DIABETICMedicine

DOI: 10.1111/j.1464-5491.2011.03480.x

Review Article
Dorothy Hodgkin Lecture 2011*
Biomarkers for diabetes prediction, pathogenesis
or pharmacotherapy guidance? Past, present and future
possibilities

N. Sattar
BHF Glasgow Cardiovascular Research Centre, Faculty of Medicine, University of Glasgow, Glasgow, UK

Accepted 6 October 2011

Abstract
An ideal biomarker should refine identification of those at risk of disease occurrence or progression, improve prediction of
complications of disease, and or guide and help tailor responses to different therapies. Biomarkers that give insights into disease
pathogenesis are also of interest. With this in mind, this review describes biomarker studies relevant to diabetes, focusing on
those conducted by the author, his colleagues and collaborators. The review highlights several points. (1) Novel biomarkers may
not improve prediction of new-onset diabetes in a meaningful way beyond what can be achieved with simple measures combined
with HbA1c, and a sensible way ahead may be to combine diabetes and cardiovascular disease prediction using HbA1c and such
measures. (2) In terms of disease pathogenesis, associations do not necessarily infer causality; potential for residual confounding
and reverse causality should always be borne in mind. The potential relevance of such issues to understanding the relationship of
some topical variables pathways, namely adiponectin, inflammation and vitamin D, with diabetes will be highlighted. (3) How
baseline and serial data on biomarkers arising from the liver have improved our understanding of the role of hepatic fat in
diabetes pathogenesis will be explored. (4) Future goals for diabetes biomarker research should focus on predicting
complications and determining subgroups who may respond better to particular therapies. (5) All novel biomarker research
(regardless of analytical platforms used) needs to be tested against information available from commonly measured variables
in clinical practice. Otherwise, many claims of clinical utility can be exaggerated. In summary, biomarker research in diabetes is
continuing apace in a number of areas, but it remains to be seen whether the promise of biomarker research to improve the care of
our patients becomes a reality.
Diabet. Med. 29, 513 (2012)
Keywords biomarkers, diabetes, fatty liver, prediction

Abbreviations 25-OHD, 25-hydroxyvitamin D; ALT, alanine aminotransferase; CRP, C-reactive protein; GGT,
gamma-glutamyl transferase; NAFLD, non-alcoholic fatty liver disease

Despite multiple studies worldwide, the biomarker field has

Introduction
Biomarkers are characteristics, often blood parameters, which
are objectively measured and tested as indicators of normal
biological processes, pathological processes and pharmaco-
logical responses. The topic of biomarkers in health and
disease, particularly in cardiovascular disease and diabetes, has
expanded rapidly in the last decade.
Correspondence to: Naveed Sattar, Professor of Metabolic Medicine, BHF
Glasgow Cardiovascular Research Centre, Faculty of Medicine, University of
Glasgow, Glasgow G12 8TA, UK. E-mail: naveed.sattar@glasgow.ac.uk
*The 2011 Dorothy Hodgkin Lecture was delivered to the Annual Professional
Conference of Diabetes UK, London, 31st March 2011
been hindered by the general lack of statistically powerful studies
addressing relevant or meaningful clinical questions [1,2].
Advances of scientific or clinical relevance require larger
studies, efficient collaboration and an eye to future clinical
translation. They also require meticulous attention to assay and
platform characteristics and cost considerations [2]. Perhaps
most importantly, however, wherever possible biomarker work
should be hypothesis-driven (including validation of findings
from hypothesis-free studies) with the aim of addressing one
of the following: (1) improved understanding of disease
pathogenesis; (2) meaningful improvement in the prediction of
incident events complications; (3) better treatment targeting;
and or (4) monitoring of treatment efficacy. In this article, I

2011 The Author.

Diabetic Medicine 2011 Diabetes UK 5
DIABETICMedicine Biomarkers for diabetespast, present and future possibilities N. Sattar

review biomarker progress made in each of these areas in the field
of Type 2 diabetes, focusing particularly on work from my group
and collaborators. I also provide suggestions of how to improve
future research in this area.
A. Biomarkers and diabetes pathogenesis
Liver fat surrogates: supporting a role for hepatic fat in
diabetes pathogenesis?
In 2004, we were amongst the first to show that raised alanine
aminotransferase (ALT) levels were predictive of new-onset
diabetes; men in the top quarter (> 29 U l) of baseline ALT vs.
those in bottom quarter (< 17 U l) had an adjusted odds ratio of
2.04 (95% CI 1.163.58) for incident diabetes in the West of
Scotland Coronary Prevention Study (WOSCOPS) [3]. Since
then, others have corroborated and extended such findings; a
subsequent meta-analysis with collaborators confirmed both
ALT and gamma-glutamyl transferase (GGT) to be predictive
of diabetes independently of age, sex, adiposity and other
commonly considered predictors [4].
Subsequently, in 2007 we were the first to show that sustained
elevation over time in ALT levels, and to a lesser extent
triglyceride levels, accompanied progression to Type 2 diabetes
in men at risk [5]. By contrast, although higher at baseline,
subsequent changes in weight, blood pressure, HDL cholesterol
and white cell count were not greater in men who developed
diabetes compared with those remaining free from diabetes.
Such findings tracking changes in measures over time provided
additional evidence for hepatic fat accumulation as a
contributing factor for conversion to diabetes.
Placing results from related studies into a wider context and
clinical translation is helpful in medical research. In 2008 we
argued [6] that many established or proposed biomarkers
associated with diabetes risk [e.g. ALT, GGT, triglyceride,
plasminogen activator inhibitor (PAI-1), tissue plasminogen
activator (t-PA) antigen, ferritin, C-reactive protein (CRP), sex
hormone-binding globulin (SHBG), etc.] arise from the liver
(albeit to differing extents) and that their circulating levels can be
variably linked to excess liver fat. The same argument, of course,
is true for fasting glucose levels, which are determined in the main
by hepatic gluconeogenesis, in turn associated with liver fat
levels.
In clinical terms, it is possible to describe a commonly seen
pattern in clinical practice indicative of excess liver fat or non-
alcoholic fatty liver disease (NAFLD): overweight, high waist
circumference, elevated fasting glucose, triglyceride and ALT
and GGT levels (Fig. 1). Such a pattern contrasts with the pattern
associated with pure alcohol excess, as we recently described in a
widely cited review [7] (Table 1). Of course, some patients at risk
of diabetes also drink to excess so they may exhibit a mixed
pattern, whereas a rising aspartate aminotransferase (AST) level
over time in those initially diagnosed with NAFLD can signal
development of non-alcoholic steatohepatitis (NASH).
Liver fat: chicken or egg for hepatic insulin resistance?
Many researchers continue to explore potential mechanistic links
between excess liver fat and diabetes. There remains debate in

