Testa MA, Simonson DC. Health Economic Benefits and Quality of Life During Improved Glycemic
Control in Patients With Type 2 Diabetes MellitusA Randomized, Controlled, Double-Blind Trial. JAMA. 1998;280(17):1490-1496. doi:10.1001/jama.280.17.1490
From the Department of Biostatistics, Harvard School of Public Health (Dr Testa), and the Department of Medicine, Brigham and Women's Hospital, Joslin Diabetes Center, Harvard Medical School (Dr Simonson), Boston, Mass.
Context.— Although the long-term health benefits of good glycemic control in patients
with diabetes are well documented, shorter-term quality of life (QOL) and
economic savings generally have been reported to be minimal or absent.
Objective.— To examine short-term outcomes of glycemic control in type 2 diabetes
Design.— Double-blind, randomized, placebo-controlled, parallel trial.
Setting.— Sixty-two sites in the United States.
Participants.— A total of 569 male and female volunteers with type 2 DM.
Intervention.— After a 3-week, single-blind placebo-washout period, participants were
randomized to diet and titration with either 5 to 20 mg of glipizide gastrointestinal
therapeutic system (GITS) (n=377) or placebo (n=192) for 12 weeks.
Main Outcome Measures.— Change from baseline in glucose and hemoglobin A1c (HbA1c) levels and symptom distress, QOL, and health economic indicators
by questionnaires and diaries.
Results.— After 12 weeks, mean (±SE) HbA1c and fasting blood
glucose levels decreased with active therapy (glipizide GITS) vs placebo (7.5%±0.1%
vs 9.3%±0.1% and 7.0±0.1 mmol/L [126±2 mg/dL] vs 9.3±0.2
mmol/L [168±4 mg/dL], respectively; P<.001).
Quality-of-life treatment differences (SD units) for symptom distress (+0.59; P<.001), general perceived health (+0.36; P=.004), cognitive functioning (+0.34; P=.005),
and the overall visual analog scale (VAS) (+0.24; P=.04)
were significantly more favorable for active therapy. Subscales of acuity
(+0.38; P=.002), VAS emotional health (+0.35; P =.003), general health (+0.27; P
=.01), sleep (+0.26; P =.04), depression (+0.25; P =.05), disorientation and detachment (+0.23; P =.05), and vitality (+0.22; P =.04) were
most affected. Favorable health economic outcomes for glipizide GITS included
higher retained employment (97% vs 85%; P<.001),
greater productive capacity (99% vs 87%; P<.001),
less absenteeism (losses=$24 vs $115 per worker per month; P <.001), fewer bed-days (losses=$1539 vs $1843 per 1000 person-days; P=.05), and fewer restricted-activity days (losses=$2660
vs $4275 per 1000 person-days; P=.01).
Conclusions.— Improved glycemic control of type 2 DM is associated with substantial
short-term symptomatic, QOL, and health economic benefits.
MOST THERAPEUTIC studies of patients with diabetes have focused on the
long-term complications of hyperglycemia1,2
or the short-term consequences of hypoglycemia.3- 7
However, other potentially beneficial outcomes of improved glycemic control
on quality of life (QOL), cognition, or symptoms for patients with either
type 11 or type 2 diabetes mellitus (DM)8- 11 generally
have been reported to be minimal or absent. Although the direct and indirect
costs of long-term diabetic complications have been established,12- 16
there has been less attention on evaluating the health economic impact of
changes in glycemic control and QOL on costs associated with work loss and
absenteeism, functional restrictions in daily activities, and use of health
care services, such as physician and home nurse visits.
