Full text of DOI:10.2165/00044011-200222090-00002

Clin Drug Invest 2002; 22 (9): 575-584
1173-2563/02/0009-0575/$25.00/0
ORIGINAL RESEARCH ARTICLE
© Adis International Limited. All rights reserved.
A ‘Real-World’ Analysis of Persistence
on and Adherence to Glipizide GITS,
Glipizide IR and Glibenclamide in
Managed Care among Patients with
Type 2 Diabetes Mellitus
Carolyn R. Harley,1 Avanish Mishra,2 Scott E. Cantor,2 Michael Nelson1 and
Alexander M. Walker1
1
2
Ingenix Inc., Eden Prairie, Minnesota, USA
Pfizer Inc., New York, USA
Abstract
Objective: This study compared persistence on and adherence to therapy with
glipizide gastrointestinal therapeutic system (GITS), glipizide immediate release
(IR) and glibenclamide (glyburide) in commercially insured patients newly
treated for type 2 diabetes mellitus. The objective of the study was to determine
if there were differences in persistence and adherence between the three second-
generation sulphonylureas.
Study design: This was a retrospective longitudinal claims data analysis for
commercial enrollees in eight independent practice model health plans. Study
subjects were 25 years of age or older with a first prescription for a study drug
from 1 January 1996 through 31 December 1999. All subjects were newly treated
with medication for diabetes mellitus, and were initiated on monotherapy. To be
included in the study, subjects had to be continuously enrolled in their health plan
6 months prior to their index claim and at least 30 days following the index claim.
Methods and subjects: Persistence was defined as the total days from the index
prescription fill date until termination, switch or augmentation of therapy.
Adherence was defined as the ratio of days supplied to total days in the treatment
period. The treatment period for the measurement of adherence was defined as
the period from index prescription fill date to run-out of days supplied of the last
filled prescription for the index drug. Cox proportional hazards analysis was used
to compare differences in persistence, and multivariate regression was used to
assess differences in adherence.
Of the 24 311 subjects, 35% filled a first prescription for glipizide GITS, 15%
for glipizide IR, and 50% for glibenclamide. Over one-half of study subjects were
male, and the average age was between 51 years for the glipizide GITS cohort
and 53 years for the glibenclamide cohort.
Main outcomes and results: By the end of the study, 79% of subjects had

576
Harley et al.
terminated therapy with their index drug. Cox proportional hazards analysis
showed that patients taking glipizide IR were 1.33 times more likely to experience
treatment change [95% confidence interval (CI) 1.25 to 1.42], and patients taking
glibenclamide were 1.16 times more likely to change therapy (95% CI 1.11 to
1.22) compared with patients taking glipizide GITS in the first 90 days following
initiation of therapy. Similar results were found upon subsequent analysis in the
1620 days following the index prescription. The analysis of adherence showed
that patients taking glipizide IR or glibenclamide were less adherent to therapy
compared with patients taking glipizide GITS (p < 0.001).
Conclusions: Glipizide GITS appears to have an advantage in persistence on
and adherence to therapy compared with glipizide IR and glibenclamide. These
differences may be related to administration frequency. Lack of persistence and
adherence has potential clinical and economic consequences.
Diabetes affected approximately 4% of Ameri-
patients, multiple doses per day are necessary to
manage their diabetes mellitus.
cans in the year 2000.[1] Direct medical expendi-
tures for diabetes mellitus were approximately
$US44 billion in 1997, and total direct and indirect
costs were approximately $US98 billion.[2] Type 2
diabetes mellitus accounts for 90 to 95% of dia-
betes cases, and is a significant factor in adult
morbidity and mortality in the USA. For the over
14 million people diagnosed with type 2 diabetes
mellitus, health risks include heart disease, stroke,
kidney disease, blindness, high blood pressure,
peripheral vascular disease and neuropathies.[3]
Intensive control of glucose limits the progres-
sion of microvascular and macrovascular compli-
cations in patients with type 2 diabetes mellitus
newly initiating therapy.[4,5] Oral hypoglycaemic
agents, including second-generation sulphonyl-
ureas, are generally the first line of pharmacologi-
cal therapy, used in approximately 64% of patients
within the first 4 years following diagnosis.[6] Most
patients are initiated on oral therapy when diet and
exercise alone provide inadequate glucose control.
