Transcript Document

Pharmacoepidemiology:
Past, present and future
Kathleen Bennett, PhD
Department of Pharmacology &
Therapeutics, Trinity College Dublin
[email protected]
3rd Sept 2009
‘A desire to take
medicine is,
perhaps, the great
feature which
distinguishes man
from other animals’
William Osler,1891
What is pharmacoepidemiology?
Defined as the study of the utilization and effects
of drugs in large numbers of people.
Pharmacoepidemiology borrows from both
pharmacology and epidemiology, a bridge
science spanning both pharmacology and
epidemiology.
Pharmacoepidemiology can also be defined
as the application of epidemiological methods to
pharmacological issues.
Pharmacoepidemiology in practice
To quantify adverse events with medicines
in the population
Patterns of drug utilisation, including
adherence
Hypothesis generating
Pharmacoepidemiology –
Observational studies
Without treatment allocation by chance,
bias due to different baseline risks for
disease in users and non-users of drugs
cannot be ruled out completely confounding by indication.
The potential for confounding is probably
larger in observational studies assessing
medications than in studies assessing
lifestyle factors.
Pharmacovigilance
There are also some areas that are altogether
unique to pharmacoepidemiology, e.g.
pharmacovigilance.
Pharmacovigilance is a type of continual
monitoring for unwanted effects and other safetyrelated aspects of drugs that are already on the
market.
Pharmacovigilance refers almost exclusively to
the spontaneous reporting systems which allow
health care professionals and others to report
adverse drug reactions to a central agency.
It relies heavily on reporting of safety events by
health professionals.
Pharmacoeconomics
Pharmacoeconomics is that branch of
health economics that focuses upon the
costs and benefits of drug therapy.
Definition: The comparative analysis of
alternative courses of action in terms of
BOTH their costs and consequences.
Pharmacoeconomic Evaluation
Pharmaceutical Expenditure Under
the Community Drug Schemes:
1991-2007
2000
1745
1800
Millions (euro)
1600
1500
1320
1400
1210
1200
1100
1000
880
800
700
560
600
430
400
200
170
190
200
210
240
260
300
350
0
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year
Source: Primary Care Reimbursement Service (PCRS)
*Pharmaceutical expenditure includes payment to pharmacies for the cost of medicines, dispensing fees and payment to wholesalers under the High Tech Drug Scheme
Pharmacoeconomics &
pharmacoepidemiology
Inter-related disciplines
Used by



Industry for marketing/price setting
HSE - reimbursement decisions
Public health - e.g. vaccination programmes
Can help identify whether particular subgroups of patients will benefit most from a
new drug and in which it is most costeffective.
Historical background
US law, Pure Food and Drug Act, passed in
1906, followed by FD and cosmetic Act in 1938.
Preclinical toxicity testing and clinical data about
drug safety required before drug marketed.
‘Thalidomide disaster’ in 1961 led to
establishment of committee of safety of
medicines in 1968 in UK and changes
elsewhere.
‘Pharmacoepidemiology’ first appeared in
medical literature (BMJ) in 1984.
Drug Toxicity
Thalidomide
Chloramphenicol and Grey Baby Syndrome
Gynaecological cancer in offspring of women
receiving Diethyl Stilboestrol
Oculomucocutaneous syndrome with
practolol
Liver disease from benoxaprofen
Valvular heart disease from Dexfenfluramine
Cardiac arrhythmias with terfenadine
Multiple drug interactions with mibefradil
What papers have shaped
Pharmacoepidemiology?
Controversies
Pharmacoepidemiology in Ireland
Since 1968 Ireland has participated in the WHO
International Drug Monitoring System and
pharmacoepidemiology was mainly in the area
of Adverse drug reaction (ADR) reporting.
The Yellow Card Scheme for ADR reporting has
operated through the Irish Medicines Board
since 1996 and the National Drugs Advisory
Board (NDAB) before that. However, ADR
reporting rates are low; <10% of all serious and
2-4% of non-serious ADRs reported.
The Pill Scare
In October 1995, the Committee of Safety of Medicines
(CSM) issued a warning on the increased risk of
thromboembolism associated with the third-generation
oral contraceptive steroids.
This ‘Pill Scare’ led to some users stopping the oral
contraceptive steroids mid-cycle and a rise
subsequently in abortions and pregnancies was noted.
The Irish Medicines Board did not advise
discontinuation of third -generation oral contraceptives ,
but advised further study and analysis of the previous
studies to evaluate the impact of biases and
confounders.
No regulatory action was taken.
Percentage uptake of oral contraceptives by
generation between January 1995 and November
1996.
2nd
60
3rd
50
Norgestimate
40
30
20
10
M
ar
nu
ar
y95
ch
-9
5
M
ay
-9
5
Se Jul
ypt
95
em
be
N
rov
em 95
be
Ja r-9
5
nu
ar
y96
M
ar
ch
-9
6
M
ay
-9
6
Se Jul
ypt
96
em
be
N
rov
em 96
be
r96
0
Ja
Percentage
1st
Month
Examples of
pharmacoepidemiology databases
GPRD – GP research database in UK
MEMO – Scottish record linkage data
Saskatchewan Health Services - Canada
Kaiser Permanente – US
Odense database- Denmark
HSE-PCRS – prescribing data only Ireland
HSE-Primary Care Reimbursement
Services (PCRS)
The PCRS is part of the HSE, and is responsible for
making payments to healthcare professionals, e.g.
doctors, dentists and pharmacists, for the free or
reduced costs services they provide to the public.
It supports the delivery of primary healthcare by
providing reimbursement services in their own
community.
There are many schemes under the HSE-PCRS but
for drug prescribing the three main ones are: GMS
medical card scheme, Drug payment (DPS) and Long
Term Illness (LTI) schemes.
HSE-PCRS pharmacy claims
database
The number of GMS eligible persons by
December 2007 was 1.28 million people
( ~ 30% of population).
Not fully representative - socially
disadvantaged persons, children and the
elderly are over-represented.
Accounts for approximately 70% of all
prescribed medicines.
Limitations – no diagnosis or outcome
data
Examples
Quality prescribing indicators

