Content uploaded by Javier T Gonzalez
Author content
All content in this area was uploaded by Javier T Gonzalez on Oct 08, 2022
Content may be subject to copyright.
Personalised nutrition: What makes you
so special?
J. A. Betts and J. T. Gonzalez
Department for Health, University of Bath, Bath, UK
Abstract Personalised nutrition can be defined as developing unique nutrition guidelines
for each individual; precision nutrition seeks to develop effective approaches
based on the combination of an individual’s genetic, environmental and lifestyle
factors. The former infers variants that underlie traits are largely fixed (i.e. stable
across time) and appeals to the notion that we are inherently different from one
another. The latter opens up the possibility that what we do and where we are
may be more important than what we are. While there are undoubtedly a specific
minority of individuals who clearly require a more personalised approach to
nutrition, several criteria must be fulfilled before we can justify or implement
personalised nutrition for the general population. These would include
identifying a desired health outcome and a valid predictor of how that outcome
changes, which can be measured with sufficient accuracy and exhibit a robust
correlation and/or causal relationship in the required direction (i.e. predictor-
response). Many attempts to personalise nutrition, such as profiling the genome
or microbiome, do not currently meet all these criteria. Therefore, we argue that
there is presently insufficient rationale for truly personalised nutrition for the
majority of people based on the inter-individual differences that separate them.
Conversely, we propose that precision nutrition based on the environmental and/
or behavioural ‘lifestyle’ variance within each person may provide a more
effective basis for adjusting diet dynamically, with recognition of varying
physiological demands and requirements over time.
Keywords: metabolic profiling, microbiota, nutrition, public health
Introduction
The concept of personalised nutrition is inherently
appealing to human egotism; simple messages recog-
nising individuality therefore resonate deeply with
consumers. This may explain the popularity of such
messages in sales and marketing, with uniquely
designed products intuitively assumed to be superior
and considered more exclusive than a ‘one-size-fits-all’
approach. Ironically, this very reasoning often culmi-
nates not in bespoke products that recognise each indi-
vidual as truly unique but rather ‘personalisation’ is
achieved by fitting groups of people into discrete cate-
gories once each individual has been branded with the
fixed label of a certain ‘type’ (i.e. stratified nutrition;
Smith 2012). Whilst the number and specificity of cat-
egories may vary (e.g. from sex differences to genetic
profiling), the ‘type’ of person you are is generally
considered to be a stable characteristic. This means
that these categories must be mutually exclusive and
represent a comprehensive list of all possible types of
people.
Correspondence: Dr. James A. Betts, FACSM, Reader in Nutrition
& Metabolism Department for Health, University of Bath, Bath
BA2 7AY, UK.
E-mail: j.betts@bath.ac.uk
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359 353
NEWS AND VIEWS DOI: 10.1111/nbu.12238
There are undoubtedly some phenotypes (albeit those
that are typically rare or transient conditions) that may
justifiably be categorised as having distinct metabolic
requirements and thus special dietary needs (e.g. con-
genital metabolic disorders or pregnancy; McNulty
1997). However, the idea that nutritional requirements
vary sufficiently between individuals in the general pop-
ulation to justify personalised nutrition for the vast
majority is an assumption that can be questioned at a
number of levels. This opinion piece will explore the
various points at which that reasoning may not neces-
sarily be supported by sound logic or empirical evi-
dence, before considering whether and how we should
identify individual differences for incorporation into
future nutritional recommendations. As a starting
point, we propose that a practically useful approach to
personalised nutrition must fulfil at least the following
inter-related but independently essential criteria:
•the ability to measure the predictive variable with
sufficient precision and accuracy;
•the ability to measure the outcome variable with
sufficient precision and accuracy;
•plus a robust and practically meaningful correlation
and/or causal relationship (in the required direction)
between the predictive and outcome variables.
Is personalised nutrition justified by small
differences between people?