Adherence to therapy is important in the main-
tenance of glucose control in patients with type 2
diabetes mellitus. Patients who are adherent to
therapy are taking their prescription medications in
accordance with the instructions provided by a
physician. In the use of second-generation oral
hypoglycaemic medications for the treatment of
type 2 diabetes mellitus, a minimum of a once-
daily dose is required for glucose control. For some
Persistence on therapy can have two separate
components. A physician switching or augmenting
initial therapy may be responding to evidence that
the initial therapy was not adequately controlling
the patient’s diabetes mellitus. Alternatively, some
patients may stop taking their medication because
they experience adverse effects or for other per-
sonal reasons. Lack of persistence suggests that the
medication the patient was on is no longer effective
in managing the diabetes mellitus, whether by
patient or physician evaluation.
This large retrospective, longitudinal claims
data analysis compares the persistence on and
adherence to glipizide gastrointestinal therapeutic
system (GITS), glipizide immediate release (IR)
and glibenclamide (glyburide) in a sample of
patients newly initiated on oral hypoglycaemic
monotherapy. While actual behaviour cannot be
observed in claims data, these types of analyses
provide an accurate picture of ‘real-world’ utilisa-
tion represented by the frequency and regularity in
which prescriptions are refilled by patients.
Study Design and Methods
Study Design
The study design was a retrospective, longitudi-
nal claims data analysis in eight UnitedHealthcare
discounted fee-for-service, independent practice
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

Analysis of Persistence and Adherence in Diabetes Treatment
577
association model health plans located in the
northeastern, southeastern and midwestern USA.
Approximately 3.8 million members were enrolled
in the eight plans at some time from 1 January 1996
through 31 December 1999. The elements of the
claims data required for the present study included
membership contract information, which provides
dates of eligibility and basic demographic charac-
teristics, and pharmacy claims, which provide the
nature, amount, days’ supply, and date dispensed
for all filled prescriptions.
Study subjects were included if they filled
a prescription for glipizide GITS, glipizide IR
or glibenclamide from 1 January 1996 through
31 December 1999. Amaryl, another second-
generation sulphonylurea, was excluded from
analysis as this drug was launched after 1 January
1996. Uptake of a new drug in the market would
have created possible bias in the analysis compar-
ing across treatment cohorts.
All subjects were at least 25 years of age to
improve the likelihood of capturing patients with
type 2 diabetes mellitus. In addition, subjects were
continuously enrolled in the health plan for the 6
months prior to their index claim and for at least
30 days following their index claim. Subjects with
a claim for an oral hypoglycaemic agent or insulin
in the 6 months prior to their index prescription
were excluded from study. All subjects were initial
monotherapy users, with no filled prescription for
any oral hypoglycaemic agent other than the index
drug or insulin during the 14 days following the
index prescription. Subjects were followed until
they disenrolled from the health plan or until the
end of the follow-up period, 30 June 2000, which-
ever occurred first.
treatment change within 1620 days was analysed.
These two different time periods were assessed to
determine if there were differences in persistence
in the short and long term. The co-variates in the
model included indicator variables to represent the
treatment groups glipizide IR and glibenclamide,
with glipizide GITS recipients treated as the refer-
ence group. Indicator variables for prescribing
physician specialties included cardiologist,
general practitioner, internist and other specialties.
Endocrinologists formed the reference group. Two
variables representing health plans in the south-
eastern and northeastern regions of the USA were
included, with the midwestern region used as the
reference region. Indicator variables represented
the year of the index prescription. The model esti-
mated the likelihood of treatment change over the
entire study period. Regression analysis was used
to compare adherence to drug therapy. Because of
heteroscedasticity in the error term, robust stand-
ard errors were estimated for the regression re-
sults.[7]
Study Sample
A total of 49 203 health plan enrollees were
identified with a claim for one of the three study
drugs during the 4-year capture period. Of these,
7733 enrollees (15.7%) were excluded as poly-
therapy users. An additional 11 290 enrollees
(22.9%) were excluded because of utilisation of an
oral hypoglycaemic agent or insulin in the 6
months prior to the index date. Finally, 5849
enrollees (11.9%) were excluded for not meeting
the minimum age and continuous enrolment
criteria. The final sample size was 24 331 subjects.
