Comment
effect change. Although a placebo control might have
confirmed outcomes, it is unethical to withhold treatment
from patients with cystic-fibrosis-related diabetes.
A possible alternative approach would be to do placebo-
controlled trials in individuals with cystic fibrosis who are
insulin insufficient but have not yet developed diabetes,
in whom placebo treatment would be ethically sound.
For now, I believe that the jury is still out on the use of oral
antidiabetes drugs for the treatment of cystic-fibrosis-
related diabetes.
Antoinette Moran
Department of Pediatrics, Division of Pediatric Endocrinology,
University of Minnesota, Minneapolis, MN 55454, USA
moran001@umn.edu
Genetics and biobanks converge to resolve a vexing
knowledge gap in diabetes
The outdated designations of type 1 diabetes as “juvenile-
onset diabetes” and type 2 diabetes as “adult-onset
diabetes” reflect the incorrect categorical paradigm that
has hindered the ability of clinicians to correctly diagnose
outlier cases, causing unnecessary delays in instituting the
appropriate therapy. The age-based distinction between
the two most common forms of diabetes has become
increasingly blurred, as type 2 diabetes becomes more
frequent at younger ages and autoimmune diabetes in
adulthood is better recognised. However, identification
of autoimmune diabetes (whether type 1 diabetes or
latent autoimmune diabetes of adults [LADA]) is largely
dependent on the astuteness of clinicians and the
availability of autoantibody testing. Thus, the diagnosis of
autoimmune diabetes is all too frequently made post-hoc,
after a patient with absolute insulin deficiency secondary
to autoimmune β-cell destruction has developed life-
threatening ketoacidosis while inadequately managed on
oral antihyperglycaemic drugs.
This historical barrier to the diagnosis of autoimmune
diabetes in adulthood is compounded by the widespread
use of insulin in severe or advanced diabetes, which
can mask the presence of an autoimmune process
in adult patients by protecting against ketoacidosis.
Furthermore, autoantibody testing is not always done at
the time of diagnosis because of its low positive predictive
value in the context of the relatively low prevalence of
www.thelancet.com/diabetes-endocrinology Vol 6 February 2018 87
I serve on a data safety monitoring board for a GLP-1 agonist study for Novo
Nordisk in young people with type 2 diabetes.
1 Schwarzenberg SJ, Thomas W, Olsen TW, et al. Microvascular complications
in cystic fibrosis-related diabetes. Diabetes Care 2007; 30: 1056–61.
2 Moran A, Becker D, Casella SJ, et al. Epidemiology, pathophysiology and
prognstic implications of CFRD: a technical review. Diabetes Care 2010;
33: 2677–83.
3 Moran A, Brunzell C, Cohen RC, et al. Clinical care guidelines for CFRD:
recommendations from the Cystic Fibrosis Foundation, the American
Diabetes Association and the Pediatric Endocrine Society. Diabetes Care
2010; 33: 2697–708.
4 Hanas R, Donaghue KC, Klingensmith G, Swift PG. ISPAD clinical practice
consensus guidelines 2009 compendium. Pediatr Diabetes 2009;
10 (suppl 12): 1–2.
5 Ballman M, Hubert D, Assael BR, et al. Repaglinide versus insulin for newly
diagnosed diabetes in patients with cystic fibrosis: a multicentre,
open-label, randomised trial. Lancet Diabetes Endocrinol 2018; 6: 114–121.
6 Moran A, Pekow P, Grover P, et al. Insulin therapy to improve BMI in cystic
fibrosis related diabetes without fasting hyperglycemia: results of the Cystic
Fibrosis Related Diabetes Therapy trial. Diabetes Care 2009, 32: 1783–88.
autoimmune diabetes in adults. Thus, the true preva­ Published Online
November 30, 2017
lence of autoimmune diabetes in adults is not known with http://dx.doi.org/10.1016/
confidence because many occult cases might exist among S2213-8587(17)30399-6
patients believed to have type 2 diabetes. See Articles page 122
In The Lancet Diabetes & Endocrinology, Nicholas Thomas
and colleagues1 describe an innovative genetic method
to estimate the prevalence of autoimmune diabetes in
people older than 30 years. Their approach was facilitated
by the explosion of genetic discovery over the past decade,
which has yielded more than 50 loci that are robustly
associated with type 1 diabetes.2 Identification of these
loci enables the construction of genetic risk scores, which
provide a fairly accurate estimate of overall genetic risk
through summation of the aggregate of type 1 diabetes
risk alleles in an individual (including variants correlated
with the HLA risk haplotypes).3–5 Indeed, in a previous
study,3 these researchers showed that a genetic risk score
for type 1 diabetes, depending on where thresholds were
set along the quantitative scale, could either exclude
95% of people without type 1 diabetes or capture 95% of
those with type 1 diabetes. Because, for complex diseases,
genetic burden has incomplete penetrance, the genetic
risk score simply serves as a probability metric.
The method used by Thomas and colleagues1 is
deployed as follows. Consider a population of people with
diabetes who are older than 30 years. This population is
probably composed of a mixture of people with type 2
 Comment
diabetes (the vast majority) and an unrecognised group
of people with autoimmune diabetes (a small minority).