A simple balance equation subtracting expenditure
from intake is often applied in relation to nutrition,
which can be a helpful method to express how needs
must somehow be properly met to prevent deficiency
(or excess) and thus maintain function/status. Based
on this model, if the ‘needs’ of a system require 10%
above average utilisation of a given substrate/nutrient,
it might be expected that a 10% higher dietary intake
of that substrate/nutrient would be necessary to meet
the elevated requirement. However, that reasoning
overlooks the capacity of physiological systems to
adapt to the naturally variable and unpredictable
availability of nutrients. Specifically, notwithstanding
extreme cases when sustained/severe deficiency can
result in harmful metabolic imbalances (e.g. famine)
or when an individual may require optimal function
beyond adequate health (e.g. elite athletes; Jeukendrup
2014), a surplus or deficit of a given nutrient can usu-
ally be tolerated or regulated via the interconversion,
synthesis, oxidation, excretion and/or safe storage of
nutrients to maintain homoeostasis.
The margins of error that can be managed in this
way can vary for particular nutrients and circumstances
but the point remains that variability in nutrient
requirements within the limits of our physiological
capacity may not necessitate any proportionate adjust-
ment in diet. Nonetheless, accepting that our capacity
to buffer such imbalances is limited, such that even a
minority of individuals might benefit from a person-
alised intake of certain nutrients, the next challenge is
to find measurements that reliably identify candidates
for personalised nutrition before symptoms of acute
nutritional imbalance and/or associated chronic diseases
manifest.
Can measuring differences between people
predict individual responses to personalised
nutrition?
Health professionals attempting to provide advice and
support regarding nutrition will be all too familiar
with the challenge of identifying the few individuals
for whom standard dietary guidelines may be inappro-
priate. Unfortunately, no single measurement tool cur-
rently exists that can take a snapshot of individual
physiology capable of identifying healthy individuals
before they become symptomatic in order to inform
appropriate dietary modifications in a preventative
manner. Profiling of individuals to estimate personal
nutritional requirements has been attempted using
numerous organismic variables (see left-hand column
of Table 1).
Two variables that ostensibly present the greatest
degree of accuracy and precision for predicting indi-
vidual nutrient requirements are based on profiling the
genome or microbiome of an individual yet, to date,
both approaches have relied heavily on the assump-
tions that: (i) inter-individual variability between peo-
ple can be mostly explained by the genome or
microbiome; and (ii) intra-individual variability within
each person is minor in relation to inter-individual
variability. On the first point, genetic profiling in par-
ticular undoubtedly identifies individual differences at
a fundamental level and remains popular despite the
realisation that most health outcomes cannot in fact
be meaningfully explained by common gene variants.
For example, the genetic contribution to risk of devel-
oping type 2 diabetes (a health outcome that is heavily
dependent on nutrition; Steven et al. 2016) is only
~10% (Morris et al. 2012). Even when considering
this contribution, using sequence-based genome-wide
association studies, genetic risk of type 2 diabetes is
much more strongly predicted by common risk vari-
ants than rare risk variants (Fuchsberger et al. 2016).
This makes sense from an evolutionary perspective, as
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
354 J. A. Betts and J. T. Gonzalez
any nutrition-related health outcome that has the
greatest health impact post-reproductive age (e.g. most
cardiovascular diseases, type 2 diabetes and some can-
cers) is unlikely to have resulted in selection pressure.
Thus, the genetic risk of most nutrition-related health
outcomes is more likely to be explained by common,
rather than rare, variants. In this sense, we are more
similar to other people than we are dissimilar. The
apparent ‘missing heritability’ is perhaps best illus-
trated by the fact that >96% of the diversity in even a
relatively objective, reliably measured and strongly
(80–90%) inherited phenotype, such as average adult
stature, cannot be explained by differences in the 40
genes that are collectively most predictive of human
height in the majority of cases (Gudbjartsson et al.
2008). Clearly, trying to predict complex real-world
phenomena based on just one or even multiple genes
can be like trying to tell the time using a clock with
no hour hand; precision and accuracy only become
useful in the wider context.
