Intake of individual saturated fatty acids and risk of coronary heart disease in US men and women: two prospective longitudinal cohort studiesBMJ 2016; 355 doi: https://doi.org/10.1136/bmj.i5796 (Published 23 November 2016) Cite this as: BMJ 2016;355:i5796
All rapid responses
Despite their detailed reply, Sun et al. did not cite the recent empirical refutation of their methods: Archer, E., G. Pavela, and C. J. Lavie. 2015. "The Inadmissibility of What We Eat in America and NHANES Dietary Data in Nutrition and Obesity Research and the Scientific Formulation of National Dietary Guidelines." Mayo Clinic Proceedings 90 (7):911-26. doi: 10.1016/j.mayocp.2015.04.009.
The above cited paper represents an existential threat to the field of nutrition epidemiology and cautions that there is no valid scientific support for any of the conclusions derived from self-reported dietary data over the past 50 years.
Nevertheless, I have a simple scientific question for Sun et al. Can these authors explain what method they used to assess the frequency with which their respondents lied, forget, mis-remembered, or mis-estimated their dietary intake and physical activity over the study period?
Or stated more simply, how do the authors know anything the respondents reported is accurate? Surely these authors do not assume that everyone is telling the truth? Or do they?
Competing interests: No competing interests
Response to Sun et al - effect of dietary carbohydrate on serum palmitate and lipoproteins casts saturated fat in a changing light.
We thank Sun et al for their intelligent and informative response to our criticisms of this paper.
It is not our position that the Praagman EPIC-Netherlands paper is more accurate regarding the true causal relationship between fats or macronutrients and heart disease than the current paper, but that both papers are likely to be affected by different errors, explaining their divergence, a position which we could have made clearer. We also find common features between both papers, despite their divergent results, which are consistently supported by other evidence and capable of explanation, a point we will return to later. Our responses to the points made by Sun et al are as follows.
1) It appears that the baseline data given in Wang et al 2016 and in the current paper do not represent the quintiles that appear in results. The NHS cohort in the lowest quintile in Wang et al has gone from being 2.3 years older than the highest at baseline to being 15 years older at results. What then is the value of the baseline data supplied, if it is not for the groupings for whom results are presented? Furthermore, if the “cumulative averaging” modelling has now created a quintile that is 15 years older than another quintile, which seems unprecedented, and contains individuals so health-conscious and risk averse that they have substantially altered their diets, then this quintile is like to be different from others in other ways, an outcome likely to increase residual confounding. Do we know what happened to the older health-conscious individuals in other quintiles?
2) Sun et al affirm the accuracy of the FFQ method by validating it against 7-day food diary results. However, this is comparing two methods of estimating intake, not validation against any objective observation, and the correlation is imprefect. There is also a risk that the cumulative averaging method introduces extra inaccuracy. In 1999 Hu et al wrote “In addition, within each model, the methods using the cumulative averages in general yielded stronger associations than did those using either only baseline diet or the most recent diet”.
3) We do not accept that the OR of 1.56 for the association of saturated fat and respiratory mortality in Wang et al is not due to a significant error or errors that will also affect the current paper. Sun et al have not explained or defended this association by providing a mechanism or supporting evidence. They have provided evidence to validate smoking recall, with impressive correlations for sensitivity of 0.86, and specificity of 0.94. However, as these subjects were the mothers of participants in NHS, this evidence does not answer the question of whether modern health professionals under-report smoking.
4) Although Praagman et al’s Rotterdam paper found that palmitic and stearic acids (C:16 and C:18) were associated with CHD, the results still did not generate a prediction that replacing saturated fat with polyunsaturated fat would reduce disease. Even in the EPIC-Netherlands paper, C:16 and to some extent C:18 have less beneficial correlations with IHD than other saturated fats or total saturated fat. In this way, both Praagman papers are consistent with the current paper – whether saturated fat seems harmful or beneficial in any given population, C:16 and C:18 are the least beneficial components of it. This is also consistent with the protective associations between types of saturated fat and risk of type 2 diabetes in the Malmö Diet and Cancer Study.
The explanation for this is simple. Modern diets are high in refined carbohydrate, and the level of C:16 and C:18 in serum is controlled by carbohydrate. The reason for this is not merely that the human body synthesizes C:16 from carbohydrate, as Sun et al correctly state, but also that the rate of C:16 oxidation is controlled by the insulin response to carbohydrate, so that C:16 drops in serum when carbohydrate is restricted, semi-independently of dietary intake.[6, 7] This interaction has been understood for 60 years and it is puzzling that it has never been incorporated into epidemiology. Mechanistically, the major harms proposed for higher saturated intakes relate to the interaction of carbohydrate and insulin with dietary C:16, either through elevated serum C:16-C:18 levels, or through increases in the atherogenicity of lipoprotein particles.[8, 9, 10, 11] This helps to explain why associations of saturated fats with disease are inconsistent between populations eating different dietary patterns.
Because the % of energy from refined carbohydrate in the current diet is very much greater than the % of C:16 and C:18, and because saturated fats can be found in nutritious foods and are associated with unsaturated fats, whereas refined carbohydrates are less desirable items of diet, and currently associated with trans fats and palm oil, this evidence supports carbohydrate restriction over fat restriction.