Outcome Measures
Methods
The outcomes of interest in the study were per-
sistence with initial monotherapy and adherence to
therapy. Persistence (time until treatment change)
was defined as the total days from the index pre-
scription fill date until the occurrence of a filled
prescription for a non-index oral hypoglycaemic
drug or insulin, or until termination of therapy with
the index drug, defined as no filled prescriptions
Cox proportional hazards analysis (survival
analysis) was used to estimate the relative risk of
treatment change for the three treatment groups.
Two time periods were analysed. Firstly, the rela-
tive hazard of treatment change within the first
90 days of treatment was estimated. Secondly, to
assess changes in the relative hazard over time,
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

578
Harley et al.
within days supplied plus 30 days. Adherence was
defined as the total days supplied for the index drug
divided by total days from index date to the end of
treatment with index drug.[8] End of treatment for
this purpose was defined as no filled prescriptions
for the index drug up to 30 days following run-out
of a prescription for the index drug. Adherence is
a proxy measure of compliance. Therefore, this
measure allowed for the addition of other oral
hypoglycaemic agents or insulin to the regimen
with the index drug.
with the other two groups. The ratio of males to
:
females was roughly 3 2 in all three treatment
groups.
Over one-half of glipizide GITS recipients had
their index prescription in 1998 or in 1999, com-
pared with less than one-half of the glipizide IR
and glibenclamide recipients. Because glipizide
GITS was newer to the market (launched in 1994),
it is possible that the difference in index years
reflects the incorporation of glipizide GITS into
physician practice patterns over time.
General practice was the most common spe-
cialty associated with the index prescription for
glipizide GITS and glibenclamide recipients, re-
presenting 49% and 42% of prescribing physicians
for these treatment groups, respectively. Glipizide
IR recipients were most likely to receive their first
prescription from an internist (41% of subjects).
A larger proportion of glipizide GITS recipients
were enrolled in health plans located in the south-
eastern USA compared with glipizide IR and
glibenclamide recipients. All three groups had
roughly equivalent proportions of subjects in
Results
Descriptive Statistics
The background variables, including age,
gender, prescribing physician specialty, index year
and health plan region are described in table I. The
mean age of glipizide GITS recipients (50.8 years)
was slightly lower than the mean age of glipizide
IR (52.1 years) and glibenclamide (52.6 years)
recipients. A larger proportion of glipizide GITS
recipients were under the age of 55 years compared
Table I. Summary of demographic and descriptive characteristics of the drug cohorts
Characteristics
Male [n (%)]
Glipizide GITS (n = 8618) Glipizide IR (n = 3580)
Glibenclamide (n = 12 133)
7030 (57.9)
4945 (57.4)
50.8 (10.1)
2022 (56.5)
52.1 (10.4)
Age [y; mean (SD)]
52.6 (10.1)
Prescribing physician specialty [n (%)]
General practitioner
Internist
4257 (49.4)
2906 (33.7)
139 (1.6)
1236 (34.5)
1469 (41.0)
102 (2.8)
5110 (42.1)
4583 (37.8)
261 (2.2)
Cardiologist
Other/unspecified
Endocrinologist
1127 (13.1)
189 (2.2)
712 (19.9)
61 (1.7)
1941 (16.0)
238 (2.0)
Health plan region [n (%)]
Southeast
2141 (24.8)
761 (8.8)
552 (15.4)
334 (9.3)
1942 (16.0)
1253 (10.3)
8938 (73.7)
Northeast
Midwest
5716 (66.3)
2694 (75.3)
Year of index prescription [n (%)]
1996
1997
1998
1999
1447 (16.8)
2434 (28.2)
2543 (29.5)
2194 (25.5)
759 (21.2)
1294 (36.1)
874 (24.4)
653 (18.2)
2896 (23.9)
4136 (34.1)
3051 (25.1)
2050 (16.9)
GITS = gastrointestinal therapeutic system; IR = immediate release; SD = standard deviation.
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

Analysis of Persistence and Adherence in Diabetes Treatment
579
point during the study period. These results were
not adjusted for co-variates.
Glipizide GITS
Glibenclamide
Glipizide IR
1.00
0.75
0.50
0.25
0.00
The correlation between adherence and persist-
ence was calculated and was found to be of
moderate size (rho = 0.191), suggesting that the
two measures are related, but measure two distinct
phenomena.
Multivariate Analysis
0
200
400
600
800
Persistence: Cox Proportional Hazards Analysis
(Survival Analysis)
1000 1200 1400 1600
Analysis time (days)
Results of the survival analysis are presented
for two time periods in table II. The relative hazard
(or incidence rate) of discontinuation, switch or
augmentation of therapy over a 90-day follow-up
period is analysed. In the second model, the rela-
tive hazard of treatment change within a 1620-day
follow-up period is analysed. These analyses ad-
justed for differences in the co-variates described
in the Methods section.