Some investigators have postulated that QOL and glycemic control are
unrelated and that patients do not perceive any short-term benefits with improved
glycemic control, thereby leading to noncompliance with complex regimens.8 Others have suggested that improvement in QOL with
better glycemic control may be offset by the adverse effects and difficulty
associated with intensive therapeutic regimens.1
However, a recent cross-sectional study of persons diagnosed as having type
2 DM found associations between high hemoglobin A1c (HbA1c) levels and a worsening in symptoms, mood, and well-being.17
Because these issues are critical for assessing the risks, benefits,
and costs of new therapeutic regimens and disease management programs for
type 2 DM, we evaluated the short-term changes in glycemic control, hyperglycemic
and hypoglycemic symptoms, and treatment side effects on QOL, work productive
capacity, functional restrictions, and health care use in patients with mild
to moderate type 2 DM participating in a multicenter, randomized, double-blind,
placebo-controlled trial of the controlled-release sulfonylurea, glipizide
gastrointestinal therapeutic system (GITS).
The trial comprised 4 phases: (1) a 1-week screening and washout period
(week 1) if the patient was receiving prior hypoglycemic therapy; (2) a 3-week
single-blind placebo period (weeks 1 to 3) prior to being randomized to either
diet and placebo or diet and glipizide GITS in a 1:2 ratio; (3) a 4-week double-blind
titration for placebo and glipizide GITS starting at 5 mg with incremental
increases to a higher dose (10 mg, 15 mg, and 20 mg) if the fasting plasma
glucose goal of levels reaching at least 6.4 mmol /L (115 mg/dL) was not achieved
(weeks 4 to 7); and (4) an 8-week maintenance on the final titration dose
(weeks 8 to 15).
Eligibility criteria included men and women with type 2 DM who were
at least 30 years old, were not pregnant, and had been treated with sulfonylurea
and diet or diet alone for a minimum of 3 months. Patients were excluded if
they had a body weight of less than 80% or more than 160% of the ideal weight
set by the Metropolitan Life Insurance tables; had received insulin for longer
than 1 week 3 months before enrollment; had ketoacidosis during the past year;
or showed evidence or had a history of significant confounding illness or
substance abuse. For the QOL assessments, patients were required to read either
English or Spanish at the sixth-grade level.
During the placebo phase, patients were randomized if they met one of
the following criteria: (1) fasting plasma glucose levels between 7.8 mmol/L
(140 mg/dL) and 13.9 mmol/L (250 mg/dL) or (2) HbA1c levels of
more than 6% (by the end of week 2), if fasting plasma glucose levels were
less than 7.8 mmol/L (140 mg/dL) and if plasma glucose levels measured 1 and
2 hours after drinking a liquid mixed meal (Sustacal, Mead Johnson & Co)
were at least 11.1 mmol/L (200 mg/dL). We received approval from institutional
review boards and obtained informed consent from participants.
During the single-blind placebo period clinicians measured HbA1c, fasting blood and plasma glucose, fasting insulin, C-peptide, Sustacal
challenge, and lipid and microalbuminuria levels. Home blood glucose monitoring
levels were also recorded. Clinical and laboratory evaluations were repeated
at 1, 2, 3, 4, 6, 8, 10, and 12 weeks after randomization. Clinicians also
assessed patients' compliance with home blood glucose monitoring, diet requirements,
and diary completion.
Quality of life, days worked, health related days missed, restricted
activity-days, and health care use were assessed at screening, randomization,
and 4, 8, and 12 weeks postrandomization (study weeks −1, 3, 7, 11,
and 15). Quality-of-life outcomes included the following: 5 visual analog
scales (VASs) of perceived health; mental and emotional health; self-reported
cognitive function; general health perceptions; and symptom distress. Assessment
rationale for QOL measurement, and analytical methods21,22
are detailed elsewhere.
Use of health services, employment history, work attendance, and short-term
morbidity as reported by the patient included hospitalizations, ambulatory
care visits, home health care visits, general household assistance, telephone
calls to health care professionals (during the previous 4 weeks), employment,
number of days worked, number of days missed, activity restrictions, and bed-days
caused by health problems (during the previous 7 days). Patients also completed
a home diary that included daily assessments of perceived hypoglycemic and
hyperglycemic reactions with an immediate subsequent home blood glucose measurement;
every-other-week assessments (for 2 days each week) of the analog overall
health rating and corresponding home blood glucose measurement taken before
breakfast and before bedtime; and a monthly assessment of 43 diabetes-specific
symptoms6,7 with corresponding
home blood glucose measurements completed 4 times a day before breakfast,
lunch, dinner, and bedtime.