One property of the type 1 diabetes genetic risk score is
that a specific cutoff essentially separates the population
without type 1 diabetes into two equal halves, whereas
the same cutoff captures more than 95% of people with
type 1 diabetes. In other words, for people without type 1
diabetes, that threshold divides the group into halves with
genetic risk below and above the median, but those with
type 1 diabetes will almost always fall above the threshold
(see figure 1 in the report1).
Enter UK Biobank. This amazing resource contains
more than 500 000 individuals in the UK who have
provided DNA and allowed their genetic data to be
linked to their electronic health record. This dataset
allows for the generation of a type 1 diabetes genetic
risk score for each participant, with particular attention
paid to people aged 30–60 years and those with
diabetes. Setting the genetic risk score threshold at the
50th centile divides the group of 379 511 participants
into two halves of 189 508 individuals with a high
type 1 diabetes genetic risk score and 190 003 individuals
with a low type 1 diabetes genetic risk score. Focusing
on the 13 250 individuals with diabetes, one sees that
7268 people are above the type 1 diabetes genetic risk
score cutoff and 5982 people are below the same cutoff,
which is significantly skewed from the 50:50 expectation
under the null hypothesis of a single population.
The investigators appropriately speculate that this
excess of 1286 individuals (nearly 10%) in the higher
type 1 diabetes genetic risk score group have autoimmune
diabetes. Although they were more likely to be diagnosed
at a younger age (58% with diabetes diagnosed aged
≤30 years), a sizeable number of people with autoimmune
diabetes were diagnosed when aged 31–60 years (about
4% of all individuals with diabetes). Among individuals
with a high type 1 diabetes genetic risk score, no major
clinical differences were seen between those diagnosed
age 30 years or younger and those diagnosed later.
This ingenious application of genetics enabled
the establishment of an approximate prevalence of
autoimmune diabetes among adults, which highlights the
need for clinicians to be alert to this possibility at the time
of diagnosis irrespective of the patient’s age. However,
it does not suggest that the type 1 diabetes genetic risk
score can be used in this manner to diagnose autoimmune
diabetes in individuals: because type 2 diabetes is much
88 www.thelancet.com/diabetes-endocrinology Vol 6 February 2018
more prevalent in adults than autoimmune diabetes,
most older adults with a high type 1 diabetes genetic risk
score will still have type 2 diabetes. However, if the type 1
diabetes genetic risk score is low, the patient is unlikely to
have autoimmune diabetes, showing how a genetic tool
(which is extremely accurate, will become inexpensive,
and only needs to be tested once in the lifetime of an
individual) might help to exclude a diagnosis more
effectively than the use of autoantibody testing or HLA
haplotyping.
The investigators focus on “genetically defined
type 1 diabetes” and do not directly address LADA,
typically defined as diabetes with positive autoantibodies
diagnosed after age 35 years and not requiring insulin
at the time of diagnosis (although many patients
with LADA do require insulin within 3 years of onset).
Whether LADA is a distinct entity, a less aggressive
form of type 1 diabetes on an autoimmune continuum,
or an artificial construct that results from a mixture
of patients with type 1 and type 2 diabetes remains
a point of contention.6–9 In view of the genetic similarities
between LADA and type 1 diabetes,10 it is likely that the
type 1 diabetes genetic risk score provides a reasonable
approximation for all autoimmune diabetes, although
this would need to be directly tested in a well phenotyped
cohort of patients with LADA.
More generally, the study by Thomas and colleagues1
shows that a concerted effort to advance genetic
discovery for complex traits can yield useful translational
results once the findings reach critical mass, genetic
material becomes available across clinical resources, data
are made available to investigators, and intelligent and
inquisitive users are empowered to test hypotheses.
Hopefully, this type of inquiry will continue to facilitate
progress across many other complex diseases.
Jose C Florez
Diabetes Unit and Center for Genomic Medicine, Massachusetts
General Hospital, Boston, MA 02114, USA; Programs in
Metabolism and Medical & Population Genetics, Broad Institute,
Cambridge, MA, USA; and Department of Medicine, Harvard
Medical School, Boston, MA, USA
jcflorez@mgh.harvard.edu
I declare no competing interests. I am a Massachusetts General Hospital
Research Scholar and am supported by US National Institutes of Health grants
NIDDK R01 DK072041, NIDDK U01 DK105554, NIGMS R01 GM117163, NIDDK
R01 DK105154, and NIDDK K24 DK110550.
Copyright © The Author(s). Published by Elsevier Ltd. This is an open access
article under the CC BY-NC-ND 4.0 license.

1 Thomas NJ, Jones SE, Weedon MN, Shields BM, Oram RA, Hattersley AT.
Frequency and phenotype of type 1 diabetes in the first six decades of life:
a cross-sectional, genetic susceptibility analysis from UK Biobank.
Lancet Diabetes Endocrinol 2018; 6: 122–129.
2 Redondo MJ, Steck AK, Pugliese A. Genetics of type 1 diabetes.
Pediatr Diabetes 2017; published online Nov 2. DOI:10.1111/pedi.12597.
