 y” (p. 894).
The Japanese Study found: “Total fat intake is associated with decreased mortality in Japanese men but not in women” (ref. 32, p. 1713).
Two studies reported weak or no associations between dietary fat and CHD mortality. Ascherio et al28 appeared to reference The Seven Countries Study with their comment, ‘These data do not support the strong association between intake of saturated fat and risk of coronary heart disease suggested by international comparisons’ (p. 84). Pietinen et al30 reported that ‘There was a significant positive association between the intake of trans fatty acids and the risk of coronary death. There was no association between intakes of saturated or cis-monounsaturated fatty acids or linoleic or linolenic acid or dietary cholesterol and the risk of coronary death’ (p. 876).
The final two studies described the relationship as unclear. Kushi et al27 concluded: ‘The nature of the association in this study between dietary lipids and the risk of mortality from coronary heart disease remains unclear’ (p. 816). The one study meeting all systematic review inclusion criteria reported: ‘We conclude that future research must be directed toward better understanding the pathway between dietary intake and coronary disease as the current diet-lipid-heart hypothesis may be overly simplistic’ (ref. 29, p. 211).
Other meta-analyses of prospective cohort studies
The aim of this systematic review, as part of a series of four,3–5 has been to evaluate the RCT and prospective cohort evidence base for two specific dietary fat guidelines (30% and 10% of energy intake in the form of total and saturated fat, respectively) at their time of introduction and the present day. This has not previously been undertaken.
A number of other systematic reviews and meta-analyses have been undertaken, which are relevant to the present review.33–35 ,46 De Souza et al46 systematically reviewed saturated not total, fat including any observational study, not just prospective cohort studies, with mortality as the end point. Siri-Tarino et al34 undertook a meta-analysis of prospective cohort studies for saturated, not total, fat, with disease as the end point. Chowdhury et al36 also examined saturated, not total, fat and with disease, not mortality, as the end point, including RCTs and prospective cohort studies in the same review. Skeaff and Miller35 were closest to this review, and the one reviewing RCTs currently available,4 having undertaken a systematic review and meta-analysis of RCTs and prospective cohort studies combined, for total and saturated fat, with CHD end points, disease and mortality. This was published in 2009.
De Souza et al46 concluded: ‘Saturated fats are not associated with all cause mortality, CVD, CHD, ischemic stroke, or type 2 diabetes, but the evidence is heterogeneous with methodological limitations’ (p. 1).
The conclusion of Siri-Tarino et al's34 review was ‘meta-analysis of prospective epidemiologic studies showed that there is no significant evidence for concluding that dietary saturated fat is associated with an increased risk of CHD or CVD’ (p. 535).
Chowdhury et al set out to summarise evidence between fatty acids and coronary disease. Their review examined saturated, monounsaturated, polyunsaturated and trans-fats, while also reviewing individual chain length fatty acids, palmitic (C16:0) and margaric (C17:0) as examples. The conclusion was ‘Current evidence does not clearly support cardiovascular guidelines that encourage high consumption of polyunsaturated fatty acids and low consumption of total saturated fats’ (ref. 36, p. 398).
Skeaff and Miller sought to summarise the evidence from cohort studies and RCTs of the relation between dietary fat and risk of CHD. Their conclusion was ‘Intake of total fat was not significantly associated with CHD mortality. Intake of total fat was also unrelated to CHD events’ (ref. 35, p. 175).
Harcombe et al3 ,5 found that the dietary fat guidelines were not supported by RCT or epidemiological evidence available at the time of their introduction. These dietary fat guidelines have prevailed until 2016 and thus the validity of their evidence base remains important to examine. UK dietary fat guidelines are unchanged. The Dietary Guidelines for Americans, issued in January 2016,47 were conspicuously silent on the subject of total fat, but reiterated that saturated fat should be restricted to no more than 10% of calorie intake. Harcombe et al4 found that the RCT evidence currently available does not support the dietary fat guidelines. This review finds that the epidemiological evidence currently available does not support the dietary fat guidelines. All systematic reviews found serious limitations with the availability of primary prevention, both sex, studies, which are the ones most likely to have generalisability for whole populations.
The conclusion of the four systematic reviews and three meta-analyses is that there was no evidence to support the dietary fat guidelines being introduced and there is no evidence currently available to support them. Public health authorities need to urgently review dietary advice.
What are the findings?
Dietary recommendations were introduced in the USA (1977) and in the UK (1983) to (1) reduce overall fat consumption to 30% of total energy intake and (2) reduce saturated fat consumption to 10% of total energy intake. We reported in previous meta-analyses that the randomised controlled trial (RCT) and prospective cohort study evidence available to the dietary committees did not support the introduction of these dietary fat guidelines. This study finds that epidemiological evidence currently available does not support the extant dietary fat guidelines.
Meta-analysis of prospective cohort studies finds no significant association between coronary heart disease (CHD) deaths and total fat consumption.
Meta-analysis of prospective cohort studies finds no significant association between CHD deaths and saturated fat consumption.
This is the final one of four systematic reviews examining the RCT and epidemiological evidence at the time dietary fat guidelines were introduced and RCT and epidemiological evidence currently available. Not one review has found evidence to support public health dietary fat guidelines.
How might it impact on clinical practice in the future?
Public health advice on dietary fat has prevailed since 1977/1983 in the absence of supporting evidence. Dietary advice in both nations need re-examination.
Protein is present in all foods, except pure fats and sucrose, and thus tends to form ∼15% of total calorie intake. Restricting total fat intake to 30% concomitantly sets a carbohydrate intake of 55%. Diabetes and obesity have increased dramatically since guidelines to restrict fat intake. This association needs examination.
Twitter Follow Zoe Harcombe at @zoeharcombe
Contributors ZH conceived of the study and was the major contributor to data extraction, writing of the manuscript and the meta-analysis. BD was involved in data extraction. All authors were involved in critical evaluation of content.
Competing interests ZH receives income from writing and from two small self-employment businesses: The Harcombe Diet Co. and Columbus Publishing.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.