The costs of lost earnings were based on the average annual wages of
male and female workers by age as reported by the Bureau of Census23 ($116 per day for men and $74 per day for women between
the ages of 35 and 65 years). Production losses for a nonemployed individual
were assumed to be 64% of that of an employed person.24
A total of 594 patients from 62 centers in the United States were randomized
to either placebo and diet management (n=201) or glipizide GITS and diet management
(n=393). In the placebo group, 30 patients (14.9%) were withdrawn prior to
the study's completion compared with 37 patients (9.4%) who were withdrawn
in the glipizide GITS group (P=.055). Of these 594
patients in whom efficacy and safety were assessed, 569 (192 taking placebo
and 377 taking glipizide GITS) met the inclusion criteria pertaining to reading
requirements for the QOL assessments (Figure
We used factor analysis to identify symptom clusters and to calculate
the QOL global outcome score. We used multivariate and univariate analyses
of variance (ANOVAs) and covariance to assess the differences between the
placebo and active therapy groups with regard to changes in the QOL scales,
subscale, and diary symptom scores. We used general linear models to evaluate
the relationship between changes in glycemic control and changes in QOL, health
resource use, work, and disability days. Exact statistical tests on proportions
and incidence rate ratios were used to compare the probability of work loss,
absenteeism, bed-days, and disability-days. Logit models were used to compare
the difference in the rate of change between baseline and end point for these
variables. Means (±SEs) are reported unless specified otherwise. All
QOL and clinical changes were computed from baseline to end point (last value
carried forward) based on intent-to-treat analysis.
Quality-of-life treatment differences are reported in SD of change (responsiveness)
units during stable treatment (calculated using weeks 11 and 15). This metric
of change has been previously calibrated to negative life events.20 The statistical power of 0.80 was based on detecting
greater than a 0.24 SD difference between placebo and glipizide GITS in the
QOL change score (2-tailed statistical test, α=.05). To maximize benefit
to the active therapy group and minimize exposure to placebo while maintaining
adequate power, 2 patients were randomized to glipizide GITS for each patient
randomized to placebo. Significance levels were unadjusted for multiple comparisons
and were all based on 2-tailed tests of significance. Randomization was performed
according to a computer-generated code (balanced in blocks of 4) prepared
prior to the start of the study by the manufacturer of glipizide GITS and
the placebo pills.
Demographic, clinical, and QOL characteristics at baseline are reported
in Table 1. Essential hypertension
(47%), dyslipidemia (27%), and osteoarthritis or unspecified arthritic disorders
(17%) were the most common concomitant illnesses. Concomitant medications
included antihypertensive drugs (39%), drugs for rheumatic diseases and gout
(35%), antibacterial drugs (21%), diuretics (17%), and lipid-lowering agents
(15%). Seventy-nine percent of patients had been taking oral hypoglycemic
therapy previously. There were no significant differences in the baseline
characteristics of the participants placed in either of the 2 randomized
The percentages of patients titrated to a final dose of 5 mg, 10 mg,
15 mg, and 20 mg by the end of the study were 4.5%, 5.8%, 5.8%, and 84.2%
for placebo and 23.8%, 15.4%, 12.1%, and 48.6% for glipizide GITS, respectively.
In the glipizide GITS group, HbA1c and fasting blood glucose levels
exhibited significant decreases from baseline to week 15 of 0.9%±0.1%
and 2.4 ± 0.2 mmol/L (43 ± 3 mg/dL), respectively (P<.001), while the placebo group increased by 0.7%±0.1% (P<.001) and 0.1±0.2 mmol/L (1±4 mg/dL)
(P=.76), resulting in statistically significant differences
in glycemic control between groups (P<.001). At
the end of the study (week 15), HbA1c and fasting blood glucose
levels for active therapy (glipizide GITS) vs placebo were 7.5%±0.1%
vs 9.3%±0.1% and 7.0±0.1 mmol/L (126±2 mg/dL) vs 9.3±0.2
mmol/L (168±4 mg/dL), respectively (P<.001).
