Vous êtes sur la page 1sur 10

KovatchevJARFall03 12/29/03 2:40 PM Page 449

Postprandial Glucose Dynamics and

Associated Symptoms in Type 2
Diabetes Mellitus
Boris Kovatchev, PhD*
Daniel J. Cox, PhD*
Kent H. Summers, PhD†
Linda Gonder-Frederick, PhD*
William L. Clarke, MD*

*University of Virginia Health System, Charlottesville, Va

Eli Lilly & Company, Indianapolis, Ind

This research was supported by Eli Lilly & Company, Indianapolis, Ind.

KEY WO R D S : postprandial glucose, limit the magnitude of post-meal BG fluc-

hyperglycemia, type 2 diabetes mellitus tuation may reduce its symptomatic and
cognitive consequences.

ABSTRACT In nondiabetic individuals postprandial glu-
Objectives: This study examined the cose (PPG) fluctuations are limited in both
dynamics of postprandial glucose (PPG) their peak value [rarely exceeding 7.8
and symptoms among adults with type 2 mmol/L (140 mg/dL)] and in their dura-
diabetes mellitus (T2DM) in their natural tion, with a peak PPG approximately 1
environment. Using a hand-held computer hour after the start of a meal, returning to
for 70 trials, 36 adults with T2DM rated preprandial levels within 2-3 hours. 1 In
symptoms, performed tests, and measured individuals with type 1 diabetes (T1DM) or
their blood glucose (BG). type 2 diabetes mellitus (T2DM) a number
of factors, such as inadequate available
Results: The mean peak PPG value was
insulin, delayed insulin action, or abnor-
11.3 mmol/L achieved 2-3 hours after meal,
malities in glucagon secretion, contribute
while highest symptom ratings and cogni-
to delayed peak PPG, and higher and pro-
tive slowing were observed within the first
longed PPG elevation.1 The American
hour after meal, at a time corresponding to
Diabetes Association Consensus Statement
the steepest slope of PPG increase. Thus,
on postprandial hyperglycemia concluded
we hypothesized that postprandial symp-
that “in general, a measurement of plasma
toms maybe related to a higher rate of BG
glucose 2 h after the start of a meal is prac-
increase, which was confirmed by high cor-
tical, generally approximates the peak value
relations (r = 0.5-0.75) between symptom
in patients with diabetes, and provides a
ratings and BG rate of increase.
reasonable assessment of postprandial
Conclusions: We conclude that postprandi- hyperglycemia.” 1
al symptom elevation is related to the rate
However, the dynamics of PPG are
of BG increase, thus treatments designed to
complex, dependent on many factors, such

