- Hypoglycemia is more common in older adults treated with insulin or sulphonylureas and can be difficult to recognize due to non-specific symptoms. Unexplained falls, unsteadiness, or nausea may indicate unrecognized hypoglycemia. - Maintaining tight glycemic control with aggressive HbA1c targets increases hypoglycemia risk in older adults. Guidelines recommend less stringent HbA1c targets for older patients based on functional status and comorbidities. - Glucose variability in addition to average blood sugar levels affects hypoglycemia risk. Measuring endogenous insulin levels may help identify patients at risk of hypoglycemia due to absolute insulin deficiency.

1. Hypoglycaemia in Older People
Suzy Hope
5th February 2016

2. Overview
• Hypos: setting the scene
• Recognition of hypos in older patients
• HbA1c targets
• Endogenous insulin levels
• Clinical implications

3. Dorothy
• 78, lives alone
• Type 2 diabetes for 19 years
• Been on insulin for 10 years
• “I’m told my diabetes is very good”
• Denies getting hypos
• On closer questioning… keeps biscuits by bed for
“when I feel a bit wobbly in the night”
• Other PMH: hypertension, osteoporosis
• Recent investigations for unsteadiness on her feet

4. What are hypos?
Hypoglycaemia = low blood glucose
• Whipple’s triad:
– Low blood glucose
– Symptoms
– Resolution of symptoms with treatment
• Hypo definitions

5. Why do hypos matter?
• Fear and quality of life
• Impact on driving
• Can cause falls, accidents
• Risk of hospitalisation
• Longer hospital stays & poorer outcomes
• Cognitive impairment & dementia
• Affects compliance & thus other treatment goals
Lundkvist et al 2005 Eur J Health Econ 6: 197-202
Jermendy et al 2008 Health Qual Life Outcomes 6: 88
Solli et al 2010 Health Qual Life Outcomes 8: 18
Signorovitch et al 2013 Diabetes Obes Metab 15: 335-41
Mayne D et al 2010 Age Ageing 39: 522-5
Johnston et al 2012 Diabetes Obes Metab 14: 634-43
Turchin et al 2009 Diabetes Care 32: 1153-7

6. Cognition and hypos
• Bidirectional relationship
• Cognitive impairment predisposes to hypos
• Episodes of severe hypos associated with graded
increase in risk of dementia (after adjustment for
age, comorbidities, HbA1c etc) eg:
– 1 hypo – 1.26x risk for dementia
– 2 hypos – 1.8x
– 3 hypos – 1.94x
Whitmer et al 2009 JAMA 301:1565-1572
Lin et al 2013 J Intern Med 273: 102-110
Aung et al 2012 Diabet Med 29: 328-336
Bruce et al 2009 Diabetologia 52: 1808-1815
Yaffe et al 2013 JAMA Intern Med 173(14): 1300-1306

7. Who is at risk?
• Diabetes = problem with glucose homeostasis
• Insulin keeps glucose levels in the normal range
• Type 1 diabetes
– absolute insulin deficiency
– need insulin treatment to survive
• Type 2 diabetes
– insulin resistance and insulin deficiency
– evolves over time and treatment intensifies

8. Simple glucose physiology

9. Hypoglycaemia
1) Insulin secretion falls

10. Hypoglycaemia
1) Insulin secretion falls
2) Glucagon released

11. Summary: hypoglycaemia
• Physiological response to increase glucose:
– 1) Insulin secretion falls
– 2) Glucagon levels increase
– 3) Adrenaline levels increase
• Symptomatic response to increase glucose
– Behaviour change - eat!

12. Summary: hypoglycaemia in
insulin-deficient diabetes
• Physiological response to increase glucose:
– Endogenous insulin secretion can’t fall
– Exogenous insulin “in the system”
– Glucagon response diminished
– Adrenaline response diminished
• Symptomatic response less marked
– Reduced opportunity for behavioural change

13. Who is at risk of hypos?
• Major risk factors
– Type 1 diabetes
– Treatment with insulin
– Treatment with sulphonylurea tablets
– Previous hypoglycaemia
• Other risk factors*
– Missing meals
– Exercise
– Tight glycaemic control
*especially when on the above treatments