The systematic review of Mensink cited by Sun et al also supports the replacement of carbohydrate with fat, even if part of that fat is saturated. A low-carbohydrate diet with 20% of energy as SFA has more potential to decrease ApoB and increase ApoA, according to the formulae supplied in that paper, than a low-fat diet that supplies 10% of energy as SFA.
This is in fact consistent with the finding of Wang et al (supplementary eTables 3, 11) that a higher total fat intake (fat replacing carbohydrate) is associated with lower mortality, 0.84 (0.81-0.88) and cardiovascular mortality, 0.86 (0.79, 0.93), despite the weak positive risk association for the saturated fat component of total fat in that paper. Yet this result, supportive of carbohydrate restriction, was never mentioned by Wang et al in the numerous interviews, press releases, and Harvard blogs generated from this study. When researchers selectively misinform the public of their findings in this way, one naturally suspects bias, and one wonders if this same bias plays a part in some of the necessary arbitrary decisions in epidemiology, such as the selection of modelling methods.
 Hu FB, Stampfer MJ, Rimm E, Ascherio A, Rosner BA, Spiegelman D, Willett WC. Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements. Am J Epidemiol. 1999 Mar 15;149(6):531-40.
 Tomeo CA, Rich-Edwards JW, Michels KB, Berkey CS, Hunter DJ, Frazier AL, et al. Reproducibility and validity of maternal recall of pregnancy-related events. Epidemiology. 1999; 10(6): 774-7.
 Praagman J, de Jonge EA, Kiefte-de Jong JC, Beulens JW, Sluijs I, Schoufour JD, et al. Dietary Saturated Fatty Acids and Coronary Heart Disease Risk in a Dutch Middle-Aged and Elderly Population. Arterioscler Thromb Vasc Biol. 2016; 36(9): 2011-8.
 Praagman J, Beulens JW, Alssema M, et al. The association between dietary saturated fatty acids and ischemic heart disease depends on the type and source of fatty acid in the European Prospective Investigation into Cancer and Nutrition-Netherlands cohort. Am J Clin Nutr2016;103:356-65.
 Ericson, U, Hellstrand, S, Brunkwall, L, Schulz, C-A, Sonestedt, E, Wallström, P, et al. Food sources of fat may clarify the inconsistent role of dietary fat intake for incidence of type 2 diabetes. AJCN 2015;114.103010v1
 Volk BM, Kunces LJ, Freidenreich DJ, et al. Effects of Step-Wise Increases in Dietary Carbohydrate on Circulating Saturated Fatty Acids and Palmitoleic Acid in Adults with Metabolic Syndrome. PLoS ONE. 2014;9(11):e113605. doi:10.1371/journal.pone.0113605.
 Lossow WJ, Chaikoff IL. Carbohydrate sparing of fatty acid oxidation. I. The relation of fatty acid chain length to the degree of sparing. II. The mechanism by which carbohydrate spares the oxidation of palmitic acid. Arch Biochem Biophys. 1955; 57(1):23-40.
 Sokolova M, Vinge LE, Alfsnes K, et al. Palmitate promotes inflammatory responses and cellular senescence in cardiac fibroblasts. Biochim Biophys Acta. 2017; 1862(2):234-245.
 Lu Z, Li Y, Brinson CW, Kirkwood KL, et al. CD36 is upregulated in mice with periodontitis and metabolic syndrome and involved in macrophage gene upregulation by palmitate. Oral Dis. 2016; Oct 18.
 Eulàlia Montell, Marco Turini, Mario Marotta et al. DAG accumulation from saturated fatty acids desensitizes insulin stimulation of glucose uptake in muscle cells. American Journal of Physiology - Endocrinology and Metabolism. 2001; 280(2): E229-E237.
 Siri-Tarino PW, Chiu S, Bergeron N, Krauss RM. Saturated Fats Versus Polyunsaturated Fats Versus Carbohydrates for Cardiovascular Disease Prevention and Treatment. Annual review of nutrition. 2015; 35:517-543.
 Mensink RP. Effects of saturated fatty acids on serum lipids and lipoproteins: a systematic review and regression analysis. Geneva: World Health Organization; 2016.
 Wang DD, Li Y, Chiuve SE, Stampfer MJ, Manson JE, Rimm EB, et al. Association of Specific Dietary Fats With Total and Cause-Specific Mortality. JAMA Intern Med. 2016; 176(8): 1134-45.
Competing interests: Grant Schofield is co-author of What The Fat?, a guide to the science and practice of low-carbohydrate eating. George Henderson is a part-time waged employee of the company that distributes What The Fat?
Re: Unresolved issues with the previous dietary fat study from the NHS and HPFS cohorts still need addressing, but concerns about palm oil may be valid.
Henderson compared our findings with those by Praagman et al.1 Nonetheless, Henderson ascribed the difference between our and Praagman et al’s studies by referring to questions for another paper from our group on saturated fatty acids (SFAs) and mortality.2 Henderson also raised questions on our estimation of energy intake, adjustment of smoking status, and self-reports of smoking status. We hereby clarified a few issues related to these questions.
