Summary

Introduction

Cardiovascular disease (CVD), specifically coronary heart disease and stroke, is the leading cause of death and disability-adjusted life years for all regions of the world (Roth et al., 2017). In 2015, there were an estimated 422.7 million cases of CVD and 17.92 million CVD deaths globally, which accounted for approximately one third of total deaths (Roth et al., 2017). Between 1990 and 2015, the age-standardized global death rate attributable to CVD dropped by 24%, mainly driven by declines in all high-income and some middle-income countries; however, the total number of CVD deaths increased by 41%, mostly attributable to population growth and aging (Roth et al., 2017). Currently, approximately 80% of CVD deaths occur in low- and middle-income countries (Roth et al., 2017), where health care infrastructures and access to treatment are limited. Therefore, effective approaches for the prevention of CVD are crucial to reduce the future disease burden and improve population health.

Among the many established risk factors for CVD, diet plays an important role (Forouzanfar et al., 2015). In the past several decades, numerous studies have enhanced our understanding of the relationship between diet and cardiovascular health. However, divergent results have created confusion among patients, health professionals, and the general public. A recent study from the Prospective Urban and Rural Epidemiology (PURE) cohort (Dehghan et al., 2017) has generated sensational headlines in the media and stirred debate in the scientific community. This study aimed to clarify the relationship between macronutrient intake and mortality, but the findings appeared to contradict those from previous studies and prevailing dietary recommendations and guidelines. By analyzing self-reported dietary data from more than 135,000 people in 18 countries and following the participants over a 7-year period, the authors concluded that “high carbohydrate intake was associated with higher risk of total mortality, whereas total fat and individual types of fat were related to lower total mortality.” In this Minireview, we described recent advances in nutritional epidemiologic studies of dietary fats and carbohydrates and cardiovascular health. We then discussed methodological issues in the PURE study. Finally, we proposed some opportunities and challenges moving forward.

Recent advances in epidemiologic study designs

Randomized clinical trials (RCTs), particularly those that are high-quality and double-blind, are the gold standard for determining causation between an exposure and an outcome and thus are the cornerstone for medical guidelines and health policy (Figure 1). A double-blind RCT can minimize or eliminate confounding, which is a major challenge in observational studies. Ideally, RCTs should be used to examine the effect of a nutrient or food on cardiovascular events, in the same way as required for approving new drugs and medical devices. In cases with supplements, RCTs are a practical choice. For example, the Alpha Omega Trial evaluated the effects of n−3 fatty acid supplementation (EPA/DHA and/or alpha-linolenic acid) and cardiovascular events among 4837 patients with myocardial infarction (Kromhout et al., 2010).

Open in a separate window

Figure 1

The typical hierarchy of study designs based on ability to infer causality

However, there are multiple reasons why RCTs are not feasible for many nutrients and foods, including macronutrients. To achieve valid results, the difference in intake levels between the treatment and control groups must be biologically meaningful and sustained for a sufficient period. For cardiovascular events, the intervention period would need to be at least several years. Additionally, a large sample size is needed to obtain sufficient statistical power. It is also difficult to maintain a high adherence rate when assigning participants to a diet that differs from their usual diet. For example, the Women’s Health Initiative (WHI) Randomized Controlled Dietary Modification Trial (Howard et al., 2006) studied the effects of decreasing total dietary fat on cancer and CVD outcomes. The study enrolled 48,835 women aged 50–79 years between 1993 and 1998 and randomly assigned them to a low-fat diet (20% of energy intake from fat) or to maintain their usual high-fat diet (37% of energy intake from fat). After a mean follow-up period of 8.1 years, most people in the low-fat group did not achieve the goal to reduce fat to 20% of energy intake. Overall, no significant effect on CVD outcomes was observed. An additional challenge of dietary interventions is that participants cannot be blinded from dietary assignment because they need to change their food choices to achieve the intervention diet. In the WHI, in addition to reducing fat, the intervention group reported significant increases in fiber, vegetables and fruits, total and whole grains, and soy. Thus, it is difficult to isolate the effects of specific macronutrients. Despite these challenges, well-conducted RCTs, if feasible, are essential to examine the effects of nutrients, foods, and dietary patterns on health outcomes and develop evidence-based recommendations. For example, the DASH trial (Dietary Approaches to Stop Hypertension) (Sacks et al., 2001) and the PREDIMED trial (Prevención con Dieta Mediterránea) (Estruch et al., 2013) have provided strong evidence to support the role of healthy dietary patterns in primary prevention of hypertension through adherence to the DASH dietary pattern and cardiovascular disease through adherence to the Mediterranean-style dietary pattern.