14. Symptoms of hypoglycaemia are
non-specific in the elderly
Autonomic:
Palpitations
Sweating
Anxiety
Neuroglycopenic:
Fatigue Irritability
Confusion Dizziness
Drowsiness Coma
Particularly in older people:
Unsteadiness
Light-headedness
All these are also common in elderly people
without diabetes
Symptoms can be different on different
occasions!
Deary et al 1993 Diabetologia 36: 771-777
Jaap et al 1998 Diabet. Med. 15: 398-401
Zammitt et al 2011 Diabetes Technol Ther 13: 571-8

15. Dorothy keeps biscuits by bed for
“when I feel a bit wobbly in the night”
Is Dorothy having hypos?

16. Pilot data suggested symptoms of hypoglycaemia are not
always recognised by the medical team or patient
• 106 patients in primary care
– insulin or sulphonylurea treatment
– HbA1c <7.5% (58.5mmol/mol)
• Retrospective review of consultations over 1 year
• % patients with >1 “hypo clue” symptom documented

17. Can non-specific symptoms associated with
hypos be important clues for recognising
hypoglycaemia in this group?

18. Methods
• Axminster Medical Practice
• Inclusion criteria:
– Patients over the age of 65
– All those on insulin
– All those on sulphonylureas
– All those just on metformin
– 50 patients who were not diabetic
• Data collected retrospectively for a one year period:
documented hypos, and potential “hypo clues”
• “Hypo clue” consultation:
– a consultation with >1 “hypo clue” symptom, where no obvious
explanation or diagnosis was recorded

19. Hypoglycaemia events reported to primary care are much more
frequent in insulin-treated patients
0
0.2
0.4
0.6
0.8
1
1.2
OVERALL
113 episodes in
334 patients
INSULIN
87 episodes in
79 patients
SULPHONYLUREA
18 episodes in
85 patients
METFORMIN
8 episodes in
120 patients
NON-DIABETIC
0 episodes in
50 patients
Hypoglycaemiaepisodes/person/year
p<0.0001 for a difference
across groups

20. “Hypo clue” consultations are common in all treatment groups
– and in patients without diabetes
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
OVERALL
367 episodes in
334 patients
INSULIN
127 episodes in
79 patients
SULPHONYLUREA
83 episodes in
85 patients
METFORMIN
119 episodes in
120 patients
NON-DIABETIC
38 episodes in
50 patients
Episodes/person/year
p=0.16

21. “Hypo clue” consultations are more frequent in
insulin-treated patients who have had a recognised episode of
hypoglycaemia
0
10
20
30
40
50
60
70
80
90
100
≥1 hypo
20/27
No hypo
21/52
≥1 hypo
2/4
No hypo
39/81
≥1 hypo
1/2
No hypo
58/119
≥1 hypo
0/0
No hypo
18/50
Insulin n=79 Sulphonylurea n=85 Metformin only n=121 Non-diabetic n=50
Proportionofpatientswith≥1
“hypoclue”consultation
p=0.004
p=NS p=NS

22. Are any particular symptoms more often seen
in “hypo clue” consultations in those patients
with documented hypos?

23. Most commonly presenting symptoms overall in those who had
also presented with >1 hypo, compared to those with no
documented hypos
Bold columns p<0.05 for a difference
0
5
10
15
20
25
30
35
Proportionofpatientsconsultingwithsymptom(%)
>=1 hypo
No hypos

24. In insulin-treated patients, falls, unsteadiness and
nausea were most notable
• 33% with a documented hypo consulted on another
occasion over the year with a fall, compared to 8% of
those without a documented hypo (p=0.008)
• 22% with a documented hypo consulted on another
occasion with unexplained nausea, compared to 2%
without (p=0.006)
• 19% with a documented hypo consulted on another
occasion with unsteadiness, compared to 4% without
(p=0.04)

25. So….
1) Hypos reported to primary care are much more
common in those who are treated with insulin
2) Overall “hypo clue” consultation rate is high across
the treatment groups, and in patients without
diabetes
3) But in patients who have had a recognised hypo,
“hypo clue” consultations are 1.5x as common
4) Unexplained falls, unsteadiness and nausea seem
more common in those with recognised hypos: could
these represent unrecognised hypoglycaemia?