1) Both Wang et al’s and our studies presented age-adjusted characteristics of participants, therefore differences in lifestyle, diet, and other factors across SFAs groups were not explained by age. The age distribution in the Table 1 of Wang et al’s publication at JAMA Internal Medicine was for baseline survey only. In the Nurses’ Health Studies (NHS) and Health Professionals Follow-up Study (HPFS), diet was updated every 2-4 years, and we calculated and used cumulative averages in data analyses, which has been clearly indicated in the publication. During follow-up, NHS participants in the first quintile were much older (difference is 15 years) than those in the highest quintile because our older participants reduced SFA intake more than their younger counterparts during follow-up. Therefore, confounding by age was strong on associations with mortality, and thus adjustment of age corrected this confounding and made the associations positive.
2) Total energy intake in absolute amounts is known to be underestimated by food frequency questionnaires (FFQs) because it is unrealistic to include all food items in FFQs. However, because the measurement errors for total energy intake and macronutrients are correlated, expressing macronutrients as % of total energy will cancel out the correlated errors. The percent of energy intake from fats did not show evidence of underestimation.3 Moreover, our validation studies demonstrated good correlations between FFQs and 7 day diet records assessments of dietary fats as well as biomarkers.
3) In most studies self-reported smoking status is reasonably accurate.4 In the NHS, women’s recalled smoking status in their early life was demonstrated to be reliable.5 In our analysis, we not only categorized participants in more categories than Praagman et al. who defined smoking as never, former, or current at baseline only,1 but also adjusted for updated smoking status every two years to control for confounding by smoking.
4) It is worth noticing that Praagman et al.’s more recent publication6 showed that the major SFA, palmitic acid, was significantly associated with a higher coronary heart disease risk, which is consistent with our study findings. The main finding of this more recent study was not consistent with that of Praagman et al.’s publication at the American Journal of Clinical Nutrition.1 Although the exact reasons underlying the discrepancy between these two studies by Praagman et al. are unknown, the differences between the two study populations and the FFQs used may play a role.6
Therefore, these suspected reasons by Henderson cannot explain the difference between the conclusions of our study and that of Praagman et al’s. In the most comprehensive analysis thus far that specifically examined effects of replacing SFAs with polyunsaturated fatty acids in a pooling project of 11 cohort studies, the conclusion is that substitution of polyunsaturated fats for SFAs may help prevent coronary heart disease.7
Henderson mentioned that epidemiological findings need to “be interpreted strictly in the light of higher quality experimental evidence”. In the light of a recent World Health Organization review of 91 clinical trials that demonstrated strong adverse effects of SFAs on blood lipids compared to monounsaturated fatty acids or polyunsaturated fatty acids,8 our findings of substitution analyses are in line with evidence from the quality experimental studies. The current United States Department of Agriculture Dietary Guidelines has emphasized the importance of improving the quality of the whole diet. Eating a healthful diet consisted of whole grains, fruits and vegetables, nuts, legumes, fish, and healthful vegetable oils (except tropical oils high in SFAs) is likely the best approach to reducing overall SFA intake. Lastly, the effects of dietary SFAs may differ from those of circulating SFAs because human body can synthesize SFAs from carbohydrates.
1. Praagman J, Beulens JW, Alssema M, Zock PL, Wanders AJ, Sluijs I, et al. The association between dietary saturated fatty acids and ischemic heart disease depends on the type and source of fatty acid in the European Prospective Investigation into Cancer and Nutrition-Netherlands cohort. The American journal of clinical nutrition. 2016; 103(2): 356-65.
2. Wang DD, Li Y, Chiuve SE, Stampfer MJ, Manson JE, Rimm EB, et al. Association of Specific Dietary Fats With Total and Cause-Specific Mortality. JAMA Intern Med. 2016; 176(8): 1134-45.
3. Willett W. Implications of Total Energy Intake for Epidemiologic Analyses. Nutritional epidemiology. 3rd ed. Oxford: Oxford University Press; 2013.
4. Patrick DL, Cheadle A, Thompson DC, Diehr P, Koepsell T, Kinne S. The validity of self-reported smoking: a review and meta-analysis. American journal of public health. 1994; 84(7): 1086-93.
5. Tomeo CA, Rich-Edwards JW, Michels KB, Berkey CS, Hunter DJ, Frazier AL, et al. Reproducibility and validity of maternal recall of pregnancy-related events. Epidemiology. 1999; 10(6): 774-7.
6. Praagman J, de Jonge EA, Kiefte-de Jong JC, Beulens JW, Sluijs I, Schoufour JD, et al. Dietary Saturated Fatty Acids and Coronary Heart Disease Risk in a Dutch Middle-Aged and Elderly Population. Arterioscler Thromb Vasc Biol. 2016; 36(9): 2011-8.
7. Jakobsen MU, O'Reilly EJ, Heitmann BL, Pereira MA, Balter K, Fraser GE, et al. Major types of dietary fat and risk of coronary heart disease: a pooled analysis of 11 cohort studies. The American journal of clinical nutrition. 2009; 89(5): 1425-32.
8. Mensink RP. Effects of saturated fatty acids on serum lipids and lipoproteins: a systematic review and regression analysis. Geneva: World Health Organization; 2016.