Despite these current limitations to genetic profiling,
there is already at least one exception to the rule that
lends hope to the promise of stratified nutrition based
on genotype in the future. Specifically, individuals
who are homozygous AA for TAP2Brs987237 lose
more weight when consuming an energy-restricted diet
that is lower in fat, whereas individuals who are
homozygous GG display greater weight loss on an
energy-restricted diet that is higher in fat (Stocks et al.
2012). In that study, however, those homozygous GG
were a small minority (4%) and the stated relationship
between genotype and body mass was not apparent
for outcomes more strongly related to adiposity and
health (e.g. waist circumference). Most importantly,
the subsequent attempt to use this particular gene
variant as a predictor of responses to high- vs. low-fat
dietary interventions was unsuccessful (Stocks et al.
2012). This finding is consistent with the results of a
recent randomised controlled trial that effectively
modified dietary behaviours by personalising nutrition
using a combination of five diet-responsive gene vari-
ants alongside phenotype and habitual diet, yet this
did not translate into enhanced health responses to
that intervention (Celis-Morales et al. 2016). While it
remains a possibility that effective stratified prescrip-
tion of diet for weight management may have been
limited by statistical power (Larsen et al. 2012; Stocks
et al. 2012), there is presently insufficient evidence to
confidently assign personalised diets according to
genotype.
With regard to microbiomic profiling, the composi-
tion of microbe species in the human gut (gut micro-
biome) is vast, with many more genes than in our
own host genome (Neish 2009). There is also evidence
for a mechanism by which gut microbes can influence
nutrition-related health outcomes, primarily through
the availability of nutrients and short-chain fatty
acids, with implications for glycaemic control, insulin
sensitivity and appetite (Byrne et al. 2015). Therefore,
it is an attractive prospect that the gut microbiome
could be used as a basis for individual level nutrition
guidance and to achieve favourable health outcomes.
An example of this was the attempted use of micro-
biomic profiling (in addition to self-reported physical
activity levels, blood glycated haemoglobin and
cholesterol concentrations) to personalise nutrition in
relation to plasma glucose responses to foods (Zeevi
et al. 2015). The lack of accurate measures of physical
activity and postprandial glycaemia mean that at least
the first two criteria for personalised nutrition pro-
posed earlier have not yet been satisfied (i.e. ‘an ability
to measure the predictive/outcome variable with suffi-
cient precision and accuracy’). With this in mind, it
has been highlighted that most of the variability in
postprandial plasma glucose responses to foods is due
to variability within individuals rather than between
individuals (Wolever 2016).
With respect to the third criterion for personalised
nutrition proposed earlier, relationships between a
Table 1 Ostensible predictors that might inform a personalised/
precision nutrition plan
Ostensible predictor
Do individual
differences predict
the benefits of
personalised nutrition?*
Are individual
differences fixed/stable
characteristics?
†
Genotype No No (epigenetics)
Gut microbiome No No
Sex No Yes
Age Yes No
Body mass Yes No
Adiposity/obesity Yes No
Pregnancy related Yes No
Physical activity Yes No
Allergy/intolerance Yes Yes
Disease state Yes Yes
*The predictive power of individual differences is based on the best mea-
surement and analytical tools currently available, so ‘yes’infers a clear justifi-
cation for standard dietary advice to be modified on the basis of this
variable in most cases.
†
Fixed/stable characteristics could inform lifelong personalised nutrition
based on a single static measurement, so ‘no’infers that personalisation
must be constantly adapted based on serial monitoring of the variable in
most cases.