The Journal of Applied Research • Vol. 3, No. 4, Fall 2003 449

KovatchevJARFall03 12/29/03 2:40 PM Page 450

as the amount and the composition of the physiological as well as epidemiologic stud-
meal, and still not well understood. Even ies that excessive post-load glucose excur-
though a linear relationship between post- sions have acute and chronic harmful
prandial and post-challenge (after a 75-g effects on the endothelium and vessel
oral glucose load) glucose 2 hours after a wall.”3 Thus, an assessment of PPG dynam-
meal was established in laboratory condi- ics in the natural environment would be a
tions,2 the dynamics of this relationship in valuable tool for evaluation of glycemic
the field is difficult to assess.3 Perhaps as a control.
result of the lack of a standard PPG assess- In addition to the long-term negative
ment, the usual clinical appraisal of effects of elevated PPG, clinical experience
glycemic control includes only better- suggests a relationship between postprandi-
defined and more stable measures, such as al hyperglycemia and acute and transient
fasting plasma glucose (FPG) and/or glyco- increases in psychological symptoms and
sylated hemoglobin (HbA1c). However, cognitive disruptions.2,16-18 However, there
FPG reflects blood glucose (BG) values have been no prospective and objective
after the effect of carbohydrate intake has investigations of the relationship of such
been eliminated and HbA1c represents the symptoms/cognitive dysfunctions with post-
average BG over a certain period of time, prandial BG parameters, especially with
which makes both of these measures insen- parameters of postprandial BG dynamics in
sitive to BG excursions throughout the the natural environment of people with dia-
day—in particular to PPG fluctuations. For betes. For example, it is unclear whether
example, a recent study4 of more than 800 the peak absolute value of PPG is responsi-
people with T2DM found that after meals, ble for triggering symptoms, or symptoms
many subjects had glucose levels >8.9 are mainly related to the speed and magni-
mmol/L (160 mg/dL) and/or glucose excur- tude of BG increase post-meal. This study
sions >2.2 mmol/L (40 mg/dL) despite investigates when BG peaks in T2DM
HbA1c <7%. This study also concluded that adults following meals in their natural envi-
HbA1c is more related to preprandial than ronment, and whether and which parame-
postprandial BG levels.4 Since there is evi- ters of BG dynamics are associated with
dence to the contrary as well,5 this conclu- experienced symptoms and cognitive dis-
sion remains unclear.6 Thus, despite being ruptions.
the gold-standard marker of glycemic con-
trol,7-8 HbA1c maybe a poor measure of MATERIALS AND METHODS
rapid BG fluctuations.
Mounting evidence, however, points to
Forty-four adults with T2DM gave
the importance of BG fluctuations. A num-
informed consent for participation in this
ber of recent studies found that postprandial
study, which was approved by our institu-
hyperglycemia is an independent factor
tion’s Investigation Review Board. Eight of
contributing to cardiovascular complica-
subjects did not complete the data collec-
tions and increased mortality, especially in
tion: 4 because of difficulty managing the
people with T2DM.9-14 The Diabetes
hand-held computer (ages 61, 63, 70, 74), 2
Intervention Study, the only prospective
because of being too busy, and 2 because of
study considering elevated PPG as a con-
other medical problems. The average age of
tributor to complications in T2DM, con-
the 36 participants was 50 years (SD = 11),
cluded that PPG, but not FPG, was an
average duration of T2DM was 10 years
independent predictor of mortality in
(SD = 9), and average BMI was 34 (SD =
T2DM.15 A recent review of studies in this
10). There were 21 females; 38% of the sub-
area concluded that “there are now compre-
jects used insulin to control their diabetes.
hensive and consistent data from patho-

