Transcript Appraisee Training Day - QResearch

The effect of the ‘Polypill’ on
mortality in general practice
Dr Julia Hippisley-Cox
Gozo
October 2004
Goals of presentation
• To given an over view of electronic
database used for the study
(QRESEARCH)
• To describe research to determine
effect of the ‘Polypill’ on mortality in
patients with CHD
GP databases in the UK
• Several different database available
– Oldest and most well known GPRD
• QRESEARCH new key databases
• Both derived from GP clinical data
• Both available for access by
researchers
• Both are an international resource
QRESEARCH
What is it?
• QR is result of a not-for-profit
partnership between UoN & EMIS
[supplies > 60% all practices]
• Patient level consolidated database
• Longitudinal data for up to 16 years
• Validated against external and
internal measures
Qresearch Coverage
•
•
•
•
468+ practices
UK wide
> 5 in every SHA
Approx 7.4 million patients including those
who died, left and still registered
• 3.3 million current patients
• Now the largest such database in the world
(unless anyone here knows otherwise!)
Data contents
•
•
•
•
•
•
•
Socio-demographics
Diagnoses
Clinical values
Laboratory tests
Prescription data
Consultation data
Category of clinician entering data
QRESEARCH
Research service
We will provide a service to
- Bona fide academics
- With original research question
- With MREC
- With freedom to publish
- Who agree to
- Acknowledge source data
- use it for purpose only
- not pass it on
What type of research can we use it for?
Best for epidemiological studies
• Case control studies
• Cohort studies
• Cross sectional survey
• Modelling
Things we cant do:
•
•
•
•
Track back to patients
Track back to practices
Randomise patients or practices
Link with external data sources eg
questionnaire, genetics
• This is where other systems come in
- THIN
- e-GRID
– UK Biobank
Aims of research project
• To determine the effect of the
‘Polypill’ on mortality in CHD
patients
• Polypill = aspirin + beta blockers +
ACE inhibitors + statins
Background to project
• Good evidence from individual trials
for benefits of individual drugs in 2’
prevention CHD
• Enthusiasm for ‘Polypill’ for all
patients over 55 years though no
direct evidence benefits stack up
• Drugs may interact with drugs or
other comorbidity
Design & setting
Design
• Open cohort study with case control
analysis
Setting
• 1.18 million patients registered with
89 QRESEARCH practices on 1st Jan
1996
• All practices had a minimum of 8
years of complete computerised data
Study cohort
• All patients with a 1st diagnosis of
CHD between 1st Jan 96 and 31 Dec
2003
• Study period chosen as statins only in
common use after Jan 96
Cases & controls
• Cases were CHD patients who died
during study period
• Controls were CHD patients who
didn’t die matched for
– Age
– Year of birth
– Sex
• Four controls per case
Index date
• Assigned index date to each case [date
of death]
• Assigned same date to matched
controls
• Reviewed medical & drug history
prior to this date
Outcomes
• Odds ratio (95% CI) for all cause mortality
in cases vs controls
• Exposure
– combinations of aspirin, statins, beta blockers and
ACE inhibitors
• Adjusted for
– Comorbidity (MI, Diabetes, hypertension, CCF)
– Smoking
– obesity
– deprivation
Statistics
• Conditional logistic regression using
STATA v8.2
• Examined for 2 way interactions
– drug*drug
– drug*disease
• Included significant interactions in
model
• Selected p < 0.01
PRELIMINARY RESTULS
study population
1.17 million patients in population
13,029 patients with CHD
2,266 deaths
So 2,266 cases + 9064 matched controls
Comparison with 4S study
• Patients on statins had 37% lower
risk of death c.f. those not on statins
• This is similar to the 30% reduction
in risk reported in 4S study
• Marginally greater benefit as our
population higher risk than trial
population
Odds ratio for death
Combination
Adjusted odds ratio*
Statin + aspirin + BB 0.63 (95% 0.51-0.79)
+ ACE
Statins + aspirin + BB 0.43 (95% 0.34-0.55)
* Adjusted for CCF, diabetes, hypertension, MI,
smoking, obesity, deprivation etc
In other words
• Despite adjustment for comorbidity,
patients on combination of :
– 3 drugs had 57% lower risk of death
– 4 drugs had 37 % lower risk of death
• Combination without ACE was better
Discussion of methodology
Case control study but
• Very large sample
• No recall bias
• No selection bias
• Misclassification unlikely as drugs all
prescribed except aspirin which is
also OTC
Possible explanations
• We found significant statistical
interaction between ACEI and statins
• Possible explanations
– ? misclassification of CCF (but no
interaction between other drugs)
– ? Chance finding
• Whatever, study challenges
assumptions that benefits
automatically stack up in ‘real life
populations’
Next Steps
Thank you for
listening
Ideas and questions
welcome!