3 Oram RA, Patel K, Hill A, et al. A type 1 diabetes genetic risk score can aid
discrimination between type 1 and type 2 diabetes in young adults.
Diabetes Care 2016; 39: 337–44.
4 Winkler C, Krumsiek J, Buettner F, et al. Feature ranking of type 1 diabetes
susceptibility genes improves prediction of type 1 diabetes. Diabetologia
2014; 57: 2521–29.
5 Redondo MJ, Oram RA, Steck AK. Genetic risk scores for type 1 diabetes
prediction and diagnosis. Curr Diab Rep 2017; 17: 129.
Has the UK really become iodine sufficient?
In December, 2016, the Iodine Global Network (IGN)
published its new map of global iodine nutrition based on
median urinary iodine concentration (mUIC) in school-
aged children.1 Notably, the status of the UK, which was
classified as mildly iodine deficient in 2014–15 (mUIC
50–99 µg/L), had become adequate by 2016 (mUIC
100–299 µg/L).1 The reason for this apparently rapid
improvement lies in the different data sources used; data
that showed mild deficiency in 2014–15 came from spot-
urine samples from 737 girls aged 14–15 years from nine
UK centres (mUIC 80·1 μg/L),2 whereas the 2016 data
were based on spot-urine samples from 458 boys and
girls aged 4–18 years, which were collected in year 6 of
the UK National Diet and Nutrition Survey (NDNS).3 The
result from the group aged 4–18 years (mUIC 138 μg/L;
table) is within the adequate range, as defined by WHO.4
However, these new mUIC data from the NDNS should
not lead to complacency; indeed, the new figures might
mask the presence of deficiency in some population
subgroups.
Considerable disparity exists between the conclusions
based on the mUIC measurements in spot-urine samples
from year 6 and those based on iodine-intake data (from
4-day food diaries) in years 5 and 6 of the NDNS (table).
Although the limitations of food-diary analysis for
iodine-intake estimation are recognised, they are less of
a concern in the UK than in countries that have iodised
salt (salt intake is hard to quantify in dietary surveys). We
would therefore expect mUIC and dietary-intake data to
tell broadly the same story. Yet, the mUIC of children aged
4–18 years is 138 μg/L, which implies adequacy, whereas
6–21% of children have an iodine intake below the lower
reference nutrient intake (LRNI),3 which implies deficiency,
especially in older children (table). The LRNI is based
www.thelancet.com/diabetes-endocrinology Vol 6 February 2018 89
Comment
6 Gale EA. Latent autoimmune diabetes in adults: a guide for the perplexed.
Diabetologia 2005; 48: 2195–99.
7 Tuomi T, Santoro N, Caprio S, Cai M, Weng J, Groop L. The many faces of
diabetes: a disease with increasing heterogeneity. Lancet 2014;
383: 1084–94.
8 Ostergaard JA, Laugesen E, Leslie RD. Should there be concern about
autoimmune diabetes in adults? Current evidence and controversies.
Curr Diab Rep 2016; 16: 82.
9 Buzzetti R, Zampetti S, Maddaloni E. Adult-onset autoimmune diabetes:
current knowledge and implications for management. Nat Rev Endocrinol
2017; 13: 674–86.
10 Mishra R, Chesi A, Cousminer DL, et al. Relative contribution of type 1
and type 2 diabetes loci to the genetic etiology of adult-onset,
non-insulin-requiring autoimmune diabetes. BMC Med 2017; 15: 88.
on the intake required to avoid goitre, and is between Published Online
April 24, 2017
50 and 70 μg per day in children aged 4–18 years.3 http://dx.doi.org/10.1016/
Where possible, IGN maps of global iodine nutrition are S2213-8587(17)30133-X
based on mUIC data from boys and girls aged 6–12 years,
since WHO thresholds are intended for use in school-
aged children. In 2014–15, in the absence of such data
for the UK, the classification was based on a UK study2
of girls aged 14–15 years. However, girls tend to have a
lower iodine intake (and thus status) than boys, and older
children (adolescents) have a lower iodine intake than
younger children (table). This difference in intake might
partly explain why the 2014–15 IGN map showed mild
iodine deficiency in the UK, whereas the 2016 IGN map,
based on boys and girls from the age of 4 years, showed
adequacy. The difference in iodine intake by age and sex
can probably be explained by the considerably lower
intake of milk, the principal source of iodine in the UK,
in teenage girls than in younger children (eg, 110 g per day
in girls aged 11–18 years vs 196 g per day in boys and girls
aged 4–10 years).5
To some extent, this disparity highlights the fact
that children aged 6–12 years might not be the most
appropriate group to represent population status. This
fact is especially true in countries such as the UK where
intake of the main source of iodine (milk) varies with age.
Furthermore, pregnant women are often susceptible to
iodine deficiency, even in countries in which the general
population is iodine sufficient.6 IGN has recognised this
issue by producing, for the first time in 2016, a separate
map of iodine status in pregnant women, again based
on mUIC.7 Unfortunately, pregnant women are not
sampled in the NDNS, but on the strength of accumulated
evidence from the past 10 years,8 the map shows that
pregnant women in the UK are iodine deficient.7 This