450 Vol. 4, No. 3, Fall 2003 • The Journal of Applied Research

KovatchevJARFall03 12/29/03 2:40 PM Page 451

Procedure (Lifescan, Milpitas, Calif.). Three precau-

Subjects completed a series of psychomet- tions were taken to encourage and monitor
ric instruments, including the Beck whether symptom entries and cognitive test-
Depression Inventory. They were then ing preceded SMBG. (1) Each HHC trial
instructed to use the Handspring Visor began with the message, “No blood sample
Platinum, (Handspring, Inc, Mountain yet.” (2) The HHC tracked the elapsed time
View, CA) hand-held computer (HHC) between the prompt “Measure your BG”
immediately before self-monitoring of and the entry of this SMBG reading. Since
blood glucose (SMBG). No specific SMBG at least 10 seconds are required for a sub-
schedule was given to the subjects; they ject to lance a finger, collect a blood sam-
were required only to complete 70 HHC tri- ple, and analyze BG level with the One
als within 3-4 weeks. The HHC was Touch Ultra, any readings entered in less
equipped with our custom-developed symp- than 10 seconds were considered invalid.
tom/behavioral assessment software. At (3) The BG readings entered by the subjects
each trial the HHC first collected data on into the HHC were compared to data in the
perceived symptoms and cognitive perform- glucometer’s memory to ensure accuracy of
ance. Then, subjects measured and entered SMBG results. An earlier version of this
their BG level. HHC routine developed for Psion 250 HHC
Symptoms. At each trial the HHC presented was used in our previous studies of symp-
in a random order 16 symptoms and toms and behaviors related to hypo-
prompted subjects to rate them on a scale glycemia.19-21
from 0 = none to 6 = extreme. There were 6 Data Analysis
physical symptoms (need to urinate; BG values were averaged across all subjects
sweet/funny taste; dry eyes, nose, mouth; in 10 time intervals post-meal and plotted
tired/fatigued; thirsty; nausea), 6 mood and compared using univariate ANOVA. To
symptoms (nervous/anxious; irritable/frus- obtain comparable estimates of average BG
trated; restless/jittery; sad/blue; across these intervals, each time interval
giddy/funny; don’t care/apathetic), and 4 was at least one half-hour in duration and
cognitive symptoms (difficulty concentrat- was required to contain at least 200 SMBG
ing; difficulty speaking; uncoordinated; readings, that is, at least 8% of all
slowed thinking). HHC/SMBG readings. This approach
Cognitive tests. The HHC presented the fol- resulted in approximately equal weights (in
lowing tests: (1) 10 mental subtraction prob- terms of number of readings) of the time
lems that used randomly generated 3-digit intervals.
numbers, with subjects entering answers on Symptoms and cognitive test performance.
a number pad; and (2) 2 levels of the Paced In order to eliminate the influence of hypo-
Serial Addition Test (PSAT) presenting a glycemia, symptom ratings and test results
sequence of single-digit numbers for which were considered only if BG was greater
the subject has to enter the sum of each pair than 6.7 mmol/L (120 mg/dL). In order to
of sequential numbers. Levels 1 and 2 of evaluate the magnitude of each individual
the test present numbers at 4-second and 2- postprandial symptom its ratings were aver-
second intervals, respectively. aged across subjects at 1-hour time inter-
Other parameters. The HHC asked subjects vals post-meal and compared using
to enter the time when “you began eating univariate ANOVA. The average symptom
your last meal” and at the end of each trial, magnitude in each category, (physical,
the subjects were prompted to measure and mood, and cognitive) was computed as well.
enter their BG. For the latter all subjects Similarly, cognitive impairment was
used One Touch Ultra glucometers assessed using the time to complete 10

The Journal of Applied Research • Vol. 3, No. 4, Fall 2003 451

KovatchevJARFall03 12/29/03 2:40 PM Page 452

Figure 1. Postprandial blood glucose (BG) levels and postprandial increase in overall symptom
reporting are plotted against elapsed time after a meal (x-axis in minutes). The postprandial BG
values (black squares) are presented in the primary y-axis and demonstrate a peak PPG value
at approximately 2.5 – 3 hours after a meal. The number of elevated postprandial symptoms
(symptoms with ratings greater than 0 over the course of PPG) is presented in the secondary y-
axis (black triangles) and demonstrates a peak symptom reporting within 1 hour after a meal.

mental subtractions and the number of cor- ject’s BG rate of increase (BGRI). This
rect additions on the faster (Level 1) and variable is similar to the previously reported
slower (Level 2) PSAT, averaged at 1-hour BG rate of change,22 but takes into account
intervals post-meal. Average postprandial only increases in BG, not overall fluctua-
BG 1, 2, and 3 hours after meal was corre- tions. Specifically, BGRI was computed as
lated with corresponding symptom ratings the average of the ratios (BG(t2)-BG(t1))/(t2-
and cognitive performance results. In order t1), where BG(t2) > BG(t1) were any two
to account for multiple tests we used increasing consecutive SMBG readings of a
Bonferroni corrections, accepting only subject taken at times t2 and t1 within the
results significant at P <0.01. same day. In essence, this computation pro-
In addition to computing average vided an estimate, for each person, of the
symptom magnitude and in order to evalu- magnitude and speed of increase of his/her
ate the overall dynamics of all symptoms BG levels in mmol/L/h. With this SMBG
combined, the average number of all symp- data set, it was only possible to compute
toms rated 1 or greater was computed in BGRI as a single measure for each individ-
several time intervals post-meal and super- ual, not as a measure specific to each indi-
imposed on the course of PPG. As with vidual’s postprandial period (the latter
PPG, each time interval was required to would require at least 2 SMBG
contain at least 200 HHC trials with readings/subject post-meal for several
BG>6.7 mmol/L. days). However, since fast BG increases are
Rate of BG increase. From each subject’s predominantly observed after meals, we
SMBG data collected concurrently with the assumed that BGRI was mostly influenced
HHC, we computed an estimate of this sub- by postprandial BG elevation.