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
Personalised nutrition 355
genotype and a health outcome are often interpreted
as causal and unidirectional. As elegantly described by
Noble (2008), the dogmatic reductionist view that the
direction of causality flows from genes–proteins–cells–
tissues–organism inadequately explains biological
processes, as reverse causality is a possibility in this
context. This is clearly the case for the gut micro-
biome, whereby the nutrition–gut microbiome causal
link flows in both directions (David et al. 2014). Gene
expression is also regulated by ‘higher’ processes, such
as by DNA methylation and histone modification
(Fraga et al. 2005; Qiu 2006). Strong associations
between a genotype and risk of obesity have been
demonstrated. For example, a common variant in the
FTO gene is associated with a 67% increased risk of
obesity, compared to not inheriting the risk allele
(Frayling et al. 2007). However, this strong associa-
tion is not consistent across different conditions. In
physically active individuals, this association between
FTO genotype and body mass is either attenuated or
abolished (Vimaleswaran et al. 2009). This beha-
vioural uncoupling of genotype with an outcome is
not uncommon; physical activity reduces the associa-
tion between genotype and obesity risk based on 12
genetic loci by 40% (Li et al. 2010). These examples
demonstrate that behavioural and environmental fac-
tors have a profound effect on energy balance and
confound congenital effects. Moreover, there is insuffi-
cient evidence of variability between people to justify
the use of personalised nutrition guidelines over com-
mon guidelines across people.
Therefore, on the basis that genotypic and gut
microbial variation in isolation may currently have
limited predictive power in relation to the health out-
comes that a nutritional intervention might be person-
alised to prevent/treat, it is perhaps understandable
why neither individual genotyping or gut microbiome
analysis alone can confidently predict the magnitude
of phenotypic response to that intervention. One
approach to personalised nutrition might, therefore,
be not to predict responsiveness to an intervention
using a static cross-sectional snapshot to profile across
a population but rather to measure within each indi-
vidual to see who actually responds positively to the
intervention over time (i.e. an individual’s initial
response upon first exposure to an intervention
becomes the predictor for their future responses).
However, this introduces a further question about
whether an individual’s observed physiological
response to intervention on one occasion can be taken
as evidence that they are a dependable ‘responder’ and
their future nutrition personalised accordingly.
Can measuring individual responses predict
who would benefit from personalised
nutrition?
Longitudinal/repeated measures studies, ideally within
a test–retest experimental design, can be a highly effec-
tive way to establish the magnitude of response to an
intervention over time at a group level. Moreover, it
has rightly become more common in recent years for
scientific experiments to be more sensitive to the consis-
tency of responses from baseline to follow-up at an
individual level (i.e. pre–post-intervention change
scores). The latter is a key step in understanding the
determinants of intra-individual responses and thus to
personalising nutrition. This may therefore partly
explain the emerging trend for such inter-individual
differences to be interpreted as evidence of distinct
‘responders’ and ‘non-responders’ in a population
(Fig. 1a), presumably to provide the required categories
for personalisation of that particular intervention.
However, in many cases, the apparent individual
differences in response are merely the result of noise/
error in measurement and/or natural test–retest oscilla-
tions in human response, as eloquently explained and
comprehensively demonstrated using a simulated data
set (Atkinson & Batterham 2015). This point is simply
illustrated by reformatting Figure 1a to a frequency
histogram which clearly is not inconsistent with the
normal distribution that might be expected for the
error in measuring naturally occurring physiological
responses (Fig. 1b). For example, the mass of most
objects can generally be measured with relative objec-
tivity and accuracy (i.e. not a ‘noisy’ measurement per
se), yet if the object in question is a living organism
(e.g. a person) then factors such as hydration, recent
voiding and food intake could introduce random
errors if not standardised pre–post-intervention, such
that for every ‘Mike’ there is a ‘Phil’ due to natural
over-/under-estimates of measurement (plus/minus ran-
dom variation) at both follow-up and baseline (the lat-
ter also introducing the possibility of statistical
regression and interactive/sensitising effects of pre-test-
ing). Therefore, based solely on one solitary group
receiving an intervention (even using a perfect mea-
surement tool), the most that might be legitimately
concluded is that Mike can be labelled as ‘responded’
(i.e. lost weight on this occasion), with no inference
whatsoever as to whether he may be a dependable
‘responder’ likely to exhibit similar effects of the same
intervention in future.
