Tayside Centre for General Practice

1
Tayside Centre for General Practice Division of
Community Health Sciences
Population Diabetes Research in Tayside
Peter T. Donnan, PhD Senior Lecturer in Medical
Statistics
2
Outline of talk

• Population research using DARTS
• Adherence oral sulphonylureas
• Predicting Coronary Heart Disease in type 2
  diabetes
• Current studies in diabetes
• Proposed new studies

3
Type 2 Diabetes in TaysidePrevalence 2.05
Prevalence of Diabetes in Tayside by age group
Morris et al, BMJ, 1997
4
Data, Data everywhere But usually not easily
linked
5
Linking the Data Vital for Seamless Care and
Research
6
(No Transcript)
7
Aims of Adherence Study
Describe patterns of adherence to oral
hypoglycaemic drugs in type 2 diabetes
Assess predictors of adherence
Test hypothesis that once daily administration is
associated with better adherence
8
Study population
All subjects with type 2 diabetes resident in
Tayside, Scotland (400,000)
First prescription for oral hypoglycaemic drug on
or after 1st Jan 1993
Follow-up to 31 Dec 1995 with at least 12 months
of prescriptions
9
Drug exposure (MEMO)

• Sulphonylurea, metformin, other
• Date of encashment
• Duration
• Dose, amount tablets, instructions

10
Outcome

• Adherence index
• Total time drug prescribed
• Total time of follow-up

Expressed as percentage or drug coverage per
annum
11
Distribution of sulphonylurea index (days of drug
coverage per annum) in those receiving
sulphonylurea monotherapy (Adherence 31)
12
Adherence defined as ? 90 or 329 days per year
Better adherence associated with

• Lower number of tablets per day
• Lower number of co-medications
• Less social deprivation
• Younger age

13
Adherence related to number of tablets per day
Adjusted OR and 95 CI
1 per day
2 per day
3 per day
Number of tablets relative to 4 per day
14
Conclusions

• In the community only one in three had adequate
  adherence (gt 90)
• Once daily dosing may be advantageous
• Benefit shown in UKPDS and large scale RCTs using
  complex drug regimens
• Adherence is a critical issue for these benefits
  to be realised in the community

15
. I hope these are easier to flush away than
the last lot .
16
(No Transcript)
17
Estimation of the absolute risk of coronary heart
disease (CHD) in the UK population with type 2
diabetes
Funded by Diabetes UK / Pfizer UK
Collaborators Louise Donnelly Statistician in
MEMO Andrew Morris Professor in Diabetic
Medicine John New Diabetologist, Salford
18
Why predict?
Having predicted those at high risk preventative
action can be taken Assumes factors are
modifiable e.g. guidelines treat everyone with
risk CHD ? 30 with statins ADA recommend treat
high risk with aspirin
19
Prediction is hazardous
The possibilities of the aeroplane have been
exhausted Thomas Edison, 1895 The computer has
no commercial future IBM Annual Report, 1948
20
Current CHD Predictors
Most prediction algorithms based on Framingham
study Only 300 with diabetes included Diabetes-re
lated risk factors not considered e.g.
HbA1c Recent UKPDS risk calculator Based on RCT
sample (65 excluded)
21
Define Coronary Heart Disease

Fatal MI non-fatal MI, CHD death
CHD (n161)
22
Statistical Methods
Proportional hazards Weibull accelerated failure
time model
Effect of factors (x) accelerates individual
towards CHD outcome
23
Results of Weibull PH Model (n3048)
Significant Factors Age
diagnosis ln (Dur. Diab.) Current Smoking Men
vs. Women Total cholesterol HbA1c SBP
24
Calculation of Absolute risk

• For a male smoker aged 65 with no previous CHD
  and 6 years duration of type 2 diabetes with SBP
  of 160 mmHg, Chol 5.8, HbA1c of 8.0, Height 1.7
  
• The absolute 5-year risk of
• CHD 18

25
Summary

• Strengths
• First risk estimation from a large unselected
  population from UK
• many diabetes related variables collected,
  well-defined diabetes
• Able to validate and extend model to include
  ethnicity

26
Next Steps

• Validate and calibrate with UKDIABS data from
  Salford and Blackburn
• Update risk calculator with 10-year follow-up
  and ethnicity
• Disseminate Excel spreadsheet calculator
• Extend study to predict stroke

27
All models are wrong, but some are useful.G. Box
28

• Current Studies
• GSK funded PhD in pharmacogentics in type 2
  diabetes statins (Louise Donnelly)
• Prediction of stroke and health economic aspects
  of prediction (Pfizer UK)
• Association of engagement with DARTS and process
  measures of management (CSO)

29

• New Proposed Studies
• Metformin and role in Cancer (CSO)
• Anaemia as a trigger for CHD in case-crossover
  design (Amgen)