During the final week, 7 (4.8%) of 146 patients in the placebo group
and 19 (6.0%) of 316 patients in the glipizide GITS group reported symptoms
of hypoglycemia (P=.67) in their diaries; however,
none had a corresponding blood glucose level less than 3.1 mmol/L (55 mg/dL).
During the same week, the 43-item symptom diaries showed more favorable changes
from baseline for patients taking glipizide GITS compared with those taking
placebo for frequency of urination (P=.048), blurred
vision (P=.049), the hyperglycemia composite symptom
scale (P=.05), and all 3 cognitive functioning scales
(P=.007) (difficulty concentrating [P=.05], inability to think clearly [P=.009],
difficulty remembering [P =.006]). Treatment difference
effect sizes (responsiveness SD units) ranged from +0.37 to +0.56 in favor
of active therapy. There were no significant differences between the 2 groups
in the hypoglycemic symptom scale, which included breathing hard, fast pulse,
pounding heart, sweating, feeling tense, trembling, and feeling lightheaded
and weak (P=.60). Similar results were obtained for
interim study weeks 7 and 11.
The 5 major QOL scales showed more favorable changes for glipizide GITS–treated
patients (P=.008, multivariate ANOVA) with significant
treatment differences ranging from +0.16 to +0.59 responsiveness SD units
(Figure 2). Treatment differences
for symptom distress (+0.59; P<.001), general
perceived health (+0.36; P=.004), cognitive functioning
(+0.34; P=.005), and overall VAS ratings (+0.24; P=.04) were more favorable for the active therapy arm than
the placebo arm. Although the direction of the treatment difference for the
mental and emotional health scale favored the active therapy arm (+0.16; P =.22), the difference did not reach statistical significance.
The associated subscales all favored the active therapy arm with significant
treatment differences or trends for cognitive acuity and memory (+0.38; P=.002), VAS emotional rating (+0.35; P=.003), general health (+0.27; P=.01), sleep
(+0.26; P =.04), depression (+0.25; P =.05), disorientation and detachment (+0.23; P=.045), vitality (+0.22; P=.04), and VAS physical
rating (+0.19; P=.09).
As shown in Table 2, the
more favorable QOL outcomes for glipizide GITS were mediated largely by the
reduction in symptom distress associated with the symptoms of hyperglycemia.
In contrast, hypoglycemic symptoms, although favoring placebo, were not
significantly different between groups.
As depicted in Figure 3, using
data from both groups, the QOL global outcome change score between baseline
and week 15 was calibrated against the interval change in HbA1c
levels. Relative to other studies19,20,25
using similar QOL scales, the calibration showed that increases in HbA1c levels of 1.0% or greater are associated with substantial decrements
in QOL, while decreases of the same magnitude showed smaller, but clinically
important, improvements in QOL. Patients taking active therapy improved significantly
more than patients taking placebo after controlling for the degree of worsening
or improvement in HbA1c levels using analysis of covariance (P=.03). For the group as a whole, improvements in the QOL
summary score were correlated with fewer bed-days (P=.02),
fewer days of restricted activity (P<.001) and
limitations in activities (P=.006), and a decreased
amount of health care services use (P=.01).
To address whether the decreases in QOL were driven mostly by the somatic
effects of symptoms rather than psychological factors such as frustration
due to persistent hyperglycemia despite "taking medication," responses to
the 3 items focusing on "worries, concerns, and fears about health," "feelings
of anxiety and frustration," and "strain, stress, and pressure" were analyzed
separately. Results indicated that both groups improved on all items as the
study progressed with no differences in the degree of improvement. The mental
health scale also showed no deterioration for the placebo group as depicted
previously in Figure 2.