26. What could this mean for Dorothy?
• Being on insulin, she is at increased risk of hypos
• Consultations with “hypo clue” symptoms may be
indicative of possible hypoglycaemia, with
unsteadiness one of the alarm bells
• “Hypo clue” symptoms should not necessarily just be
put down to hypos – other diagnoses should be
considered too!

27. Does “very good” mean her HbA1c is
low?
Is Dorothy actually being “over-
treated”?
I’m told my
diabetes is
very good!

28. HbA1c
• WHO 2011 guidelines:
– diagnosis of diabetes in asymptomatic patients
• >2 readings of 48mmol/mol
• repeated two weeks apart from each other
• “3 months’ average glucose” - HbA1c value typically comes
– 50% from the previous month’s red blood cells (RBCs)
– 32% from the month before
– 18% from the month before that
2011: http://www.who.int/diabetes/publications/report-hba1c_2011

29. Factors affecting HbA1c
• Increases with age
• Will be higher if low RBC turnover ie
disproportionate numbers of old RBCs
• iron deficiency anemia
• vitamin B12 deficiency anemia
• folate deficiency anemia
• chronic renal failure
• alcoholism
• asplenia
• Will be lower if high RBC turnover ie
more young RBCs
• haemolysis
• haemorrhage
• blood transfusions
• treatment for iron, vitamin B12, or folate
deficiency
Kilpatrick et al 1996 QJM 89(4):307-12
Gallagher et al 2009 J. Diabetes 1:9-17
r=0.49

30. HbA1c treatment targets
• Increased appreciation of perceived risks in older
patients with too stringent HbA1c targets
• Little evidence but discussions and consensus
statements
• Adjustment of QOF guidelines over time
• National/international guidelines now qualify HbA1c
targets for older adults according to comorbidities –
eg IDF, AGS/ADA, ADA/EASD guidelines and
consensus reports
2013: http://www.idf.org/guidelines-older-people-type-2-diabetes.pdf
2012: http://www.americangeriatrics.org/files/documents/ADA_Consensus_Report.pdf
2012: ADA/EASD Consensus report. Diabetes Care 35(6): 1364-1379

31. IDF glycaemic targets for older adults
Category HbA1c (%) HbA1c (mmol/mol)
Functionally
independent
7-7.5% 53-59
Functionally
dependent
7-8% 53-64
- frail
Up to 8.5% may be
appropriate
Up to 70
- dementia
Up to 8.5% may be
appropriate
Up to 70
End of life care
Avoid symptomatic
hypoglycaemia
2013: http://www.idf.org/guidelines-older-people-type-2-diabetes

32. All-cause mortality by HbA1c deciles
A: Metformin + sulphonylureas B: Insulin therapies
Adjusted hazard ratios
HbA1c deciles with 1289 – 3513 people per group
Vertical bars 95% confidence intervals
Currie et al 2010 Lancet 375: 481-89

33. Could Dorothy be having hypos?
HbA1c = 69mmol/mol (8.5%)

34. Are hypos or “hypo clue” symptom consultations
more frequent in patients with lower HbA1c?

35. Proportion of insulin or sulphonylurea-treated
patients (per HbA1c group)
who had at least one “hypo clue” visit
0
20
40
60
80
100
<6%
n=5
6-6.5%
n=9
6.5-7%
n=32
7-7.5%
n=34
%patientswithatleastone
“hypoclue”consultation

36. Proportion of insulin or sulphonylurea-treated
patients with >1 “hypo clue” consultation is similar
across HbA1c groups
0
10
20
30
40
50
60
70
80
90
100
<6%
n=5
6-6.5%
n=9
6.5-7%
n=32
7-7.5%
n=34
7.5-8%
n=32
8-8.5%
n=17
8.5-9%
n=12
9-9.5%
n=7
>9.5%
n=16
Proportionofpatientswithatleastone
“hypoclue”consultation
HbA1c
p=0.42

37. Dorothy
• On insulin
• Unsteadiness, and sometimes “wobbly” at night
• HbA1c 8.5%
• Could she be having hypos?
• What else could be happening?