FBG or frank
Subcutaneous diabetes
fat stores
saturated? Trigs

ALT > AST

GGT

Process accelerated in at-risk groups? e.g.

Men
If family history of diabetes
Certain ethnicities, e.g. South Asians
If history of gestational diabetes/PCOS

FIGURE 1 Hepatic fat excesssigns suggestive of non-alcoholic fatty liver disease (NAFLD)? This simple illustration proposes that ectopic fat accumulation
will occur in individuals near or at a point when their subcutaneous fat storage capacity is nearly exhausted. If correct, the concept which emerges is one whereby
individuals, dependent on their age, gender, ethnicity and family history of diabetes, will have differing capacity to store subcutaneous fat with, for example, an
Asian man with a family history of diabetes having a very low capacity such that he would develop diabetes at much lower BMI, and lower average age. At the
point of subcutaneous fat saturation, continued excess calories will contribute to fat gain in the liver along with other organs (including the pancreas) with a
resultant increase in hepatic insulin resistance, continued gluconeogenesis and rise in fasting sugar levels. Excess liver fat will also reveal itself by its greater
production of triglyceride-rich lipoproteins and so circulating triglyceride levels are increased (representing another example of ectopic fat). In addition, liver
enzymes, in particular alanine aminotransferase (ALT) and gamma-glutamyl transferase (GGT) levels, will be higher. Future research should help determine
what factors dictate subcutaneous fat storage capacity and also to what extent ectopic fat gain in the pancreas contributes to beta-cell dysfunction and thus
diabetes and is reversible? AST, aspartate aminotransferase; FBG, fasting blood glucose; PCOS, polycystic ovary syndrome; Trigs, triglycerides.

