Implementing food-based dietary guidelines for - United Nations ...

S36
of uncertainty factors. Furthermore, ADME data
could be used to explore the dose–response curve at
the lower extreme of intakes to set safe lower levels as
part of a risk–benefit analysis [16]. The construction of
biological models for dose–response curves and CSAFs
would need different gradations of intakes from those
traditionally used for toxicologic studies.
In some instances, high nutrient intakes have already
been associated with phenomena that correspond with
stage 2 or 3 markers for adverse health effects. These
are metabolic interactions among nutrients (for example,
those among iron, zinc, and copper) in situations in
which imbalanced intakes compromise the specificity
of individual metabolic pathways [1].
The value of using markers in nutrient risk assessment
would be enhanced if the totality of the evidence
supporting the biological validity for each marker
could be explicitly evaluated in the context of overall
causation, incorporating the strength of the association;
consistency across all lines of evidence; specificity; temporal
relationship; a demonstrable relationship between
intake and a functional or health effect response for the
marker (i.e., for a homeostatic stage 2 marker, an indication
of an adaptive phenomenon, rather than a linear
response that might reflect exposure rather than a specific
adaptive health effect [5]); and plausibility, coherence,
and experimental support from other sources
(e.g., animal models [17]). It is doubtful whether much
information is available to support potential markers
for nutrient risk assessment.
These principles are also relevant to the early detection
of inadequate intakes. Some initiatives have considered
whether it would be possible to have a common
approach to assessing nutrient deficiency and excess.
This has been explored for essential trace elements [2]
and as a risk–benefit analysis for micronutrients in
general [16]. Recently a human health dose–response
risk assessment has been used to explore the dual
response curve risk assessment for copper [11], and
a spectrum from copper deficiency to copper toxicity
has been compiled with the use of data from studies on
humans and animal models. The exercise provided an
opportunity to explore several approaches to dose– or
intake–response modeling, including the benchmark
dose and categorical regression. Existing data allowed
for these approaches, but the development of a biologically
based dose–response risk assessment and of
CSAFs was limited by the quality and amount of the
data [11]. Many of the individual studies that were
reviewed during hazard identification and characterization
in this exercise had been designed to demonstrate
the effects of prolonged exposures to single measured
and usually very high or very low concentrations of
copper in diets. These studies were not designed to
generate intake–response curves or to examine risk.
Most just reported the copper contents of diets fed to
animals and gave no indication of actual intakes; these
had to be estimated, albeit imperfectly, from knowledge
of animal weights and data from other reports
on animal food consumption. After such reports were
excluded during the systematic literature search, the
residual database was very scant. This experience cautions
against having high expectations of being able
soon to address nutrient risk assessment through a
biologically based response approach. The situation for
amino acids, where systematic metabolic studies using
tracers are improving the generic understanding of the
application of kinetic and dynamic studies to homeostasis,
may be more encouraging [3–5].
Summary and conclusions
P. J. Aggett
There is a need for a transparent model for nutrient risk
assessment that would enable key elements of nutrient
metabolism and function, and gastrointestinal and
systemic adaptive phenomena in response to excess
intakes (i.e., above the “physiological requirements”),
to be identified and used as markers of excessive exposure.
Such a biologically based dose–response model
to determine ULs for nutrients could also be used to
explore lower levels of intake and thereby enable the
setting of lower levels of reference intakes.
Nutrient hazard identification and characterization is
an iterative process. It needs to be supported by a complete
compilation and review of the available literature
and data, i.e., an evidence-based systematic review with
predefined search and summary strategies and transparent
criteria for rating, including and excluding individual
studies and their data. As with risk assessment of
non-nutrient chemicals, published systematic reviews
may be useful. However, they should provide a means
to access primary data and to rate their quality, and the
bases of their systematization should not be allowed
to prejudice the nutrient hazard identification and
characterization. The critical intermediate outcome of
this process is the agreed selection of an adverse health
effect from which a UL, as a health-based guidance
value, can be derived for the protection of the public’s
health. The approach proposed in this paper for identification
of the adverse health effect, namely, the use of
health effects and markers that occur relatively earlier
or at lower intakes on the pathogenic response curve
than classic toxicologically adverse effects, necessitates
specific data search strategies that make greater use of
ADME, biokinetic, and biodynamic data and that will
probably need specific research. The advantage of using
adverse health effects, or markers thereof, that occur
at such lower intakes is that such research should be
feasible in human participants. The disadvantages at
the moment are the overall paucity of data, the nonsystematic
and opportunistic nature of most of the
relevant data, and the fact that most systematic data
are derived from animal models.