Competing interests: No competing interests
Several readers including Archer questioned the validity of the food frequency questionnaires (FFQs) in assessing dietary intakes and the value of observational studies. Nutrition epidemiology is a science that studies the relationship between diet and disease risk in human populations. It is challenging to quantify individuals’ diet in an absolutely precise and accurate term, especially when collecting dietary data in large-scale epidemiological studies. However, it is well-established in nutritional epidemiology that discriminating individuals in terms of their diet by ranking intakes in relative terms is sufficient to examine relationships between dietary factors and chronic diseases.2-4 FFQs have been demonstrated to be a useful and valid dietary assessment instrument for the purpose to rank intakes in relative terms when they are appropriately developed and used in rigorously-designed and well-conducted epidemiological studies. Labeling the use of FFQs as pseudo-scientific therefore reflects the ignorance of human nutrition research. Herein we briefly review the strengths and limitations of major dietary assessment instruments and illustrate the evidence supporting the value and validity of well-designed FFQs.
Traditional dietary assessment methods primarily include 24-hour dietary recalls, diet records, and FFQs. The 24-hour recalls are typically conducted by trained interviewers to solicit detailed information about meals participants had in the previous day. This method is one of the primary methods used in the National Health and Nutrition Examination Survey (NHANES) to assess diet. While it helps to estimate the average diet of a population, this method is subject to significant random measurement errors due to day-to-day variation of the diet (unless used repeatedly over an extended period to estimate the diet of an individual), as well as short-term recall error.3 Ascher et al’s work calculated average energy expenditure using an equation that accounts for age, physical activity indicator, weight, height, age, and gender, and then compared it with energy intake estimated from a single 24-hour recall in NHANES,5 essentially ignoring the impact of within-person variability. Diet records are designed to allow trained participants to record diet at the time foods are eaten so that the impact of inaccurate recall can be minimized. Multiple 7 day diet records (7DDR) spanning an extended period of time can reduce day-to-day variability of diet and serve as a reflection of usual or long-term diet. However, multiple diet records administered a few times within a year will result in significant burden to study participants and non-compliance. In addition, recording diet using 7DDRs can lead to changes in diets by the participants.6 Nonetheless, 7DDRs are used in small-scale studies to assess diet or as a gold-standard to validate dietary assessments by other dietary instruments.3
In contrast to 24-hour recalls and 7DDRs, FFQs have the advantages of being less burdensome to participants and being able to assess long-term, usual diet. FFQs are typically composed of a structured food list and a frequency response section for participants to report how often each food was eaten in the past year. The development of the FFQs that we used was built on extensive research regarding the food list, frequency response, portion size, and administration. For example, when considering food items to be included in the FFQs, three criteria were considered: the food has to be consumed often and commonly enough; the food contains a significant amount of nutrient(s); and there is a reasonable variability in food intake. Over four decades of research in the Nurses’ Health Studies (NHS) and Health Professionals Follow-up Study (HPFS), we have demonstrated the validity of using FFQs to assess diet in epidemiological studies.7-9 In a validation study in the NHS (1980-1981), Pearson correlation coefficients (rs) between multiple 7DDRs and FFQs data were 0.49-0.59 for saturated fatty acids (SFAs) and 0.48-0.53 for total fats. In a similar validation study in HPFS (1986-1988), rs were 0.63-0.75 for SFAs and 0.53-0.67 for total fats. In our most recent validation study conducted among participants aged 61-80 years, we still observed good correlations for SFAs, which were 0.69 for total SFAs, 0.70 for 16:0, and 0.68 for 18:0.8 The validity of FFQs in assessing diet was further illustrated using valid biomarkers of diet that are not related to memory. For example, the rs between fatty acid biomarkers and dietary fatty acid intakes were 0.42 for trans fats, 0.41 for long-chain n-3 polyunsaturated fatty acids (PUFAs), and 0.34 for n-6 PUFAs.2 In the most recent validation study, rs between fatty acid biomarkers and FFQ assessments were 0.42 for α-linolenic acid, 0.64 for long-chain n-3 PUFAs, and 0.25 for linoleic acid, and 0.46 for trans fats (unpublished). Of note, these fatty acid biomarkers are also subject to measurement errors that result from laboratory assessments, metabolism within the human body, and influence of other fatty acids that can be synthesized de novo. Moreover, although biomarkers are useful as objective measurements of diet, in comparison with FFQs, biomarkers are too expensive to use as dietary assessment tools in large-scale epidemiological studies and can only help to quantify a limited number of dietary factors.
The validity of the FFQs is further supported by the ability of dietary intakes to predict long-term disease risk. Using trans fats as an example, using both FFQ assessments of trans fat intake and plasma biomarkers of trans fat, we reached the same conclusion that trans fat intake is detrimental to cardiovascular health,10, 11 consistent with randomized clinical trials showing the deleterious effects of trans fats on blood lipids.12
To further alleviate random and systematic measurement errors of FFQs, we applied several strategies in statistical analyses, including excluding participants with existing diseases at study baseline, calculating and using cumulative averages of multiple dietary assessments to represent long-term intake, and stopping updating diet upon the diagnosis of severe diseases to avoid reverse causation bias. We use a prospective study design to ensure that the recall of diet is not influenced by the incidence of disease under investigation.
Because of the unique challenges related to assessing the complex human diet, diet assessments in nutritional epidemiology have been a constant research topic for decades. Many nutritional scientists have dedicated their research to develop new dietary assessment tools and improve traditional methods. FFQs have been considered as the best approach to measuring habitual diet in a large population, in terms of cost, representativeness, and intensity of participant involvement compared to other methods. Based on these considerations, we believe that the accusation that FFQs are not a valid tool for assessing diet is erroneous and misleading.