2011 The Author.

6 Diabetic Medicine 2011 Diabetes UK
 Review article
Table 1 Common features recognised in non-alcoholic fatty liver disease
(NAFLD) and alcoholic liver disease (adapted from ref 7)
Alcoholic liver
Feature NAFLD disease
ALT
AST
ALT AST ratio ALT AST > 1.0 ALT AST < 1.0
(may be reversed in
hepatocellular
necrosis)
GGT
Mean corpuscular
volume
Weight
Fasting plasma
glucose
HDL cholesterol
History of alcohol
intake
Triglyceride or
Typical clinical and biochemical patterns in NAFLD and
alcoholic liver disease. Some patients can, of course, show
mixed patterns. This guide illustrates that non-alcoholic fatty
liver disease is often suspected even before any imaging
necessarily takes place.
ALT, alanine aminotransferase; AST, aspartate aminotransfer-
ase; GGT, gamma-glutamyl transferase. NAFLD, non-alcoholic
fatty liver disease
somequartersoverwhetherexcessliverfatissimplyaconsequence
of peripheral insulin resistance driving hyperinsulinaemia, in turn
driving de novo lipogenesis in the face of continued caloric excess,
or whether excess calories, perhaps particularly excess
carbohydrate, directly drive de novo hepatic lipogenesis, with
subsequent deteriorationin hepatic insulin resistance. Support for
liver fat being a cause of insulin resistance is increasing and the
reader is referred to some recent excellent reviews that discuss
mechanisms in more detail [8,9]. Irrespective of mechanisms, it is
increasingly clear that excess liver fat is common in the majority of
patientswithType 2diabetes(10),andthatNAFLDisastrongrisk
factor for the development of diabetes. Furthermore, an emerging
concept, whereby individuals with a lower subcutaneous storage
capacity develop NAFLD at lower average BMI, is being
examined, suggesting that individuals have differing set points
for accumulating ectopic fat stores. Such a concept would fit
broadly with certain ethnic groups (e.g. South Asians) being at
higher risk for diabetes at any given BMI [11], or indeed men being
at higher risk of Type 2 diabetes than women for a given BMI, an
observation we recently confirmed using Scottish Diabetes
Research Network data [12].
NAFLD as a risk indicator?
This is an area of major study given the rising prevalence of
NAFLD and the concerns that NAFLD is not only a risk factor
for diabetes but also liver cirrhosis and hepatocellular carcinoma.
2011 The Author.
Diabetic Medicine 2011 Diabetes UK
DIABETICMedicine
We reviewed the broad risks for conversion to such conditions
over time in the aforementioned review [7] and have
subsequently suggested that all patients with NAFLD should be
screened for diabetes, if not already screened [13]. Using GGT as
a biochemical surrogate of liver fat, we were also amongst the
first to show that greater levels of physical activity may help keep
liver fat down [14]. More recently, we showed higher fibre intake
was related to lower risk of developing diabetes, and that this
association was attenuated by adjusting for GGT levels [15].
These observations fit with the broad concept that factors which
lower diabetes risk (i.e. better diet, greater activity) will also
lower NAFLD risk or progression. Several ongoing clinical trials
are examining the relevance of lifestyle changes to NAFLD
progression.
Finally, some researchers have tried to make the case that, in
addition to being a strong diabetes risk factor, NAFLD is also a
strong cardiovascular disease risk factor. In reality the evidence
for this assertion lacks clear proof of an independent association.
Thus, we have suggested that rather than labelling all NAFLD
patients at high risk of cardiovascular disease, they should
undergo risk screening using established cardiovascular disease
risk scores [13]. This suggestion makes clinical sense as
prescribing a statin in patients with NAFLD at low
cardiovascular disease risk, which may number many, may
even accelerate the risk of developing diabetes [16] but achieve
little absolute cardiovascular disease risk reduction.
Inflammation
Using the valuable West of Scotland Coronary Prevention Study
data resource, our group were amongst the first to show that
CRP levels predict incident diabetes [17], a finding since
corroborated by many others. Subsequently, we have reported
a positive association between interleukin 6 (IL-6) and new-
onset diabetes that appears to be independent of obesity and
insulin resistance [18]. This observation is important as CRP
appears not to be causally related to insulin resistance [19] or
obesity [20]. Whether upstream cytokines such IL-6 are causally
linked to diabetes therefore deserves further investigation. The
present review is not able to comprehensively examine the
inflammationdiabetes arena. Nevertheless, the important point
to emphasize is that the extent to which inflammatory processes
occurring in several tissues (adipose tissue, liver, vasculature,
heart and pancreas, etc.) are causally related to diabetes remains
unclear. Indeed, presently we cannot reject the alterative
hypothesis that insulin resistance, obesity and diabetes drive
inflammatory perturbances, which subsequently do not, or only
negligibly, contribute to further glucose dysregulation (Fig. 2).
The best way to dissect out causality in this complex minefield
of pathways is to intervene with novel anti-inflammatory
agents in patients with diabetes, and here several agents using
generic (e.g. salsalate [21]) or more focused anti-inflammatory
agents (e.g. IL-1 pathways blockers) are being tested in the
metabolic field. Genetic studies can also help in this regard, but
the acid test will be intervention studies.
7
DIABETICMedicine Biomarkers for diabetespast, present and future possibilities N. Sattar
Model 1
Risk factors Inflammatory Type 2
obesity, pathways/markers
social deprivation
CRP, IL-6, IL-1etc diabetes
ethnicity
Model 2
Inflammatory
pathways
Risk factors
obesity, Type 2
social deprivation
Insulin
ethnicity resistance or diabetes
relative insufficiency
FIGURE 2 Inflammation: cause or consequence of insulin resistance and
diabetes? Whilst plentiful evidence indicates associations linking
inflammation markers perturbances in a variety of tissues, including, in
particular, adipose tissue and diabetes or insulin resistance, the causal
direction of this association remains uncertain. Model 1 proposes such
inflammatory perturbances directly contribute to Type 2 diabetes, which
could occur via effects on insulin resistance or b-cell function. By contrast,
in model 2, inflammatory pathways are dysregulated as a consequence of
metabolic perturbance (i.e. insulin resistance, relative insulin insufficiency or
dysglycaemia) rather than the reverse pathway. Future focused
inflammatory-based interventions and genetic studies will help better
unravel directional associations and clinical relevance of such associations.
Presently, as described in the text, measures of inflammation, in particular
C-reactive protein (CRP), are not needed to help predict diabetes. IL,
interleukin.
Other pathways relevant to diabetes pathogenesis?
A fuller account of other pathways potentially relevant to
diabetes pathogenesis is not possible here, but the reader is
referred to our review from 2008 as a starting point for further
relevant reading [6]. However, two parameters deserve particular
mention as they illustrate the complexities in attributing causality
in diabetes pathogenesis on the basis of epidemiological
data alone. These are adiponectin and vitamin D.
Adiponectin
Adiponectin is a protein released from fat cells and commonly
considered to provide a signal to peripheral tissues to improve
insulin action. In line with its presumed effects, circulating
adiponectin is not only lower in obese individuals, but it also
correlates with insulin sensitivity and related metabolic
phenotypes (e.g. lower triglyceride, higher HDL cholesterol,
lower CRP, etc.) and does so independently of BMI [15]. Higher
levels also predict lower diabetes risk, suggesting an important
insulin-sensitizing role for adiponectin. However, in a well-cited
paper we showed that higher adiponectin levels actually predict
higher, not lower, all-cause and vascular mortality in men above
60 years of age [22]. The same finding has subsequently been
reported in several other patient populations, including those
with renal disease, Type 1 diabetes, angina and heart failure, as I
recently reviewed [23]. Based on further work from the British
Regional Heart Study [24], we recently hypothesized a potential
contributing mechanism for this paradoxical observationnam-
ely, that brain naturetic peptide (BNP) released from stressed
8 Diabetic Medicine 2011 Diabetes UK
myocardium in the elderly or those at vascular risk may promote
adiponectin release from adipose tissue, and thus lead to higher
circulating adiponectin levels. In this way, high adiponectin
may serve as a compensatory signal to vascular stress (Fig. 3). Of
course, other mechanisms may also contribute to this
adiponectin paradox, but clearly these studies remind us that
epidemiological findings in isolation cannot determine causality.
To further enhance complexity in this area, other researchers
have elegantly shown that adiponectin in man (as opposed to
findings in animals) is more likely to be a downstream inverse
signal of hyperinsulinaemia and, in turn, of insulin resistance,
rather than an upstream negative regulator [25]. Such a finding
explains why adiponectin levels are higher in Type 1 diabetes.
The adiponectin area is therefore ripe for further detailed study
including more robust genetic studies to determine causal
associations, or otherwise.
Vitamin D
There are ample studies to suggest low levels of
25-hydroxyvitamin D (25OHD) predict incident diabetes in
later life, and plentiful mechanisms have been suggested to
explain such a link. However, some recent prospective studies do
not confirm an inverse association between 25OHD levels and
Insulin
resistance Insulin sensitivity
Hyper HDL cholesterol, CRP
insulinaemia
+ve Diabetes
ve
ve
Adiponectin
+ve
+ve
BNP Total and
CVD
mortality
FIGURE 3 Adiponectin, diabetes and cardiovascular disease: a complex
picture. Most readers understand adiponectin is associated with higher
insulin sensitivity and associated metabolic pattern [lower triglyceride,
higher HDL cholesterol, lower C-reactive protein (CRP), lower weight] in
humans (green lines). Indeed, multiple prospective studies have linked higher
adiponectin with lower risk for incident diabetes. However, two factors
less well appreciated, but which add complexity, must be considered. First,
there is now a wealth of evidence, including data from our own group,
demonstrating higher adiponectin levels predict higher all-cause and
vascular mortality in at-risk groups, including the elderly (red lines).
Additional work from our group suggests brain naturetic peptide (BNP)
released from stressed myocardium may provide a mechanism underpinning
this latter association: released brain naturetic peptide may stimulate
adiponectin release from adipose tissue. This area clearly requires further
study. Secondly, the orange lines depict findings questioning whether
adiponectin is genuinely insulin sensitizing in humans? Indeed, most data (as
reviewed in [25]) would suggest hyperinsulinaemia suppresses adiponectin
levels, i.e. adiponectin suppression reflects insulin resistance rather than the
reverse pathway. CVD, cardiovascular disease.
2011 The Author.
 Review article DIABETICMedicine