Fig. 1. Kaplan-Meier survival curves. GITS = gastrointestinal thera-
peutic system; IR = immediate release.
northeastern health plans, although the proportion
of glipizide GITS recipients in northeastern plans
was significantly smaller than the proportion of
glibenclamide recipients in northeastern plans. A
smaller proportion of glipizide GITS recipients
were enrolled in midwestern plans (66%) com-
pared with glipizide IR (75%) and glibenclamide
recipients (74%).
Within the first 90 days, 9261 subjects (38.1%)
discontinued treatment. By 1620 days, 19 255 sub-
jects (79.1%) had discontinued treatment. Of
patients taking glipizide GITS, 3063 subjects
(35.5%) discontinued in the first 90 days and 6593
subjects (76.5%) discontinued in the first 1620
days. Of patients taking glipizide IR, 1533 subjects
(42.8%) ended treatment in the first 90 days and
2977 subjects (83.2%) ended in the first 1620 days.
Of patients taking glibenclamide, 4665 subjects
(38.5%) discontinued treatment in the first 90 days
and 9685 subjects (79.8%) did so in the first 1620
days. Tests of proportionality did not detect any
change in the relative rates of change over time
between the different treatment groups.
Patients taking glipizide IR and glibenclamide
had higher change rates in the 90-day study period
compared with patients taking glipizide GITS.
Patients taking glipizide IR were 1.3 times more
likely to experience change, and patients taking
glibenclamide were 1.2 times more likely to
change compared with those taking glipizide
GITS. Older subjects were less likely to change
than were younger subjects. For every 10-year
increase in age, the rate of treatment change de-
creased by 17%. Males were less likely to experi-
ence treatment change compared with females.
Subjects with a general practitioner or internist as
their prescribing physician were also less likely to
change in the first 90 days of treatment compared
with subjects with an endocrinologist as their
prescribing physician, with the relative rate change
17% and 20% lower, respectively. Subjects in
southeastern health plans were approximately 1.2
times more likely to change compared with sub-
jects in midwestern plans. However, subjects in
northeastern plans were less likely to change com-
pared with subjects in the midwestern plans. All
three index-year indicator variables were signifi-
cantly related to the risk of change compared with
Figure 1 shows the Kaplan-Meier survival
curves for the three treatment groups over time.
This figure shows that glipizide GITS has a higher
probability of continuation over time compared
with glibenclamide and glipizide IR. Glipizide IR
has the lowest probability of continuation for every
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

580
Harley et al.
the reference group, subjects with index dates in
1999.
Both glipizide IR recipients and glibenclamide
recipients were more likely to experience treat-
ment change in the 1620-day analysis period com-
pared with glipizide GITS recipients.
(table III). Older age increased adherence to ther-
apy. Having a general practitioner or internist as
the prescribing physician decreased adherence to
therapy compared with subjects with an endocri-
nologist as their prescribing physician. Compared
with having the index prescription in 1996, having
the index prescription in 1999 increased adherence
to therapy. Subjects in a southeastern health plan
had lower adherence compared with subjects in a
midwestern plan.
Adherence: Regression Analysis
Regression analysis was performed estimating
the effects of treatment group on the ratio of days
supplied to total possible treatment days (adher-
ence ratio), controlling for age, gender, geographic
region, prescribing physician specialty and year of
index prescription. Both glipizide IR and gliben-
clamide recipients had decreased adherence to
therapy compared with glipizide GITS recipients
Discussion
The purpose of this study was to compare the
persistence on and adherence to treatment with
glipizide GITS, glipizide IR and glibenclamide.