1. Zong G, Li Y, Wanders AJ, Alssema M, Zock PL, Willett WC, et al. Intake of individual saturated fatty acids and risk of coronary heart disease in US men and women: two prospective longitudinal cohort studies. BMJ. 2016; 355: i5796.
2. Willett W. Reproducibility and Validity of Food Frequency Questionnaires. Nutritional epidemiology. 3rd ed. Oxford: Oxford University Press; 2013. p. ix, 529 p.
3. Satija A, Yu E, Willett WC, Hu FB. Understanding nutritional epidemiology and its role in policy. Adv Nutr. 2015; 6(1): 5-18.
4. Hebert JR, Hurley TG, Steck SE, Miller DR, Tabung FK, Peterson KE, et al. Considering the value of dietary assessment data in informing nutrition-related health policy. Adv Nutr. 2014; 5(4): 447-55.
5. Archer E, Hand GA, Blair SN. Validity of U.S. nutritional surveillance:National Health and Nutrition Examination Survey caloric energy intake data, 1971-2010. PloS one. 2013; 8(10): e76632.
6. Willett W. 24-Hour Recall and Diet Record Methods. Nutritional epidemiology. 3rd ed. Oxford: Oxford University Press; 2013. p. ix, 529 p.
7. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. American journal of epidemiology. 1985; 122(1): 51-65.
8. Yuan C, Spiegelman D, Rimm EB, Rosner BA, Stampfer MJ, Barnett JB, et al. Nutrient Validation Utilizing Self-Reported Dietary Assessment Methods in Women’s Lifestyle Validation Study. American journal of epidemiology. 2016 ( in press).
9. Rimm EB, Giovannucci EL, Stampfer MJ, Colditz GA, Litin LB, Willett WC. Reproducibility and validity of an expanded self-administered semiquantitative food frequency questionnaire among male health professionals. American journal of epidemiology. 1992; 135(10): 1114-26; discussion 27-36.
10. Sun Q, Ma J, Campos H, Hankinson SE, Manson JE, Stampfer MJ, et al. A prospective study of trans fatty acids in erythrocytes and risk of coronary heart disease. Circulation. 2007; 115(14): 1858-65.
11. Willett WC, Stampfer MJ, Manson JE, Colditz GA, Speizer FE, Rosner BA, et al. Intake of trans fatty acids and risk of coronary heart disease among women. Lancet. 1993; 341(8845): 581-5.
12. Mensink RP, Zock PL, Kester AD, Katan MB. Effects of dietary fatty acids and carbohydrates on the ratio of serum total to HDL cholesterol and on serum lipids and apolipoproteins: a meta-analysis of 60 controlled trials. The American journal of clinical nutrition. 2003; 77(5): 1146-55.
Competing interests: No competing interests
Re: Concerns regarding the Validity of Observational Study Findings, Overlapping of Results with Previous Publications, and Conflict of Interest.
Wheatley questioned the value of observational studies, and yet used results of meta-analyses and systematic reviews which summarized data of observational studies to refute our study. Wheatley ignored a few critical issues here. First, the quality and validity of meta-analysis results depend on the quality of individual studies included in the meta-analysis. One of the primary limitations of meta-analyses is that data analysis strategy is not standardized across studies. It is critical to understand that the health effects of saturated fatty acids (SFAs) depend on which other macronutrients they replace in an isocaloric setting. In the meta-analyses, individual studies often did not specify the macronutrients that SFAs replaced. As a result, the associations for SFAs represent mixed effects of replacing various macronutrients with diverse health effects.1-4 Pooling projects, in which a priori statistical analysis plan is followed by all contributing studies, are not subject to this limitation and can thus provide more meaningful evidence. In the largest pooling project consisting of 11 prospective cohort studies, investigators demonstrated clear benefits of replacing SFAs with polyunsaturated fatty acids (PUFAs).5 Wheatley also ignored that fact that our study findings are in line with evidence from clinical trials that tested the effects of SFAs on blood lipids.6 Both observational studies and clinical trials have their own strengths and limitations and often well-conducted observational studies and trials generate evidence that is complementary to each other.
Wheatley failed to notice that the aim of our study is to investigate individual SFAs and coronary heart disease (CHD) risk, which is different from two previous investigations in the cohorts: one compared total SFAs with other types of fats/carbohydrates regarding associations with CHD risk, and the other investigated types of fats with both total and cause-specific mortality. Wheatley further ignored that 6-8% risk reduction was based on 1% energy from SFAs replaced by healthy energy sources, when the average % energy of SFAs during follow-up was 9%-13%. For an average individual with 2100 kcal/day energy intake and 15% of energy from SFAs, reducing SFA intake to 10% of energy will lead to 27%-34% lower CHD risk. Despite our prospective study design, Wheatley criticized our study for not meeting the assumption of temporality. He also preferred to observe a straight line to imply the existence of dose-response relationship, which is not the most physiologically plausible associations between nutrients and health outcomes per se.