incident diabetes. For example, we recently showed that 25OHD
levels measured near the first antenatal visit did not predict
gestational diabetes [26]. Perhaps more critically, small dietary
vitamin D intervention studies do not show any clear benefit on
glycaemia measures, as recently reviewed [27]. The area
therefore requires much larger studies to confirm or refute a
causal relationship. Furthermore, the vitamin D story offers
some lessons for medical researchers. Firstly, the possibility of
reverse causality, whereby disease occurrence leads to lower
vitamin D levels, must be considered. Equally, residual
confounding, whereby other lifestyle factors generating higher
diabetes risk, but not able to be completely adjusted for, must be
considered. For example, we know well that sedentary behaviour
increases diabetes risk and, as such individuals would spend less
time outdoors, this leads in turn to less sun exposure and lower
circulating 25OHD levels. Also, 25OHD levels appear to exhibit
acute-phase response behaviour, so that patients with higher
inflammatory status will have lower levels. These broad concepts
are important to remember and pertain to careful consideration
of vitamin Ds potential role, or otherwise, in a wide range of
conditions beyond diabetes (e.g. autoimmune conditions as we
recently reviewed [28], see Fig. 4).
Future pathogenic insights from biomarker research?
The emergence of omics-based technologies will no doubt
throw up many new potential pathways towards diabetes
development, but in each case the same principles must be
taken into account, namely consideration of residual
confounding or reverse causality. Also, it is important to
remember that obesity or excess weight gain remains the major
upstream driver for development of Type 2 diabetes working via
some of the mechanisms alluded to herein. That noted, improved
omics technologies combined to better genetic research should
help us better understand the molecular pathways linking obesity
to diabetes in populations at differing risk and could, in time,
result in development of novel therapies. Such studies should of
course not detract from public health or governmental measures
aimed at addressing the nations expanding waistlines.
B. Biomarkers for diabetes prediction?
Diabetes prediction is an area which has received considerable
attention, but some recent developments have changed the
landscape such that the role of any new blood biomarker being
clinically useful in prediction is perhaps now difficult to imagine.
Furthermore, other pragmatic considerations must dictate how
we proceed with diabetes screening in clinical practice. I have
listed four points which are often less well considered but deserve
attention in this area:
No need for perfection. To better target those at greater
risk of a heart attack or stroke or death from such events,
cardiovascular disease risk prediction has to be as good as
possible. By contrast, as diabetes diagnosis is defined on a
blood test rather than an event, and diabetes risk
assessment can be repeated over time as an individuals
risk escalates or becomes more apparent, prediction of
diabetes is perhaps less critical to perfect.
Simple risk scores can get us far. There are many
published risk scores for diabetes, but most share the same
few major components; namely, age, gender, BMI and or
waist circumference, family history of diabetes, blood

Known 25OHD determinants Possible 25OHD determinants

Age Ethnicity Smoking Diet BMI Latitude Physical activity Social class Systemic inflammation Other?

Sunlight exposure

B: Reverse
causality C: Residual
Serum 25OHD
confounding
Sedentary behaviour
chronic pain, institutional A: Causality
care, etc less time
outdoors

Outcomes
Incident diabetes
Poor cancer prognosis
Incident CVD
Worsening rheumatoid
Mortality

FIGURE 4 Associations of 25-hydroxyvitamin D (25OHD) with other risk pathways, lifestyle factors and with risk of disease outcomes, including
Type 2 diabetes. This diagram, modified from Welsh et al. (2011) [28], with permission from John Wiley & Sons, shows associations of circulating
25-hydroxyvitamin D with outcomes that may represent true causal pathways (pathway A), be subject to reverse causality (pathway B) or to residual
confounding (pathway C), or some combination of the three possibilities. The point of this diagram is to show that epidemiological results in isolation should
not necessarily be taken as representing causal associations. CVD, cardiovascular disease.