Table II. Survival analysis – persistence: treatment change in 90 days and in 1620 days
Model – 90 days
Model – 1620 days
hazard ratio
hazard ratio
95% CI
95% CI
Cohort
Glipizide IR
Glibenclamide
Glipizide GITS
1.332a
1.164a
1.000
1.251-1.417
1.111-1.220
1.261a
1.122a
1.000
1.207-1.318
1.086-1.158
Male
Age
0.922a
0.983a
0.885-0.961
0.981-0.985
0.929a
0.986a
0.903-0.956
0.984-0.987
Prescribing physician specialty
General practitioner
Internist
0.833b
0.803c
0.876
1.022
1.000
0.725-0.956
0.699-0.923
0.719-1.067
0.885-1.180
0.855c
0.850c
0.923
1.001
1.000
0.774-0.945
0.769-0.940
0.804-1.060
0.902-1.110
Cardiologist
Other/unspecified
Endocrinologist
Health plan region
Southeast
1.227a
0.855a
1.000
1.167-1.291
0.793-0.922
1.137a
0.851a
1.000
1.096-1.179
0.809-0.895
Northeast
Midwest
Year of index prescription
1996
1997
1998
1999
0.935b
0.891a
0.894a
1.000
0.878-0.996
0.841-0.944
0.842-0.950
1.169a
1.135a
1.094a
1.000
1.117-1.223
1.088-1.183
1.047-1.142
a
b
c
p < 0.001.
p < 0.05.
p < 0.01.
CI = confidence interval; GITS = gastrointestinal therapeutic system; IR = immediate release.
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

Analysis of Persistence and Adherence in Diabetes Treatment
581
Table III. Regression analysis – adherence ratio
Coefficient
95% CI
Cohort
Glipizide IR
Glibenclamide
Glipizide GITS
–0.005a
–0.006a
1.000
–0.009
–0.008
–0.002
–0.003
Male
Age
0.000
–0.002
0.001
0.002
0.001
0.001a
Prescribing physician specialty
General practitioner
Internist
–0.005
–0.004
–0.002
–0.007
1.000
–0.012
–0.011
–0.013
–0.015
0.002
0.003
0.008
0.000
Cardiologist
Other
Endocrinologist
Health plan region
Southeast
Northeast
–0.008a
–0.005b
1.000
–0.011
–0.008
–0.006
–0.001
Midwest
Year of index prescription
1997
1998
1999
1996
–0.000
–0.000
0.006a
1.000
–0.003
–0.003
0.003
0.003
0.003
0.010
Constant
0.911a
0.902
0.920
a
b
p < 0.001.
p < 0.05.
CI = confidence interval; GITS = gastrointestinal therapeutic system; IR = immediate release.
Accounting for the length of time study subjects
remained on therapy, the relative risk of a change
in therapy was the focus of the persistence ana-
lysis. Medication refill patterns were used to
measure adherence to therapy, a surrogate for com-
pliance, where regular medication refills imply
that a patient is being adherent with their recom-
mended treatment regimen. If adherence to treat-
ment is more likely with one drug compared with
others, the prospect for better health outcomes is
increased.
This study demonstrated that among new users
of oral hypoglycaemic monotherapy, patients re-
ceiving glipizide GITS were less likely to experi-
ence treatment change over time compared with
those receiving glipizide IR or glibenclamide,
controlling for age, gender, prescribing physician
specialty, health plan location, and year of index
prescription. Adherence to therapy was also asso-
ciated with glipizide GITS use, as glipizide IR and
glibenclamide recipients were less adherent to
therapy compared with glipizide GITS recipients.
Poor adherence and lack of persistence have
potential economic implications. Therapy modifi-
cation generally involves a physician visit, moni-
toring costs and drug costs, including drug wastage
and the costs of additional medications. A second
drug added to a patient’s regimen might add com-
plexity to the patient’s diabetes mellitus regimen,
which may make adherence more difficult to
achieve. In another study using administrative
claims data from a managed-care organisation, to-
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

582
Harley et al.
tal healthcare costs, emergency room visits and
hospitalisations were higher for patients less than
75% compliant when compared with patients with
better compliance.[9]
the other drugs since patients taking this drug were
younger on average. Despite having a young user
population, the change rate for glipizide GITS
recipients was lower than for glipizide IR and
glibenclamide recipients. Adjustment for baseline
differences between the groups served to further
clarify the superiority of glipizide GITS with re-
spect to its utilisation over time.
Having a general practitioner or internist as the
prescribing physician was associated with a lower
likelihood of treatment change compared with sub-
jects with an endocrinologist as the prescribing
physician. Subjects with an endocrinologist as the
prescribing physician may have more serious risks,
have more severe hyperglycaemia, or be more
closely monitored at the onset of therapy. These
patients may be more likely to have augmented
therapy or experienced an early switch to other
agents compared with patients with general practi-
tioners as their initial prescribing physician.