Wheatley correctly stated that circulating SFAs can come from different sources, including diet, lipogenesis from carbohydrates, and microbiota metabolism, although he erroneously asserted that dietary SFAs and circulating SFAs represent the same exposures. Wheatley argued that it is incorrect to conclude that all SFAs are harmful, but our analysis focused on the major dietary saturated fatty acids, rather than all SFAs. 15:0 and 17:0 are minor fatty acids (<2% of total fat) even in their major food sources, such as fat from dairy, beef, veal, lamb, and mutton.7 According to Praagman et al., the sum of 15:0 and 17:0 contributed to 0.3% of total energy (compared to the 15% by total SFAs) among participants from the Netherlands whose animal fat intake is higher.8 In this regard, we have focused on even-chain SFAs as what was indicated in the title. The same is true for short- and medium-chain SFAs, since 4:0, 6:0, 8:0, and 10:0 collectively contributed to about 0.5% of total energy.8, 9
It is important to examine and understand the health effects of both nutrients and foods. The current dietary recommendation encourages people to avoid certain foods that are high in SFAs, such as animal fats, red meats and processed meats, full-fat dairy, and tropical oils, rather than olive oils, nuts, vegetable cooking oils, and other healthful sources of fats or oils.
We would like to take this opportunity to address the concern of conflict of interest raised by some readers. We have acknowledged in our publication that Unilever funded Dr. Geng Zong with a postdoctoral fellowship to conduct the analysis. Our dietary and lifestyle data collection and disease follow-up and ascertainment are exclusively supported by funding from the NIH. As a global company, Unilever has many products with very diverse profiles of fats. Our previous research on trans fats showed that many of Unilever’s products, especially margarine, were harmful to cardiovascular health. The company responded quickly to scientific findings by leading the industrial efforts of removing trans fats from margarine and other products. Apparently, the company is not wedded to any specific types of fats. The collaborations with Unilever scientists are academic. Dr. Peter Zock, for example, has published extensively on the adverse effects of SFAs and trans fats on blood lipids and cardiovascular health. He is a co-author of a landmark meta-analysis and systematic review that summarized evidence of dietary fats on blood lipids and suggested that individual SFAs may have different effects on blood lipids.6, 10 This is the primary rationale for us to perform the current investigation.
1. Wang DD, Li Y, Chiuve SE, Stampfer MJ, Manson JE, Rimm EB, et al. Association of Specific Dietary Fats With Total and Cause-Specific Mortality. JAMA Intern Med. 2016; 176(8): 1134-45.
2. Li Y, Hruby A, Bernstein AM, Ley SH, Wang DD, Chiuve SE, et al. Saturated Fats Compared With Unsaturated Fats and Sources of Carbohydrates in Relation to Risk of Coronary Heart Disease: A Prospective Cohort Study. Journal of the American College of Cardiology. 2015; 66(14): 1538-48.
3. de Souza RJ, Mente A, Maroleanu A, Cozma AI, Ha V, Kishibe T, et al. Intake of saturated and trans unsaturated fatty acids and risk of all cause mortality, cardiovascular disease, and type 2 diabetes: systematic review and meta-analysis of observational studies. BMJ. 2015; 351: h3978.
4. Chowdhury R, Warnakula S, Kunutsor S, Crowe F, Ward HA, Johnson L, et al. Association of dietary, circulating, and supplement fatty acids with coronary risk: a systematic review and meta-analysis. Ann Intern Med. 2014; 160(6): 398-406.
5. Jakobsen MU, O'Reilly EJ, Heitmann BL, Pereira MA, Balter K, Fraser GE, et al. Major types of dietary fat and risk of coronary heart disease: a pooled analysis of 11 cohort studies. The American journal of clinical nutrition. 2009; 89(5): 1425-32.
6. Katan MB, Zock PL, Mensink RP. Effects of fats and fatty acids on blood lipids in humans: an overview. The American journal of clinical nutrition. 1994; 60(6 Suppl): 1017S-22S.
7. Hooper L, Martin N, Abdelhamid A, Davey Smith G. Reduction in saturated fat intake for cardiovascular disease. The Cochrane database of systematic reviews. 2015; (6): CD011737.
8. Praagman J, Beulens JW, Alssema M, Zock PL, Wanders AJ, Sluijs I, et al. The association between dietary saturated fatty acids and ischemic heart disease depends on the type and source of fatty acid in the European Prospective Investigation into Cancer and Nutrition-Netherlands cohort. The American journal of clinical nutrition. 2016; 103(2): 356-65.
9. Zong G, Li Y, Wanders AJ, Alssema M, Zock PL, Willett WC, et al. Intake of individual saturated fatty acids and risk of coronary heart disease in US men and women: two prospective longitudinal cohort studies. BMJ. 2016; 355: i5796.
10. Mensink RP, Zock PL, Kester AD, Katan MB. Effects of dietary fatty acids and carbohydrates on the ratio of serum total to HDL cholesterol and on serum lipids and apolipoproteins: a meta-analysis of 60 controlled trials. The American journal of clinical nutrition. 2003; 77(5): 1146-55.
Competing interests: No competing interests
Vos first paid attention to the distribution of confounding factors (hypercholesterolemia, use of multivitamin, smoking, and physical activity) according to saturated fatty acid (SFA) quintiles, and suggested that those with lower SFA intake were healthier. This explained the attenuation of positive associations in model 2 of Table 2 when these confounding factors were adjusted, but positive association remained in this model. Although we did not fully understand his comments on the distribution of macronutrients, but carbohydrates from whole grains and plant proteins turned to be lower among people with high SFA intake in both cohorts, whereas non-plant proteins were higher. This is expected because animal foods are major sources of SFAs.