Due to the constraints of RCTs, prospective cohort studies are the preferred observational study design to examine the association between diet and CVD (Satija et al., 2015) (Figure 1). With this design, researchers follow a large population of participants for a certain period of time and do not intervene to change dietary habits. In many cohort studies, dietary intake data are collected at baseline when participants are enrolled and CVD and other health outcomes are assessed during follow-up. Dietary data are typically calculated from validated food-frequency questionnaires (FFQs) or 24-hour dietary recalls and are sometimes collected from multiple-day dietary records. The advantages and disadvantages of different dietary measurement methods are extensively discussed elsewhere (Satija et al., 2015; Willett, 2013). Self-reported data are subject to both random and systematic measurement errors. The random errors are typically present in 24-hour recalls and derived from the large day-to-day variation in dietary intake in free-living individuals, which could substantially attenuate the strength of the associations (e.g., correlation coefficients, regression coefficients and relative risks) (Willett, 2013). Therefore, studies often use repeated 24-hour recalls on nonconsecutive days to reduce this type of error (Willett, 2013). FFQs typically have fewer random errors but are subject to systematic errors due to omission of food items and inaccurate nutrient composition databases. One common solution is to use energy-adjusted nutrients in the analysis. Because both the specific nutrients and total energy are derived from the same foods in the FFQ, the measurement errors (over-reporting or under-reporting) in the nutrients and total energy are likely to be correlated (Willett, 2013). Stastistically, the correlated errors can be reduced by adjusting for total energy intake. To address concerns that single-time measures do not represent long-term patterns, more studies are now collecting repeated dietary measures during follow-up. An example is the Nurses’ Health Study, where diet was evaluated every four years using validated semi-quantitative FFQs during more than three decades of follow-up (Bao et al., 2016). Furthermore, repeated dietary measures with validation data also reduce the impact of measurement errors and misclassifications. Detailed methods of measurement error corrections are described elsewhere (Spiegelman, 2016).

Results from prospective cohort studies and RCTs may not be always consistent. For example, a number of cohort studies have reported that antioxidant vitamins (e.g., vitamin C and E) are inversely associated with CVD (Knekt et al., 2004) but most RCTs have indicated that supplementation with these vitamins does not provide beneficial effects (Myung et al., 2013). This discrepancy may be due to the fact that RCTs usually select high-risk populations and the duration of the supplementation is relatively short (3–5 years) compared to observational studies. In addition, residual confounding by socioeconomic status and a more healthy diet and lifestyle in observational studies may partially explain an inverse association between antioxidant vitamins and risk of CVD (Lawlor et al., 2004). It is also likely that antioxidants from high dose supplements have different effects than those from dietary sources where these vitamins come together with other beneficial nutrients such as fiber, minerals, and polyphenols.

Confounding needs to be carefully considered in all observational studies, including prospective cohorts, as participants with different dietary patterns may also differ in other characteristics. All relevant confounders should be carefully identified before study initiation and systematically collected during the study. Residual confounding can be reduced by multivariate regression models with proper adjustment for confounders as well as by stratified analyses or restricting analyses to a specific subgroup. Compared to other observational methods, such as retrospective case-control or cross-sectional studies, prospective cohort studies are less affected by reverse causation, recall bias and selection bias. Reverse causation is reduced because exposures are measured years before the ascertainment of the outcomes. The participants are disease-free at baseline, which reduces recall bias. Incident cases are documented through follow-up, which minimizes selection bias because the cases and non-cases are obtained from the same population. However, biases due to non-differential loss to follow-up and other reasons still need to be considered in cohort studies.

Many prospective cohort studies have been conducted in the USA and Europe, for example, the Framingham Heart Study (Mahmood et al., 2014), Nurses’ Health Study (Bao et al., 2016), and UK Biobank study (Collins, 2012). More cohort studies are now being conducted in developing countries and in populations undergoing rapid nutrition and economic transitions. An example is the China Kadoorie Biobank study which recruited 512,891 adults in 10 regions of China beginning in 2004 (Chen et al., 2011). The study aims to investigate lifestyle, environmental, and genetic determinants of chronic diseases in the Chinese population..

Recently, many cohort studies have formed consortia or initiated pooling projects to increase sample size, evaluate heterogeneity across different populations, and increase generalizability of their findings (Smith-Warner et al., 2006). Multi-center cohort studies are another type of international collaboration which are conducted in various locations under the same or similar protocols. One example is the European Prospective Investigation into Cancer and Nutrition (EPIC) study (Bingham and Riboli, 2004), which recruited half a million participants across 10 European countries and followed them for almost 15 years.