Adherence improves with once-daily admini-
stration compared with three-times-daily adminis-
tration in patients with type 2 diabetes mellitus.[6,10]
In the study by Paes et al.,[11] underconsumption,
or missed doses, occurred more frequently with
diabetes mellitus drugs administered more than
once per day. If missed doses occur with sufficient
frequency, they can lead to a reduced level of glu-
cose control and a perceived lack of effectiveness
on the part of the patient and prescriber. Lower
adherence with dose timing was documented with
twice- or three-times-daily administration re-
gimens when compared with once-daily regimens.
Initial administration for glipizide IR and
glibenclamide can be once per day, but patients
requiring higher dosages of glipizide IR (>10
mg/day) or glibenclamide (>15 mg/day) may be
switched to a twice-daily administration sched-
ule.[12] Glipizide GITS, approved for use in 1994,
is an extended-release form of glipizide and is
taken once a day for 24-hour glucose control.
Because of its extended-release formulation, glipi-
zide GITS provides more stable drug concentra-
tions than the immediate-release formulation.[13]
The convenience of once-daily administration with
The adherence effect, while significant, was
relatively small. It is not known if a small im-
provement in adherence translates into clinically
meaningful improvements in patients with type 2
diabetes mellitus. It is unlikely that the positive
benefits of a small improvement in adherence
would be evident in a short-term observation
period. However, a small improvement in adher-
ence maintained over time may realise long-term
clinical and economic benefits.
The findings from the persistence analysis were
more striking, with glipizide GITS showing a sig-
nificant advantage. Differences in persistence may
be explained by clinically unobserved differences
in treatment groups. For example, initial therapy
with a longer-acting product such as glipizide
GITS may be targeted for more stable patients
whereas shorter-acting drugs such as glipizide IR
and glibenclamide may be preferentially pre-
scribed to less stable patients. By selecting newly
treated patients, this study addressed some of the
potential confounding problems related to chan-
nelling. Other factors, such as clinical complica-
tions and level of metabolic control at treatment
start, may influence selection of initial therapy.
However, these factors were either unavailable or
unmeasured in this study. Economic factors such
as the cost of the medications may influence the
choice of initial therapy. However, all of the study
drugs were included on the health plan’s preferred
drug list during the entire study period, limiting
patient liability to a level co-payment for all three
drugs.
Other factors relating to persistence on therapy
were of interest in this study. Age was negatively
associated with treatment change, suggesting that
older subjects are more likely to continue treatment
over time. This is consistent with other diabetes
mellitus studies that have evaluated predictors of
therapy adherence in diabetes mellitus. This would
appear to disadvantage glipizide GITS relative to
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

Analysis of Persistence and Adherence in Diabetes Treatment
583
glipizide GITS may lead to better adherence and
therefore improve glucose control, leading to more
persistent use of this drug.
management strategy in diabetes mellitus.[15]
Non-adherence with oral diabetes mellitus medi-
cations can occur in the form of dose-taking
(under- and overconsumption) and dose-timing
(not taking at the prescribed interval).[11] Non-
adherence effectively changes the amount of daily
drug consumed, and therefore impacts on sul-
phonylurea concentrations and blood glucose con-
trol. Each of these forms of non-adherence could
lead to suboptimal glucose control, and sub-
sequently lead to modification in therapy by the
prescriber.
Non-adherence to diabetes mellitus therapy has
potential clinical implications for patients, in the
form of reduced drug effectiveness, hyperglycae-
mia and cost. While this study does not establish a
direct link between non-adherence and the need for
therapy modification, it does identify an associa-
tion between glipizide GITS and higher level of
adherence and longer time to therapy modifica-
tion. The results suggest that there are ‘real-life’
advantages with the use of glipizide GITS, a once-
daily time-release sulphonylurea, over glipizide IR
and glibenclamide.
This study had several important findings that
must be considered within the limitations of the
data and study design. Firstly, this study analysed
a sample of managed-care enrollees, and findings
from this study should be interpreted with caution
when considering the impact in a non–managed-
care population. Secondly, this study was con-
ducted using administrative claims data. Actual
drug-taking behaviour could not be observed, and
the pattern of filled prescriptions had to serve as a
surrogate marker for patient behaviour, including
adherence. While using refill patterns as a measure
of adherence is less accurate than observed pill
consumption or pill-counting procedures, use of
refill patterns from claims data may offer some
improvement over self-reported measures of
medication-taking behaviour.