There is increasing awareness amongst researchers
that reporting only measures of central tendency and
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
356 J. A. Betts and J. T. Gonzalez
variability (e.g. mean and standard deviation) can be
uninformative and ‘miss’ important patterns in data
(Weissgerber et al. 2015). Thankfully, modern pub-
lishing and archiving practices now enable raw data to
be made available as supplementary or linked files, so
interested readers have the opportunity to fully
explore inter-individual responses. However, these
advances in awareness and data sharing must be cou-
pled with recognition that certain methods of express-
ing and/or interpreting individual data can be
misleading for people who are receptive to messages
highlighting that they may be one of a minority of
possible ‘responders’. Most consumers ultimately just
want to know will this product or intervention work
for me?, and some may be uncertain or even sceptical
about whether a group average applies to them. Of
course, by definition, the appropriate summary statis-
tics do provide a relatively accurate descriptor for the
majority of people (e.g. 68–95% within 1–2 standard
deviations of the mean). Therefore, before any individ-
ual can be identified as an outlier for whom standard
dietary guidelines may not be appropriate, we must:
(i) be confident that a more personalised approach is
justified by the proper analyses applied to the correct
variables; and (ii) empirically demonstrate that this
approach provides a benefit over standard approaches.
Implications and possible solutions
As discussed earlier, current methods of profiling
across a population to fit groups of people into dis-
crete categories are largely ineffective for identifying
inter-individual differences that genuinely warrant
personalised nutrition. Part of the problem with this
approach might well be that ‘the role of genotypic
variation in common non-communicable diseases is
essentially overwhelmed by environmental, cultural
and behavioural factors’ (Joyner & Prendergast 2014).
Indeed, a key confounding factor when assessing the
role of congenital differences between groups of peo-
ple appears to be lifestyle variance within each indi-
vidual. Therefore, rather than trying to control or
adjust for this confounding influence, it may be more
prudent to take full advantage of recent advances in
monitoring such lifestyle factors and use them (within
the constraints of measurement error and biological
variability) as the primary determinants informing
personalisation.
Table 1 lists a number of organismic variables that
have been used to try and personalise nutrition (first
column). The second column summarises how neither
profiling the genome or microbiome nor even sex dif-
ferences justify a clear and consistent deviation from
standard dietary guidelines for any particular category
within those variables (e.g. in general, men and
women can avoid ill health equally well without the
need to personalise nutrition according to sex differ-
ences). Thereafter, a number of additional variables
are listed which could inform personalised nutrition in
order to meet a substantially altered requirement and/
or favour more positive health outcomes. Some of
those listed are self-evident, such as changes in nutri-
ent synthesis/requirements over the lifespan (i.e. age);
greater absolute requirements for most nutrients to
sustain a larger overall body mass vs. advisable reduc-
tions in energy intake in cases of excessive adiposity
(i.e. body/fat mass); the well-established need for
increased intake of certain nutrients prior to, during
Figure 1 Hypothetical sketch illustrating how the common presentation of ‘responders’vs. ‘non-responders’(a) can be reformatted to a frequency histogram
showing normally distributed measurement error (b).
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
Personalised nutrition 357
and after pregnancy; and clear dietary restrictions with
certain allergies, intolerances or disease states. Physical
activity on the other hand remains poorly understood
yet represents the most variable component of energy
expenditure (both between and within individuals)
and therefore the single factor with greatest potential
to adjust both energy requirements and metabolic sub-
strate availability (e.g. via muscle and liver glycogen
depletion). However, beyond widespread recognition
that a physically active individual may need to con-
sume more food in general and often in the form of
carbohydrate (Williams 1989), there appears to be
limited information available about other specific con-
siderations. For example, are any links between diet-
ary sugars and poor health outcomes offset if the
ingested sugar is immediately metabolised during exer-
cise; how much are vitamin B requirements elevated
by various types and patterns of physical activity; and
do exercise-induced sweat sodium losses justify
exceeding recommended limits of salt intake?