2011 The Author.

Diabetic Medicine 2011 Diabetes UK 9
DIABETICMedicine Biomarkers for diabetespast, present and future possibilities N. Sattar

pressure and ethnicity. Broadly speaking, all such scores
perform quite well as indicators of diabetes risk, achieving
receiver operating characteristic (ROC) scores around
0.80 [6]. They can therefore direct healthcare
professionals to individuals at higher diabetes risk and
help select those who could be better targeted for blood
testing for diabetes. Interestingly, on this basis we recently
showed that, if subjects in the top 40% of a simple risk
score were selected for blood testing, then only 3% of
those classified as low risk on the simple score would have
been reclassified as high risk (> 6% risk) on the basis of
blood markers [29]. In other words, few subjects would be
misclassified as low risk based on questionnaire alone.
This is not a major concern as some of the 3%
misclassified might be picked up in other ways as their
risk escalates over time. Thus, a simple risk score could
function well as an initial filter to determine who should be
tested for future or present risk of diabetes using blood
tests. Of course, it should be noted that simple risk scores
for diabetes can be easily determined in any given country
and that, once taken up, the key is to determine what the
cut-points are for moving forward to a blood test.
HbA1c enters the arena. Acceptance of HbA1c use for
diabetes diagnosis is gathering pace and values
48 mmol mol (6.5%) are considered diagnostic
according to American Diabetes Association (ADA)
criteria. Upon acceptance and implementation, the next
obvious question is how we define those at elevated risk for
developing diabetes. A simple pragmatic approach is to
base subsequent risk only on HbA1c level, with one option
being that individuals with HbA1c values from 42 to
47 mmol mol (6.0 to 6.4%) should be considered at high
risk of future diabetes and offered lifestyle intervention.
We can do betterthan this of course. In the aforementioned
British Region Heart Study [29], we showed that HbA1c
combined with other simple variables already gathered in a
simple risk score enhances risk prediction for future
diabetes. Interestingly, GGT levels but not CRP improved
prediction even further, albeit modestly, suggesting liver
tests may be of value in a more detailed risk score.
However, in this more complex approach, the physician
would be required to enter relevant results (age, gender,
BMI, family history of diabetes, HbA1c, possibly GGT,
etc.) into a risk algorithm calculator to determine future
diabetes risk, in much the same way cardiovascular disease
risk screening is conducted.
A need to combine cardiovascular disease and diabetes
risk screening. Whatever process is taken up, it is clear that
diabetes screening cannot occur without attention to
cardiovascular disease risk screening. We have recently
argued the rationale for combined cardiovascular and
diabetes risk screening and suggested two ways this could
be conducted [30] (Fig. 5). One option is a two-stage
process, as described above, whereby HbA1c is only added
to blood tests for lipids, liver function and renal measures
in individuals in the top 40% or so of diabetes risk on the

Combined CVD/diabetes screening:

age, gender, prior CVD, family history of CVD, smoking,
ethnicity, social class

Simple measurements:
Blood pressure, BMI

ADD HbA1c to non-fasting

lipids, urea and electrolytes, liver function tests

< 42 mmol/mol (6.0%) 48 mmol/mol (6.5%)

Low risk:
General advice Diabetes
Rescreen in 3 years 4247 mmol/mol (6.06.4%)

High risk:
Lifestyle advice, weight loss
Rescreen in 1 year

FIGURE 5 Combined cardiovascular and diabetes screening in future? This figure, adapted from Preiss et al. (2011) [30], with permission from John Wiley &
Sons (supplied by Diabetic Medicine) shows a one-step process (population-based strategy) for diabetes health check in which HbA1c is measured in all
individuals who are being targeted for cardiovascular risk screening. Of course, a two-step process to target high-risk individuals for diabetes health checks,
whereby HbA1c is only measured after risk scoring in high-risk individuals, may be preferable. However, the best pragmatic process to select these high-risk
individuals needs to be put in place and may differ in differing regions or countries. CVD, cardiovascular disease.