In U.S. diet, n-6 polyunsaturated fatty acids (PUFAs) accounted for the vast majority of total PUFA intake. We and others have reported inverse associations between dietary intakes and circulating biomarkers of linoleic acid with lower coronary heart disease (CHD) risk in prospective cohorts studies.1 In our most recent study regarding types of fats with mortality, n-6 PUFAs have been associated with lower total and cardiovascular disease mortality when replacing SFA intake, and hazard ratios (95% confidence intervals) were 0.93(0.91-0.96) and 0.89 (0.85-0.94), respectively, for 5% energy substitution. Relative risks for the same replacement by n-3 PUFAs were 0.95 (0.93-0.96) and 1.01 (0.97-1.05), respectively.2
1. Lucas M. Letter by Lucas Regarding Articles, "Dietary Linoleic Acid and Risk of Coronary Heart Disease: A Systematic Review and Meta-Analysis of Prospective Cohort Studies" and "Circulating Omega-6 Polyunsaturated Fatty Acids and Total and Cause-Specific Mortality: The Cardiovascular Health Study". Circulation 2015; 132(3): e21.
2. Wang DD, Li Y, Chiuve SE, et al. Association of Specific Dietary Fats With Total and Cause-Specific Mortality. JAMA Intern Med 2016; 176(8): 1134-45.
Competing interests: No competing interests
Concerns about a study linking intake of individual saturated fatty acids with risk of coronary heart disease
We read the article by Zong et. al.,1 “Intake of individual saturated fatty acids and risk of coronary heart disease in US men and women: two prospective longitudinal cohort studies” in BMJ with great interest. We appreciate the effort required to perform such a large prospective cohort study on the association between intake of individual saturated fatty acids (SFA) and the risk of coronary heart disease (CHD). Here, we would like to make four points that may have significant consequences for the results of this cohort study.
First, the initiation point of the two cohort studies used in the analysis by Zong et al. differed by two years. Are there any reasons for this difference? We are concerned that high saturated fat intake between 1984 to 1986 in the Nurses’ Health Study could have positively distorted the results. Could the authors perform a sensitivity analysis by excluding the data from 1984 to 1986 in the Nurses’ Health Study?
Second, the authors have not included important information about the flow of participants throughout the course of the study. In particular, the authors need to disclose the number of participants who dropped out of the cohorts. Also, because diabetes mellitus, stroke, and cancer are associated with high intake of SFA2-5, it is important that the authors indicate the number of participants who reported a diagnosis of these conditions during the follow-up periods.
Third, evaluation of dietary information with regard to participants who reported a diagnosis of diabetes, stroke, or cancer during the study period was performed using data obtained before disease onset to prevent reverse causation bias. However, this method might lead to overestimation of the association between CHD and SFA. We suggest that the authors should perform a sensitivity analysis by including the SFA data across the entire study period.
Finally, we would like to point out a number of additional potential confounding factors, including drug use (statins, antihypertensive agents, bisphosphonates, and contraceptive pills) and comorbidities (chronic kidney disease, osteoporosis, familial hypercholesterolemia) 6-8. Although these were not included in the study analysis, we suggest that the authors readjust their analysis to include these factors, if possible.
1 Zong G, Li Y, Wanders AJ, et al. Intake of individual saturated fatty acids and risk of coronary heart disease in US men and women: two prospective longitudinal cohort studies. BMJ 2016;355:i5796. doi:10.1136/bmj.i5796
2 Forouhi NG, Koulman A, Sharp SJ, et al. Differences in the prospective association between individual plasma phospholipid saturated fatty acids and incident type 2 diabetes: the EPIC-InterAct case-cohort study. Lancet Diabetes Endocrinol 2014;2:810–8. doi:10.1016/S2213-8587(14)70146-9
3 Yamagishi K, Iso H, Kokubo Y, et al. Dietary intake of saturated fatty acids and incident stroke and coronary heart disease in Japanese communities: the JPHC Study. Eur Heart J 2013;34:1225–32. doi:10.1093/eurheartj/eht043
4 Kurahashi N, Inoue M, Iwasaki M, et al. Dairy product, saturated fatty acid, and calcium intake and prostate cancer in a prospective cohort of Japanese men. Cancer Epidemiol Biomarkers Prev 2008;17:930–7. doi:10.1158/1055-9965.EPI-07-2681
5 Thiébaut ACM, Kipnis V, Chang S-C, et al. Dietary fat and postmenopausal invasive breast cancer in the National Institutes of Health-AARP Diet and Health Study cohort. J Natl Cancer Inst 2007;99:451–62. doi:10.1093/jnci/djk094
6 Tonelli M, Lloyd A, Clement F, et al. Efficacy of statins for primary prevention in people at low cardiovascular risk: a meta-analysis. CMAJ 2011;183:E1189–202. doi:10.1503/cmaj.101280
7 Siegerink B, Maino A, Algra A, et al. Hypercoagulability and the risk of myocardial infarction and ischemic stroke in young women. J Thromb Haemost 2015;13:1568–75. doi:10.1111/jth.13045
8 Pittman CB, Davis LA, Zeringue AL, et al. Myocardial infarction risk among patients with fractures receiving bisphosphonates. Mayo Clin Proc 2014;89:43–51. doi:10.1016/j.mayocp.2013.08.021
Competing interests: No competing interests
Individual saturated fatty acids and risk of coronary heart disease in US men and women study: possible bias
Several responses to the article highlight the fact that the study suffers from bias on several counts. We also re-iterate the bias related to the validity of the food frequency questionnaire which is used as the primary source of fatty acid intake for evaluating against the occurrence of coronary heart disease.