Of all these factors with the potential to predict the
benefits of personalised nutrition, the third column
then indicates which one can generally be considered a
fixed or stable characteristic of an individual. The
implication is therefore that only sex, allergy/intoler-
ance and disease state can generally be assessed/diag-
nosed on the basis of a single, static measurement that
can then inform future treatment thereafter. In con-
trast, other interrelated factors such as age, body/fat
mass and even pregnancy all vary naturally through-
out life and might therefore require transient and con-
stantly adapted personalisation of nutrition. Again,
physical activity is a unique case here as this variable
fluctuates much more acutely on a day-to-day basis
and so epitomises the suggestion that rather than per-
sonalising what certain types people should eat every
day we might do better to personalise what every per-
son should eat on certain types of day. Not only
would this dynamic approach be more consistent with
the observation that the fixed differences between us
are generally smaller than the modifiable daily vari-
ance within each person but would also be conducive
to general recommendations for a more varied diet.
In this way, a diet personalised according to activity
levels could be based on a wider framework of overall
energy balance and recognise that the most effective
health-gain strategies involve simultaneous modifica-
tion of both diet and lifestyle, partly due to interactions
whereby adequate physical activity supports appropri-
ate appetite regulation (Blundell & King 1999; Hill
et al. 2012). An optimal diet can therefore be person-
alised not only to what an individual is currently doing
but to what they should be doing. To return to our ear-
lier example, recent trials exploring the relative merits
of modified dietary intake of carbohydrates or sugars
in relation to obesity are almost exclusively conducted
in the context of the general population who typically
fall far short of health guidelines for taking physical
activity or exercise. The conclusions drawn may then
be specific to individuals who are sedentary (which can
be worse for health than obesity per se), in which case
the recommendations made may be inappropriate for
individuals leading an otherwise healthy and active life-
style (potentially even causing lower energy expendi-
ture if inadequate carbohydrate availability reduces the
motivation or ability to partake in physical exertion).
Conversely, tailoring nutrition not to the person (i.e.
personalised nutrition) but dynamically as most appro-
priate to the varying demands and requirements we
face on a daily, weekly, seasonal and lifespan basis
might represent a more effective method of what might
be better referred to as precision nutrition.
Conclusions
Whilst the idea of personalised nutrition might appeal
to our intuition, in this article we have applied logical
reasoning to several examples where the underlying
science calls for a second look at that way of thinking
(i.e. an ‘intuition pump’; Dennett 2014). This is not to
say that we are all the same and that inter-individual
differences between us do not exist; certainly some are
even of sufficient magnitude that they can be measured
and cannot be either tolerated or rectified by our physi-
ology, so may genuinely justify a personalised (or at
least stratified) approach to nutrition. Nonetheless, the
key point put forward in this opinion piece is that these
cases are a rare minority, such that the inter-individual
differences that separate most people are smaller and
less important than the day-to-day variance within each
of us. Accordingly, precision nutrition is a more attrac-
tive prospect than personalised nutrition. In other
words, what makes you special with regard to nutrition
is not who you are but what you do.
Conflict of interest
The authors have no conflict of interest to disclose.
References
Atkinson G & Batterham AM (2015) True and false interindividual
differences in the physiological response to an intervention. Exper-
imental Physiology 100: 577–588.
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
358 J. A. Betts and J. T. Gonzalez
Blundell JE & King NA (1999) Physical activity and regulation of
food intake: current evidence. Medicine and Science in Sports and
Exercise 31: S573–S583.
Byrne CS, Chambers ES, Morrison DJ et al. (2015) The role of short
chain fatty acids in appetite regulation and energy homeostasis.
International Journal of Obesity 39: 1331–1338.
Celis-Morales C, Livingstone KM, Marsaux CFM et al. (2016)
Effect of personalized nutrition on health-related behaviour
change: evidence from the Food4me European randomized con-
trolled trial. International Journal of Epidemiology. doi:10.1093/
ije/dyw186.
David LA, Maurice CF, Carmody RN et al. (2014) Diet rapidly and
reproducibly alters the human gut microbiome. Nature 505: 559–
563.
Dennett DC (2014) Intuition Pumps and Other Tools for Thinking.
WW Norton & Co: New York.