452 Vol. 3, No. 4, Fall 2003 • The Journal of Applied Research

KovatchevJARFall03 12/29/03 2:40 PM Page 453

Table 1. Average Postprandial Symptom Ratings* and Cognitive Test Performance.

Pre-prandial Postprandial Ratings F (P)

1h 2h 3h

Need to urinate 1.0 1.7 1.3 1.2 7.7 (<0.001)

Sweet/funny taste 1.0 1.8 1.3 0.7 38.8 (<0.001)
Dry eyes, nose, mouth 1.3 2.2 1.8 1.5 13.0 (<0.001)
Tired/fatigued 1.8 2.1 1.9 1.9 1.0 (NS)
Thirsty 1.5 2.4 1.9 1.7 11.2 (<0.001)
Nausea 0.6 1.5 1.0 0.8 13.8 (<0.001)
Average magnitude of
physical symptoms 1.11 1.94 1.56 1.30 18.9 (<0.001)
Nervous/anxious 0.6 1.2 0.9 1.0 7.8 (<0.001)

Irritable/frustrated 0.9 1.2 1.0 1.1 1.4 (NS)

Restless/jittery 0.6 1.1 1.0 1.1 9.1 (<0.001)
Sad/blue 0.4 0.8 0.6 0.7 5.5 (<0.001)
Giddy/funny 0.1 0.5 0.5 0.2 10.6 (<0.001)
Don’t care/apathetic 0.4 0.8 0.5 0.6 8.2 (<0.001)
Average magnitude of
mood symptoms 0.50 0.93 0.76 0.79 9.5 (<0.001)

Difficulty concentrating 0.5 0.9 0.8 0.7 5.2 (<0.001)

Difficulty speaking 0.2 0.5 0.4 0.3 6.6 (<0.001)
Uncoordinated 0.4 0.7 0.6 0.6 4.7 (<0.001)
Slowed thinking 0.5 0.9 0.8 0.7 8.0 (<0.001)
Average magnitude of
cognitive symptoms 0.39 0.75 0.65 0.61 7.9 (<0.001)

Time to complete
10 mental subtractions (sec.) 104 130 110 99 14.9 (<0.001)
PSAT – Level 1
Number correct answers 3.8 3.4 3.2 3.9 2.4 (NS)
PSAT – Level 2
Number correct answers 2.4 1.6 2.1 2.5 11.7 (<0.001)

Using a scale from 0 = none to 6 = extreme.

NS = not significant; h = hour.

RESULTS preprandial and postprandial BG was 2.5

mmol/L (45 mg/dL). One-way ANOVA
Postprandial BG Dynamics
showed that the PPG values were signifi-
Figure 1 presents the average (across sub- cantly different between the time intervals
jects) course of BG fluctuations. The peak post meal (F = 10.3, P <0.001) (see Figure
PPG value of 11.3 mmol/L (203 mg/dL) 1, primary axis).
was achieved between 2.5 and 3 hours after
a meal. The average difference between

The Journal of Applied Research • Vol. 3, No. 4, Fall 2003 453

KovatchevJARFall03 12/29/03 2:40 PM Page 454

Table 2. Correlations of Average Symptom Ratings* and Cognitive Test Performance with
Blood Glucose Rate of Increase.