As shown in Table 3, baseline
rates of employment, group productive capacity, health-related absenteeism,
bed-days, and days of restricted activity were similar for the 2 groups.
By week 15, improved glycemic control for patients in the active therapy
group was associated with greater improvement in overall work and disability
outcomes compared with patients in the placebo group. Active therapy patients
experienced higher retention of employment (97% vs 85%) and a greater retained
productive capacity (99% vs 87%) compared with the placebo group, even beyond
what could be attributed simply to the difference in the dropout rate. Changes
in absenteeism, bed-days, and days of restricted activity also were statistically
different between the 2 groups. By week 15, absenteeism rose 8.1% for the
placebo group and decreased slightly by 0.8% for the active therapy group.
The absenteeism risk ratio (missed work cases per person-days worked) was
4.8 (95% confidence interval, 2.0-11.9), indicating a nearly 5-fold increase
in risk of absenteeism for placebo vs active therapy. The number of patients
reporting that they stayed in bed for a half day or more per week rose 4.4%
for the placebo group but slightly declined 0.4% for the active therapy group.
Rates of bed-days and days of restricted activity (half day or more) were
both more favorable for active therapy group. These differences resulted
in higher estimated production losses for the placebo group compared with
active therapy group (Table 4).
The rate of hospitalizations was low at baseline and throughout the
treatment period (0.08 days per person) and was comparable between the 2 groups.
Home nurse visits were used by less than 1% of both groups, while use of general
household assistance was reportedly used by 7.8% of those in the placebo group
and 4.9% of those in the active therapy group during the last 4 weeks of treatment
(P=.22). The number of patients reporting 1 or more
nonstudy ambulatory care visits (clinic, emergency department, or physician
office) were comparable at baseline but decreased more by week 15 for patients
taking glipizide GITS (38.9% to 27.7%; P =.002) compared
with placebo (36.7% to 34.0%; P=.65), yielding an
estimated savings of $11 per patient per month assuming a cost of $66 an ambulatory
visit. A summative index of health care use correlated negatively with the
change in the QOL global outcome score (P=.01), so
as QOL improved with better glycemic control, overall health care use decreased.
Although many studies have focused on the long-term benefits of glycemic
control in reducing diabetic complications, this study demonstrated the shorter-term
symptomatic relief, improvements in QOL, and health economic benefits associated
with improved glycemic control in patients with type 2 DM. The QOL improvements
were both consistent and substantial across all major domains and were due,
in part, to the reduction in adverse symptoms associated with hyperglycemia
with no measurable increase in hypoglycemia or its associated symptoms.
The power of this study to detect a strong relationship between glycemic
control and QOL while other studies have not1,8- 12
is probably due to a number of factors, including a low incidence of hypoglycemia,
which would not offset gains in QOL from improved glycemic control; longitudinal
evaluations, which guard against the confounding typical in cross-sectional
studies; a placebo-control group with substantial hyperglycemia; and comprehensive
disease-specific evaluations focusing on patient self-perceptions of symptoms
In contrast to previous studies of type 1 DM,1,11
diary reports of perceived hypoglycemia and associated autonomic symptoms
were relatively uncommon and not significantly different between the 2 arms.
However, patients in the placebo group reported significantly more distress
with hyperglycemia-related symptoms than patients receiving glipizide GITS.
The diary results suggest that using a reporting period as brief as 24 hours,
treatment differences favoring patients in good control are evident. More
substantiating evidence was obtained from the clinic assessments that showed
that improved glycemic control was associated with improvements in all QOL
domains with only 1 scale, mental and emotional health, not quite reaching
The largest and most dramatic treatment difference occurred in the symptom
distress index and the symptom clusters commonly related to hyperglycemia,
as well as other nonspecific symptoms. Although a cluster of symptoms commonly
associated with hypoglycemia was detected, the treatment difference was much
smaller and not statistically different between groups.
