Low Carb or High Carb? Everything in Moderation … Until Further Notice

Russell J. de Souza; Mahshid Dehghan; Sonia S. Anand


Eur Heart J. 2019;40(34):2880-2882. 

In this issue of European Heart Journal, Mazidi et al.[1] report an analysis of the prospective National Health and Nutrition Examination Survey (NHANES) cohort from the USA, set in the context of other prospective studies with a meta-analysis, examining the relationship between lower carbohydrate diets and mortality. From NHANES, they report that high adherence to a lower-carbohydrate diet increases the risk of death from any cause by 32%, cancer death by 35%, coronary heart disease death by 51%, and cerebrovascular disease death by 50%. The highest adherence to a diet jointly lower in carbohydrate and higher in protein increases the risk of death from any cause by 21%, cancer death by 22%, coronary heart disease death by 44%, and cerebrovascular disease death by 41%. These findings are supported by an accompanying meta-analysis: a pooled analysis of primary and secondary prevention cohorts with 5–30 years' follow-up, conducted in the USA, Sweden, Greece, and Japan, most of which show associations in the same direction as NHANES [22% for all-cause mortality (24% when limited to primary prevention studies); 13% for cardiovascular mortality; and 8% for cancer mortality].

NHANES is a nationally representative USA survey. In this study, 24 825 participants were followed for 144 months, over which time 3432 total deaths were recorded, including 827 cancer deaths, 709 heart disease deaths, and 228 cerebrovascular disease deaths. Trained staff measured diet using two 24-h dietary recalls with the automated multiple-pass method, the survey's standard practice since 2002.[2] This technique requires short-term memory, is less burdensome, and less likely to alter eating behaviour than food records, and can be used with diverse populations and people with low levels of literacy.[3] Because two recalls are administered, usual dietary intake distributions can be estimated after adjusting for day-to-day variation. The authors control for several confounding factors including age, sex, race, education, marital status, poverty to income ratio, total energy intake, physical activity, smoking, alcohol consumption, body mass index (BMI), waist circumference, hypertension, serum cholesterol, and diabetes.

The limitations of the NHANES analysis include use of self-reported dietary intake which typically underestimates total energy intake[2,3] and some macro- and micronutrients.[4] Under-reporting bias in energy intake is proportionate to reported total energy intake, and is likely greater in overweight and obese people, and women.[5] This concern is mitigated in the present analysis by adjustment for dietary energy using the residual method.[6] Another limitation is that participants who complete the recalls are probably not representative of those who do not, leading to potential selection bias. In the NHANES study, those in the lowest carbohydrate score quartile were younger, male, more likely to be Mexican-American and less likely to be Black, less educated, more impoverished, less physically active, consumed more alcohol, more likely to smoke, consumed less polyunsaturated fat, saturated fat, and fibre, and were more likely to have diabetes. Thus, a lower carbohydrate diet may serve as a proxy for any of these factors, which are involved in multiple pathways to mortality. Although the authors carefully adjusted for cholesterol, blood pressure, and other risk factors in the final analyses, the possibility of residual confounding by unmeasured confounders, imprecise measurement, or misclassification cannot be ruled out.

The association between lower carbohydrate diets and death may not hold across all regions of the globe, notably in studies where total carbohydrate contributes a larger percentage of energy than in Western countries. Because the authors' primary aim was to examine the joint effects of low carbohydrate and high protein, they excluded from their meta-analysis studies which looked at these macronutrients individually.[7] In NHANES, the lowest carbohydrate group consumed 39% of energy from carbohydrate, and the highest, 66%. In the meta-analysis, the lowest carbohydrate groups consumed 26–52% carbohydrate, and the highest carbohydrate groups, 54–73%. Compared with the highest carbohydrate group, the lowest carbohydrate group had a 22% [95% confidence interval (CI) 6–39%] increased risk of total mortality. These results can be compared with the Prospective Urban Rural Epidemiology (PURE) study of 135 335 middle-aged men and women from 21 countries, in which diet was assessed by country-specific validated food-frequency questionnaires. In the PURE study, the lowest carbohydrate group consumed 48% of energy from carbohydrate (18% protein, 34% fat) and the highest, 76%.[7] The lowest carbohydrate group compared with the highest carbohydrate group has a hazard ratio for total mortality of <1.0 (Mahshid Dehghan, personal communication). Interestingly, four of eight studies included in the meta-analysis with compatible data show a U-shaped association between carbohydrate and mortality: both lower and higher carbohydrate intake is associated with increased mortality. (Figure 1A).

Figure 1.