2011 The Author.

10 Diabetic Medicine 2011 Diabetes UK
 Review article
basis of a simple pre-blood test diabetes risk score. The
second option is to perform HbA1c testing on all
individuals targeted for cardiovascular risk screening
regardless of baseline diabetes risk, but this would
involve a greater number of HbA1c tests, a potential cost
concern for laboratories across the country.
Diabetes risk screening has been discussed for many years, but
rarely implemented given its complexity and the need for
multiple repeated tests. However, the advent of HbA1c as a
diagnostic test for diabetes has opened up the possibility to
advance this process to the next stage. Many primary care
physicians are already considering using HbA1c in their practice
to test for diabetes in patients they judge to be at risk. Given
HbA1c testing does not require fasting and, given that
non-fasting lipid profiles perform as well as fasting profiles
for cardiovascular risk prediction [31], opportunities to screen
for diabetes throughout the day arise. Interestingly, the
ADDITION study did not prove earlier diabetes screening
lessens cardiovascular disease risk, in part because routine
cardiovascular risk care raised its performance in the control
group to a near similar level as the intervention group, but it did
show relative stability of HbA1c over the 5 years of the study in
the intervention and active groups, as well as minimal BMI
change over this same period [32]. This observation suggests that
earlier identification of diabetes may increase patients chances
of adopting sustainable lifestyle changes. This suggestion is
speculative, but, clearly, picking up diabetes sooner after HbA1c
crosses the diagnostic threshold could have multiple advantages
for patients.
C. Biomarkers for complications, treatment
guidance?
Biomarker and cardiovascular risk in diabetes
There are many cardiovascular risk scores in diabetes, the most
widely used being the UK Propsective Diabetes Study (UKPDS)
risk engine. However, to what extent cardiovascular risk scoring
is required or should be used in patients with Type 2 diabetes
remains debatable. Currently, patients with diabetes aged
40 years and older in the UK are considered, albeit incorrectly,
as having a coronary heart disease risk level equivalent to
those with existing vascular disease. Nevertheless, lifetime
cardiovascular disease risk in diabetes is still likely to be high
and statin use in diabetes has been shown to be cost-effective [33].
Newer risk calculators are being determined and validated in
several parts of the world and the strengths and weaknesses of
differing approaches require further discussion. Suffice to say
most patients with diabetes above 40 years of age will continue
to be offered statins, with more intensive statin therapy restricted
to those with existing cardiovascular disease.
In a recent analysis of mortality and cardiovascular event rates
in participants with diabetes in large clinical trials, we noted that
history of either cardiovascular disease or any evidence of
2011 The Author.
Diabetic Medicine 2011 Diabetes UK
DIABETICMedicine
proteinuria at baseline in patients with diabetes is associated with
substantially higher rates of death and cardiovascular disease
events compared with those with predominantly uncomplicated
diabetes [34]. In this way, we highlighted proteinuria as a very
strong signal for death and thus our work would suggest that
those with proteinuria also deserve more potent statin therapy.
This may be increasingly achievable in economic terms given that
atorvastatin is shortly coming off patent.
Our group has also studied many other risk factors, but most
notably we have helped advance data on CRP in general [35] and
in diabetes, an area which continues to attract considerable
debate. More recently, we showed that N-terminal pro-brain
natriuretic peptide (NT-proBNP) correctly reclassified
cardiovascular disease risk in approximately three times more
elderly men than did CRP [36], a finding broadly consistent with
data in a recent meta-analysis [37]. Such data suggest brain
naturetic peptide (or NT-proBNP) demands further study as a
potentially valuable cardiovascular risk factor, including in
patients with diabetes.
Biomarkers and microvascular complications?
Perhaps most gains for future biomarker research in diabetes will
come for the prediction of microvascular complications. Current
data are sparse but recent studies using omics technologies
suggest proteomic signatures can help identify individuals at risk
of chronic kidney disease [38]. However, such work is currently
in its infancy and, as yet, not guaranteed to reach clinical practice.
To speed up work in these areas, the following points are
relevant:
Development of large, well-phenotyped cohorts (ideally
with serial samples) and robust ascertainment of
complications is a priority.
Ideally combine novel markers, genetics and omics, and
compare these results to the best available routine tests as
the current clinical benchmark. This approach avoids
exaggerating the value of newer platforms.
Wherever possible, use robust quality-controlled
measurements and robust statistics, moving beyond
simple associations by using robust reclassification
metrics.
Wherever possible, biomarker research should consider
relevant clinical questions and aim to improve upon
current practice.
Biomarkers and differential response to therapies?
This is a complex area and will require major collaborations.
Identifying subgroups of patients who would benefit better from
one line of therapy compared with another would be helpful,
particularly as the choice for second-line therapy expands. Of
course, here the definition of benefit needs careful attention, and
recent experience has taught us not to rely solely on glucose
reductions but to consider other effects, including weight,
vascular risk and other side effects. Indeed, if biomarkers could
11
DIABETICMedicine Biomarkers for diabetespast, present and future possibilities N. Sattar
also help to predict those who would respond adversely to a
particular therapy, this would be helpful. Such work will not be
easy to conduct and also there is no guarantee the findings will