Thirdly, there are many potential reasons for
therapy modification. Claims data are insufficient
to establish the reason for treatment failure.
Fourthly, we assumed that patients with sub-
optimal management had an equal likelihood of
detection by their provider. It is likely that some
patients with non-adherence were unrecognised
because of a lack of monitoring. Prescriber spe-
cialty was adjusted for in the analysis to attempt to
account for differences in monitoring and treat-
ment of diabetes mellitus by specialists. Ideally,
differences in adherence that result in reduced
effectiveness or in increased hypoglycaemic
events would be confirmed via evaluation of clin-
ical observations. In this retrospective database
study, these data were not available.
Acknowledgements
Funding for this study was provided by Pfizer, Inc.
References
1. Boyle JP, Honeycutt AA, Narayan KMV, et al. Projection of
diabetes burden through 2050. Diabetes Care 2001; 24: 1936-
40
2. American Diabetes Association. Economic consequences of
diabetes mellitus in the US in 1997. Diabetes Care 1998; 21:
296-309
3. National Diabetes Data Group. Diabetes in America. 2nd ed.
Washington (DC): US Govt Printing Office, 1995. Report
No.: NIH publ. no. 95-1468
4. UK Prospective Diabetes Study Group. Intensive blood glucose
control with sulphonylureas or insulin compared with con-
ventional treatment and risk of complications in patients with
type 2 diabetes (UKPDS 33). Lancet 1998; 352: 837-53
5. UK Prospective Diabetes Study Group. Association of glycae-
mia with macrovascular and microvascular complications of
type 2 diabetes: prospective observational study (UKPDS
35). BMJ 2000; 321: 405-12
6. Fertig BJ, Simmons DA, Martin DB. Therapy for diabetes. In:
National Diabetes Data Group, editors. Diabetes in America.
2nd ed. Washington (DC): US Govt Printing Office, 1995.
Report No.: NIH publ. no. 95-1468
Conclusion
Dose response studies have demonstrated a sig-
nificant relationship between plasma sulphonyl-
urea concentrations, fasting blood glucose, and
glycosylated haemoglobin within the range of
recommended doses.[14] For this reason, adherence
to the chosen medication regimen is important
to the long-term effectiveness of this disease
7. Stata Reference Manual, Release 7, Volume 3 Q-St. College
Station (TX): Stata Press, 2001
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)

584
Harley et al.
8. Rizzo JA, Simons WR. Variations in compliance among hyper-
ide and immediate release glipizide in patients with NIDDM.
Diabetes Care 1994; 17: 1460-4
tensive patients by drug class: implications for health care
costs. Clin Ther 1997; 19: 1446-57
14. Simonson DC, Kourides IA, Feinglos M, et al. Efficacy, safety,
and dose-response characteristics of glipizide gastrointestinal
therapeutic system on glycemic control and insulin secretion
in NIDDM: results of two multicenter, randomized, placebo-
controlled clinical trials. Diabetes Care 1997; 20: 597-606
15. Diehl AK, Bauer RL, Sugarek NJ. Correlates of medication
compliance in non-insulin-dependent diabetes mellitus.
South Med J 1987; 80: 332-5
9. White TJ, Chang EY, Vanderplas AM. Impact of compliance
on health care cost and utilization in patients with diabetes
mellitus and cardiovascular disease [abstract]. 36th ASHP
Midyear Clinical Meeting; 2001 Dec 2-6; New Orleans.
American Society of Health - System Pharmacists, Bethesda,
MD, 2001: PPR-7
10. Eisen SA, Miller DK, Woodward RS, et al. The effect of
prescribed daily dose frequency on patient medication com-
pliance. Arch Intern Med 1990; 140: 1881-4
11. Paes AH, Bakker A, Soe-Agnie CJ. Impact of dosage frequency
on patient compliance. Diabetes Care 1997; 20: 1512-7
12. Drug Facts and Comparisons. St. Louis (MO): Facts and Com-
parisons, 2001
13. Berelowitz M, Fischette C, Cefalu W, et al. Comparative effi-
cacy of a once-daily controlled release formulation of glipiz-
Correspondence and offprints: Carolyn R. Harley, 12125
Technology Drive, MN002-0258, Eden Prairie, MN 55344,
USA.
E-mail: carolyn.harley@ingenix.com
© Adis International Limited. All rights reserved.
Clin Drug Invest 2002; 22 (9)