Correlation Coefficients**
1h 2h 3h
Need to urinate 0.48 -0.02 0.02
Sweet/funny taste 0.39 0.01 0.08
Dry eyes, nose, mouth 0.46 -0.05 -0.02
Tired/fatigued 0.41 0.21 0.18
Thirsty 0.33 0.08 0.03
Nausea 0.42 0.02 0.12
Average magnitude of physical
symptoms 0.52 0.06 0.09

Nervous/anxious 0.69 0.48 0.43

Irritable/frustrated 0.56 0.28 0.33
Restless/jittery 0.50 0.52 0.45
Sad/blue 0.68 0.53 0.44
Giddy/funny 0.50 0.07 -0.06
Not care/apathetic 0.66 0.54 0.37
Average magnitude of mood
symptoms 0.70 0.49 0.44

Difficulty concentrating 0.60 0.39 0.30

Difficulty speaking 0.75 0.34 .25
Uncoordinated 0.76 0.52 .22
Slowed thinking 0.58 0.34 .31
Average magnitude of cognitive
symptoms 0.74 0.44 0.30

Time to complete 10 mental

subtractions (sec.) 0.27 0.06 0.02
PSAT – Level 1
Number correct answers -0.21 -0.21 0.18
PSAT – Level 2
Number correct answers -0.26 -0.02 0.15

* Using a scale from 0 = none to 6 = extreme.

** With this sample size correlations above 0.37 yield P-levels below P= 0.05, while correlations above
0.47 are significant at P = 0.01. The latter are indicated in bold.

Postprandial Symptoms and Cognitive (Table 1).

Slowing Specifically, 5 of the 6 physical symp-
Most physical, mood, and cognitive symp- toms were rated higher post-meal: subjects
toms displayed similar patterns of highest reported a greater need to urinate (F = 7.7,
average rating within the first hour after P <0.001), sweet taste (F = 38.8, P<0.001),
meal and a significant decrease thereafter dry eyes/nose/mouth (F = 13.0, P<0.001),

454 Vol. 4, No. 3, Fall 2003 • The Journal of Applied Research

KovatchevJARFall03 12/29/03 2:40 PM Page 455

thirst (F = 11.2, P<0.001), and nausea (F = cantly with PPG values averaged at the
13.8, P< 0.001). One-way ANOVAs across same time intervals, with the exception of a
time ranges were highly significant for all few correlation coefficients of approximate-
comparisons. Similarly, all mood symptoms ly 0.4 (approximate P = 0.02), which were
(nervous/anxious; irritable/frustrated; rest- not considered significant taking into
less/jittery; sad/blue; giddy/funny; don’t account the large number of simultaneous
care/apathetic) were significantly elevated tests.
within the first hour after a meal, but In order to more precisely assess the
remained elevated somewhat longer than relationship between PPG and overall post-
the physical symptoms. All but one mood prandial symptoms, we superimposed the
symptom (giddy/funny) went up, then number of symptoms with ratings greater
down, then up again with a second smaller than 0 over the course of PPG (Figure 1,
peak between 2 and 3 hours after meal. secondary axis). This superposition con-
This fluctuation, however, was not statisti- firmed that symptom occurrence was not
cally significant. All four cognitive symp- related to highest PPG and in general pre-
toms (difficulty concentrating; difficulty ceded the peak PPG values. Figure 1 also
speaking; uncoordinated; slowed thinking) implied that more symptoms occurred when
became significantly elevated during the PPG increase was fastest, that is, within the
first hour after meal and receded thereafter first-to-second hour post-meal, where the
(Table 1). The average magnitudes in all slope of the PPG curve was steepest. Thus,
symptom categories—physical, mood, and we formulated the hypothesis that postpran-
cognitive—were significantly higher in the dial symptoms are primarily related to a
first hour post-meal (Table 1). higher rate of BG increase, and less to the
Mental calculations within 1 hour post- absolute value of extreme BG.
meal were almost 30% slower (slowing
Postprandial Symptoms and BGRI
from a baseline of about 100 to 130 sec-
In an attempt to test this hypothesis, we
onds to complete 10 subtractions), with this
used BGRI as an estimate of the steepness
effect disappearing within 2 hours (F =
of his slope of BG increase throughout the
14.9, P<0.001). The correctness of the
day for each subject. The average BGRI
mental calculations, however, remained
was 0.78 (SD = 0.60) mmol/L/h and the
constant throughout all postprandial periods
median BGRI was 0.68 mmol/L/h. The
(about 90% correct subtractions). The easi-
BGRI correlated significantly with all
er PSAT–Level 1 test did not demonstrate
mood and cognitive symptoms at 1-hour
significant impairment post-meal; however,
post-meal. The correlation with physical
the more demanding Level 2 resulted in
symptoms was weaker. Table 2 presents the
approximately 30% fewer correct answers
correlation coefficients with BGRI of all
within 1 hour after a meal (from 2.4 to 1.6
postprandial symptom ratings and cognitive
correct answers). This effect vanished with-
performance results at 1, 2, and 3 hours
in 2 hours post-meal (F = 11.7, P<0.001)
(See Table 1, cognitive tests).
Taken by symptom category within the
Table 1 demonstrates that most symp-
first hour post-meal, mood and cognitive
toms achieved their highest magnitude
symptoms displayed high correlations with
within the first hour after meal, before the
BGRI (r = 0.70 and r = 0.74, respectively).
PPG reached its peak. Thus, symptom ele-
The correlation of the average magnitude of
vation was not clearly related to extreme
physical symptoms with BGRI was weaker
PPG. Indeed, average postprandial symp-
(r = 0.52), but still statistically significant
tom magnitudes and test results 1, 2, and 3
at P <0.01 (Table 2). For all symptoms
hours after a meal did not correlate signifi-
(individual and by category) the correla-