prove clinically useful. However, with rising numbers of
prospective cohorts and sophisticated platforms generating
huge numbers of measurements, there will soon be a plethora
of data addressing relevant issues. When this work is conducted,
researchers must remain cognisant of the possibility that, rather
than complex biomarkers, simple phenotypic variables may
achieve sufficient success in determining potential risks and
benefits, so that any gains from biomarkers may be limited.
Again, careful attention to platforms and statistics is needed in
such studies.
Predict declining b-cell function progression to insulin?
This is an area also less well studied. Why some patients can be
maintained for years on oral therapies whereas others move
rapidly towards insulin is often unclear, so, if biomarkers can
help better identify likely rapid progressors to insulin, such
work would be helpful, especially if it also reveals mechanisms to
slow progression.
Conclusion
In reviewing the work I and my colleagues and collaborators have
undertaken on diabetes biomarkers, I hope it is apparent that
biomarkers have helped improve knowledge on diabetes
pathogenesis, with particular gain of knowledge in relation to
non-alcoholic fatty liver disease (via multiple inter-related
biomarkers) and its links to diabetes. Future pathogenesis work
will require more joined-up research, with parallel genetic and
phenotypic linkages. In terms of predicting diabetes, I have
suggested that novel biomarkers are unlikely to meaningfully
advance clinical practice in this area. Rather, with the advent of
HbA1c in diagnostics algorithms, we have an opportunity to
develop simple and pragmatic combined diabetes and
cardiovascular disease risk screening. A number of questions
remain if we are to adopt routine diabetes screening, such as a
need for, and implementability of, a simple risk score to direct
relevant blood tests (i.e. HbA1c) to higher-risk subjects. Such
questions could be readily addressed by expert groups to allow
this area to move forward more rapidly. Finally, the real future
action for biomarkers in diabetes is likely to be prediction of
complications and for treatment guidance. Once again, this field
will require careful attention to a number of factors, most
important being robust collection of phenotypic information
from prospective studies, robust measurement of relevant
biomarkers, and careful and comprehensive statistical analyses,
taking into account prediction already achievable by simpler
measures. In all such studies, there is a need for collaboration and
validation, recognizing that only with larger studies (and greater
power) and a consistency of evidence will we determine robust
associations of potential clinical relevance. In short, biomarker
research in diabetes is very much in its infancy. Given rising
12
diabetes levels worldwide, continued biomarker research along a
number of avenues as reviewed herein seems appropriate.
However, only in time will we determine if the promise of
biomarker research to improve the care of our patients becomes a
reality.
Competing interests
Nothing to declare.
Acknowledgements
The author is grateful to Dr David Preiss for his valuable
comments on this article and many relevant contributions. He
would also like to acknowledge the excellent biomarker expertise
of Dr Paul Welsh and the technical skills of Dr Lynne Cherry,
Elaine Butler and colleagues of the Clinical Biochemistry
Department of Glasgow Royal Infirmary. He is grateful to his
many collaborators throughout the UK, with particular
recognition in the area of diabetes biomarkers to Professor
Goya Wannamethee, Dr Nita Forouhi and Professor Deborah
Lawlor. Finally, he is grateful to Diabetes UK, the Medical
Research Council, The Wellcome Trust, Chest Heart Stroke,
Scotland and the British Heart Foundation for their funding of
relevant studies.
References
1 Ioannidis JP, Panagiotou OA. Comparison of effect sizes associ-
ated with biomarkers reported in highly cited individual articles
and in subsequent meta-analyses. J Am Med Assoc 2011; 305:
22002210.
2 Sturgeon C, Selby P. Development and validation of new biomar-
kers: major opportunities for health-care scientists. Ann Clin
Biochem 2010; 47: 499502.
3 Sattar N, Scherbakova O, Ford I, OReilly DS, Stanley A, Forrest E
et al. Elevated alanine aminotransferase predicts new-onset type 2
diabetes independently of classical risk factors, metabolic
syndrome, and C-reactive protein in the West of Scotland Coronary
Prevention Study. Diabetes 2004; 53: 28552860.
4 Fraser A, Harris R, Sattar N, Ebrahim S, Davey Smith G, Lawlor
DA. Alanine aminotransferase, gamma-glutamyltransferase, and
incident diabetes: the British Womens Heart and Health Study and
meta-analysis. Diabetes Care 2009; 32: 741750.
5 Sattar N, McConnachie A, Ford I, Gaw A, Cleland SJ, Forouhi NG
et al. Serial metabolic measurements and conversion to type 2
diabetes in the West of Scotland Coronary Prevention Study:
specific elevations in alanine aminotransferase and triglycerides
suggest hepatic fat accumulation as a potential contributing factor.
Diabetes 2007; 56: 984991.
6 Sattar N, Wannamethee SG, Forouhi NG. Novel biochemical risk
factors for type 2 diabetes: pathogenic insights or prediction
possibilities? Diabetologia 2008; 51: 9269240.
7 Preiss D, Sattar N. Non-alcoholic fatty liver disease: an overview of
prevalence, diagnosis, pathogenesis and treatment considerations.
Clin Sci (Lond) 2008; 115: 141150.
8 Taylor R. Pathogenesis of type 2 diabetes: tracing the reverse route
from cure to cause. Diabetologia 2008; 51: 17811789.
9 Stefan N, Kantartzis K, Haring HU. Causes and metabolic conse-
quences of fatty liver. Endocr Rev 2008; 29: 939960.
2011 The Author.
Diabetic Medicine 2011 Diabetes UK
 Review article
10 Tolman KG, Fonseca V, Dalpiaz A, Tan MH. Spectrum of liver
disease in type 2 diabetes and management of patients with diabetes
and liver disease. Diabetes Care 2007; 30: 734743.
11 Forouhi NG, Sattar N. CVD risk factors and ethnicitya homo-
geneous relationship? Atheroscler Suppl 2006; 7: 1119.
12 Logue J, Walker JJ, Colhoun H, Leese GP, Lindsay RS, McKnight
JA et al.. on behalf of The Scottish Diabetes Research Network
Epidemiology Group. Do men develop type 2 diabetes at lower