Another advancement in nutritional epidemiology is the ability to measure dietary biomarkers in stored biospecimens (blood, urine or other biofluids) in cohort studies and analyze their relationship with disease outcomes. This may be the only option for estimating intake for nutrients that are not available in food composition tables or that vary widely within the same foods due to variability in production, storage, processing, or preparation. Biomarkers are established for some nutrients (e.g., protein intake by urinary nitrogen, salt intake by urinary sodium/potassium, and essential dietary fatty acids in plasma or erythrocytes); however, there are no specific biomarkers for most individual foods or for total dietary fat or carbohydrate. Additionally, while biomarker measurements are objective and avoid recall bias, these short-term measures may not adequately represent long-term dietary exposure. Therefore, dietary biomarker measurements are complementary to, rather than a replacement for, the self-reported methods in nutritional epidemiologic studies (Satija et al., 2015).

Cardiovascular health effects of dietary fat and carbohydrate

Methodological issues in the PURE study

When interpreting the PURE study findings, it is crucial to consider the study methodology as well as the importance of macronutrient quality and dietary patterns in the context set above. Beginning in 2003, the PURE study collected self-reported dietary data using FFQs at baseline from 135,000 participants in 18 countries. After tracking participants’ health over an average of seven years, researchers found that those with the highest intake of dietary fat (35% of daily calories on average) were 23% less likely to have died than those with the lowest intake (10% of daily calories on average). Oppositely, for carbohydrates, those with the highest intake (77% of daily calories on average) were 28% more likely to have died than those with the lowest intake (46% of daily calories on average). The authors concluded that higher fat intake was associated with decreased mortality while higher carbohydrate intake was associated with increased mortality. Although this study had a large sample size and included participants from many countries, the results need to be interpreted with caution due to the following methodological issues.

First, grouping all carbohydrates together is overly simplistic because different types of carbohydrates have opposing effects on health. Studies have consistently demonstrated that refined carbohydrates are associated with increased risk whereas whole grains are associated with decreased risk of type 2 diabetes (Aune et al., 2013; Hu et al., 2012) and CVD (Zong et al., 2016).

Second, in many developing countries such as those included in the PURE study, people subsist on a “poverty diet” which contains little fat and many refined carbohydrates, such as white rice and refined wheat products. Therefore, when attempting to compare high versus low saturated fat intake, the study actually compared participants with high saturated fat intake to those with high refined carbohydrate intake. The authors reported that saturated fat was not significantly associated with CVD, which is not surprising given that saturated fat was essentially compared with refined carbohydrates in the analyses. In addition, the association between carbohydrate intake and mortality risk was mainly driven by non-CVD mortality, such as deaths from respiratory, liver and infectious diseases. Low-income people in developing countries tend to bear the burden of these types of diseases, as well as consume high-carbohydrate diets. Therefore, an association between high carbohydrate intake and non-CVD mortality is likely to be confounded by poverty, food insecurity, undernutrition and lack of medical care access.

Third, the PURE study population was highly heterogeneous, consisting of participants from rural and urban areas of 18 countries with different levels of socioeconomic development. The dietary patterns, food choices, lifestyle habits, cultural backgrounds, and disease patterns differed considerably across various countries and regions. Therefore, country-specific analyses, region-specific analyses and urban-rural stratified analyses are needed to better understand the observed associations.

Fourth, the macronutrient compositions presented in the PURE study as the healthiest (roughly 50% carbohydrate, 30–35% fat, 15–20% protein) are very close to current dietary recommendations (Food and Agriculture Organization of the United Nations, 2010). Therefore, the PURE study’s conclusion that “global dietary guidelines should be reconsidered in light of these findings” is not supported by its own data.

Finally, the dietary intake data for the Chinese participants, who made up one third of the PURE study sample, were not consistent with data from other studies conducted in China. Fat intake in China has been increasing since the 1990s, coupled with a decrease in carbohydrate intake. In parallel with these dietary changes, rates of obesity, hypertension, and diabetes have increased considerably (Figure 2). National surveys reported that the Chinese population currently consumes an average of 30% of their daily calories from fat. However, the PURE study reported an average total fat intake of 17.7% in the Chinese participants. The PURE study used the same FFQ as the 2002 China National Nutrition and Health Survey. This FFQ has been validated among 271 healthy adults between 1999 and 2001 by comparing the data from the FFQ and 18 days of 24-hour recalls throughout one year. In that validation study, the energy contribution from carbohydrate, fat and protein was 58.8%, 27.3% and 12.9%, respectively. The large discrepancies in dietary data in the PURE study compared to the FFQ validation study and other Chinese national surveys casts doubt on the validity of the PURE study’s dietary assessment procedures. Surprisingly, vegetable oils, which are major sources of dietary fat in Chinese adults, were not on the PURE study’s list of top food contributors for dietary fat.