Elderly population exposed to potentially
inappropriate medications; generic prescribing
European drug utilisation studies

Statins, PPIs, SSRIs, ACE/ARBs
Various disease areas including adherence
to tamoxifen in breast cancer
Prevalence of potentially inappropriate
medication use in those 65+ years
The 1-year risk of receiving at least one PIM using the combination of Beers and
McLeod criteria was 20.6% using the national prescribing data (2004)
Criteria- PIM
Proportion of all
patients (n=367,260)
Proportion of
patients in the ERHA
(n=101,935)
Proportion of patients
in ERHA nursing
homes (n=2085)
9515 (2.59%)
2340 (2.30%)
43 (2.06%)
Sedatives
Long-acting
benzodiazepines
39059 (10.64%)
11963 (11.74%)
332 (15.92%)
Antibacterial
Nitrofurantoin
7028 (1.91%)
2277 (2.23%)
87 (4.17%)
Anticoagulants and
aspirin
4943 (1.35%)
1588 (1.55%)
28 (1.34)
Anticoagulants and
NSAIDs
6370 (1.73%)
1767 (1.73%)
35 (1.68%)
Antidepressants
amitriptyline,
doxepin, flouxetine
Potential Drug-Drug
Interactions
% of prescriptions dispensed
generically on the GMS in 2008
60.0%
56.8%
50.0%
40.0%
30.0%
24.9%
20.0%
15.9%
10.0%
2.4%
0.0%
Unbranded generic
Branded generic
Proprietary drug with a
generic equivalent
Proprietary drug with no
generic equivalent
International Comparisons of
Drug Utilisation Statistics
% utilisation generic omeprazole vs. all PPIs
120
2001
2004
2007
100
% Utilisation
80
60
40
20
Country
Sp
ai
n
Sw
ed
o
Sl
Se
r
Po
l
Po
rt
Sc
o
or
N
Li
th
Ita
l
Ire
e
Ic
er
m
t
G
Es
En
g
ro
C
Au
0
Differences may reflect demographics and reimbursement issues
BMJ 2005
Aliment Pharm Therap 2009
On September 30, 2004, Merck and Co.
voluntary withdrew rofecoxib (Vioxx) due
to increased risk of CV events
B J Clin Pharm 2007
Hormone treatment in cancer
Hormone therapy is designed to alter hormone
production in the body so that cancer cells stop
growing or are killed completely.
Breast cancer hormone therapies often focus on
reducing estrogen levels (a common drug for this
is tamoxifen) and prostate cancer hormone
therapies often focus on reducing testosterone
levels.
In addition, some leukemia and lymphoma cases
can be treated with the hormone cortisone.
Early Discontinuation
of Tamoxifen
T I Barron1 R M Connolly2 K Bennett1
M J Kennedy2 J Feely1
1 Dept
Pharmacology & Therapeutics,
Trinity College Dublin.
2 Academic Unit of Clinical & Molecular Oncology
Trinity College Dublin & St. James’s Hospital.
Email: [email protected]
Background
Adherence & Persistence
Definitions
Adherence (Synonym: Compliance) is the extent to which a
patient takes medication in accordance with the prescribed
interval and dose.1
Persistence is the accumulation of time from initiation to
discontinuation of therapy.1
Significance?
Substantial worsening of disease and mortality.
Increased healthcare costs.
Healthy drug user effect.
1
ISPOR Medication Compliance & Persistence Special Interest Group.
http://www.ispor.org/sigs/MCP_accomplishments.asp#definition
Background
Tamoxifen
5 years of treatment reduces the relative breast
cancer recurrence risk by 46% and the relative risk
of death by 26%.1
Comparisons of treatment durations indicate that
women receiving less than 5 years of treatment
have significantly higher breast cancer recurrence
rates and mortality.1,2
1
Lancet 1998; 351: 1451-67;
2
Lancet 2005; 365: 1687-717
Aims
Aims
To evaluate persistence with tamoxifen therapy, in
women aged 35 years or older, using prescription
refill data from a national prescribing database.
Prescription refill data provide accurate and objective
estimates of medication use in large populations over
long periods of time.
This is the first study to evaluate tamoxifen
persistence using this method.
Methods
HSE-PCRS 2000-2005
All women over the age of 35 years , commenced
on tamoxifen as initial hormonal therapy between
January 2001 and January 2004.
Measurement of tamoxifen persistence
A period of 180 consecutive days of no tamoxifen
supply.
Patients with no prescription for any item in the 12
months following non-persistence were reclassified
as lost to follow up.
Patients starting alternative hormonal therapy
before 180 days of no tamoxifen were reclassified
as treatment switchers.
Methods
Potential determinants of non-persistence
Age at initiation of tamoxifen.
Co-morbidities.
Number of pharmacological agents received.
Use of antidepressant, antipsychotic, or
anxiolytic/hypnotic agents.
Treatment for cognitive or functional impairment
(Parkinson’s disease or dementia)
Cox proportional hazards model.