The authors have indicated that there is a correlation with the food frequency questionnaire and seven day diet records and 24-hr recall. The analysis was statistically supported, however, we should also note that there is no absolute correlation i.e. 0.71, 0.69 which also points towards the lacunae in the data collection process. This is further strengthened by the fact that dietary intake pattern was evaluated four yearly, which is a long period and there is a likelihood that dietary pattern will have changed during this period. Usual intake since four years does not, thus, reflect the actual intake pattern. This limitation may actually influence the study results. The recall bias here acts as a strong confounding variable and it may not be corrected by the statistical modelling. Hence, the study results should not be considered as a pointer towards causal association but with caution.
Competing interests: No competing interests
The recent paper by Zong et al should be retracted because their methods are pseudo-scientific and their data and results are inadmissible as scientific evidence.
In a series of publications,1-4 my colleagues and I empirically refuted the memory-based dietary assessment methods (M-BMs; e.g., food frequency questionnaires, FFQs) employed by Zong et al. We unequivocally demonstrated that M-BM-data were physiologically implausible (i.e., incompatible with life 5) and therefore could not be representative of dietary consumption. More importantly, our work established that because M-BM-data were derived from non-empirical phenomena (i.e., memories), the errors associated with M-BM-data were non-quantifiable due to false memories, omissions (i.e., forgetting), intentional misreporting (i.e., lying), and gross misestimation. It is important to note that the distinction between quantifiable and nonquantifiable errors is the demarcation between valid scientific (i.e., falsifiable) data and data that are pseudo-scientific and “inadmissible.”2
Furthermore, FFQs lack any semblance of face validity (i.e., FFQs cannot possibly measure what they were intended to measure.). During the study period there were over 100,000 items in the US Food Supply available for consumption by Zong et al.’s participants. It defies logic and common sense to speculate that data from FFQs with <150 total items (i.e., a list with <0.2% of available foods and beverages) could be representative of actual dietary consumption.
Specifically, the USDA Food Composition Databases lists >77,000 items and a simple search using the terms provided by Zong et al. yields results that are orders of magnitude greater than the number of items on their FFQs. For example, using the search term “oatmeal” returns 522 items; the term “brown rice” produces 339 items, and “whole wheat bread” produces 721 items. Therefore, each item on Zong et al.'s FFQs returns hundreds of possible foods and beverages that vary in energy and nutrient composition. Nevertheless, Zong et al. credulously assumed their FFQs were representative of dietary consumption. This is inane. Given the inestimable number of physiologic interactions between the <0.2% of the foods and beverages that were available for consumption and actually on the FFQs, and the 99.8% of potential dietary components that were not on the FFQs (and therefore not estimated nor quantified), Zong et al.’s use of FFQs is pseudo-scientific, illogical, and lacking in both scientific and common sense.
Stated simply, Zong et al.'s data, results, and conclusions are meaningless. As such, their paper should be retracted.
Edward Archer, PhD, MS
Chief Science Officer
1. Archer E, Hand GA, Blair SN. Validity of U.S. nutritional surveillance: National Health and Nutrition Examination Survey caloric energy intake data, 1971-2010. PLoS One 2013;8(10):e76632. doi: 10.1371/journal.pone.0076632
2. Archer E, Pavela G, Lavie CJ. The Inadmissibility of What We Eat in America and NHANES Dietary Data in Nutrition and Obesity Research and the Scientific Formulation of National Dietary Guidelines. Mayo Clinic Proceedings 2015;90(7):911-26. doi: 10.1016/j.mayocp.2015.04.009 [published Online First: 2015/06/14]
3. Archer E, Pavela G, Lavie CJ. A Discussion of the Refutation of Memory-Based Dietary Assessment Methods (M-BMs): The Rhetorical Defense of Pseudoscientific and Inadmissible Evidence. Mayo Clinic Proceedings 2015;90(12):1736-38.
4. Archer E, Blair SN. Reply to LS Freedman et al. Advances in nutrition (Bethesda, Md) 2015;6(4):489-90. doi: 10.3945/an.115.009183 [published Online First: 2015/07/17]
5. Ioannidis JPA. Implausible results in human nutrition research. BMJ 2013;347 doi: 10.1136/bmj.f6698
Competing interests: No competing interests
The latest paper by Harvard TH Chan School of Public Health/Unilever observed 10 non-significant associations and 6 very small statistically significant associations between coronary heart disease and energy substitution for individual saturated fatty acids and alternative nutrients.
This followed searches for patterns in two prospective cohort studies (the Nurses' Health Study and the Health Professionals Follow-up Study), from which more than 30 papers for 2016 alone have been generated with Walter C Willet and Frank B Hu as co-authors.
Surely an institution as esteemed as Harvard, would know that epidemiological studies can merely suggest associations, which should then be tested for causation with randomised controlled trials (RCTs). On the subject of dietary fat, epidemiology and dietary interventions have been running in parallel since the 1950s. If the observations being repeatedly made by Harvard had any value, they would have been supported by RCTs and they have not. So when will Harvard admit this and stop this endless stream of non-sense?
Competing interests: ZH writes and publishes books and web content in the field of diet and health.