38. Glucose variability
Siegelaar et al 2010
DeVries 2013
HbA1c represents the mean glucose
But these graphs represent the same means
Potential for
more hypos
Self-monitoring
can miss
fluctuations

39. Detecting glucose variability
Continuous glucose monitoring can reveal extra details
- Glucose reading taken every 5 minutes
- Can wear up to 7 days
- Summary graphs and statistics obtained
Glucose variability higher in T1D than T2D
Due to insulin deficiency and impaired counter-regulation
Basal bolus regimens attempt to minimise variability
Medtronic – iPro2 Professional continuous glucose monitor

40. Glucose variability increases with increasing treatment
intensity in T2D
• Calculated from individual 72-h continuous glucose monitoring tracings
• Between-treatment group differences statistically significant, p < 0.001
• Intensification of treatment in T2D is due to progressive insulin deficiency
• Heterogeneity within even insulin-treated population of T2D
White columns:
patients with T2D treated with
- diet (DIET)
- metformin (MET)
- a-glucosidase inhibitor (AGI)
- sulphonylurea (SU)
- thiazolidinedione (TZD)
- conventional insulin therapy (INSct)
- intensified insulin therapy (INSict)
Hatched column:
patients with T1D
Kohnert et al 2013 Diabetes Tech & Therapeutics 15(6): 448-454

41. Could measuring insulin levels help
stratify hypoglycaemia risk?

42. C-peptide
• C-peptide = measure of
endogenous insulin levels
• Recent validation of more practical methods
– blood C-peptide – fasting or random
– Urinary c-peptide creatinine ratio (UCPCR)
• Absolute insulin deficiency in T1D (DCCT):
– C-peptide <200pmol/L
McDonald et al 2009 Clinical Chemistry 55(11): 2035–2039
Besser et al 2011 Diabetes Care 34: 607–609
Jones et al 2011 Diabetic Medicine 28(9): 1034-1038
Bowman et al 2012 Diabetic Medicine 29: 90–93
McDonald et al 2012 PLoS ONE 7(7): e42084

43. Progressive insulin deficiency in T2D can result
in absolute insulin deficiency
• 3% insulin-treated patients with a clinical diagnosis
of T2D had absolute insulin deficiency (AID)
– UCPCR screening
– Confirmation in MMTT
• Anecdotally patients with AID found glycaemic control
difficult – both high and low
• Treatment regimes may suggest clinicians finding it
difficult too - eg only 27% those with absolute insulin
deficiency were on a basal bolus regime
Hope et al 2013 Diabetic Medicine 30(11): 1342–1348

44. Are these people with T2D and absolute insulin
deficiency at the same risk of complications as
those with T1D?
Can c-peptide be used as a biomarker to predict
glucose variability and hypoglycaemia risk?

45. Random non-fasting C-peptide (rCP):
Correlation between rCP and 90 minute blood stimulated C-peptide (sCP) in
the mixed meal tolerance text for 50 patients
Spearman’s rho correlation
coefficient=0.93, p<0.0001
Hope et al, submitted

46. Do people with T2D but severe insulin deficiency have
increased glucose variability?
• Two matched groups of insulin-treated participants
with clinical diagnosis of T2D (diagnosed >35yrs, took
at least 2yrs to start insulin)
– Severe insulin deficiency – rCP <200pmol/L
– Retained insulin levels – rCP >500pmol/L
• Matched for glycaemia (HbA1c), age, diabetes
duration, BMI
• Continuous glucose monitoring (mean 4.1 days)
• Clarke’s hypoglycaemia questionnaire
rCP = random C-peptide

47. Characteristics of matched participants

48. Glucose variability on continuous glucose monitoring is much
higher in the low C-peptide group
23456
C-peptide <200 pmol/L C-peptide >500 pmol/L
p = 0.0004

49. Midnight Midday Midnight

51. Proportion of patients with >1 hypoglycaemia episode is much
higher in the low C-peptide group
0
10
20
30
40
50
60
70
80
90
100
<=200 >500
Proportionwithatleastoneepisode
Random non-fasting C-peptide
3 to 4
2.2 to 3
<=2.2
p=0.002

52. In people with clinically diagnosed T2D:
Low c-peptide measured on routine blood
samples is associated with increased glucose
variability and increased risk of hypoglycaemia