body mass indices than women? Diabetologia (In Press) [Epub
ahead of print].
13 Ghouri N, Preiss D, Sattar N. Liver enzymes, nonalcoholic fatty
liver disease, and incident cardiovascular disease: a narrative review
and clinical perspective of prospective data. Hepatology 2010; 52:
11561161.
14 Lawlor DA, Sattar N, Smith GD, Ebrahim S. The associations of
physical activity and adiposity with alanine aminotransferase and
gamma-glutamyltransferase. Am J Epidemiol 2005; 161: 1081
1088.
15 Wannamethee SG, Tchernova J, Whincup P, Lowe GD, Rumley A,
Brown K et al. Associations of adiponectin with metabolic and
vascular risk parameters in the British Regional Heart Study reveal
stronger links to insulin resistance-related than to coronory heart
disease risk-related parameters. Int J Obes (Lond) 2007; 31: 1089
1098.
16 Sattar N, Preiss D, Murray HM, Welsh P, Buckley BM, de Craen AJ
et al. Statins and risk of incident diabetes: a collaborative meta-
analysis of randomised statin trials. Lancet 2010; 375: 735742.
17 Freeman DJ, Norrie J, Caslake MJ, Gaw A, Ford I, Lowe GD et al.
C-reactive protein is an independent predictor of risk for the
development of diabetes in the West of Scotland Coronary
Prevention Study. Diabetes 2002; 51: 15961600.
18 Wannamethee SG, Lowe GD, Rumley A, Cherry L, Whincup PH,
Sattar N. Adipokines and risk of type 2 diabetes in older men.
Diabetes Care 2007; 30: 12001205.
19 Timpson NJ, Lawlor DA, Harbord RM, Gaunt TR, Day IN, Palmer
LJ et al. C-reactive protein and its role in metabolic syndrome:
Mendelian Randomisation Study. Lancet 2005; 366: 19541959.
20 Welsh P, Polisecki E, Robertson M, Jahn S, Buckley BM, de Craen
AJ et al. Unraveling the directional link between adiposity and
inflammation: a bidirectional Mendelian randomization approach.
J Clin Endocrinol Metab 2010; 95: 9399.
21 Goldfine AB, Fonseca V, Jablonski KA, Pyle L, Staten MA, Shoelson
SE. The effects of salsalate on glycemic control in patients with
type 2 diabetes: a randomized trial. Ann Intern Med 2010; 152:
346357.
22 Wannamethee SG, Whincup PH, Lennon L, Sattar N. Circulating
adiponectin levels and mortality in elderly men with and without
cardiovascular disease and heart failure. Arch Intern Med 2007;
167: 15101517.
23 Sattar N. Adiponectin and raised mortality in type 1 diabetes: any
credible explanatory mechanisms? J Intern Med 2011; 270: 339
342.
24 Wannamethee SG, Welsh P, Whincup PH, Sawar N, Thomas MC,
Gudnarsson V et al. High adiponectin and increased risk of car-
diovascular disease and mortality in asymptomatic older men: does
NT-proBNP help to explain this association? Eur J Cardiovasc Prev
Rehabil 2011; 18: 6571.
25 Cook JR, Semple RK. Hypoadiponectinemiacause or conse-
quence of human insulin resistance? J Clin Endocrinol Metab
2010; 95: 15441554.
26 Makgoba M, Nelson SM, Savvidou M, Messow CM, Nicolaides K,
Sattar N. First-trimester circulating 25-hydroxyvitamin D levels
and development of gestational diabetes mellitus. Diabetes Care
2011; 34: 10911093.
2011 The Author.
Diabetic Medicine 2011 Diabetes UK
DIABETICMedicine
27 Sattar N. Vitamin D deficiency as a cause of diabetes? A critique of
the current evidence base. Diabetes Digest 2010; 9.
28 Welsh P, Peters MJ, Sattar N. Is vitamin D in rheumatoid arthritis a
magic bullet or a mirage? The need to improve the evidence base
prior to calls for supplementation. Arthritis Rheum 2011; 63:
17631769.
29 Wannamethee SG, Papacosta O, Whincup PH, Thomas MC,
Carson C, Lawlor DA et al. The potential for a two-stage diabetes
risk algorithm combining non-laboratory-based scores with sub-
sequent routine non-fasting blood tests: results from prospective
studies in older men and women. Diabet Med 2011; 28: 2330.
30 Preiss D, Khunti K, Sattar N. Combined cardiovascular and
diabetes risk assessment in primary care. Diabet Med 2011; 28: 19
22.
31 Di Angelantonio E, Sarwar N, Perry P, Kaptoge S, Ray KK,
Thompson A et al. Major lipids, apolipoproteins, and risk of vas-
cular disease. J Am Med Assoc 2009; 302: 19932000.
32 Griffin SJ, Borch-Johnsen K, Davies MJ, Khunti K, Rutten GE,
Sandbaek A et al. Effect of early intensive multifactorial therapy on
5-year cardiovascular outcomes in individuals with type 2 diabetes
detected by screening (ADDITION-Europe): a cluster-randomised
trial. Lancet 2011; 378: 156167.
33 Raikou M, McGuire A, Colhoun HM, Betteridge DJ, Durrington
PN, Hitman GA et al. Cost-effectiveness of primary prevention of
cardiovascular disease with atorvastatin in type 2 diabetes: results
from the Collaborative Atorvastatin Diabetes Study (CARDS).
Diabetologia 2007; 50: 733740.
34 Preiss D, Sattar N, McMurray JJ. A systematic review of event rates
in clinical trials in diabetes mellitus: the importance of quantifying
baseline cardiovascular disease history and proteinuria and impli-
cations for clinical trial design. Am Heart J 2011; 161: 210
219 e1.
35 Sattar N, Murray HM, McConnachie A, Blauw GJ, Bollen EL,
Buckley BM et al. C-reactive protein and prediction of coronary
heart disease and global vascular events in the Prospective Study of
Pravastatin in the Elderly at Risk (PROSPER). Circulation 2007;
115: 981989.
36 Wannamethee SG, Welsh P, Lowe GD, Gudnason V, Di Angelan-
tonio E, Lennon L et al. N-terminal pro-brain natriuretic peptide is
a more useful predictor of cardiovascular disease risk than
C-reactive protein in older men with and without pre-existing
cardiovascular disease. J Am Coll Cardiol 2011; 58: 5664.
37 Di Angelantonio E, Chowdhury R, Sarwar N, Ray KK, Gobin R,
Saleheen D et al. B-type natriuretic peptides and cardiovascular risk:
systematic review and meta-analysis of 40 prospective studies.
Circulation 2009; 120: 21772187.
38 Mullen W, Delles C, Mischak H. Urinary proteomics in the
assessment of chronic kidney disease. Curr Opin Nephrol Hyper-
tens 2011; 20: 654661.
Supporting Information
Additional Supporting Information may be found in the online
version of this article:
Appendix 1. PowerPoint Presentation.
Please note: Wiley-Blackwell are not responsible for the content
or functionality of any supporting materials supplied by the
authors. Any queries (other than for missing material) should be
directed to the corresponding author for the article.
13