The other consistent finding was the improvement in perceived cognitive
function with improved glycemic control. Loss of cognitive ability is feared
by patients with diabetes as one of the potential negative symptoms of hypoglycemia
often associated with tighter glucose control. Paradoxically, cognitive functioning
showed a much more beneficial profile for patients randomized to the sulfonylurea
glipizide GITS than placebo.
We also observed a strong relationship between improved glycemic control
and the beneficial changes in QOL, suggesting that the rate of QOL deterioration
due to increasing symptoms is progressive with worsening glycemic control.
This relationship also supports the hypothesis that it is the relief from
negative symptoms of hyperglycemia that facilitates the improvement in QOL
seen herein. However, the finding that glucose lowering alone could not account
entirely for the more beneficial QOL profile observed for patients taking
glipizide GITS suggests that other factors, eg, smaller daily fluctuations
in blood glucose levels, may also be involved.
That subjects were not blinded to their home blood glucose readings
may have contributed to the associations observed. However, the symptom distress
index was highly sensitive and specific for identifying the most relevant
disease-specific effects of glycemic control to an extent that cannot be solely
attributed to knowledge of glucose levels (Table 2). In addition, even though fasting blood glucose levels
were identical from baseline to end point for the control group, their corresponding
QOL scores declined substantially, indicating that individuals were reporting
deterioration in QOL even with the knowledge of constant blood glucose readings.
This further suggests that the negative effects of poor control might be
cumulative and progressive.
Previous studies have demonstrated that treatment differences greater
than 0.10 to 0.15 SD units or more are clinically meaningful.20,25
Twenty-three symptoms evaluated by the symptom distress index showed treatment
differences favoring the glipizide GITS group of 0.2 SD units or greater compared
with the placebo group (Table 2).
The treatment difference for the global symptom distress scale was 0.59 SD
units and for cognitive function and perceived health status was approximately
0.35 SD units. For comparison purposes, patients who undergo hip arthroplasty
typically improve between 0.32 and 0.47 SD units, depending on the QOL scale
When evaluating the cost-effectiveness of therapeutic and management
programs, it is important to consider that the changes in QOL found herein
were significantly correlated with use of medical and health care services
as well as work days lost, sick days, and days of reduced activities. Furthermore,
improved glycemic control positively affected employment, absenteeism, productive
capacity, bed-days, and days of restricted activity in addition to ambulatory
medical services. The direct and indirect costs associated with the observed
treatment differences could be considered substantial relative to the typical
capitation allotments for the patient with type 2 DM. However, these savings
must be balanced against the increased costs of improved glycemic control,
such as medication and home glucose monitoring supplies.
The usual methods of evaluating the economic benefits of diabetes treatment
typically focus on lifetime costs and savings using Monte Carlo simulations
and decision analytic modeling techniques.12,27,28
Assuming that the treatment differences observed in this clinical trial could
be maintained throughout the lifetime of this cohort, long-term health and
economic benefits might indeed be substantial. However, patients, employers,
and health care providers might be more motivated to implement and comply
with education and disease management programs if immediate QOL and health
economic benefits could be anticipated.
When assessing the benefit of therapies and disease management programs
for formularies and managed care organizations, emphasis should be placed
on the patient's health perceptions and QOL in addition to the more common
measures of objective health status and process-oriented or report card criteria
(eg, convenience, waiting time, and number of laboratory tests and special
tests performed). Health perceptions may be a more sensitive reflection of
overall patient satisfaction with health care services than such process-oriented
measures. For the patient with type 2 DM, even a moderate worsening in HbA1c levels was shown to affect negatively the patient's QOL and overall
well-being. For the payer, the lost productivity, increased absenteeism, and
increased use of health resources associated with poor glycemic control should
provide a strong incentive for demanding more comprehensive diabetes management
from the health care provider organization than what is currently offered.
We conclude from our findings that the day-to-day QOL in patients with type
2 DM can be enhanced by better glucose control and should be considered when
implementing disease management programs, evaluating health outcomes and patient
satisfaction, and estimating the costs and benefits of therapeutic regimens
and prevention programs.