The Journal of Applied Research • Vol.3, No. 4, Fall 2003 455

KovatchevJARFall03 12/29/03 2:40 PM Page 456

tions with BGRI were highest within 1 hour design of the study was that the BG rate of
post-meal, and gradually decreased at 2 and increase could not be computed strictly for
3 hours post-meal. Since PPG increase pre-to-postprandial periods, that is, the
slowed down at 2 and 3 hours post-meal, slope of PPG increase could not be estimat-
and the PPG curve gradually flattened, this ed directly. This was due to the lack of suf-
was precisely the effect to be expected if ficient number of BG readings within a
symptoms were related to PPG increase and pre- and corresponding postprandial peri-
not to extreme PPG values. With this sam- od; at least two, pre-and postprandial, read-
ple size (N = 36) correlations above 0.47 ings on several days per subject are needed
yielded P-levels below 0.01 and were con- in order to estimate PPG slope.
sidered significant. Several correlation Instead, we confirmed a less specific
coefficients were above 0.6. For example, hypothesis: higher postprandial symptoms
the average magnitude of “difficulty speak- are related to higher overall BG rate of
ing” and “uncoordinated” 1 hour post-meal increase. The BGRI was based on a previ-
produced a correlation coefficient of 0.75- ously reported measure, BG rate of
0.76 with BGRI (see Table 2). change,22 but it took into account only con-
secutively increasing SMBG readings, not
DISCUSSION any two consecutive readings. As computed
This study used new monitoring tech- here, BGRI was not specific to postprandial
nology (HHC and custom software) to time periods; however, we assumed that it
assess in-the-field dynamics of postprandial was most influenced by the largest (per
BG and associated symptoms and cognitive hour) BG increases that subjects experi-
impairment in adults with type 2 diabetes enced post-meal. Indeed, it was illogical to
mellitus. Our data allowed for a reconstruc- expect that large and fast BG elevations
tion of PPG dynamics in our subjects’ natu- could have occurred with no relationship to
ral environment and confirmed laboratory a preceding significant carbohydrate intake.
observations and common knowledge that Thus, we could speculate that the process of
PPG is most elevated 2-3 hours after meal. 1 PPG elevation contributed to symptoms
In addition, we demonstrated that after more than the absolute values of postpran-
meals self-reported physical symptoms and dial hyperglycemia. While this speculation
moods became significantly elevated, and may be confirmed (or rejected) by future
that objectively determined cognitive slow- studies, here we were able to clearly
ing of approximately 30% was apparent. observe two related properties: (1) symp-
Surprisingly, however, symptom eleva- tom ratings were higher during the first
tion and cognitive slowing did not follow hour post-meal and this was the period of
the course of absolute PPG values. Instead, steepest slope of PPG increase, and (2)
for most symptoms highest symptom rat- postprandial symptoms were highly corre-
ings were observed within 1 hour after a lated with subject’s overall rate of BG
meal (Table 1), not when PPG was at its increase.
peak 2-3 hours post-meal. Overall, plotting In fact, the correlations between mood
a summary of all symptoms and PPG and cognitive symptoms and BGRI were
against the time elapsed after a meal very high, in some cases higher than 0.7
revealed that symptoms generally occurred (Table 2). Given that symptom ratings were
during the time of steepest slope of PPG derived from behavioral self-assessment,
increase (Figure 1). Thus, we hypothesized while BGRI was derived from concurrent
that postprandial symptoms occurred during but quite different sets of SMBG data
times when BG increase was fastest. With downloaded from the subjects’ glucometer
our data, however, we could not definitely memories, correlations of that magnitude
confirm this notion. One limitation in the imply very strong [linear] relationships.