Open in a separate window

Figure 2

Trends in energy intake from carbohydrate and fat and corresponding increase in obesity-related chronic diseases in China (1992–2012).

Data on percentage of carbohydrate and fat intake as well as the prevalence of overweight and obesity are from Chinese nutrition and health surveillance reports in 2002 (National Health and Family Planning Commission of the People’s Republic of China, 2004) and 2012 (Chang and Wang, 2016). The surveillance reports used the Chinese cut-off point defining overweight as 24≤BMI<28, and obesity as BMI ≥28. The data extracted here were among adults (>18 years). The prevalence of hypertension in adults was 13.6% in 1991 (Tao et al., 1995), 17.7% in 2002 (Wu et al., 2008), and 33.5% in 2010 (Li et al., 2012). The prevalence of diabetes in adults was 2.5% in 1994 (Pan et al., 1997), 5.5% in 2000 to 2001 (Gu et al., 2003), and 10.9% in 2013 (Wang et al., 2017).

National Health and Family Planning Commission of the People’s Republic of China (2004). The nutrition and health status of the Chinese people [Chinese]. Available at http://www.moh.gov.cn/wsb/pzcjd/200804/21290.shtml.

Chang, J.L., Wang, Y. (2016). Chinese Nutrition and Health Surveillance: 2010–2013 Report [Chinese] (Beijing, China: Peking University Medical Press).

Tao, S., Wu, X., Duan, X., Fang, W., Hao, J., Fan, D., Wang, W., Li, Y. (1995). Hypertension prevalence and status of awareness, treatment and control in China. Chin. Med. J. 108, 483–489.

Wu, Y., Huxley, R., Li, L., Anna, V., Xie, G., Yao, C., Woodward, M., Li, X., Chalmers, J., Gao, R., et al. (2008). Prevalence, awareness, treatment, and control of hypertension in China: data from the China National Nutrition and Health Survey 2002. Circulation 118, 2679–2686.

Li, Y.C., Wang, L.M., Jiang, Y., Li, X.Y., Zhang, M., Hu, N. (2012). Prevalence of hypertension among Chinese adults in 2010 [Chinese]. Chin. J. Prev. Med. 46, 409–413.

Pan, X.R., Yang, W.Y., Li, G.W., Liu, J. (1997). Prevalence of diabetes and its risk factors in china, 1994. National diabetes prevention and control cooperative group. Diabetes Care 20, 1664–1669.

Gu, D., Reynolds, K., Duan, X., Xin, X., Chen, J., Wu, X., Mo, J., Whelton, P.K., He, J.; InterASIA Collaborative Group. (2003). Prevalence of diabetes and impaired fasting glucose in the Chinese adult population: international collaborative study of cardiovascular disease in Asia (InterASIA). Diabetologia 46, 1190–1198.

Wang, L., Gao, P., Zhang, M., Huang, Z., Zhang, D., Deng, Q., Li, Y., Zhao, Z., Qin, X., Jin, D., et al. (2017). Prevalence and ethnic pattern of diabetes and prediabetes in China in 2013. JAMA 317, 2515–2523.

In summary, despite the large study population from multiple countries, the PURE results need to be interpreted in the context of the existing literature and various methodological issues in data collection, analysis and interpretation. Overall, the study results did not appear to challenge the existing recommendations that are focused on healthy sources of fats and carbohydrates.

Opportunities and challenges

Conclusions and Future Perspectives

Nutritional epidemiology has greatly enhanced our understanding of the complex relationship between diet and cardiovascular health. Due to the constraints in conducting RCTs with nutrients and foods, large prospective cohort studies have been the predominant nutritional epidemiologic study design. Understanding the unique design and analytic issues in the field is crucial when interpreting findings and determining the validity of findings from different types of studies. Biases and confounding issues need to be carefully considered and the totality of evidence should be thoroughly evaluated to reach a consensus and inform policy. To achieve better cardiovascular health, existing evidence strongly supports an overall healthy dietary pattern that emphasizes quality and healthy sources of fats and carbohydrates rather than the absolute amounts of fats or carbohydrates in the diet. Building a strong evidence base through high-quality observational and intervention studies is critical for effective policy changes, which can greatly improve the food environment and population health. With recent advances in study design, dietary assessment methods, omics technologies, and big data analytics, the field of nutritional epidemiology is evolving rapidly and will continue to play an indispensable role in furthering our understanding of the role of diet in cardiovascular health and in translating nutritional sciences into policies and practice.

Acknowledgments

Footnotes

References