Stepwise selection with criteria for entry of p <0.1.
Results
Study Cohort (n=2816)
Table 1: Characteristics of patients starting tamoxifen on the HSE-PCRS database (January 01January 04)
n
%
n
%
Cognitive/functional impairment †
Age (years)*
35-44
262
9.3
45-54
509
18.1
55-64
592
21.0
65-74
575
20.4
>75
878
31.2
Prescription Drug Use †
Parkinson’s Disease
81
2.9
Dementia
25
0.9
0
1906
67.7
≥1
910
32.3
Number of co-morbidities †
Number of cognitive/functional impairments †
Benzodiazepine anxiolytic/hypnotic
1163
41.3
0
2710
96.2
Antidepressant
534
19.0
≥1
106
3.8
Benzodiazepine related hypnotic
390
13.8
Antipsychotic
375
13.3
Co-morbidities †
Mean number of pharmacological agents per month †
≤1
951
33.8
>1 ≤3
920
32.7
Respiratory disease
385
13.7
>3 ≤5
502
17.8
Cardiovascular disease
599
21.3
>5
443
15.7
Diabetes
117
4.2
* At tamoxifen initiation; † In 12 months prior to tamoxifen initiation
Results -Tamoxifen non-persistence
Fig 1: Outcomes for patients starting tamoxifen on the HSE-PCRS database (January 01 – January 04)
Study cohort of 2816 women ≥35yrs commenced on tamoxifen as initial hormonal therapy
(January 2000 – January 2004)
470 (16.7%)
Lost to follow up
716 (25.4%)
Stop tamoxifen & switch hormonal therapy before 180 days
34 (1.2%)
Stop tamoxifen & switch hormonal therapy after 180 days
143 (5.1%)
Stop tamoxifen & restart tamoxifen after180 days
569 (20.2%)
Stop tamoxifen & no subsequent hormonal therapy
884 (31.4%)
Persist with tamoxifen
Non-persistent
Population 746
(26.5%)
Results - Determinants of non-persistence
Table 2: Crude & adjusted hazard ratios (HR) for potential determinants of tamoxifen non-persistence
Crude HR
95% CI
Adjusted HR
95% CI
35-44
1.36
1.01 to 1.82
1.36
1.01 to 1.83
45-54
Reference
-
-
-
55-64
1.07
0.84 to 1.37
1.11
0.86 to 1.42
65-74
1.23
0.96 to 1.57
1.27
0.98 to 1.61
>75
1.42
1.14 to 1.77
1.46
1.16 to 1.83
Benzodiazepine anxiolytic/hypnotic
0.99
0.86 to 1.15
-
-
Antidepressant
1.31
1.10 to 1.56
1.41
1.18 to 1.70
Benzodiazepine related hypnotic (ZZZ)
0.81
0.65 to 1.01
-
-
Antipsychotic
1.16
0.95 to 1.43
-
-
0
Reference
-
-
-
≥1
1.04
0.90 to 1.22
-
-
0
Reference
-
Reference
-
≥1
1.78
1.29 to 2.46
1.72
1.24 to 2.39
≤1
Reference
-
Reference
-
>1≤3
0.89
0.74 to 1.05
0.84
0.71 to 1.00
>3≤5
0.87
0.70 to 1.07
0.76
0.61 to 0.94
>5
0.90
0.72 to 1.12
0.72
0.58 to 0.92
Age (years)
Prescription Drug Use
Number of co-morbidities
Number of cognitive/functional comorbidities
Mean number of pharmacological agents per
month
* At tamoxifen initiation; † In 12 months prior to tamoxifen initiation; ‡ Adjusted
for age, antidepressant use, number of cognitive or functional impairments and
Discussion
Non-persistence with tamoxifen in clinical
practice is higher than previously reported.
35.2% of women discontinue tamoxifen by 3.5 years.
22.1% discontinue tamoxifen by 1 year.
The determinants
Extremes of age, treatment with an antidepressant with nonpersistence
Increasing numbers of prescribed medications associated
with better tamoxifen persistence.
Conclusion
Persistence with tamoxifen cannot be assumed
Raises concerns about persistence with other oral hormonal
therapies/anti-neoplastics in general
Pharmacoepidemiology –
Methodological developments
Studies based on large health care
utilization databases tend to use data
collected for reasons unrelated to the
research hypothesis thus lacking data on
all important confounders.
To address unmeasured confounding and
bias in database studies, several
approaches have been proposed.
Some examples of methodological
developments
STROBE and ISPE good practice guideline
for conduct of PE studies
Propensity scores
Optimal selection of controls
Time dependency in cohort studies
Immortal time bias
Validation Studies
Case-crossover design
Future research
Linkage studies