53. Clarke’s hypoglycaemia questionnaire (Q5&6):
frequency of episodes <3.5 with or without symptoms,
by C-peptide group

54. Does this association between random C-
peptide and hypoglycaemia risk hold true on a
larger scale?
• All insulin-treated patients in DARE invited to
participate
• Answered Clarke’s hypoglycaemia questionnaire
• System set up so a random C-peptide could be
measured when a routine HbA1c from consenting
patients was sent into the lab
• Results analysed for people where a rCP was
obtained within a year of questionnaire completion

55. Results
• 480 participants
• Given a Type 1 or 2 diagnosis according to
RCGP diagnosis guidelines

56. Participant characteristics

57. Distribution of random C-peptide by
C-peptide deciles and diabetes diagnosis
Approx 48 patients per decile

58. More self-estimated episodes of blood glucose <3.5mmol/L
overall, the lower the C-peptide decile
Approx 48 patients per decile

59. … despite HbA1c being similar across the deciles

60. Median estimated number of episodes <3.5mmol/L was higher
the lower the C-peptide decile,
regardless of clinical diagnosis

61. More episodes reported,
both with and without symptoms,
in those with C-peptide <200pmol/L

62. In summary:
1) Need to recognise severe insulin deficiency - which cannot
always be clinically obvious
2) Need to be given right treatment (review!)
3) Need to have education to go with it – coping with a more
complex insulin regimen, self-monitoring, effect of diet and
exercise, driving…
4) Learning (complex) new concepts more challenging in older
age
Those with long-standing T1D already have strategies to cope
with all the above… need to think about those with “type 2”
diabetes carefully too!

63. Knowing C-peptide level
can be helpful
C-peptide
45pmol/L
C-peptide
1200pmol/L

64. Conclusion
• People with T2D but low endogenous insulin levels
are at significantly higher risk of hypos than those
with more substantial residual endogenous insulin
• Simple C-peptide measurement in people with
insulin-treated diabetes may help with risk
stratification, education and management regimens
for older patients, carers and healthcare
professionals

65. Acknowledgements
Study participants
Exeter NIHR Clinical Research team especially Professor
Andrew Hattersley, Drs Bev Shields, Angus Jones and Bea
Knight, and Anita Hill and Tina Libretto
Dr Phil Taylor and Axminster Medical Practice team, and
Professor Willie Hamilton (Exeter)
Drs Pratik Choudhary (King’s) and Kai Tan Horng (Plymouth)
Northcott Devon Medical Foundation for funding the
continuous glucose monitoring study

66. RCGP diabetes classification guidelines

67. Method
• 601 adults recruited, with
– insulin-treated diabetes
– diabetes duration >5y
• Home urine sample collected for C-peptide measurement (UCPCR test)
• Performance of clinical diagnostic criteria assessed and other criteria
explored using ROC curves
“Gold-standard” Type 1 defined as
UCPCR<0.2nmol/mmol (absolute insulin deficiency)
AND
continuous insulin treatment within 3 years of diagnosis
“Gold-standard” Type 2
all other patients (insulin-treated and duration >5y)

68. Number of patients classified as having T1D or T2D according to the RCGP
guidelines, and proportions classified correctly or incorrectly compared to
the gold standard
0
50
100
150
200
250
300
350
400
Age Diag<35 &
TTI<6m
Age Diag>=35 &
TTI=0m
Age Diag<35 &
TTI>=6m
Age Diag>=35 &
TTI>0m
Type 1 Type 2
Numberofpatients
Misclassified
Correctly classified
44% 77%
93%
TTI = time to insulin treatment from diagnosis
87%

69. ROC curve for discriminating between Type 1 and Type 2 diabetes based on
the gold standard definition
Red: time to insulin from diagnosis (AUC=0.904); black: age at diagnosis (AUC=0.871);
blue: BMI at diagnosis (AUC=0.824); green: BMI at recruitment (AUC=0.715)

70. Conclusions
• RCGP guidelines are clinically useful
• Correctly classified 86% insulin-treated patients >5y
from diagnosis
• In those diagnosed >35y and on insulin from diagnosis,
37/66 (56%) were misclassified as T1D
• Time to insulin & diagnosis age performed best in
predicting long-term endogenous insulin production,
but altering guidelines with optimal cut-offs did not
significantly improve guideline accuracy
• BMI was not a clinically significant predictor