456 Vol. 3, No. 4, Fall 2003 • The Journal of Applied Research

KovatchevJARFall03 12/29/03 2:40 PM Page 457

Thus, the magnitude and speed of BG 2. Wolever TMS, Palmason C, Chiasson J, et al.
Variation of postprandial plasma glucose, palata-
increase post-meal may be the single most bility, and symptoms associated with a standard-
significant determinant of postprandial ized mixed test meal versus 75 g oral glucose.
symptoms. Diabetes Care. 1998;21:336-340.
In contrast to symptoms, reduced per- 3. Hanefeld M. Postprandial hyperglycemia: nox-
ious effects on the vessel wall. Int J Clin Pract.
formance on cog n i t ive tests, although well 2002;129(suppl):45-50.
expressed in the first hour after meal (Table
4. Bonora E, Calcaterra F, Lombardi S, Bonfante N,
1, cog n i t ive tests), could not be explained by et al. Plasma glucose levels throughout the day
absolute PPG peak, or by BGRI. Thus, alter- and hba1c interrelationships in type 2 diabetes:
native mechanisms of postprandial cognitive implications for treatment and monitoring of
metabolic control. Diabetes Care. 2001;24:2023-
slowing must be considered. For example, 2029.
there may be a BG threshold, above which
5. Avignon A, Radauceanu A, Monnier L.
these effects occur in a stepwise fashion, or Nonfasting plasma glucose is a better marker of
some disruption of metabolic homeostasis diabetic control than fasting plasma glucose in
may be responsible for the observed cog n i- type 2 diabetes. Diabetes Care. 1997;20:1822-
tive slowing. Further studies will be needed
6. Caputo S, Pitocco D, Ruotolo V, Ghirlanda G.
to address these issues. What is the real contribution of fasting plasma
The results of this study imply that in glucose and postprandial glucose in predicting
people with T2DM the most effective time HbA1c and overall blood glucose control?
Diabetes Care. 2001;24:2011-2011.
to sample symptoms and cognitive func-
7. The Diabetes Control and Complications Trial
tioning postprandially is during the first 1-2 Research Group. The effect of intensive treat-
hours after a meal. This is also the time ment of diabetes on the development and pro-
when fastest BG elevation was observed in gression of long-term complications in
insulin-dependent diabetes mellitus. N Engl J
this study and documented by others.0 Since Med. 1993;329:977–986.
the BGRI was a better correlate of post-
8. UK Prospective Diabetes Study (UKPDS)
prandial symptoms than absolute PPG Group. Intensive blood-glucose control with
peaks, treatment regiments designed to sulphonylureas or insulin compared with conven-
reduce the magnitude and the speed of PPG tional treatment and risk of complications in
patients with type 2 diabetes (UKPDS 33).
fluctuations, such as rapid-acting insulin Lancet. 1998;352:837–853.
analogs or complex carbohydrate diets, may
9. Hanefeld M, Temelkova-Kurktschiev T. The post-
reduce the negative symptomatic and cogni- prandial state and the risk of atherosclerosis.
tive consequences of postprandial glucose Diabetic Med. 1997;14(suppl 3): S6-11.
excursions. 10. Soonthornpun S, Rattarasarn C, Leelawattana R,
Setasuban W. Postprandial plasma glucose: a
good index of glycemic control in type 2 diabetic
ACKNOWLEDGMENTS patients having near-normal fasting glucose lev-
The authors would like to thank els. Diabetes Res Clin Pract. 1999:46:23–27.
Pamela Erickson, Haya Ascher-Savanum, 11. Gavin JR III. The importance of postprandial
PhD, and James Malone, MD, for their hyperglycaemia. Int J Clin Pract. 1999;107
helpful reviews and suggestions on previous
drafts of this manuscript. 12. Hanefeld M. Post-prandial hyperglycaemia and
vascular disease. Int J Clin Pract. 2000;
13. Rajmohan L, Mohan V, Ramanujam TR.
1. American Diabetes Association. Postprandial Postprandial hyperglycaemia—the real challenge
Blood Glucose: Consensus Statement. Diabetes in diabetes. J Assoc Physicians India. 2001;
Care. 2001; 24:775-778.2. Wolever TMS, 49:357-360.
Palmason C, Chiasson J, et al. Variation of post-
prandial plasma glucose, palatability, and symp- 14. Haffner S. The importance of postprandial hyper-
toms associated with a standardized mixed test glycaemia in development of cardiovascular dis-
meal versus 75 g oral glucose. Diabetes Care. ease in people with diabetes. Int J Clin Pract.
1998;21:336-340. 2001;123(suppl):24-26.