Scatterplot of most-adjusted relative risk estimates from eight cohorts [Prospective Urban Rural Epidemiology (PURE) study, National Health and Nutrition Examination Survey (NHANES), National Integrated Project for Prospective Observation of Non-communicable Disease and Its Trends in the Aged 1980 (NIPPON80), Atherosclerosis Risk in Communities (ARIC), Nurses' Health Study (NHS), Health Professionals' Follow-up Study (HPFS), and Västerbotten Intervention Program (VIP, M and F)]. (A) Polynomial functions fit for lines of best-fit to each study separately (Poly-Cohort). (B) Polynomial functions fit for lines of best-fit to all data points together (Poly-Cohort).

The only Asian study in the meta-analysis, the NIPPON study (Japan),[8] reported a 13% reduced risk of all-cause mortality and 22% reduced risk of cardiovascular disease (CVD) mortality in the lowest compared with the highest quartile of carbohydrate consumption. These findings are similar to those of the PURE study in which half of the participants were from Asian countries. In China, India, and Japan, women and men typically consume >60% of energy as carbohydrate, ~10–15% more than that reported in Western populations. Why would Asian populations with a higher carbohydrate intake show a different association between low carbohydrate and mortality? The answer may lie in the amount and/or food sources of the nutrient, or other factors including socio-economic status and/or health behaviours. For example, if a lower carbohydrate diet in Western countries is achieved by decreasing carbohydrate-containing foods such as whole grains, fruits, and vegetables and increasing animal protein and fats; and a lower carbohydrate diet in Asian countries is achieved by decreasing foods such as white rice, white bread, and consuming more nutrient-dense foods such as fish, chicken, dairy, or eggs, differential health effects could be expected. However, it is also plausible that differences in other social or health factors that are correlated with specific diet patterns explain the apparent associations with health outcomes. In the NHANES study, low carbohydrate consumers were also of lower socio-economic status and had more cardiovascular risk factors (i.e. less physically active, more smoking, more diabetes) than high carbohydrate consumers. In the NIPPON study, low carbohydrate consumers consumed less rice, and more flour products, fruits, vegetables, fish and shellfish, meat, eggs, and fibre. They were also more likely to be executives or professionals, more likely to smoke and drink alcohol, and had less hypertension than higher carbohydrate consumers.

Though the meta-analysis presented by Mazidi et al.[1] is largely consistent with previous work that has looked at low-carbohydrate, high-protein diets using scores, it is worth noting that the influence of any given diet on health outcomes must consider not only macronutrient ratios, but the quality of the foods that contribute to these ratios. For example, a simple low-carbohydrate diet score makes no allowance for the quality of carbohydrate (i.e. whole grain vs. refined), protein (i.e. plant vs. animal), or fat (i.e. saturated vs. unsaturated) consumed. In cohort studies, low-glycaemic index, whole-grain carbohydrates reduce chronic disease risk compared with refined carbohydrate,[9,10] though this has not always been substantiated in randomized trials of events or biomarkers of CVD.[11,12] In cohort studies, replacing refined carbohydrates (starch and sugars) with whole grains has been shown to reduce cardiovascular disease risk;[13] and, in randomized trials, replacement of carbohydrate with protein or fat improves cardiovascular risk factors.[14] Further, the type of protein consumed (animal vs. plant) may determine its influence on health. In the present meta-analysis, the highest quintile of animal low-carbohydrate score was associated with 23% higher all-cause, 14% higher cardiovascular, and 28% higher cancer mortality; in contrast, one of the US cohort studies included in the meta-analysis reported that a higher vegetable low-carbohydrate score was associated with 20% lower all-cause and 23% lower cardiovascular mortality.[15] However, no differential effects of animal-based and plant–fish-based low-carbohydrate diet were observed in the NIPPON study.[8]

The Acceptable Macronutrient Distribution Ranges (AMDRs) published by the US Institute of Medicine[16] are largely supported by data from NHANES, the included pooled analyses, and the PURE cohort study. The AMDRs suggest that a diet with 45–65% carbohydrate, 10–35% protein, and 20–35% fat is associated with reduced risk of chronic disease while providing intakes of essential nutrients. Data from the PURE study demonstrate increased mortality risk beginning at >65% carbohydrate, and data from NHANES suggests increased mortality risk beginning at <40% carbohydrate (42% fat) (Figure 1A). A U- or J- shaped relationship between carbohydrate intake and mortality may exist (Figure 1B), although this needs to be confirmed by individual person data meta-analyses or pooling projects, and additional larger, longer term studies. Taking all the studies into account, the message of moderation is perhaps the most convincing one of all—diets that focus too heavily on a single macronutrient, whether extreme protein, carbohydrate, or fat intake, may adversely impact health. Until these ratios have been studied in the general population for long periods of time or in long-term randomized trials, the best advice seems to be to select whole foods from a variety of sources and avoid dietary extremism. For now, for carbohydrates, everything in moderation carries the day.