50 / 52
88
VOLUME 15 NUMBER 2 • NOVEMBER 2018
DIABETES NEWS
SA JOURNAL OF DIABETES & VASCULAR DISEASE
Evolving evidence about diet and health
N
utritional research initially focused
almost entirely on conditions of
nutritional deficiencies (e.g. scurvy, beriberi,
pellagra). By the 1950s, with the increase
in coronary heart disease in high-income
countries, attention shifted to a range of
so-called diet–heart hypotheses.
These included the putative and harmful
effects of fats (especially saturated fats)
and the protective effects of the so-called
Mediterranean diet to explain why
individuals in the USA, northern Europe
and the UK were more prone to coronary
heart disease, whereas those in European
countries around the Mediterranean (or
Japan) seemed to have lower risks.
Some of the initial studies were
enormously influential while undergoing
limited scrutiny as to the rigor of their
methods. The lack of replication of these
early claims should have prompted caution
and re-examination of whether fats
(especially saturated fats) were indeed
harmful.
More recently, studies using standardised
questionnaires, careful documentation
of outcomes with common definitions,
and contemporary statistical approaches
to minimise confounding have generated
a substantial body of evidence that
challenges the conventional thinking that
fats are harmful. Also, some populations
(such as the US population) changed their
diets from one relatively high in fats to one
with increased carbohydrate intake. This
change paralleled the increased incidence
of obesity and diabetes.
The focus of nutritional research has
recently shifted to the potential harms of
carbohydrates. Indeed, higher carbohydrate
intake can have more adverse effects on
key atherogenic lipoproteins (e.g. increase
the apolipoprotein B-to-apolipoprotein A1
ratio) than can any natural fats. Additionally,
in short-term trials, extreme carbohydrate
restriction led to greater short-term weight
loss and lower glucose concentrations
compared with diets with higher amounts
of carbohydrate.
Robust data from observational studies
support a harmful effect of refined, high-
glycaemic-load carbohydrates on mortality.
The realisation that cardiovascular disease
is a global epidemic, with most cases
occurring in developing countries, has
also stimulated studies involving multiple
countries at different economic levels.
Last year, the Prospective Urban Rural
Epidemiology (PURE) study of 135 335
individuals from 18 countries in five
continents showed that a diet high in
carbohydrates (more than approximately
60% of energy) but not high in saturated
fats, was associated with higher risk of
death. However, in PURE, even the group
with the highest level of fats (i.e. quintile
5; mean total fat intake 35% of energy,
and saturated fat intake 13% of energy)
was not as high as the average in studies
from Finland (37 and 20%, respectively),
Scotland (37 and 17%, respectively), or the
USA (38 and 16%, respectively), done in
the 1960s and 1970s.
Therefore, a marked reduction in fat
intake in several countries might have
occurred over the past few decades in
several countries. It is not clear that further
reductions in dietary fat intake will lead to
reductions in incidence of disease.
In countries (or individuals) with high
carbohydrate intakes, limiting intake could
be beneficial. In a recent issue of
The
Lancet Public Health
, Sara Seidelmann and
colleagues examine the 25-year follow-up
data from the Atherosclerosis Risk in
Communities (ARIC) study and place their
findings in the context of a meta-analysis
of published studies about carbohydrate
intake.
The authors concluded that the
epidemiological association between
carbohydrate intake and death is
U-shaped, with the lowest risk occurring
with a carbohydrate intake of 50–55% of
energy, and with both lower and higher
intakes being associated with higher risk
of death (hazard ratio 1.20, 95% CI: 1.09–
1.32 for low carbohydrate consumption;
1.23, 1.11–1.36 for high carbohydrate
consumption). Such differences in risk
associated with extreme differences in
intake of a nutrient are plausible, but
observational studies cannot completely
exclude residual confounders when the
apparent differences are so modest.
Based on first principles, a U-shaped
association is logical between most essential
nutrients versus health outcomes. Essential
nutrients should be consumed above
a minimal level to avoid deficiency and
below a maximal level to avoid toxicity. This
approach maintains physiological processes
and health (i.e. a so-called sweet spot).
Although carbohydrates are technically not
an essential nutrient (unlike protein and
fats); a certain amount is probably required
to meet short-term energy demands during
physical activity and to maintain fat and
protein intakes within their respective
sweet spots.
On the basis of these principles, moderate
intake of carbohydrate (e.g. roughly 50% of
energy) is likely to be more appropriate for
the general population than are very low
or very high intakes. This would translate
to a generally balanced diet that includes
fruit, vegetables, legumes, whole grains,
nuts, fish, dairy and unprocessed meats, all
in moderation.
The findings of the meta-analysis should
be interpreted with caution, given that
so-called group thinking can lead to biases
in what is published from observational
studies, and the use of analytical
approaches to produce findings that fit in
with current thinking. The ideal approach
to meta-analysis would be a collaboration
involving investigators of all the large
studies ever done (including those that
remain unpublished) that have collected
data about carbohydrate intake and clinical
events, and pool the individual data using
transparent methods. This approach is likely
to provide the best and most unbiased
summary of the effects of carbohydrates on
health, rather than reliance on the results
of any single study.
Future observational studies should
also consider new methods, which include
triangulation, to assess whether there is a
coherent pattern of information about the
links between consumption of a nutrient,
such as carbohydrates, with a panel of
physiological or nutritional biomarkers and
clinical outcomes. When appropriate, this
approach should be complemented by large
and long-term clinical trials investigating
the effects of different dietary patterns
(constructed from information about the
effects of individual nutrients and foods),
because the effect of individual nutrients
is likely to be modest. When coherent
information emerges from different
approaches and is replicated, this will form
a sound basis for robust public health
recommendations.
The Lancet Public Health
2018;
3
(9): e408–409.