Fraga MF, Ballestar E, Paz MF et al. (2005) Epigenetic differences
arise during the lifetime of monozygotic twins. Proceedings of the
National Academy of Sciences of the United States of America
102: 10604–10609.
Frayling TM, Timpson NJ, Weedon MN et al. (2007) A common
variant in the FTO gene is associated with body mass index and
predisposes to childhood and adult obesity. Science 316: 889–894.
Fuchsberger C, Flannick J, Teslovich TM et al. (2016) The genetic
architecture of type 2 diabetes. Nature 536:41–47.
Gudbjartsson DF, Walters GB, Thorleifsson G et al. (2008) Many
sequence variants affecting diversity of adult human height. Nat-
ure Genetics 40: 609–615.
Hill JO, Wyatt HR & Peters JC (2012) Energy balance and obesity.
Circulation 126: 126–132.
Jeukendrup A (2014) A step towards personalized sports nutrition:
carbohydrate intake during exercise. Sports Medicine (Auckland,
N. Z.) 44(Suppl 1): S25–S33.
Joyner MJ & Prendergast FG (2014) Chasing Mendel: five questions
for personalized medicine. Journal of Physiology 592: 2381–2388.
Larsen LH, Angquist L, Vimaleswaran KS et al. (2012) Analyses of
single nucleotide polymorphisms in selected nutrient-sensitive
genes in weight-regain prevention: the DIOGENES study. Ameri-
can Journal of Clinical Nutrition 95: 1254–1260.
Li S, Zhao JH, Luan J et al. (2010) Physical activity attenuates the
genetic predisposition to obesity in 20,000 men and women from
EPIC-Norfolk prospective population study. PLoS Medicine.
doi:10.1371/journal.pmed.10003322
McNulty H (1997) Folate requirements for health in women. Pro-
ceedings of the Nutrition Society 56: 291–303.
Morris AP, Voight BF, Teslovich TM et al. (2012) Large-scale asso-
ciation analysis provides insights into the genetic architecture and
pathophysiology of type 2 diabetes. Nature Genetics 44: 981–990.
Neish AS (2009) Microbes in gastrointestinal health and disease.
Gastroenterology 136:65–80.
Noble D (2008) Claude Bernard, the first systems biologist, and the
future of physiology. Experimental Physiology 93:16–26.
Qiu J (2006) Epigenetics: unfinished symphony. Nature 441:143–145.
Smith R (2012) Stratified, personalised, or precision medicine. Avail-
able at: http://blogs.bmj.com/bmj/2012/10/15/richard-smith-strati-
fied-personalised-or-precision-medicine/ (accessed 14 September
2016).
Steven S, Hollingsworth KG, Al-Mrabeh A et al. (2016) Very low-
calorie diet and 6 months of weight stability in type 2 diabetes:
pathophysiological changes in responders and nonresponders.
Diabetes Care 39: 808–815.
Stocks T, Angquist L, Banasik K et al. (2012) TFAP2B influences
the effect of dietary fat on weight loss under energy restriction.
PLoS ONE 7: e43212.
Vimaleswaran KS, Li S, Zhao JH et al. (2009) Physical activity
attenuates the body mass index-increasing influence of genetic
variation in the FTO gene. American Journal of Clinical Nutrition
90: 425–428.
Weissgerber TL, Milic NM, Winham SJ et al. (2015) Beyond bar
and line graphs: time for a new data presentation paradigm. PLoS
Biology 13: e1002128.
Williams C (1989) Diet and endurance fitness. American Journal of
Clinical Nutrition 49: 1077–1083.
Wolever TM (2016) Personalized nutrition by prediction of gly-
caemic responses: fact or fantasy? European Journal of Clinical
Nutrition 70: 411–413.
Zeevi D, Korem T, Zmora N et al. (2015) Personalized nutrition by
prediction of glycemic responses. Cell 163: 1079–1094.
© 2016 British Nutrition Foundation Nutrition Bulletin,41, 353–359
Personalised nutrition 359