The Journal of Applied Research • Vol. 3, No. 4, Fall 2003 457

KovatchevJARFall03 12/29/03 2:40 PM Page 458

15. Hanefeld M, Fisher S, Julius U. Risk Factor for 19. Kovatchev BP, Cox DJ, Gonder-Frederick LA, et
myocardial infarction and death in newly detect- al. Stochastic model of self-regulation decision
ed NIDDM: the Diabetes Intervention Study, 11- making exemplified by decisions concerning
year follow-up. Diabetologia. hypoglycemia. Health Psychol. 1998;17:277-284.
20. Cox DJ, Gonder-Frederick LA, Kovatchev BP, et
16. Van der Does FE, De Neeling JN, Snoek FJ, et al: Biopsychobehavioral model of severe hypo-
al. Symptoms and well-being in relation to glycemia. II. Understanding the risk of severe
glycemic control in type II diabetes. Diabetes hypoglycemia. Diabetes Care. 1999;22: 2018-
Care. 1996;19:204-210. 2025.
17. De Sonnaville JJ, Snoek FJ, Colly LP, et al. Well- 21. Kovatchev BP, Cox DJ, Robeva RS, et al.
being and symptoms in relation to insulin thera- Quantifying bio-behavioral determinants of risk
py in type 2 diabetes. Diabetes Care. for severe hypoglycemia in Type 1 diabetes. J
1998;21:919-924. Appl Res. 2001;1:16-23.
18. Cox DJ, Gonder-Frederick LA, McCall A, et al. 22. Kovatchev BP, Cox DJ, Gonder-Frederick LA,
The effects of glucose fluctuation on cognitive WL Clarke. Methods for quantifying self-moni-
function and QOL: the functional costs of hypo- toring blood glucose profiles exemplified by an
glycaemia and hyperglycaemia among adults examination of blood glucose patterns in patients
with type 1 or type 2 diabetes. Int J Clin Pract. with Type 1 and Type 2 Diabetes. Diabetes
2002;129(suppl):20-26. Technol Ther. 2002; 4:295-303.

458 Vol. 3, No. 4, Fall 2003 • The Journal of Applied Research