Electronic record linkage now established between
NCRI and HSE-PCRS for GMS patients (probabilistic
matching of breast, colorectal cancers and continuing
for all cancer sites).
Every GMS eligible patient on NCRI will have
prescribing data available from 2000 onwards.
Will enable future studies
Outcomes from prescribing (protective or otherwise)
Drug utilisation before and after diagnosis of cancer
Pharmacoeconomics of cancer treatments
Hypothesis generating, etc.
Current proposals
Adherence to oral hormonal therapies and
outcomes in breast cancer.

Building on existing work but linked to breast cancer
outcomes.
Breast cancer and digoxin


There is recent laboratory work linking cardiac
glycosides to hypoxia inducible factor (HIF-1) inhibition
which has been show to inhibit tumour growth and
cancer metastasis.
Considering this effect it may be worthwhile to examine
the effect of digoxin (cardiac glycoside) on disease
progression and mortality in women who already have
a diagnosis of breast cancer.
Future scientific development
Application of new epidemiological
methods to this area
Quality of life studies linked to prescribing
Individualising drug therapy
Pharmacogenomics
The rapidly advancing field of pharmacogenomics,
have created important new opportunities in cancer
research and control.
Pharmacogenomics involves identifying an
individual's response to a drug based on his or her
genetic, genomic, and/or proteomic profile.
The concept that each patient's cancer or risk of
cancer has a unique genetic and/or molecular
profile – this may provide new ways to personalize
an individual's therapy for increased clinical benefit.
Example
Co-prescription rates of
tamoxifen and CYP2D6 inhibitors



Co-prescription rate of tamoxifen and
CYP2D6 inhibitors is 13.2%
Literature suggests link to reduced breast
cancer outcome
Avoidance of moderate and potent CYP2D6
inhibitors advised where possible
Conclusion
Pharmacoepidemiology is relatively new science in
Ireland, but of increasing significance as drug
licensing and marketing take on a European
dimension.
From a public health perspective, it is important to
assess the impact that vaccines and drugs have on
the overall patterns of disease in the population.
Finally, society is now more sensitive to the costs of
medical care and in particular drug use.
Pharmacoeconomics can help predict the economic
implications of drug use prior to their marketing.
The young physician starts life with 20 drugs
for each disease, and the old physician
ends life with one drug for 20 diseases.
William Osler
Acknowledgements
HSE-PCRS for supply of data for research purposes.
Dr Ian Barron, Dr Lesley Tilson